High Voltage Engineering PDF

IET Power and Energy Series 40
Advances in High
Voltage Engineering
Edited by A. Haddad and D. Warne
lEI P0WEk AN0 ENEk0 5EklE5 40
Series Editors: Prof. A.T. !ohns
D.F. Warne
Advances in High
voltage Engineering
0tber volumes in tbis series:
volume 1 Power circuit breaker tbeory and design C.H. Flurscheim (Editor)
volume 4 lndustrial microwave beating A.C. Metaxas and R.!. Meredith
volume 7 lnsulators for bigb voltages !.S.T. Looms
volume 8 Variable freguency AC motor drive systems D. Finney
volume 10 5F6 switcbgear H.M. Ryan and C.R. !ones
volume 11 Conduction and induction beating E.!. Davies
volume 13 5tatistical tecbnigues for bigb voltage engineering W. Hauschild and
W. Mosch
volume 14 Uninterrugtable gower sugglies !. Platts and !.D. St Aubyn (Editors)
volume 15 0igital grotection for gower systems A.T. !ohns and S.K. Salman
volume 16 Electricity economics and glanning T.W. errie
volume 18 Vacuum switcbgear A. Creenwood
volume 19 Electrical safety: a guide to causes and grevention of bazards
!. Maxwell Adams
volume 21 Electricity distribution network design, 2nd edition E. Lakervi and
E.!. Holmes
volume 22 Articial intelligence tecbnigues in gower systems K. Warwick, A.C. Ekwue
and R. Aggarwal (Editors)
volume 24 Power system commissioning and maintenance gractice K. Harker
volume 25 Engineers' bandbook of industrial microwave beating R.!. Meredith
volume 26 5mall electric motors H. Moczala
volume 27 AC-0C gower system analysis !. Arrill and .C. Smith
volume 29 Higb voltage direct current transmission, 2nd edition !. Arrillaga
volume 30 Flexible AC Iransmission 5ystems {FACI5) Y-H. Song (Editor)
volume 31 Embedded generation N. !enkins
volume 32 Higb voltage engineering and testing, 2nd edition H.M. Ryan (Editor)
volume 33 0vervoltage grotection of low-voltage systems, revised edition P. Hasse
volume 34 Ibe ligbtning asb v. Cooray
volume 35 Control tecbnigues drives and controls bandbook W. Drury (Editor)
volume 36 Voltage guality in electrical gower systems !. Schlabbach
volume 37 Electrical steels for rotating macbines P. eckley
volume 38 Ibe electric car: develogment and future of battery, bybrid and fuel-cell
cars M. Westbrook
volume 39 Power systems electromagnetic transients simulation !. Arrillaga and
N. Watson
volume 40 Advances in bigb voltage engineering M. Haddad and D. Warne
volume 41 Electrical ogeration of electrostatic grecigitators K. Parker
volume 43 Ibermal gower glant simulation and control D. Flynn
volume 44 Economic evaluation of grojects in tbe electricity suggly industry H. Khatib
volume 45 Progulsion systems for bybrid vebicles !. Miller
volume 46 0istribution switcbgear S. Stewart
volume 47 Protection of electricity distribution networks, 2nd edition !. Cers and
E. Holmes
volume 48 Wood gole overbead lines . Wareing
volume 49 Electric fuses, 3rd edition A. Wright and C. Newbery
volume 51 5bort circuit currents !. Schlabbach
volume 52 Nuclear gower !. Wood
volume 905 Power system grotection, 4 volumes
Advances in High
voltage Engineering
Edited by
A. Haddad and D.F. Warne
The Institution of Engineering and Technology
Published by The Institution of Engineering and Technology, London, United Kingdom
First edition 2004 The Institution of Electrical Engineers
Reprint with new cover 2007 The Institution of Engineering and Technology
First published 2004
Reprinted 2007
This publication is copyright under the erne Convention and the Universal Copyright
Convention. All rights reserved. Apart from any fair dealing for the purposes of research or
private study, or criticism or review, as permitted under the Copyright, Designs and Patents
Act, 1988, this publication may be reproduced, stored or transmitted, in any form or by
any means, only with the prior permission in writing of the publishers, or in the case of
reprographic reproduction in accordance with the terms of licences issued by the Copyright
Licensing Agency. Inquiries concerning reproduction outside those terms should be sent to
the publishers at the undermentioned address:
Michael Faraday House
Six Hills Way, Stevenage
Herts, SC1 2AY, United Kingdom
www.theiet.org
While the author and the publishers believe that the information and guidance given in this
work are correct, all parties must rely upon their own skill and judgement when making use
of them. Neither the author nor the publishers assume any liability to anyone for any loss or
damage caused by any error or omission in the work, whether such error or omission is the
result of negligence or any other cause. Any and all such liability is disclaimed.
The moral rights of the author to be identihed as author of this work have been asserted by
him in accordance with the Copyright, Designs and Patents Act 1988.
The author thanks the International Electrotechnical Commission (IEC) for permission to
reproduce information from its Technical Report IEC 60479-1. All such extracts are copyright
of IEC, Ceneva, Switzerland. All rights reserved. Further information on the IEC is available
from www.iec.ch. IEC has no responsibility for the placement and context in which the
extracts and contents are reproduced by IET; nor is IEC in any way responsible for the other
content or accuracy therein.
The following hgures from Chapter 8: 'variation of resistivity with salt (a), moisture (b)
and temperature (c), 'Diagonal voltage prohle for a 16-mesh square grid with 9 rods in a
two layer soil' and 'Potential distribution above an earth grid with a potential ramp and
without ramp' have been reprinted with permission from IEEE STD 81 1983, "IEEE Cuide for
Measuring Earth Resistivity, Cround Impedance and Earth Surface Potentials of a Cround
System" 1983, and IEEE STD 80 2000, "IEEE Cuide for Safety in AC Substation Crounding",
by IEEE. The IEEE disclaims any responsibility or liability resulting from the placement and
use in the described manner.
8ritisb Library Cataloguing in Publication 0ata
Advances in high voltage engineering. - (Power & energy series)
1. High voltages
I. Haddad, Manu II. Warne, D. F. III. Institution of Electrical Engineers
621.3'1913
l58N {10 digit) 0 85296 158 8
l58N {13 digit) 978-0-85296-158-2
Typeset in India by Newgen Imaging Systems (P) Ltd, Chennai, India
Printed in the UK by MPC ooks Ltd, odmin, Cornwall
Reprinted in the UK by Lightning Source UK Ltd, Milton Keynes
Contents
Contributors xvii
Introduction xix
1 Mechanisms of air breakdown 1
N.L. Allen
1.1 Introduction 1
1.1.1 Beginnings 1
1.1.2 Basic breakdown processes 2
1.2 Physical mechanisms 3
1.2.1 Avalanche development 4
1.2.2 Avalanche properties 6
1.2.3 The critical avalanche and the critical volume 8
1.2.4 Streamer formation 10
1.2.5 Streamer development 12
1.2.6 Corona 13
1.2.7 The streamer trail 14
1.2.8 The leader 15
1.2.9 Negative discharges 20
1.3 Applications 21
1.3.1 Sparkover under lightning impulse voltage 22
1.3.2 Sparkover under slow front impulse voltage 24
1.3.3 The influence of field configurations: the gap factor 25
1.3.4 Atmospheric effects 28
1.3.5 Corona at low air density 30
1.3.6 Sparkover over insulator surfaces 31
1.4 Note 32
1.5 References 32
2 SF
6
insulation systems and their monitoring 37
O. Farish, M.D. Judd, B.F. Hampton and J.S. Pearson
2.1 Introduction 37
2.2 Ionisation phenomena in SF
6
38
vi Contents
2.3 Breakdown mechanisms in low divergence fields 40
2.3.1 Streamer breakdown 41
2.3.2 Quasi-uniform fields (coaxial cylinders) 42
2.3.3 Effect of surface roughness 43
2.4 Non-uniform field breakdown in SF
6
45
2.4.1 Corona stabilised breakdown 46
2.4.2 Leader breakdown 47
2.5 Breakdown in GIS 48
2.5.1 Streamer-controlled breakdown 48
2.5.2 Leader breakdown 49
2.5.3 Particle-initiated breakdown 50
2.6 Possible improvements in SF
6
insulation 51
2.6.1 Use of additives or gas mixtures 51
2.6.2 Improved spacer formulation and construction 51
2.6.3 Particle control 52
2.7 Partial discharge diagnostic techniques for GIS 52
2.7.1 Introduction 52
2.7.2 The range of diagnostic techniques for PD detection 54
2.7.3 Comparison of the techniques 56
2.7.4 Overview of UHF technology 58
2.8 The generation and transmission of UHF signals in GIS 59
2.8.1 Introduction to UHF theory 59
2.8.2 Excitation 59
2.8.3 Propagation 60
2.8.4 Extraction 60
2.8.5 Waveguide modes and UHF propagation 61
2.8.6 Attenuation of UHF signals 64
2.9 Application of UHF technique to PD detection in GIS 65
2.9.1 Design and testing of UHF couplers 65
2.9.2 Design of a PDM system for GIS 69
2.9.3 Display and interpretation of PD data 70
2.9.4 AI diagnostic techniques 72
2.9.5 Service experience 73
2.10 References 74
3 Lightning phenomena and protection systems 77
R.T. Waters
3.1 From Franklin to Schonland 77
3.2 Phenomenology of lightning 79
3.2.1 Characterisation of the flash 79
3.2.2 Incidence 79
3.2.3 Polarity 82
3.2.4 Flash components 82
3.2.5 Peak current 83
Contents vii
3.2.6 Current shape 84
3.2.7 Electric fields 85
3.2.8 Spatial development 87
3.3 Physics of lightning 88
3.3.1 Long sparks in the laboratory 88
3.3.2 Lightning leader propagation 97
3.4 Lightning termination at ground 100
3.4.1 Striking distance 100
3.4.2 Geometric models and lightning standards 101
3.4.3 Electrogeometric models 102
3.4.4 Generic models 107
3.4.5 Positive lightning 114
3.5 Risk factors and protection 116
3.5.1 Risk assessment 116
3.5.2 Standard procedure for the calculation of
risk factor 118
3.5.3 Electrogeometric calculation of risk factor 119
3.5.4 Generic modelling of risk factor for
a negative flash 120
3.5.5 Protection of overhead power lines 125
3.5.6 Protection of electronic equipment 126
3.5.7 Strikes to aircraft and space vehicles 128
3.6 Note 130
3.7 References 130
4 Partial discharges and their measurement 139
I.J. Kemp
4.1 Introduction 139
4.2 Partial discharge degradation mechanisms 139
4.2.1 Particle impact stress 140
4.2.2 Thermal stress 143
4.2.3 Mechanical stress 144
4.2.4 Chemical stress 145
4.2.5 Electrical stress 149
4.2.6 Synergetic interaction of stresses 149
4.3 Partial discharge measurement 150
4.3.1 Electrical detection 150
4.3.2 Acoustic detection 167
4.3.3 Thermography and other camera techniques 170
4.3.4 Chemical detection 170
4.3.5 Comparison among different PD measurement
techniques relative to type of plant under
investigation 177
4.3.6 Other items of plant 179
4.4 Concluding remarks 183
viii Contents
4.5 Note 184
4.6 References 185
5 ZnO surge arresters 191
A. Haddad
5.2 Evolution of overvoltage protection practice 192
5.2.1 Simple spark gaps 192
5.2.2 Valve-type arresters 192
5.2.3 Surge arresters with active gaps 193
5.2.4 Metal oxide surge arresters 193
5.2.5 Existing applications of ZnO surge arresters 194
5.3 Basic properties of ZnO material 195
5.3.1 Composition and effect of additives 195
5.3.2 Fabrication process 195
5.3.3 Microstructure 197
5.3.4 Conduction mechanism in ZnO varistors 199
5.4 Thermal performance of ZnO surge arresters 201
5.4.1 Background 201
5.4.2 Heat dissipation capability and thermal stability
of ZnO surge arresters 201
5.4.3 Thermal runaway 204
5.4.4 Thermal runaway critical condition 205
5.4.5 Dynamic stability of ZnO surge arresters 206
5.4.6 Simulation of thermal characteristics of ZnO surge
arresters 206
5.5 Degradation and ageing of ZnO surge arresters 207
5.5.1 Differences between degradation and thermal
runaway 207
5.5.2 Factors affecting rate of degradation 208
5.5.3 Destruction mechanism 210
5.6 Life estimation of ZnO surge arresters 212
5.6.1 Long term accelerated ageing tests 212
5.6.2 DakinArrhenius plots of life span 214
5.6.3 Alternative methods of life estimation 215
5.7 Test procedures for the characterisation of ZnO arrester 215
5.7.1 Prebreakdown regime of conduction: AC and
DC tests 215
5.7.2 Breakdown regime of conduction and up-turn
region: impulse tests 216
5.7.3 Voltage distribution along arrester columns 217
5.8 Characteristics of ZnO surge arresters 221
5.8.2 Frequency response of ZnO material 221
5.8.3 Impulse response 224
Contents ix
5.8.4 Combined stress response 225
5.8.5 Equivalent circuit of ZnO material 230
5.9 Monitoring of ZnO surge arresters 233
5.10 Standards and application guidelines 234
5.10.1 Standard definitions of important parameters 234
5.10.2 Classification of ZnO surge arresters 236
5.10.3 Other important arrester characteristics 237
5.10.4 Standard tests 238
5.10.5 Recommended arrester identification 238
5.11 Selection of gapless metal oxide surge arresters 239
5.12 Location and protective distance of surge arresters 242
5.12.1 Effect of distance on protective level 242
5.12.2 Calculation of separation distance 242
5.12.3 Calculation of arrester protective zones 243
5.13 Note 244
5.14 References 244
6 Insulators for outdoor applications 257
D.A. Swift
6.2 Role of insulators 258
6.3 Material properties 259
6.4 Examples of design 260
6.4.1 Cap and pin insulators 261
6.4.2 Longrods 268
6.4.3 Posts 268
6.4.4 Barrels 271
6.5 Flashover mechanisms 272
6.5.1 Surface wettability 272
6.5.2 Hydrophilic case 274
6.5.3 Hydrophobic case 278
6.5.4 Ice and snow conditions 278
6.6 Electrical characteristics 279
6.6.1 Performance under natural pollution 280
6.6.2 Performance under artificial pollution 283
6.6.3 Interrupter head porcelains 288
6.6.4 AC versus DC 289
6.6.5 Transient overvoltages 289
6.6.6 Iced insulator 291
6.6.7 Snow on insulators 292
6.7 Selection and dimensioning 292
6.8 Supplements, palliatives and other mitigating measures 296
6.8.1 Booster sheds 296
x Contents
6.8.2 Shed extender 297
6.8.3 Shed protector 298
6.8.4 Coatings 298
6.8.5 Washing 300
6.8.6 Provisions for ice and snow 300
6.9 Miscellaneous 300
6.9.1 Cold switch-on and thermal lag 300
6.9.2 Semiconducting glaze 301
6.9.3 Live line working 301
6.9.4 Visual annoyance, audible noise and electromagnetic
compatibility 302
6.9.5 Electric field distributions 302
6.9.6 Interphase spacers 303
6.9.7 Compact and low profile lines 303
6.9.8 Financial and related matters 303
6.10 References 304
7 Overvoltages and insulation coordination on transmission
networks 309
D.M. German and A. Haddad
7.2 System overvoltages 311
7.2.1 External overvoltages 312
7.2.2 Internal overvoltages 314
7.3 Network simulation and analysis 318
7.3.1 Transmission lines 318
7.3.2 Cables 326
7.3.3 Circuit breakers 326
7.3.4 Transformers 327
7.3.5 Network reduction 327
7.4 Computed switching overvoltages 327
7.5 Insulation coordination 333
7.5.1 Analytical expressions for F(U) and P(U) 336
7.5.2 Risk of failure 338
7.5.3 Simplified method 338
7.5.4 Withstand voltage 339
7.6 Compact transmission lines 340
7.6.1 Insulation 341
7.6.2 Surge arresters 341
7.6.3 Comparison between compact and conventional
network 341
7.7 Acknowledgement 344
7.8 Note 344
7.9 References 345
Contents xi
8 Earthing 349
H. Griffiths and N. Pilling
8.2 Earthing system components and system earthing methods 351
8.2.1 Transmission system 351
8.2.2 Distribution system 352
8.2.3 Methods of system earthingtreatment
of neutral 353
8.2.4 Application of different system earthing
methods 355
8.3 Earth resistivity and measurement techniques 355
8.3.1 Conduction mechanisms and resistivity 355
8.3.2 Resistivity data of soils and rocks 355
8.3.3 Site investigation and measurement techniques
of earth resistivity and structure 357
8.4 Power frequency performance of earthing systems 365
8.4.1 Standards recommendations 365
8.4.2 Earth impedance 367
8.4.3 Interactions between fault currents and earthing
systems 370
8.4.4 Measurement of earth impedance and potentials 372
8.4.5 Maintenance and integrity testing of earthing
systems 377
8.4.6 Special installations 379
8.5 Electrocution hazards and safety issues 380
8.5.1 Step and touch potentials 380
8.5.2 Computation of tolerable voltages 384
8.5.3 Methods for limiting hazardous potential differences
and dimensioning of earthing systems 389
8.5.4 Risk management approach to earthing safety 391
8.6 Impulse performance of earthing systems 394
8.6.1 Standard guidelines for transient earthing 394
8.6.2 Soil ionisation 395
8.6.3 Models of concentrated earth electrodes exhibiting
soil ionisation 398
8.6.4 Models of earthing systems under high frequency and
transient conditions 398
8.6.5 Simulations of earthing system performance under
transient and high frequency conditions 400
8.7 References 402
9 Circuit breakers and interruption 415
H.M. Ryan
xii Contents
9.2 Circuit interruption characteristics, arc control and
extinction 422
9.2.1 Principles of current interruption in HV systems 427
9.3 Distribution switchgear systems 428
9.4 Substation layouts and control aspects 429
9.4.1 Substation layouts 429
9.4.2 Intelligent networks 434
9.4.3 Need for more intelligence 441
9.4.4 Network protection and future developments 441
9.5 Substation control in the system control
(CIGRE. WG.39.01) 441
9.6 Planning specification and testing of controlled HVAC
switching systems 445
9.6.2 Specification of controlled switching installations 446
9.6.3 Concluding remarks 447
9.7 Dielectric and global warming considerations 448
9.8 Some examples of modern switchgear 448
9.8.1 SF
6
live-tank and dead-tank switchgear 448
9.9 Equipment life expectancy: condition monitoring
strategies 451
9.9.1 Evaluation of solid/gaseous dielectric systems
for use in HV switchgear 451
9.9.2 Open market: revised optimal network structure 454
9.9.3 Condition monitoring strategies 456
9.9.4 General discussion 470
9.10 Summary 473
9.11 Acknowledgements 474
9.12 References 474
10 Polymer insulated power cable 477
J.C. Fothergill and R.N. Hampton
10.1.1 Structure 477
10.1.2 Voltage ratings 477
10.1.3 Uses of cables 479
10.1.4 AC and DC 479
10.1.5 Cable types 480
10.2 The components of the polymeric cable 481
10.2.1 Conductor 481
10.2.2 Semicon 484
10.2.3 Insulation 484
10.2.4 Metal sheath 487
10.2.5 Oversheath ( jacket) 488
Contents xiii
10.3 Cable manufacture 488
10.3.1 Stages of cable manufacture 488
10.3.2 Methods of core manufacture 489
10.4 Failure processes 491
10.4.1 Extrinsic defects 492
10.4.2 Wet ageing water trees 496
10.4.3 Dry ageing thermoelectric ageing 498
10.5 Mathematical design models for cables 500
10.6 Direct current transmission 503
10.6.1 Economics 503
10.7 Testing 505
10.7.1 Prequalification and development tests 505
10.7.2 Type approval testing 505
10.7.3 Sample testing 506
10.7.4 Routine testing 506
10.7.5 Future trends in testing 506
10.8 Notes 507
10.9 References 507
11 Numerical analysis of electrical fields in high
voltage equipment 511
A.E. Baker, T.W. Preston and J.P. Sturgess
11.2 Which numerical method? 512
11.2.1 The finite-difference method 512
11.2.2 The finite-element method 513
11.2.3 The boundary-element method 513
11.2.4 Comparative summary 515
11.3 Formulation of the finite-element equations in two and three
dimensions 515
11.3.1 General 515
11.3.2 Forming the functional equation 517
11.3.3 The energy functional illustrated 517
11.3.4 Numerical representation 518
11.4 Variations on the basic formulation 519
11.4.1 General 519
11.4.2 Representation of foils 519
11.4.3 Directional permittivity 521
11.4.4 Modelling resistive/capacitive systems 521
11.4.5 Modelling partially conducting tapes and paints 522
11.4.6 Space charge modelling 523
11.4.7 Time variation 524
11.4.8 Open boundary problems 524
11.5 Applications 525
11.5.1 General 525
xiv Contents
11.5.2 High voltage transmission line 525
11.5.3 Foils in high voltage bushings 525
11.5.4 Modelling the effect of contamination on an
insulating system 527
11.5.5 Stress grading of high voltage windings 530
11.6 The choice of the order of the finite-element approximation 536
11.6.1 General 536
11.6.2 First-order elements 538
11.6.3 Higher-order elements 538
11.7 Assessment of electrical stress distribution 539
11.7.1 General 539
11.7.2 Mathematical singularities 539
11.7.3 Relationship between stress and breakdown 540
11.8 Pre and post processor developments 540
11.8.1 General 540
11.8.2 Description of the problem geometry 540
11.8.3 Creation of a discretisation from the problem
geometry 541
11.8.4 Assigning material properties 541
11.8.5 Post processor developments 541
11.8.6 Design optimisation 542
11.9 Notes 542
11.10 References 542
12 Optical measurements and monitoring in high voltage
environments 545
G.R. Jones, J.W. Spencer and J. Yan
12.2 Fundamental optical principles 547
12.2.2 Optical intensity 547
12.2.3 Spectroscopy 547
12.2.4 Light scattering 551
12.2.5 Optical fibre propagation 557
12.3 Optical equipment and systems 560
12.3.1 High speed imaging 560
12.3.2 Spectrometer systems 561
12.3.3 Light scattering systems 564
12.3.4 Optical fibre sensing systems 566
12.4 Examples of test results 575
12.4.1 High speed photography 575
12.4.2 Spectroscopic results 575
12.4.3 Coherent scattering results 577
12.4.4 Incoherent scattering results 578
Contents xv
12.4.5 Optical fibre transducer results 580
12.4.6 Time and wavelength response of optical fibre and
free space techniques 585
12.5 Conclusions 585
12.6 Acknowledgements 587
12.7 Note 587
12.8 References 588
13 Pulsed power principles and applications 591
J.E. Dolan
13.2 Pulsers and topologies 594
13.2.1 Capacitive discharge 595
13.2.2 Charging supplies 595
13.2.3 Capacitors 596
13.2.4 Voltage multiplication: the Marx bank 597
13.2.5 Compact, fast rise time Marx banks 598
13.2.6 Pulse compression 598
13.2.7 The Melville line magnetic pulse
compressor 599
13.2.8 Transmission line circuits 600
13.2.9 Charge lines 600
13.2.10 The Blumlein circuit 602
13.2.11 Inductive voltage adders 602
13.2.12 Inductive energy storage 604
13.2.13 The Tesla transformer 605
13.3 Semiconductor switching 605
13.3.2 Thyristor 606
13.3.3 Bipolar transistor avalanche mode switching 607
13.3.4 MOSFET Marx 608
13.3.5 MOSFET switching stacks 608
13.3.6 MOSFETs with inductive coupling 609
13.3.7 General semiconductor switching design issues 610
13.3.8 Novel semiconductor devices 610
13.3.9 Applications of novel semiconductors 611
13.3.10 Electro-optic switching 612
13.3.11 Conclusions on semiconductor switching 612
13.4 Non-linear transmission lines 613
13.5 Pulsed power applications 615
13.5.2 Ion beam materials treatment 616
13.5.3 Air treatment and pollution control 616
13.5.4 Pulsed corona precipitators 617
13.5.5 Biological applications 617
xvi Contents
13.5.6 Biofouling and ballast water treatment 618
13.5.7 Food processing 618
13.5.8 Water purification 619
13.5.9 Mechanical applications of spark discharges 619
13.5.10 Medical applications 620
13.5.11 Ultrawideband and HPM applications 621
13.5.12 X-ray simulators 622
13.6 Conclusions 622
13.7 References 623
Index 633
Contributors
N.L. Allen
Department of Electrical Engineering
and Electronics, UMIST,
Manchester, England, UK
A.E. Baker
AREVAT&D Technology Centre,
Stafford, England, UK
J.E. Dolan
Optics & Laser Technology Department
Advanced Technology Centre Sowerby
BAE SYSTEMS
P.O. Box 5, FPC 30 Filton
Bristol, UK
O. Farish
Institute for Energy and Environment,
Strathclyde University, Glasgow,
Scotland, UK
J.C. Fothergill
Department of Engineering,
University of Leicester,
Leicester, England, UK
D.M. German
Cardiff University, Cardiff, Wales, UK
H. Griffiths
A. Haddad
B.F. Hampton
Diagnostic Monitoring Systems Ltd,
Glasgow, Scotland, UK
R.N. Hampton
Borealis AB
Stenungsund,
Sweden, 444 86
G.R. Jones
Centre for Intelligent Monitoring
Systems, Department of Electrical
Engineering and Electronics,
University of Liverpool, England, UK
M.D. Judd
Institute for Energy and Environment,
Strathclyde University, Glasgow,
Scotland, UK
I.J. Kemp
Glasgow Caledonian University,
J.S. Pearson
Diagnostic Monitoring Systems Ltd,
N. Pilling
xviii Contributors
T.W. Preston
H.M. Ryan
McLaren Consulting and University of
Sunderland, England, UK
J.W. Spencer
University of Liverpool, England, UK
J.P. Sturgess
AREVAT&D Technology Centre,
Stafford, England, UK
D.A. Swift
Cardiff University, Cardiff,
Wales, UK
R.T. Waters
Cardiff University, Cardiff,
Wales, UK
J. Yan
University of Liverpool,
England, UK
Introduction
In 1984 a book entitled Les proprits dilectriques de lair et les trs hautes tensions
was published. It represented a collection of research studies undertaken at EdF and
other major world utilities over a significant period, and it has been widely used by
manufacturers and users of high voltage equipment and systems, and by academic
groups working in the area.
Undeniably, there have been a large number of developments in the high voltage
field over the intervening years, but no such reference work has since been produced.
There have been a number of key advances in materials. Polymeric insulators
are still under major trials for transmission voltages but have become widely used
in overhead lines at distribution levels. Cable insulation has moved from paper sys-
tems to various polymer-based materials. SF
6
has taken over as an efficient and
reliable insulation medium in high-voltage switchgear, and gas-insulated substations
are now frequently the preferred option for new substations especially in urban areas.
All of these changes have not been direct substitutions; they have each required
a reconsideration of equipment configuration, and new installation and operational
techniques.
ZnO surge arresters are also widely used worldwide. Their excellent overvoltage
protection characteristics have allowed design of modern compact systems to become
more reliably achievable. The propagation of lightning and fault currents through
earthing arrangements can now be modelled and predicted in much more reliable
fashion. Much of this improvement in understanding has come from better use of
numerical methods such as boundary elements, finite element modelling and others,
which has in turn put pressure on the experimental derivation of key material and
system properties. The understanding of basic physical processes has also developed.
The nature of breakdown in air and in other materials is nowbetter characterized than
it was twenty years ago.
Improvements have been made in instrumentation and experimental techniques.
Who would have predicted thirty years ago, for instance, that it would nowbe possible
to measure the distribution of space charge in solid insulation? So, the case seemed
compelling for a new book reviewing, once again, the advances that have been made
in high voltage engineering.
xx Introduction
This project has had some sense of urgency because of the demographic trends in
the research field. The age profile of researchers and experts in the field is disturbingly
skewed, a number of experts having retired in the past few years or being about to
retire. In many of the key areas of research and development, there is no obvious
succession of expertise and a lack of any ongoing group within which the accumulated
knowledge might be stored, protected and further developed. It was important then
to encapsulate the accumulated wisdom and experience resident in these experts.
It is hoped, therefore, that this work on the one hand meets a need to update
the advances that have undoubtedly been made in high voltage engineering during
the past twenty years and, on the other hand, captures succinctly an accumulated
knowledge which might otherwise be rather difficult to unearth through individual
papers dispersed through the literature.
Edited by
A. Haddad and D.F. Warne
Chapter 1
Mechanisms of air breakdown
N.L. Allen
1.1 Introduction
1.1.1 Beginnings
Studies of air breakdownbeganinthe eighteenthcentury. Twonames are pre-eminent:
Franklin [1] and Lichtenberg [2], although contemporaries were active. Franklins
work grew out of his interest in lightning a long spark while Lichtenberg drew
tree-like discharges, now called corona, across the surface of a large cake of resin.
These two men defined two broad approaches to the study of breakdown which are
perpetuated to this day in experimental and theoretical work. In the late 19th century,
the emergence of modern physics, exemplified by the work of Townsend [3] and his
successors, permitted knowledge of the process of ionisation to be applied to these
phenomena. The two approaches were thus linked and another concept from the 18th
century, the electric field, became established as paramount in all discussions of the
subject.
Indeed, many of the quantities used in discussion of the processes in discharge
physics, such as ionisation and attachment coefficients, electron and ion temperatures,
diffusioncoefficients andsoon, have beenmeasuredandare quotedinterms of electric
field. Usually, it is assumed that the electric fields being considered are uniform.
In fact, this is very rarely true in practice and the use of these quantities is always
subject to such modifications as are dictated by non-uniformelectric field conditions.
This was true in the times of Franklin and Lichtenberg, and it will be assumed in most
of the discussion in this chapter since practical engineers rarely enjoy the luxury of
simple, uniform field configurations.
The concept of breakdown will be assumed to signify the collapse of the dielectric
strength of air between two electrodes, which is in practice defined by the collapse
of the voltage that had previously been sustained between them.
2 Advances in high voltage engineering
1.1.2 Basic breakdown processes
Over many decades, research has identified concepts which contribute to the
formation of a basic picture of breakdown in air.
1.1.2.1 Primary electrons
Free electrons exist only for very short times in air that is not subject to a high
electric field; normally they are trapped, after creation by cosmic rays, background
radiation and so on, to form negative ions. These have a density commonly of the
order of a fewhundredper cubic centimetre. However, electrons canbe detachedagain
from negative ions by acceleration and resulting collisions with neutral molecules in
a strong electric field.
1.1.2.2 Ionisation
The electrons so liberated can themselves accelerate in the field, collide with neutral
molecules and settle down to a constant average drift velocity in the direction of
the field. When sufficiently energetic, the collisions may liberate a further electron,
so leaving behind a positive ion. The process is cumulative, quantified initially by
Townsend [3], and resulting in the formation of avalanches of electrons. The growth
in numbers of electrons and positive ions imparts a small conductivity to the air,
which does not lead immediately to a breakdown, that is, to a collapse of voltage.
1.1.2.3 Excitation
Where electrons are sufficiently energetic to cause ionisation, there is usually a plen-
tiful supply with lower energies that can excite neutral atoms without liberating
electrons. When returning to the ground state, these atoms emit radiation as visible
or ultra-violet light. This property is widely used in research to indicate the presence
of ionisation.
1.1.2.4 Other electron processes
The electrons created by the growth of ionisation may be trapped, as described above,
and so removed from the ionisation process. This is the attachment process; a net
growth of electron and ion population occurs only when the field is sufficiently high
for the rate of ionisation to exceed the rate of attachment. Subsequent detachment of
electrons fromnegative ions occurs at the same time, through collisions with neutrals,
with free electrons or by interaction with photons. Recombination between electrons
and positive ions and between positive and negative ions is a further element in the
competing processes that are active in an ionised gas.
1.1.2.5 Regeneration
Initially, Townsend postulated that the positive ions could also ionise, a process now
recognised as insignificant. Also, that they move towards the negative electrode to
release further electrons by secondary emission, so that the ionisation process could be
sustained and grow indefinitely until breakdown occurred. Experiment later showed
Mechanisms of air breakdown 3
that breakdown could occur much more quickly than this process would allow. The
solution lay in postulating that the positive ions, created by ionisation, are sufficient
to create an electric field which, when added to the applied field, intensifies the
ionisation process [4, 5]. Additional initiatory electrons are assumed to be created
by ultra-violet radiation from the excited molecules in the electron avalanches in
which ionisation takes place [6]. They will also be created by photo-emission from
the negative electrode. In a sufficiently intense field, these events are cumulative and
can occur very rapidly [7, 8]. The current density rises, heating the gas and reducing
its density, leading to a rapid increase in energy input and conductivity. This results
in a discharge of very low impedance and causes voltage collapse.
1.1.2.6 Reduced electric field
Common sense suggests that the above processes, which all depend on an applied
electric field, are determined by the energy that electrons and ions acquire between
collisions. Thus the ratio E/N of electric field EV/cmto the gas density N mols/cm
3
,
known as the reduced field, is now widely used as the reference variable when mea-
suring values of fundamental quantities. The unit of this ratio is the Townsend (Td),
which has the numerical value 10
17
Vcm
2
. Older work used the equivalent ratio
E/p where E was in V/cm and p in torr.
They are related as follows:
E
N
Td = 3.03
E
p
Vcm
1
torr
1
when temperature is not a variable. It will be noted that custom has determined that
c.g.s. units are still used for these quantities.
The breakdown mechanisms are now examined in more detail. In most of what
follows, the discussion will assume that the non-uniform field occurs at a positive
polarity electrode (i.e. diverging lines of force), with only a brief description of ion-
isation processes at a negative (converging field) electrode. The reason is that the
processes in a diverging field lead more readily to breakdown than those in a con-
verging field, so that, in engineering practice, the dielectric strength of a gap is lower
when the more sharply radiused electrode is positive rather than negative. Thus, for
example, positive surge voltages are frequently more dangerous to a power system
than negative surge voltages.
1.2 Physical mechanisms
Discussion will first be general, in which physical processes are described in relation
to the electric field or E/N value which sustains them. Later, differences will be
discussed when the ionisation growth originates in the field at either a positive or
a negative electrode.
1.2.1 Avalanche development
Analytic treatment of ionisation by collision assumes a continuous process in which
the number of electronion pairs created by an electron of a given average energy
in a given electric field is proportional to the distance that it travels in that field.
The number of new electrons dn created in distance dx is thus:
dn = dx (1.1)
where is a proportionality constant which is the number of electronion pairs created
by the electron per unit of distance in the direction of the field. Including now the
electron progeny created by the first pair, all of which are drifting in the field at the
same average rate and with the same ionising efficiency, then at any point x along its
path some number n of electrons will enter an element of length dx and the number
of new electrons created is:
dn = n dx (1.2)
Over a distance d, starting with one electron at the origin, the total number N of ion
pairs created becomes
N
a
= exp d (1.3)
Here N
a
is the avalanche number. Whereas integration of Equation (1.1) shows that
the originating single electron produces a number of ion pairs which increases linearly
with its path length in the field, and which is always a relatively small number, use
of Equation (1.2) shows that when the effect of its progeny is taken into account, the
number of ion pairs increases to a very large number.
The quantity embraces complex physical processes which include several types
of collision, including electronic excitations of neutral molecules and the subsequent
production of radiation which may itself aid the ionisation process. These in turn
depend on the electric field in relation to the air density or pressure, E/N or E/p,
for this is an indicator of the energy gained by electrons between collisions with
gas molecules. Indeed, it determines the function governing the distribution of the
energies of the electrons in their gas-kinetic-type drift motion in the electric field.
In round terms, the average energy required for an electron to ionise is of the order of
20 electron volts (eV). At a given instant, the electrons in the distribution have only
a very small probability of gaining sufficient energy to ionise. The mean energy is
in fact much less than 20 eV in the typical case of the field required for breakdown,
referred to above: here, a value in the order of 2 eV has been determined from the
small amount of data that is available [9, 10].
Some gases, of which the most important in electrical engineering are air and SF
6
,
exhibit a strong affinity for electrons. This is the process of attachment to a neutral
molecule toforma negative ion. It is describedbyanattachment coefficient, , defined
as the number of attachments per unit length per electron moving in the direction of the
electric field. This is analogous in form to the definition of the ionisation coefficient,
, describing the rate of loss of free electrons per unit length, rather than the rate of
increase.
Where attachment is significant, the reverse process of detachment can also occur.
Two mechanisms are possible:
(i) by collisions with neutral molecules in a high electric field
(ii) by interaction with radiation.
Process (i) [11] is the mechanism by which it is assumed that primary free electrons
are liberated at the outset of the ionisation process (section 1.1.2.1). Since the time
betweencollisions is of the order of a nanosecond, sucha mechanismis able toproduce
free electrons either for the initiation of avalanches or to restore to an avalanche some
of the electrons that have been trapped by attachment. Adiscussion of the process is
given in References 10 and 11.
Photodetachment, mechanism(ii), can take place only when avalanches are form-
ing, as a result of the excitations occurringat the same time as the ionisation. It has been
proposed by Boylett and Williams [12] as a possible mechanism in the propagation
of a corona discharge, to be discussed later in this chapter.
Where the processes of ionisation, attachment and detachment exist together, the
basic ionisation growth Equation (1.1) becomes:
dn = ( +)n dx (1.4)
where is the detachment coefficient, so that:
N
a
= exp( +)d (1.5)
Data on the values of ( + )/N in air, as a function of E/N, is available in
References 9 to 11, 13 and 14.
As an example of the use of Equations (1.4) and (1.5), we take a value of electric
field needed to break down air at normal temperature and pressure. This is in the order
of 3 MV m
1
, where E/N is about 121 Td and E/p is about 40 Vcm
1
torr
1
. Here,
the value of ( + ) is about 1800 per metre. Thus, in traversing a gap between
electrodes of 10
2
metre, one electron creates about 18 ion pairs (Equation (1.4)).
This may be compared with the total number of collisions made by the electron in
crossing the gap, which is in the order of 10
5
. However, Equation (1.5) shows that
when the similar ionising power of the initial electrons progeny is taken into account,
the total number of ion pairs created is about 10
8
. Strictly, the number of positive
ions created is exp( +)d 1, since the integration takes into account the fact
that one electron exists before any ionising collision has occurred.
In this example, where the net value of ( +) is about 1800 per electron per
metre, the value of for the same condition of E/N is about 1000 per electron per
metre [10, 15]. The effect of electronegative attaching molecules in the gas is therefore
considerable. Atmospheric air contains two elements that are electronegative, namely
oxygen and water vapour. In the latter case, it is the complex O
2
(H
2
O)
n
where
1 < n < 5 that is most active in attaching an electron. The affinity is of the order
of 1 eV and the energy given up in the attachment process is released as radiation,
given to a third body as kinetic energy, or produces dissociation of the host molecule.
The effective reduction in the ionisation coefficient caused by attachment means that
an electronegative gas tends to have a higher breakdown strength than one which
does not show this property. The concentration of water vapour in air has significant
consequences on the formation of corona and the sparkover stress, as will be discussed
later.
1.2.2 Avalanche properties
In an electric field, electrons move at much higher velocities than ions, so that in the
ionisation process outlined in section 1.2.1, the positive ions can, on the timescale
of avalanche formation, be regarded as remaining stationary. As the concentration
of positive ions increases, so the effect of the electric field due to their own space
charge increases. The electrons at the head of a developing avalanche thus experience
a reduced net field, being the vector sum of the applied and the space charge fields.
Likewise, the resultant field is increased behind the avalanche head.
Some important avalanche properties are determined, however, by the diffusion
of the electrons during their ionising progress. For many purposes, it is a satisfactory
approximation to picture the electron cloud, while growing in numbers according to
Equation (1.5), as having diffused to the same extent as the same number of electrons
starting at the same time as the single electron that initiates the avalanche. Then,
diffusion theory states that the mean square radius of the cloud, assumed spherical,
at time t is given by:
r
2
= 6Dt (1.6)
where D is the coefficient of free diffusion of the electrons.
In time t , the centre of the electron cloud has progressed a distance x, so that:
x = vt (1.7)
where v is the mean drift velocity of the centre of the cloud.
Then:
r
2
=
6Dx
v
(1.8)
and the volume traced out by avalanche growth is approximately a paraboloid.
As the avalanche develops, however, effects of the field due to the positive ion
space charge become important. In the axial direction, the reduced field at the head
of the avalanche and the enhanced field behind (Figure 1.1) affect the symmetry of
the diffusion process that has hitherto been assumed. Estimation of these effects is
difficult in any simple picture and a proper treatment requires computation using
the continuity equations for both the electron and positive ion components. Lateral
diffusion of the cloud is also affected, but here a simple model can utilise the fact that
lateral diffusion of electrons will be reduced to the level of that of the positive ions
when the potential energy of the electrons in the field of the ions is of the same order
as their random kinetic energy. This equality is formalised, in the case of a plasma
of roughly equal densities of electrons and ions, by the expression for the Debye
limit of free
diffusion
E
E
s
E
s
E
s
E
s
+
r
+ + +
Figure 1.1 Outline of avalanche development, showing space charge field E
s
in
relation to applied field E. Also mean square radius r of avalanche
electron cloud, with transition from free diffusion to diffusion limited by
positive ion space charge
length L
d
over which the equality is achieved:
L
d
=
_
0
kT
ne
2
_
1/2
(1.9)
Here, n is the electron density, e is the electron charge, k is Boltzmanns constant
and T is the electron temperature [9].
If the square of the Debye length is now equated with the mean square radius of
the avalanche, the avalanche length x at which free diffusion ceases and is replaced
by ambipolar diffusion is obtained from:
e
exp( +)x
4
0
=
_
9
_
D
_
3
x
E
_
1/2
(1.10)
where is the electron mobility and E the applied electric field. After this point,
the insignificant rate of diffusion of the positive ions ensures that the radius of the
avalanche remains almost constant thereafter (Figure 1.1). There is also a redistribu-
tion of electron density due to the internal field of the ion space charge; this will not
be discussed here.
The drift of electrons in the space between electrodes is recorded in the external
circuit by virtue of the displacement current existing between the moving electrons
and the electrodes. The problem has been discussed by Shockley [16] for the case
of a charge q between parallel plane electrodes. Here, the current I recorded due
to charge q moving at velocity v is:
I =
qv
d
(1.11)
where d is the distance between the electrodes. For the more general case of the
charges due to electrons, positive and negative ions and including the effect of electron
diffusion, the equation due to Sato [17] can be used. Ionic currents are of course small
compared with those of the electrons, due to the disparity in velocities.
More rigorous modelling of the avalanche and its effect on external circuitry
requires the solution of the continuity equations for the electrons and ions, with
Poissons equation. An example is the work of Morrow and Lowke [14], where the
circuit current due to avalanche growth is computed taking into account the effects of
electron diffusion and of photoionisation. For the conditions of electric field taken, the
calculations showa typical current for a single avalanche rising to a fewnanoamperes
and lasting for a few tenths of a nanosecond.
1.2.3 The critical avalanche and the critical volume
The importance of the space charge field set up by the positive ions has already
been mentioned. It reinforces the applied electric field, in which the avalanche has
been created, behind its head of positive ions, that is, in the region from which the
generating electrons have just come. It is thus able to extend the region of the applied
field over which further avalanches might be initiated. Clearly, the extent to which this
happens depends on the magnitude of the applied field itself, since this will determine
both the size of the avalanche and the space charge field that it creates.
Experiment has shown, however, that breakdown occurs in air when the applied
field reaches a value of about 310
6
Vm
1
implying that ionisation reaches a critical
stage at this field. It is also known that the minimum field at which net ionisation can
occur is about 2.6 10
6
Vm
1
, for at this lower value the sum of the coefficients
and in Equation (1.5) exactly balances the attachment coefficient , so that there is
no net gain or loss of electrons. Between these two field values, ionisation can occur,
but not the development to breakdown, which requires the higher value.
This has led to the concept that the positive ion space charge field plays a critical
role: if it reaches a certain value, it extends the volume around the electrode in which
a successor avalanche can be started. Meek [7] proposed that the value of the space
charge field at the boundary of the positive ions should be equal to that of the critical
field of 3 10
6
Vm
1
. Thus, the avalanche number exp( + )d of ions must
be contained within the roughly spherical region created by the advancing electron
cloud. Experiment, again, indicates that the critical avalanche number is of the order
10
8
, that is, a charge of 16 picocoulombs, with a diameter of the order of 100 microns.
This model leads to the expectation that, once this critical avalanche size has been
achieved, avalanches can form repeatedly in a direction away from the electrode.
Before considering this, however, it is necessary to introduce a further concept; that
of the critical volume of field around the electrode.
Obviously, ionisation occurs more efficiently the closer an initiating electron
appears to the electrode surface. However, if it starts within only a few ionising free
paths of the electrode, there is insufficient distance available for exponential growth
to the critical avalanche number of positive ions. The electrons are quickly absorbed
by the anode and the positive ions gradually disperse. Evidently, there can be defined
a contour around the electrode within which no critical avalanche can be formed.
The distance of this contour from the electrode will depend upon the voltage at the
electrode and also on its radius.
Similarly, there is an outer contour at which the electric field falls to the critical
value of 2.6 10
6
Vm
1
. Beyond this boundary, a free electron has insufficient net
ionising power to initiate an avalanche. At any point between the outer and inner
contours, a critical avalanche can be formed. The volume between the two contours
is termed the critical volume. An example is the critical volume shape around the tip
of a rod of radius 1 cm, which is of the order of a few cubic centimetres when the
breakdown voltage for a rodplane electrode gap of 1 m is applied. An example of
such a critical volume is given in Figure 1.2 [18].
160
140
120
100
80kV
R=1cm
Figure 1.2 Right section showing the growth of the critical volume around the tip
of a hemispherically ended rod of diameter 2 cm as voltage levels rise
to 180 kV (Allen et al. [18])
Within this volume, an electron must be detached from a negative ion in order to
start an avalanche. The concentration of ions has already been noted as a fewhundred
per cc. The probability of detachment per unit of time is not known with certainty.
Where direct or alternating voltages are applied, the time required for a detachment
to occur is of no importance, but where a rapidly rising impulse voltage is applied,
the statistical uncertainty of the detachment process means that there is a variable
statistical time lag before a free electron can start the avalanche, which is responsible
for resultant statistical lags in corona formation with consequent effect on the time to
the final breakdown. The effect of local atmospheric conditions on the statistical lag
is discussed in Reference 18.
1.2.4 Streamer formation
It has already been noted that the space charge field of the critical avalanche is
sufficient, when added to the applied field, to initiate a succeeding avalanche. If this
development is considered to be along the axis of the system, in a simple case, then
the successor is formed in a weaker applied field than the original. This raises two
questions:
(i) are there sufficient free electrons created in this direction to allow this process
to proceed?
(ii) what is the weakest applied field in which replication can occur?
Regardingcase (i), the formationtime of the avalanche is less than1ns, but the positive
ion space charge, having a mobility of 10
4
metres per second per volt per metre of
field, would have a velocity of only 300 ms
1
in a field of 3 10
6
Vm
1
. Thus,
under the influence of this charge, a significant time is available for the detachment
of electrons from negative ions. But if the first avalanche critical head occurs very
close to the electrode surface, as might be expected, the additional critical volume
which it adds to that already established by the voltage at the electrode is extremely
small so that the probability of the observed rapid avalanche replication by this means
(see section 1.2.5) is negligible.
For this reason, the alternative mechanism of photo-emission of electrons from
neutral molecules within the avalanche critical volume is generally considered to be
the most likely cause of further development. The photons are generated by excitation
in the first avalanche and work reported on the number of photons created as a
proportion of the number of ion pairs [6, 19] has allowed estimates to be made of the
density of photoelectrons produced as a function of distance from the photon source.
As an example, at a distance of 1 mm from the source in an avalanche, the number of
photoelectrons generated is about 0.1 per ionising collision in the avalanche. Since the
number of ionising collisions is very large, the number of photoelectrons produced
in the critical volume of the avalanche space charge is also large.
The sequence of events is shown in Figure 1.3(i), taken from Reference 20,
where the case of streamers emanating from a negative electrode is also dealt with
(see section 1.2.9) [20]. Here, progression from the initiatory electron to the critical
avalanche and subsequent streamer growth is illustrated in a to e.
(i) (ii)
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ +
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ + +
+
+
+
+ +
+
+ +
+
+
+
+
+
+
+ +
+
+
+
+
+

E
o
a
b
c
d
e
+
+ + + +
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ + + +
+ + + +
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ +
+
+
+
++
+
+
+
+
+
+
+ +
+
+
+
+
+
++
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

+
+
+
+
+
+
+
+
+
+
E
o
a
b
c
d
e
Figure 1.3 Models of streamer growth near a rod electrode
(i) positive streamer development from free electron a, avalanche
b, c, streamer initiation d to growth e
(ii) negative streamer development from free electron a, avalanche
b, c, streamer and further avalanche d, to space stem e (see
section 1.2.8) (courtesy of CIGREs Electra, (74), pp. 67216,
Paris)
An alternative hypothesis is that the necessary free electrons are created by
photodetachment [12] from the negative ions. This proposal has the attraction that
the energy required is of the order of only 1 eV, whereas the energy required for
photoionisation is of the order 10 eV so that a much larger flux of active photons
is available. The hypothesis suffers from the same difficulty as that for collisional
detachment, namely that the number of negative ions in the existing critical volume
may be insufficient to maintain a rapid rate of replication. It is, however, consistent
with several aspects of streamer propagation and branching.
As the number of successive avalanches increases, so the critical volume, by virtue
of the potential at the end of the string of positive space charge heads, moves in the
direction of propagation. Although the space charge head is of very small diameter,
of the order 100 m, the perturbation to the applied electric field, calculated in
Reference 14 is significant, perpetuating to a great extent the field profile introduced
by the rod itself (Figure 1.4). Thus the critical volume remains of appreciable size,
comparable with that around the rod, but the known statistical lags associated with
such a critical volume appear to rule out collisional detachment and photodetachment
as a means of free electron generation, able to allow rapid replication of the heads.
Photoionisation thus appears to be the most likely mechanism.
The concept of repeated formation of critical avalanches, leading to advancement
of intense ionisation across the space between electrodes, led to the formulation of the
b a cathode
Z
,

m
m
R, mm
anode
20 10 0 10 20
120
100
80
60
40
20
0
20 10 0 10 20
0
20
40
Z
,

m
m
R, mm cathode
anode
60
80
100
120
Figure 1.4 Potential distributions a before, b during development of a positive
streamer from an electrode. Anode at 20 kV, equipotential intervals
1 kV. Streamer has progressed 23 mm in 54 ns (after Morrow and
Lowke [14])
streamer theory first clearly enunciated by Meek [7], although supported by several
earlier experimental observations, for example References 21 and 22, which indicated
some of the principles involved.
1.2.5 Streamer development
The picture so far presented is of successive avalanches progressing into regions of
decreasing electric field, thereby prompting the second of the two questions posed
earlier (page 10).
Meek propounded the criterion that for propagation the electric field at the bound-
ary of an assumed spherical positive space charge is put equal to the applied field
at the electrode. The further criterion has been noted that the resultant field at the
boundary of the critical volume must be at least equal to that needed to sustain ion-
isation. More exactly, the sum of the applied and space charge fields in this latter
region must be sufficient to supply the rate of energy input for ionisation to produce
a critical avalanche.
This approach was adopted by Gallimberti [23], who considered the parameters
involved in the formation of a single avalanche which was, itself, the successor of
a previous avalanche in the electric field. It was concluded, from theoretical con-
siderations, that an applied field of about 700 kVm
1
was necessary for continuous
propagation of critical avalanches. This conclusion was subjected to practical tests at
later dates, initially by Phelps and Griffiths [24], who initiated streamers at a point
but propagated them in a uniform electric field, so removing any ambiguities due
to the non-uniformity of field in the models discussed above. In this work, both
the pressure (and therefore density) and the humidity of the air were varied, but
the authors concluded that at the standard normal temperature and pressure (NTP)
condition and at a standard humidity of 11 grams of moisture per cubic metre of
air the minimum applied field needed to sustain streamer propagation was close to
500 kVm
1
.
The result has since been generally confirmed by other measurements [2528]
in a variety of experimental arrangements and the value of 500 kVm
1
has been
incorporated into the IEC Standard 60060-1 (1989) concerned with the effects of
atmospheric changes on sparkover voltages in air (section 1.3.4).
The velocity of streamers, as a function of electric field, is an important parameter
that has been investigated [28, 29]. At fields just above the minimum, the velocity
is about 2 10
5
ms
1
, but this increases faster than linearly as the electric field
increases and is commonly greater than 10
6
ms
1
at the start of propagation in the
non-uniform field exemplified by the rod plane gap referred to above.
It is also worth noting here that at a given condition of air density and humidity, the
streamer properties of minimumpropagation field and of velocity of growth at higher
fields are very precisely defined. It has been found that both of these quantities can be
quoted with an uncertainty of less than one per cent. The statistical nature of sparkover
measurements is thus due to other factors, but the precision of streamer growth has
found application in the use of the rodrod gap as a standard for measurement of
direct voltage [30, 31].
1.2.6 Corona
Practical experience shows that streamers branch, after propagating short distances,
to form what is generally termed corona which can then develop and extend to cause
a sparkover of the gap. Branching is assumed to be caused by development of side
avalanches in the field of an initial avalanche. An example is shown in Figure 1.5. The
mechanism of branching is not fully understood. It might be thought that the flux of
photons sufficient to cause rapid extension of the streamer by photoionisation would
be sufficient to cause a much higher density of branches than is shown in Figure 1.5.
The lack of a favourable combination of applied and space charge fields may be
responsible, but it has also been postulated that the observed pattern may result from
a reliance on photodetachment from surrounding, but relatively rare, negative ions in
order to produce the necessary free electrons [12].
Many tests with various electrode configurations show that streamers, although
originating in very high fields, can propagate into regions in which the field is much
less than 500 kVm
1
. Figure 1.5 is an example where streamers in the corona progress
to the plane, where the electric field is of the order 100 kVm
1
. Direct measurement
has shown, however, that in such a case, the field at the plane, at the instant of arrival
of the streamers, was about 450 kVm
1
[26, 32]. The largest component of the field
was thus due to the sum of the positive space charges at the heads of large numbers
0.4m
plane
rod
Figure 1.5 Image of positive corona obtained by sheet film placed along the
axis of a rodplane gap 0.4 m long. Voltage 170 kV. Note
evidence of secondary emission when streamers reach the plane (after
Hassanzahraee, PhD thesis, University of Leeds, 1989)
of streamers which combined to allow propagation in a comparatively weak applied
field. This phenomenon allows propagation to take place over distances of the order
of metres.
1.2.7 The streamer trail
The electrons produced in each replicating avalanche pass through the positive space
charge head left by the preceding avalanche and drift towards the anode along the ionic
debris left by earlier avalanches. Some recombination occurs but a more important
process is attachment to electronegative molecules, such as O
2
or O
2
(H
2
O)
n
, to form
negative ions. The use of known attachment coefficients, of the order 100 to 500 per
metre per electron [10], indicates an e-folding distance for the rate of loss of free
electrons of the order of a few centimetres in air, a result that has been confirmed
by recent experiment [33]. Thus, after drifting from the streamer head towards the
anode, most of the free electrons are soon lost and the remainder of the streamer trail
is made up of positive ions which are nearly matched, in density, by negative ions.
Since the mobility of positive and negative ions is low, of the order
10
4
m
2
V
1
s
1
, the conductivity of the streamer trail is low, even though a small
proportion of free electrons is able to reach the anode. Typically, the resistance of the
trail is of the order of megohms. Thus, the existence of an assembly of streamers, as
in a corona, does not constitute a breakdown, even though the corona may bridge
the electrode gap.
1.2.8 The leader
When the electric field is sufficiently high, a new development can occur within
the corona: the formation of one or two distinct highly conducting channels having
properties different from those of the streamer channels. This is the leader, first
photographed as a precursor to the main flash of the lightning discharge by Schonland
and Collens [34] and later in laboratory studies by Allibone and Schonland [35].
In contrast to the streamer channel, the light emission is in the visible range of the
spectrumand it has been observed to growfromboth positive and negative electrodes,
although more extensively studied in the former case, usually under impulse voltages.
1.2.8.1 Formation of the leader
In an extensive corona, having many branches over a length of several tens of cen-
timetres, the number of electrons able to reach the stem of the streamer trails at the
point of origin, close to the anode, becomes considerable. Extensive detachment of
electrons from negative ions in that region is also believed to occur. As a result of
energy exchange between the energetic electrons and neutral gas molecules, the rate
of ohmic loss of energy increases and significant heating of the channel can occur.
The temperature of the neutral gas thus rises, as a result of which it expands; the
gas density falls. The quantity E/N therefore increases and ionisation becomes more
efficient. The process is cumulative and a transition takes place to a highly ionised
arc-like channel of high temperature and relatively high conductivity. The channel has
now been transformed into a leader which proceeds to grow in the general direction
of the electric field.
An idealised model of leader development is given in Figure 1.6, in which a rod
rod gap is imagined to be subjected to an impulse voltage with a time to peak of
the order of a few hundred microseconds. The voltage at the positive electrode rises
until the field at the tip exceeds 3 10
6
Vm
1
at time t
1
when a streamer corona
forms. A burst of current is detected in the circuit. The corona injects a net positive
charge into the region, so reducing the field at the tip and inhibiting further streamer
formation for a dark period until the voltage has increased. Asecond corona (often
termed a secondary corona) then occurs at time t
2
which may be followed by others at
short intervals of time thereafter (omitted for clarity). After another corona is formed,
at t
3
, sufficient heating has occurred in the streamer stem at the anode for a leader
channel to form. Where the diameter of the electrode is relatively large, the leader
may formimmediately out of the secondary corona at time t
2
. In either case, it extends
in length across the gap towards the opposite electrode. Since the leader channel is
highly conducting the potential of its tip remains high and a streamer corona forms
ahead. Thus, the avalanches at the heads of the streamers provide the ionisation
and, therefore, the electron current and consequent heating needed for further leader
development.
The streamer coronas have formed more or less continuously as the leader has
extended across the gap. Since the positive ions deposited remain immobile on
the timescale involved, a roughly cylindrical volume of remanent positive charge
surrounds the leader channel along the whole of its length.
250
t, s

t
4
t
1
+
t
2
t
3
U
cr
U
,
k
V
Figure 1.6 Simplified picture of streamer and leader development to breakdown in
a rodrod gap under an impulse voltage rising to peak in 250 s
When the field at the negative electrode exceeds the order of (56) 10
6
Vm
1
at time t
4
negative streamers (section 1.2.9) can form and develop towards the anode,
but, as these require an ambient field of about 10
6
Vm
1
to progress, they do not
extend as far into the gap as do the positive streamers. Moreover, by contrast with
the process at the positive electrode, the electrons are moving into a reducing electric
field. At a later stage, transformation to a leader occurs but these processes occur at
higher fields than is the case with the positive counterparts, and the distance traversed
is smaller. When the two leader systems meet, a conducting channel bridges the gap
and a low voltage arc can complete the breakdown of the gap.
A simple semiempirical argument can be used to relate the streamer and leader
lengths and the respective average electric gradients needed for their propagation to
the sparkover voltage V
s
of the gap. For at the instant at which the systems meet,
4
,
we can write down the sum of the voltages across the gap:
V
s
= E
+
s
L
+
s
+E
+
l
L
+
l
+E
s
L
s
+E
l
L
l
(1.12)
where E
+
s
, E
s
, E
+
l
and E
l
are the gradients for the positive and negative streamers
and leaders, and L
+
s
, L
s
, L
+
l
and L
l
are the corresponding lengths. Note also that
the gap length d is:
d = L
+
s
+L
+
l
+L
s
+L
l
(1.13)
Certain of these quantities, such as E
+
s
, E
s
are known from independent measure-
ments and, in specific experiments, lengths of streamers and leaders estimated, so that
other quantities in Equation (1.12) can be estimated from a sparkover measurement.
It may be expected from the foregoing descriptions that only a relatively large
streamer corona is likely to develop into a leader, where fields near the positive
Figure 1.7 Photograph of leaders developing in a 4 m rodrod gap [38]. Note that
the leader does not develop along the line of maximum field between the
rods; also bifurcation at the tip of the rod
electrode are high, detachment of electrons is rapid and large numbers of electrons
are found in that region. Thus, leader initiation will occur most readily in large gaps,
usually 0.5 m or more, where the sparkover voltage is high. Indeed, Reference 36
shows that the charge in the initial streamer corona, which subsequently produces
a leader in a five metre gap, is several tens of microcoulombs.
The description of leader growth in Figure 1.6 is much simplified. For example,
the leader rarely follows the axial path between electrodes. It tends to go off-axis in
a path which may be much longer than the direct one, particularly if the electrode
diameter is relatively large. Moreover, a bifurcation often occurs, originating at the
electrode, although one of the two branches is usually dominant. An example of
a leader in a long gap is shown in Figure 1.7. Waters [37] gives a more detailed
description of leader development under impulse voltage in a practical case. Ross
et al. [36], quoting results obtained by the Les Renardieres Group, give an interesting
summary of quantities associated with leader development in a 10 m gap under the
50 per cent sparkover impulse voltage. For example, the total energy dissipated during
the growth of the leader is about 25 J per metre of its length; the current is taken as
0.6 A, so that if the length of streamers in the leader corona, having a gradient of
500 kVm
1
, is 1 m, then the power input to the leader corona is 300 kW. Since the
Figure 1.8 Two examples of leader development in a short (0.6 m) rodplane gap
showing on left, streamer formation (leader corona) at the leader tips.
Impulse voltage 330 kV (after Hassanzahraee, PhD thesis, University
of Leeds, 1989)
system advances at the rate of 2 cms
1
, then the energy dissipated in the leader
corona alone in 1 m of advance is about 15 J. The energy available to heat the leader
channel itself is estimated at about 10 Jm
1
. As will be shown below, these are
average values, indicating general features only, since the potential gradient in the
leader varies along its length and the leader corona is also variable.
The simple heating process described above suggests also that leaders can be
produced in short gaps provided a sufficiently high electric field can be created at the
electrode. This can indeed be achieved by impulse overvolting of a gap; Figure 1.8
shows an example obtained in a gap of 0.4 m, to which an impulse of twice the
threshold sparkover voltage was applied.
1.2.8.2 Leader properties
Much experimental data on leader properties has come from the work of the Les
Renardieres Group [36, 38]. Although there is general agreement that the foregoing
represents the main features of leader initiation and propagation, there are many
aspects that are not fully understood by theory. Waters [37] gives a summary of the
models of leader dynamics proposed by various authors up to 1978, and a refinement
based on later experimental work is given in Reference 38. We now discuss those
properties of the leader that are of the greatest significance to the breakdown process.
Several parameters have been measured. Of primary importance in breakdown is
the velocity of propagation of the leader which, in laboratory tests, has been found
to be of the order of 2 10
4
ms
1
. This value increases roughly in proportion
to the current, but changes only slowly with the voltage across the gap. Indeed,
the leader velocity is generally insensitive to changes in the electrical conditions or
gap configuration, although it increases with increasing atmospheric humidity. By
contrast, streamer velocities increase with electric field, usually exceeding 10
6
ms
1
at their point of creation, falling to 10
5
ms
1
at the ends of their trajectories. Thus, a
streamer system, that is, the leader corona, ahead of the leader tip has time to form
and progress along with the tip of the extending leader.
The charge injected by the leader includes not only the charge in the leader stem
but also the charge in the streamer corona which everywhere precedes the leader tip.
Thus, the charge injected per metre of length of the actual path is about 40 C. This
is independent of the diameter of the electrode, although the total charge injected
increases with increasing rod diameter, for a given gap length [36]. This may be
a reflection of a corresponding increase in the leader length. The charge injected in
a long gap can thus be very appreciable and, when impulse testing, the inception of
the leader can sometimes be recognised by a sudden drop in the voltage waveform
measured by a capacitor potential divider.
The mean current in the leader is measured in a 10 m gap to be of the order of
1 A and estimates of the channel width of 5 to 10 mm lead to a value of current
density of up to 10 A/cm. The mean electric field in the leader has been estimated by
fluxmeter and probe measurements and depends upon its length [36]. At the instant
of transition from streamer to leader, at the electrode tip, the gradient is nearly equal
to the gradient of the streamer fromwhich it develops, namely about 500 kV/m. After
a few tens of microseconds, that is, after several tens of centimetres growth, the
effect of heating has greatly reduced the local resistivity and, hence, the gradient.
Thus, the mean gradient is reduced as the leader progresses. The problem has been
considered by Ross et al. [36] who, by relating ionisation and attachment rates and
the energy input to the channel, showed that the gradient averaged over the whole
time of growth of a leader decreases rapidly with time and, therefore, with length.
The predicted result, which depended on an assumed channel radius, was verified by
later experiments [38] and a comparison between the two is shown in Figure 1.9.
400
500
300
200
100
E
,

k
V
/
m
0
0 1 2
L
l
, m
3
Figure 1.9 Average electric field along the leader channel as a function of its
length [38]
The formation of the relatively highly conducting leader can be likened to an
extension of the metal anode itself, for the potential drop along it has become com-
paratively small. The high field region at the electrode tip has now been transferred
to the tip of the leader where conditions for the formation of further streamers now
exist. Thus the leader corona is formed, which itself now provides the condition for
further leader formation at its stem. In this way, the leader is able to grow outwards
to a distance from the anode which is determined by the combination of applied and
space charge electric fields.
The comparatively low velocity of advance of the leader is determined not only
by the rate of heating of the channel; it is also checked by the amount of positive
space charge produced in the leader corona. Where this is large, it may reduce the
field around the leader tip to reduce heating effects or even to choke off leader devel-
opment completely. On the other hand, high humidity tends to increase the rate of
leader growth. In this case, streamer corona development is inhibited (section 1.3.4),
resulting in a lower charge in the leader corona and a higher resultant field at the tip
of the leader, so facilitating progressive development.
1.2.9 Negative discharges
Negative discharges are of generally lesser importance in the breakdown process than
discharges in the diverging fields close to electrodes at positive polarity. The differ-
ence arises because, in the negative case, electrons are moving into an electric field
of decreasing intensity, whereas the reverse is the case near a positive electrode. As
a result, the drift velocity and ionisation efficiency decrease with increasing distance
from the electrode. It is more difficult for a discharge to spread across the gap, unless
a higher electric field is applied. Also, diffusive spread of an avalanche increases as it
progresses down the field gradient so that the space charge field of the positive ions
decreases.
As a consequence, a negative corona has a different, more complex, structure than
a positive corona and tends to be more localised around the electrode.
Auseful picture of the formation and propagation of negative streamers has been
given by Hutzler et al. [20], shown in Figure 1.3(ii) where the contrast between
positive and negative streamer mechanisms is drawn. Again, a single free electron a
is postulated to start the first avalanche, which progresses into a decreasing electric
field. The electron space charge so produced causes a local increase of electric field
ahead, so that an electron liberated by ultra-violet radiation c can produce a further
avalanche d. The electrons from the first avalanche move into the (almost stationary)
positive ions left by the second avalanche, so producing a channel with conducting
properties behind the leading avalanche e. This process can continue, provided the
electric field is sufficiently high.
The dipolar channel illustrated in e has been termed a space stem, within which
a reduced electric field component, due to the space charges, exists.
A moments thought will show that, for a given magnitude of electric field, the
diverging field at an anode will provide more favourable conditions for ionisation
than the converging field at a cathode, since, in the former case, the electrons are
1m
Figure 1.10 Streak photograph of development of a negative leader system, showing
negative streamers ahead of the leader tip and retrograde streamers
towards the anode [20]
always moving into an increasing field. Thus, a negative streamer requires a higher
field for propagation than a positive streamer; the ratio is, in fact, about two to one.
Figure 1.3(ii) indicates that an augmented field also exists between the elec-
trode and the positive space charge deposited by the first avalanche. This provides
conditions entirely analogous to those of Figure 1.3(i) which are favourable to the
formation of a positive streamer. Thus, in addition to the anode-directed negative
streamer, a retrograde, cathode-directed streamer can also be set up. This has, in fact
been observed, by means of image convertor streak photography, Figure 1.10 [20].
This shows a fairly diffuse initial streamer corona, less well defined than its positive
counterpart, which quickly at its head appears concentrated and which then gives rise
to filaments extending in both directions.
The space stem acts as a location for the onset of the negative leader; it appears
as regions of high luminosity in Figure 1.10, gradually moving across the gap. The
anode and cathode-directed streamers appear on either side, and the leader develops
after some delay. It is assumed that the streamerleader transition occurs by heating
in the same way as in the positive case, and as in that case also, its rate of growth is
determined by the corona at its head.
The velocity of negative streamers is not known. The negative leader grows
towards the anode with a velocity less than 10
4
ms
1
, which is significantly less than
that of the leader in a long positive rodplane gap. However, the cathode-directed
streamer has been shown to have the same velocity as in a positive discharge.
1.3 Applications
The applications discussed here are relevant mainly to high voltage power systems
and associated apparatus. Most power systems employ alternating voltages, although
direct voltage operation is becoming significant in special situations. However, the
insulation against breakdown may be required to be at least twice that needed for
the operating voltage, since transient overvoltages of higher levels may occur. These
may be caused by lightning strikes, near to or upon an overhead line, by switching
operations in the power system, or by temporary increases in the level of the operating
voltage. Of these, the first two are the most frequent and dangerous to the system and
the insulation is designed with this in view.
The lightning impulse approximates to the disturbance caused by a lightning
strike: it is characterised by a relatively fast rising impulse reaching its peak in a
few microseconds and decaying in a few tens of microseconds. The switching, or
slow front impulse is considered to rise in a few hundred microseconds, decaying in
a few thousand microseconds. In both cases, these impulses may be reflected from
discontinuities in the transmission line, with consequent increase in voltage. For this
reason, testing of high voltage apparatus is carried out under an impulse voltage
related to, but much higher than, the proposed operating voltage.
1.3.1 Sparkover under lightning impulse voltage
The IEC definition of a standard lightning impulse voltage specifies a rise time of
1.2 s and a decay to half the peak value in 50 s [31]. Atolerance of +30 per cent
is allowed on the rise time and of +20 per cent on the decay time (Figure 1.11).
The duration of voltage around the peak is thus short compared with the times
required for a leader to advance a significant distance, but the time scales are long
compared with that needed by streamers to propagate. The leader thus has a negligible
role in the breakdown process. Taking as an example the positive rodplane gap, the
critical volume is expanding around the tip of the electrode during the rise time of
the impulse voltage. Negative ions are attracted to the volume in this time, but their
density is not sufficient to allow a significant number to enter the critical volume in
the time available. Also, the number located in the critical volume for detachment of
electrons to occur is subject to significant statistical variation. Thus, avalanches and
1.0
0.9
0.5
0.3
0
0
1
T
T
T
1
T
2
A
B
U
time, t
T
1
= 1.67T
T = 0.3T
1
= 0.5T
Figure 1.11 Parameters of the lightning impulse voltage [31]. The front of 1.2 s
is defined as the time 1.67T; the tail of 50 s is defined as the time T
2
to half the peak voltage
1
streamer corona will occur at statistically variable times during the rise of voltage
to the peak. This first group of streamers may not be sufficiently extensive to cause
immediate breakdown. Since appreciable leader growth cannot progress, breakdown
can occur only if a second group of streamers crosses to the cathode, around or soon
after peak voltage, producing sufficient secondary emission of electrons to cause
increased ionisation and form a conducting channel across the gap.
It is evident, from this qualitative picture, that the breakdown stress in a gap
with a single region of non-uniform field under a lightning impulse voltage depends
directly upon the stress needed for streamers to cross the gap. This follows from
Equation (1.12) where the following simplifications can be made:
there is no leader development at either electrode
there is no streamer development at the plane cathode.
Thus, Equation (1.12) reduces to:
V
s
= E
+
s
d (1.14)
and the sparkover voltage increases linearly with the gap length.
This has been shown experimentally to be true with rodplane gaps up to 8 m
[39, 40]. For a gap with two non-uniform field regions, typified by the rodrod gap,
some negative streamer growthfromthe cathode is likelytooccur. Here, the approach-
ing positive streamer corona will generally set up a sufficient space charge field to
enhance the field at the cathode to a higher value than that set up by the applied
voltage alone. Thus, again noting the absence of leader formation, Equation (1.12)
reduces to:
V
s
= E
+
s
L
+
s
+E
s
L
s
(1.15)
so that the voltage at sparkover is now increased with respect to that in the rod
plane case.
Two important properties make the rodplane gap, under positive impulse, a
valuable reference in high voltage testing:
(i) it has the lowest sparkover voltage of any gap configuration of the same length
(from Equation (1.14))
(ii) it shows a linear increase of sparkover voltage with gap length.
Both of these properties arise from the lack of any negative discharge growth at the
plane.
The absence of significant leader growth and consequent lack of ambiguities
makes the positive lightning impulse an important test voltage, recommended in
standards, to be applied to high voltage components and apparatus, such as bushings,
insulators, transformers and so on.
Under negative impulse voltage, applied to a rodplane gap, the sparkover voltage
increases with gap length slightly less rapidly than linearly, but the magnitudes are
a factor of nearly two greater than those for the positive case. These differences
may be expected from the differences in the discharge propagation modes, discussed
in section 1.2.9.
1.3.2 Sparkover under slow front impulse voltage
The IECStandard 60060-1 (1989) [31] defines a standard slowfront impulse having a
nominal rise time of 250 s and a decay time to half value of 2500 s. In practice,
a variety of rise and decay times is encountered and it is necessary to consider,
in a general way, the nature of the processes involved in breakdown.
Consideration of Figure 1.6 and the associated qualitative arguments indicates at
once that where the voltage rises relatively slowly, there is a relationship between the
voltage rise time, the leader growth rate and the gap length which will determine the
voltage at which breakdown occurs. Asimple example will illustrate.
Consider a rodplane gap of, say, 2 m, to which a positive slow front impulse
voltage is applied. At a velocity of 2 10
4
ms
1
a leader would take of the order
of 100 s to cross the full gap (in fact somewhat less because its own leader corona
would occupy a significant length). Taking a range of rise times from 1 s to, say,
250 s, it is clear that as they increase, the leader can traverse progressively larger
proportions of the gap. The breakdown voltage V
s
is correspondingly reduced on
account of the fact that the gradient of the leader is lower than that of streamers.
Experiment shows that for further increase in rise time, the trend reverses and the
breakdown voltage increases. The reason is that successive streamer coronas develop
during the relatively slowrise of voltage, none of which has injected sufficient charge
at a sufficiently high stress to cause a leader to form. The resulting positive space
charge reduces the local field at the anode so that a larger stress is ultimately needed
to form the leader. Thus, a higher voltage is required and there is a minimum in the
curve of V
s
against time to peak. It follows from this argument that the time at which
the minimum occurs depends also on the gap length.
This behaviour results in the so-called U-curve which has been established by
testing over a wide range of gaps. It is illustrated in Figure 1.12, where U-curves
obtained in rodplane gaps in the range 1 < d < 25 m are given. The time to peak
impulse voltage at which the minimum occurs for a given gap is called the critical
time to peak and it defines, therefore, a limiting minimum voltage at which that gap
can be broken down under an impulse.
It is found that the critical time to peak shows a closely linear increase with the
length of gap. Since the rate of leader growth has been shown to be approximately
independent of its length [36], the U-curve minimumfor a given gap is thus identified
with optimum leader growth. For shorter times to peak, the reduced time for leader
growth means that a higher voltage is needed for breakdown. For longer times to peak,
the succession of streamer coronas also requires a higher stress for leader initiation,
so that the breakdown voltage rises again. It follows from these facts that, where an
impulse having a critical time to crest is applied to the gap, the breakdown occurs at
or near the crest of the impulse.
Experimentally, it has been found [41] that the sparkover voltage of the rodplane
gap at the critical time to crest is given by the formula:
V =
3400
1 +8/d
(1.16)
0
100 200 300
T
1
, js
400 500 600
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
2400
2600
2800
3000
3200
3400
3600
25m
15m
10m
6m
5m
4m
3m
2m
1m
9m
k
V
Figure 1.12 A selection of U-curves for rodplane gaps in the range 1 to 25 m.
Note the trend for the minima to occur at longer times to peak as the
front of the impulse increases
where d is the gap length. This formula must be modified for other gap configurations
(section 1.3.3) and also when variations in atmospheric conditions must be taken into
account (section 1.3.4).
The existence of the U-curve was first realised about 1960 [41, 42] and it is
now widely used in high voltage technology, for instance in determining minimum
clearances in high voltage equipment such as overhead lines and in general questions
of insulation coordination.
1.3.3 The influence of field configurations: the gap factor
Discussion so far has assumed the non-uniform field electrode to be of small radius,
so that high fields exist in its vicinity, leading to ready formation of corona. When the
radius is increased, so reducing the per unit field at the electrode boundary, higher
5000
4000
3000
2000
1000
1 10 100 1000
T
cr
, s
U
s
o
,

k
V
cone
sphere
Figure 1.13 Effect of the change in profile of the energised electrode (plane earthed)
on the U-curve; gap = 10 m [43]
cone electrode of small tip radius
sphere electrode of radius 0.75 m
voltages are needed for the initial ionisation, resulting in a higher ultimate voltage
for the final breakdown. The effect is shown in Figure 1.13 where the U-curves are
shown for a 10 m air gap with two electrode geometries, namely a cone and a sphere.
The former shows the U-curve minimum clearly, but with the sphere; production of
a viable streamer corona requires a higher voltage in order that the transition to leader
can occur. Therefore, the minimum becomes much flatter. For shorter times to peak,
where leader formation is minimal, the two curves come together.
It is evident that, since the corona initiation depends on the field configuration
around highly stressed electrodes, the variety of gap geometries that is inherent in
practical situations will impose variations in the highly stressed regions and, therefore,
in breakdown strength. This effect is particularly marked where there is a highly
stressed region around the negative, as well as the positive, electrode. An argument
is presented in the CIGRE guidelines [43] in which it is shown that the total voltage
across the gap for maintenance of the predischarges in mid-gap prior to sparkover
is increased by the insertion of a high field region around a pointed cathode, when
compared with the case of a plane cathode. It follows also from Equation (1.12)
that where significant negative streamer development occurs (which has gradient for
propagation of the order of twice that for positive streamers), the voltage across the
gap at the instant of breakdown is higher than it would have been in the absence of
negative streamers.
Similar considerations apply for gaps of other geometries that may arise in
practice. This has led to the concept of the gap factor k, which is defined as the
ratio of the sparkover voltage of a particular gap to the positive rodplane air gap
sparkover voltage, for gaps of the same length and subjected to the same switching
impulse. It is of value because the ratio holds good for nearly all lengths of gap that
are of practical interest. Thus, adapting Equation (1.16) for the U
50
sparkover voltage
d
d
d
d
d
H
H
H
H
H
H
rodplane
configuration k
1
1.1 to 1.15
1 + 0.6
(1.1 to 1.15) exp 0.7
k
o
exp 0.7
H
H
rodrod
conductorplane
conductorrod
protrusions
*sign + for protrusions from the negative electrode
sign for protrusions from at the positive electrode
k
o
: gap factor without protrusions
H
H
H
H
k >1
*
Figure 1.14 Some gap factors for a selection of simple electrode geometries [43]
at the critical time to peak, the sparkover voltage of any gap of gap factor k is:
V =
3400 k
1 +8/d
(1.17)
Gap factors have been determined from experiment with a number of basic geome-
tries and a summary, taken from Reference 43, is shown in Figure 1.14. More
detailed empirical formulae for estimation of gap factors for a number of practical
configurations are also given in the same reference.
Since gap factors are always >1, the value of k may be assumed to be an indicator
of the increase in stress needed for leader formation, compared with that in the
rodplane gap. It would then be expected that the gap factor must influence the
critical time to peak; the empirical formula below shows this to be the case:
T
p
= [50 35(k 1)]d (1.18)
where T
p
is in s and d is in m [43].
The presence of an insulator in the gap also affects the gap factor, since the
presence of its surface affects both streamer and leader properties (section 1.3.5). The
following formula has been given [43]:
k
i
= [0.85 +0.15 exp (k 1)]k (1.19)
where k
i
and k are the gap factors with and without the insulator in place.
It is frequently possible to approximate a practical gap to a simple one for which
the gap factor is already known and the concept is, therefore, of considerable value
to power engineers concerned with design and insulation coordination.
1.3.4 Atmospheric effects
All insulation problems which involve air as a primary medium must take account of
the variations in atmospheric conditions. These conditions are of pressure, tempera-
ture and humidity, but it is important to note that fog, coastal salt laden air and, in
areas of high pollution, particulate matter in the atmosphere can dominate. Pressure
and temperature changes are manifest in changes of air density; altitude can reduce
the air density, in regions where power must be delivered, to as little as 0.65 of that at
sea level [44]. Temperature changes occur, between 40
, with resulting changes in

air density to 0.9 and 1.1, respectively, of that at the standard atmospheric tempera-
ture of 20
C. These changes are expressed in terms of the relative air density (RAD)
which is unity at 101.3 kPa pressure and 293 K. At all other conditions of pressure, b,
and temperature, T , the relative air density is:
=
b
101.3
293
T
(1.20)
Absolute humidity can vary between 1 gm
3
and 30 gm
3
; the standard atmospheric
humidity is taken as 11 gm
3
. Changes in both air density and humidity result in
changes in ionisation efficiency and, therefore, in sparkover voltage characteristics.
The prebreakdown processes have been described in sections 1.2.1 to 1.2.7.
The effects of atmospheric changes are on avalanche initiation and development
and the streamer trail. In the case of density variation, where caused by altitude
change, the effect on the density of atmospheric negative ions, available for detach-
ment in the critical volume (section 1.1.2.1), is not known. Indeed, the density of
negative ions is more likely to be affected by local variations in geology and indus-
trial activity, as well as by local weather, than by altitude. On the other hand, an
increase of humidity increases the density of the O
2
(H
2
O) ionic complexes which,
therefore, increases the probability of detachment of electrons in a high electric field.
The effect of air density change is considered first.
The most significant changes occur in avalanche development. As density
decreases, ionisation efficiency can notionally be maintained at a constant level at
constant E/N (section 1.1.2.6) so that the electric field can be reduced in proportion
to N. However, electron diffusion effects, depending on N as well as E/N, become
important, so affecting changes in avalanche radius and the conditions for streamer
formation.
The foregoing remarks assume that the kinetic processes of ionisation, attachment
and electron mobility remain unchanged at constant E/N as the density N changes.
However, there is evidence that if N is reduced by increase of temperature at constant
pressure, both the ionisation coefficient and the electron mobility are reduced at con-
stant E/N [45]. This implies that, together with changes in diffusion already referred
to, a further change will take place in the condition for achieving a critical avalanche.
The subject of avalanche formation as a function of density N requires solution of the
continuity equations for electron flowin which experimentally determined parameters
can be used.
Experiment has shown that the electric field required for stable propagation of
a streamer varies with relative air density as
1.3
[24, 28]. This appears to be true
whether arising from pressure or temperature change within the range 0.71.0. Tests
with long rodplane air gaps (>1 m) have established, however, that under light-
ning impulse, where the sparkover voltage depends on the streamer gradient, this
voltage varies linearly with [46]. This fact has been adopted as a reference in the
IECStandard 60060-1(1989) [31] which specifies howhigh voltage measurements of
dielectric strength shall be adjusted to take account of density variations. Under slow
front impulse, however, where there is significant leader growth, the dependence
of sparkover voltage on density is less strong since the leader properties are less
dependent than those of streamers upon relative air density. However, the conditions
for leader inception and, therefore, development, do depend on the streamers from
which it originates. The Standard has set out an empirical procedure for adjustment
of sparkover voltages which implicitly takes account of the extent of leader growth;
it is, however, beyond the scope of this chapter.
Where relative air density is linked to temperature change, the situation is less
clear. Testing in long gaps, over an appreciable range of temperatures, is impractica-
ble in the open atmosphere and this has precluded measurements in which significant
leader development could be expected. In practice, the range of atmospheric RADs
encountered is smaller than that arising frompressure variations. Therefore, the Stan-
dard adopts an ad hoc adjustment procedure which makes no distinction between the
two. Researches in short gaps (<25 cm) over large temperature ranges have shown
conflicting results despite the fact that the breakdown process involves streamers
only. The results, again, conflict with those obtained with variable pressure [47, 48].
Further work is needed for a resolution of these problems.
Humidity change also affects avalanche and streamer development, since the
attachment coefficient is increased relative to the ionisation coefficient [10]. Much
more data is available in this area, from outdoor testing and from laboratory experi-
ments, than is the case with density variation. Researches have shown that the electric
field required for streamer propagation increases at the rate of about 1 per cent per
gram of moisture content per cubic metre of air at the standard pressure and temper-
ature of 101.3 kPa and 293 K [23, 50]. This is also the rate at which the sparkover
of the rodplane gap increases under lightning impulse [49] where, as noted earlier,
the breakdown is determined solely by streamer growth in the gap.
Where significant leader growth occurs, the effect of humidity is smaller since
humidity, while increasing the leader velocity, does not change the leader gradient.
Thus, inspection of Equation (1.15) indicates that the relatively large effect on the
sparkover voltage of the humidity effect on streamers is offset by the smaller effect
on the leader. This results in a humidity coefficient of less than 1 per cent per gm per
cubic metre when the leader occupies a significant part of the gap.
Again, the IEC procedure for adjustment of sparkover voltages for humidity
change is given in IEC60060-1(1989) to which the reader is referred for details.
1.3.5 Corona at low air density
Some of the changes occurring to the prebreakdown corona at reduced air density
have already been alluded to in the preceding section. However, it is likely that
there will be future interest in the incidence of corona and breakdown under the
low density conditions encountered in the aerospace arena, where electrical equip-
ment may be called upon to operate in relative air densities of the order of <0.3.
Much work has been carried out in the past on corona and breakdown at low pres-
sures (see, for example, References 50 to 52), but a different focus is needed at high
altitude. In this case, simultaneous large changes in pressure, temperature and humid-
ity occur. Little data is available, but some relevant physical factors can be briefly
presented.
The minimumvoltage for corona inception is an important parameter. As pressure
is reduced at roomtemperature, it has been shown that the inception voltage decreases
approximately linearly with air density [47]. With increasing temperature, corona
inception voltage decreases much more slowly, according to a relationship given
in Reference 53. However, the effect of decreasing temperature in the presence of
decreasing pressure is not known.
The inception voltage depends upon the E/N condition for ionisation as the den-
sity decreases and this, in turn, determines the volume and boundaries of the critical
volume around an electrode, as discussed in section 1.2.3. However, the probability of
formation of an avalanche also depends upon the probability of an initiating electron
being found within the critical volume. This in turn depends on the local density of
atmospheric negative ions. Discussion of the influence of atmospheric ion densities
on corona initiation was given in Reference 54, and the effects of a change in the
ambient atmospheric conditions were measured in Reference 18. The increase in ion
density with altitude resulting from cosmic ray activity is well known and must be
considered as a further variable in determining corona onset.
The change with air density of corona development after inception is also to be
taken into account. Measurements made under both positive [50] and negative [52]
polarity corona studies show that the charge injected in corona tends to increase with
decreasing air density, even though the inception voltage decreases. This result is
true whether the coronas are produced under impulse [50] or direct [56] voltage. The
reasons for this trend have not yet been elucidated.
Finally, design of high altitude systems having relatively small dimensions of
the order of centimetres must take into account the fact that Paschens Law, relating
sparkover voltage of a gap to the product of pressure and electrode separation ( pd )
[57], also applies to the onset of corona. Briefly, the Law indicates that where pd is
small so that there is relatively little gas between electrodes, production of sufficient
ionisation to initiate corona or breakdown requires that either the voltage needed to
start corona must be raised or, if the electrode arrangement permits an alternative path
of greater length, the corona can develop at a lower voltage over the longer path.
The use of power systems at very high altitudes, albeit on a relatively small
scale, thus involves novel applications of the fundamental knowledge that is presently
extant.
1.3.6 Sparkover over insulator surfaces
Conductors between which high potential differences are maintained are usually
separated by solid insulators, frequently porcelain, glass or polymers. Insulators for
outdoor installations such as overhead lines, while maintaining necessary clearances
between conductors, must be designed primarily to cope with rain, fog and pollution
conditions which dictate detailed design (see chapter 6 and Looms [58] for more
details). However, under more specialised requirements for indoor insulation, choice
of materials and their dielectric properties become important.
The development of avalanches and streamers in close proximity to insulating
surfaces is likely to be affected in several ways. In addition to air processes, the
electrons in an avalanche may have a probability of attaching to the insulator surface,
with consequent removal from the ionisation process. This has been postulated to
account for the fast rate of decay of current in a corona which is propagated over
a surface [55]. Nevertheless, it has been established that streamers propagate more
rapidly over a surface than in air, and this has been related to photoelectron emission
from the surface, so effectively increasing the ionisation coefficient and shortening
the lengths of avalanches required to achieve a critical condition (section 1.2.3).
Possibly overriding these influences is the effect of the increase in electric field
around the head of a streamer due to the relative permittivity of the material. In the
streamer trail, it is likely that negative or positive ions will attach to the surface,
depending on the chemical nature and activation energies of the material.
Data is available on some of these effects. Thus, Verhaart et al. [59] have demon-
strated that electrons are, indeed, photo-emitted at an insulating surface to an extent
that depends on the quality of the ultra-violet light emitted by avalanches in the partic-
ular gas surrounding the insulator. In the experiments described, these photoelectrons
had the effect of augmenting the avalanches that were initiated on the insulator sur-
face. The UV light from SF
6
proved to be effective in causing photo-emission from
PTFE, but no significant emission was detected when repeated with carbon dioxide.
However, there was no attenuation of the avalanches in this case, indicating that
unless counterbalanced by photo-emission, attachment was negligible. Generally,
similar effects were obtained by Jakst and Cross [60].
The attachment of ions to the surface is, of course, very well studied by the
Lichtenberg figure technique, which has been made the subject of quantitative
measurement by several authors [5, 61, 62]. Very recently, however, scanning mea-
surements by electrostatic probe have been integrated over the whole discharge region
to show that, with positive impulse corona, the net charge detected on the surface
is only a few per cent of the total charge injected from the circuit [55]. The result
suggests that ions of both polarity settle on the surface after the passage of the dis-
charge current, with a slight preponderance of the same polarity as the electrode
producing the corona.
Further measurements have shown that the ambient electric field required to
sustain streamer propagation over an insulating surface is of the order of 20 per cent
higher than that needed in air. This appears at variance with the generally accepted
reduction in the dielectric strength of an insulator surface, compared with that in air
alone. However, the reduction is due to effects at a highly stressed electrode, where
the combination of a (usually) imperfect contact between electrode and insulator in air
(the so-called triple junction) produces very high localised fields in which lowenergy
discharges trigger the larger discharge between electrodes. This subject is beyond the
scope of this chapter.
1.4 Note
1
Figure from British Standards reproduced with the permission of BSI under licence
number 2003 SK/0157. British Standards can be obtained from BSI Customer
Services, 389 Chiswick High Road, London W4 4AL (Tel +44 (0) 20 8996 9001).
1.5 References
1 FRANKLIN, B.: Experiments and observations on electricity made at
Philadelphia (Cave, London, 1751)
2 LICHTENBERG, G.C.: Novi Comment, 1777, 8, pp. 168177
3 TOWNSEND, J.S.: The conductivity produced in gases by the motion of
negatively charged ions, Phil. Mag. Servi., 1901, 1, pp. 198227
4 LOEB, L.B.: The problem of mechanism of spark discharge, J. Franklin Inst.,
1930, 210, pp. 115130
5 MERRIL, F.H., and von HIPPEL, A.: The atom physical interpretation
of Lichtenberg Figures and their application to the study of gas discharge
phenomena, J. Appl. Phys., 1939, 10, pp. 873887
6 TEICH, T.H.: Emission gasioniseierender Strahlung aus Elektronenawinen,
Zeits. fur Phys., 1967, 199, pp. 378394
7 MEEK, J.M.: The electric spark in air, JIEE, 1942, 89, pp. 335356
8 RAETHER, H.: Arch. Elektrotech., 1940, 34, p. 49
9 BROWN, S.C.: Basic data of plasma physics (Technology Press and Wiley,
New York, 1959)
10 BADALONI, S., and GALLIMBERTI, I.: The inception mechanism of the
first corona in non-uniform gaps, University of Padova, 1972, UPee 72/05,
pp. 3138
11 ROCHE, A.E., and GOODYEAR, C.C.: Electron detachment from negative
oxygen ions at beam energies in the range 3 to 100 eV, J. Phys. B, At. Mol.
Phys., 1969, 2, pp. 191200
12 BOYLETT, F.D.A., andWILLIAMS, B.G.: The possibility of photodetachment
in the impulse breakdown of positive point plane gaps in air, Br. J. Appl. Phys.,
1967, 18, pp. 593595
13 CROMPTON, R.W., HUXLEY, L.G.H., and SUTTON, D.J.: Experimental
studies of the motion of slowelectrons in air, with application to the ionosphere,
Proc. R. Soc. Lond. A, Math. Phys. Sci., 1953, 218, pp. 507519.
14 MORROW, R., and LOWKE, J.J.: Streamer propagation in air, J. Phys. D,
Appl. Phys., 1997, 32, pp. 614627
15 GALLIMBERTI, I., MARCHESI, G., and NIEMEYER, L.: Streamer corona at
insulating surface. Proceedings of 5th international symposiumon High voltage
engineering, Dresden, 1991, paper 41.10
16 SHOCKLEY, W.: Currents to conductors induced by a moving point charge,
J. Appl. Phys., 1938, 10, pp. 635636
17 SATO, N.: Discharge current induced by the motion of charged particles,
J. Phys. D, Appl. Phys., 1980, 13, pp. L3L6
18 ALLEN, N.L., BERGER, G., DRING, D., and HAHN, R.: Effects of humidity
on corona inception in a diverging electric field, IEE Proc. A Phys. Sci. Meas.
Instrum. Manage. Educ., 1981, 128, pp. 565570
19 BURCH, D.S., SMITH, S.J., and BRANSCOMB, L.M.: Photodetachment
of O, Phys. Rev., 1959, 112, pp. 171175; correction, Phys. Rev., 1961,
114, p. 1952
20 HUTZLER, B., et al.: LES RENARDIERES GROUP: Negative discharges
in long air gaps at Les Renardieres, 1978, Electra, 1981, (74), pp. 67216,
published by CIGRE, Paris
21 vonHIPPEL, A., andFRANCK, J.: Der Elektrische durchschlagundTownsends
Theorie, Zeits. fur Phys., 1929, 57, pp. 696704
22 RAETHER, H.: Gasentladungen in der Nebelkammen, Zeits. fur Phys., 1935,
94, pp. 567573
23 GALLIMBERTI, I.: A computer model for streamer propagation, J. Phys. D,
Appl. Phys., 1972, 5, pp. 21792189
24 PHELPS, C.T., and GRIFFITHS, R.F.: Dependence of positive streamer
propagation on air pressure and water vapour content, J. Appl. Phys.,
1976, 47, pp. 29292934
25 ACKER, F.E., and PENNEY, G.W.: Some experimental observations of the
propagation of streamers in low-field regions of an asymmetrical gap, J. Appl.
Phys., 1969, 40, pp. 23972400
26 GELDENHUYS, H.J.: Positive streamer gradient and average breakdown
voltage as functions of humidity. Proceedings of 5th international symposium
on High voltage engineering, Braunschweig, 1987, paper 14.02
27 ALLEN, N.L., and BOUTLENDJ, M.: Study of electric fields required for
streamer propagation in humid air, IEE Proc. A, Sci. Meas. Technol., 1991,
138, pp. 3743
28 ALLEN, N.L., and GHAFFAR, A.: The conditions required for the propagation
of a streamer in air, J. Phys. D, Appl. Phys., 1995, 28, pp. 331337
29 ALLEN, N.L., and MIKROPOULOS, P.N.: Dynamics of streamer propagation
in air, J. Phys. D, Appl. Phys., 1999, 32, pp. 913919
30 FESER, K., and HUGHES, R.C.: Measurement of direct voltage by rodrod
gap, Electra, 1988, (117), pp. 2334
31 INTERNATIONAL ELECTROTECHNICS COMMISSION IEC 60060-1:
High voltage testing, 1989
32 ALLEN, N.L., and DRING, D.: Effect of humidity on the properties of corona
in a rodplane gap under positive impulse voltages, Proc. R. Soc. Lond. A,
Math. Phys. Eng. Sci., 1985, 396, pp. 281295
33 FAIRCLOTH, D.C., and ALLEN, N.L.: High resolution measurements of
charge densities on insulator surfaces, IEEE Trans. Dielectr. Electr. Insul.,
2003, 10, pp. 286290
34 SCHONLAND, B.F.J., and COLLENS, H.: Progressive lightning, Nature,
1933, 132, p. 407
35 ALLIBONE, T.E., and SCHONLAND, B.F.J.: Development of the spark
discharge, Nature, 1934, 134, pp. 735736
36 ROSS, J.N., et al.: LES RENARDIERES GROUP: Postive discharges in long
air gaps at Les Renardieres, Electra, 1977, (53), pp. 31153, Table 3-1
37 WATERS, R.T.: in MEEK, J.M., and CRAGGS, J.D. (Eds): Electrical
breakdown of gases, (Wiley, Chichester, 1978), Chap. 5, pp. 385532
38 WATERS, R.T. (Ed.): LES RENARDIERES GROUP: Double impulse tests of
long air gaps, Proc. IEE A, Phys. Sci. Meas. Instrum. Manage. Educ. Rev.,
1986, 133, pp. 393483
39 PARIS, L., and CORTINA, R.: Switching and lightning impulse discharge
characteristics of large air gaps and long insulator strings, IEEE Trans., 1968,
PAS-87, pp. 947968
40 CARRARA, G.: Investigation on impulse sparkover characteristics of long
rod/rod and rod/plane air gaps. CIGRE report no. 328, 1964
41 BAZELYAN, E.M., BRAGO, E.N., and STEKOLNIKOV, I.S.: The large
reduction in mean breakdown gradients in long discharge gaps with an oblique
sloping voltage wave, Dokl. Akad. Nauk SSSR, 1960, 133, pp. 550553
42 HUGHES, R.C., and ROBERTS, W.J.: Application of flashover characteristics
of air gaps to insulation coordination, Proc. IEE, Part A, 1965, 112, pp. 198202
43 CIGRE: Guidelines for the evaluation of the dielectric strength of external
insulation. Technical brochure no. 72, 1992
44 ZHANG, X., LHADIAN, and WANG, X.: Effect of humidity on the positive
impulse strength at the high altitude region. Proceedings of 8th international
symposium on High voltage engineering, Yokohama, 1993, paper 40.01
45 FRIEDRICH, G.: Temperature dependent swarm parameters in N
2
and air.
Proceedings of 9th international conference on Gas discharges and their
applications, Venezia, 1988, pp. 347350
46 BOUTLENDJ, M., and ALLEN, N.L.: Assessment of air density correction for
practical electrode systems, Eur. Trans. Electr. Power, 1996, 6, pp. 267274
47 DAVIES, A.J., MATALLAH, M., TURRI, R., and WATERS, R.T.: The effect
of humidity and pressure on corona inception in a short air gap at breakdown
voltage levels. Proceedings of 9th international conference on Gas discharges
and their applications, Venezia, 1988, pp. 185188
48 ALLEN, N.L., LAM, D.S.K., and GREAVES, D.A.: Tests on the breakdown
of air in non-uniform electric fields at elevated temperatures, IEE Proc., Sci.
Meas. Technol., 2000, 147, pp. 291295
49 FESER, K., and PIGINI, A.: Influence of atmospheric conditions on the
dielectric strength of external insulation, Electra, 1987, (112), pp. 8395
50 DAVIES, A.J., DUTTON, J., TURRI, R., and WATERS, R.T.: Predictive
modelling of impulse corona in air at various pressures and humidities.
Proceedings of 9th international conference on Gas discharges and their
applications, Venice, 1988, pp. 189192
51 LOEB, L.B.: Electrical coronas (University of California Press, 1965)
52 SIGMOND, R.S.: Corona dischargesinMEEK, J.M. andCRAGGS, J.D. (Eds):
Electrical breakdown of gases (John Wiley, Chichester, 1978), Chap. 4
53 KONG, J., and ALLEN, N.L.: Temperature effect on air breakdown under
switching impulse voltages. Proceedings of 14th international symposium on
High voltage engineering, Delft, 2003
54 WATERS, R.T., JONES, R.F., and BULCOCK, C.J.: Influence of atmospheric
ions on impulse corona discharge, Proc. IEE, 1965, 112, (7), pp. 14311438
55 ALLEN, N.L., and FAIRCLOTH, D.C.: Corona propagation and charge depo-
sition on a PTFE surface, IEEE Trans. Dielectr. Electr. Insul., 2003, 10,
pp. 295304
56 ALLEN, N.L., and KONG, J.: Effect of temperature on corona onset charac-
teristics. Proceedings of 14th international conference on Gas discharges and
their applications, Liverpool, 2002, 1, pp. 271274
57 DUTTON, J.: Spark breakdown in uniform fields, in MEEK, J.M., and
CRAGGS, J.D. (Eds): Electrical breakdown of gases (Wiley, 1978), Chap. 3
58 LOOMS, J.S.T.: Insulators for high voltages (IEE Power Engineering Series,
No. 7, Peter Peregrinus, London, 1988)
59 VERHAART, H.F.A., TOM, J., VERHAGE, A.J.L., and VOS, C.J.: Avalanches
near solid insulators. Proceedings of 5th international symposium on High
voltage engineering, Braunschweig, 1987, paper 37.01
60 JAKST, A., and CROSS, J.: The influence of a solid dielectric spacer on electron
avalanches in Nitrogen at atmospheric pressure, Can. Electr. Eng. J., 1981, 6,
pp. 1420
61 NASSER, E.: The nature of negative streamers in spark breakdown,
Dielectrics, 1963, pp. 110117
62 MUROOKA, Y., TAKADA, T., and HIDAKA, K.: Nanosecond surface
discharge and charge density evaluation, Part 1, Review and experiments,
IEEE Electr. Insul. Mag., 2001, 17, pp. 616
Chapter 2
SF
6
insulation systems and their monitoring
O. Farish, M.D. Judd, B.F. Hampton and J.S. Pearson
2.1 Introduction
Sulphur hexafluoride is used as an insulant in a wide range of power system
applications, including switchgear, gas-insulated substation (GIS) components,
transformers and gas-insulated cables. SF
6
is chemically stable, non-toxic and
non-flammable, and has a high vapour pressure (21 bar at room temperature).
At pressures typical of GIS applications (5 bar) it can be used down to 35
C
without liquefaction occurring. Its wide use in power equipment is promoted by
the fact that, in addition to high dielectric strength, SF
6
has good heat transfer
characteristics and excellent arc-quenching properties.
Although its dielectric strength is three times that of air and, at a pressure of 6 bar
is approximately the same as that of transformer oil, SF
6
is a brittle medium, in that
ionisation builds up very rapidly if a critical field strength is exceeded. In practical
applications, this can happen in the vicinity of any small defects such as electrode
surface protrusions or free conducting particles. Depending on the nature of the defect,
the partial discharges (PD) which occur at these local regions of field enhancement
may result in breakdown of the insulation system. There is, therefore, increasing
interest in the use of techniques for monitoring the PD activity in SF
6
-insulated
equipment with a view to identifying critical defects before they lead to failure.
This chapter reviews the basic ionisation processes which occur in SF
6
, the
streamer mechanism which controls breakdown under relatively uniform field condi-
tions, and the influence of electrode surface roughness on breakdown at high pressure.
The characteristics of the partial discharges (corona discharges) which occur under
the non-uniform field conditions associated with certain types of defect are then dis-
cussed. Following a discussion of the various PDdiagnostic techniques that have been
proposed for use in GIS, an account is given of the principles of the UHF technique
for detection of PD in metal-clad equipment. Finally, the design and calibration of
the sensors used in UHF monitoring are discussed and an explanation given of the
interpretation of the PD patterns recorded in practical UHF monitoring systems.
2.2 Ionisation phenomena in SF
6
The high dielectric strength of SF
6
is due to its property of electron attachment. In
this process, a free electron moving in the applied field, which collides with a neutral
molecule, may be attached to form a negative ion:
SF
6
+e (SF
6
)
This process competes with that of collisional ionisation, by which an electron

with sufficient energy can remove an electron from a neutral molecule to create
an additional free electron:
SF
6
+e (SF
6
)
+
+2e
Ionisation is a cumulative process and, provided that the field is high enough,
successive collisions can produce ever increasing numbers of free electrons which
can result in electrical breakdown of the gas. On the other hand, the heavy, slowly
moving negative ions which are formed by attachment are unable to accumulate the
energy required to cause ionisation and the attachment process therefore effectively
removes electrons and inhibits the formation of avalanches of electrons which might
lead to breakdown.
Some gases, such as nitrogen, hydrogen and argon, do not form negative ions;
others, such as oxygen and CO
2
, are weakly attaching. SF
6
, in common with a
number of other gases containing fluorine or chlorine, is highly electronegative
(i.e. exhibits strong electron attachment). Some gases have a dielectric strength sig-
nificantly greater than that of SF
6
(Table 2.1). However, most present problems of one
kind or another, including toxicity, limited operating pressure range or production of
solid carbon during arcing. SF
6
is therefore the only dielectric that is accepted as suit-
able for GIS applications, although some high strength gases have been considered
for use in mixtures with SF
6
.
In later sections of the chapter, the mechanisms governing the formation of partial
discharges and breakdown in compressed SF
6
insulation are reviewed. It is therefore
useful at this stage to consider howthe (net) rate of electron production in SF
6
depends
on the gas pressure and the applied field.
The average energy attained by charged particles moving in a gas varies linearly
with both the applied field E and the average mean free path between collisions.
Since is proportional to the reciprocal of the number density N, the energy varies
directly with E/N or, if the temperature is constant, with E/p, where p is the gas
pressure.
The frequency of collisions made by a charged particle of a given energy increases
with number density (or pressure). Thus, for an electron moving in a gas, for example,
the probability of ionisation for a single collision will be a function of the energy
parameter E/p, and the rate of ionisation will depend both on pressure and on E/p.
SF
6
Table 2.1 Dielectric strength of certain
gases relative to SF
6
[1]
Gas Relative
strength
H
2
0.18 non-attaching
air 0.3
weakly attaching CO
2
0.3
CO 0.4
C
2
F
8
0.9
strongly or very
strongly attaching
CCl
2
F
2
0.9
SF
6
1.0
C-C
4
F
8
1.3
C-C
4
F
6
1.7
C
4
F
6
2.3
Considering a swarm of electrons moving in a gas under a constant field, the
growth of ionisation rate is defined in terms of the number of ionising collisions per
electron per cm travel in the gas parallel to the applied field. This parameter is known
as the ionisation coefficient, , where = pf
1
(E/p), that is:
p
= f
1
_
E
p
_
Similarly, the removal of electrons from the swarm is determined by an attachment
coefficient, , defined as the number of attachments per electron per cmtravel. Hence:
p
= f
2
_
E
p
_
The net ionisation then depends on the balance between ionisation and attachment.
If > , attachment dominates over ionisation and no discharge growth is possible;
if > , cumulative ionisation can occur.
Figure 2.1 shows the net (pressure-reduced) ionisation coefficient ( )/p as a
function of E/p for air and SF
6
. It can be seen that the critical reduced field strength
at which ( ) = 0 is about 89 kV/cm bar in SF
6
, compared with only 27 kV/cm
bar in air. This explains the high dielectric strength of SF
6
relative to air as no build-up
of ionisation can occur until the reduced field exceeds the critical value (E/p)
crit
.
It is worth noting the steep slope of the curve of ( )/p versus E/p in SF
6
.
This means that SF
6
is a relatively brittle gas in that, once (E/p)
crit
is exceeded,
the growth of ionisation is very strong. This is significant in situations where stress-
raising defects are present in gas-insulated equipment as intense ionisation activity
will occur in the regions where E/p > (E/p)
crit
and this may initiate complete
breakdown of the insulation.
300
200
100
air
50
27 kV/cm bar
100
89 kV/cm bar
E/p
SF
6
100
200
Figure 2.1 Effective ionisation coefficients in air and SF
6
Also worthy of note is the fact that the net ionisation coefficient in SF
6
can be
represented by the linear relationship:

p
= A
_
E
p
_
B
where A = 27.7 kV
1
and B = 2460 bar
1
cm
1
. The critical reduced field strength
is therefore:
_
E
p
_
crit
=
Bp
A
= 88.8 kV/cm bar
This simple relationship is useful in estimating onset voltages in SF
6
insulation.
2.3 Breakdown mechanisms in low divergence fields
As discussed above, the build-up of ionisation in SF
6
is possible only under con-
ditions where the (pressure-reduced) field exceeds a critical value (E/p)
crit
of
89 kV/cm bar.
For highly divergent fields (as, e.g., for the case of a sharp protrusion on a high
voltage conductor) ionisation will be confined to a critically-stressed volume around
the tip of the protrusion. In this situation localised PD, or corona, will be the first
phenomenon observed as the applied voltage is increased. Breakdown under these
conditions is a complex process, because of the effects of the space charge injected
by the prebreakdown corona.
As any stress-raising defect in gas-insulated equipment will result in PD activity,
it is important to understand non-uniformfield discharge mechanisms. However, GIS
SF
6
are designed for relatively low field divergence and it will be useful first to consider
the simple case of breakdown in SF
6
under uniformfield conditions, before reviewing
the phenomena associated with particulate contamination or other defects.
2.3.1 Streamer breakdown
For a perfectly uniform field (planeplane electrode geometry), no ionisation activity
can occur for reduced fields less than the critical value. Above this level, ionisation
builds up very rapidly and leads to complete breakdown of the insulation (formation
of an arc channel).
The first stage of the breakdown involves the development of an avalanche of
electrons. The growth of this avalanche froma single starter at the cathode can readily
be found by computing the net electron multiplication. Considering a swarm that has
grown to contain n(x) electrons at position x in a gap of width d; then, in travelling
a further incremental discharge dx, these will generate a net new charge:
dn(x) = ( )n(x) dx = n(x) dx
as a result of ionising and attaching collisions with neutral molecules, where is the
net ionisation coefficient.
Integration over the interval 0 to x gives the number of electrons in the avalanche
tip at that stage in its growth:
n(x) = exp
__
x
0
dx
_
= exp( x)
In crossing the whole gap, an avalanche of exp( d) electrons is created.
In itself, the occurrence of avalanches does not constitute breakdown. For exam-
ple, if conditions were such that = 5 then, in a 1 cm gap at 1 bar, the current gain
would be e
5
150. The normal low background conduction current density (due
to collection of free charges present in the gap) would be increased as a result of
ionisation from 10
13
A/cm
2
to 10
11
A/cm
2
, but the gap would still be a very
good insulator. However, as illustrated in Figure 2.1, increases very quickly when
the reduced field exceeds (E/p)
crit
and the multiplication can rapidly reach values
of 10
6
or greater, with most of the charge confined to a very small region at the head
of the avalanche (approximately a sphere of typically 10 m radius).
The bipolar space charge generated by the ionisation process results in local
distortion of the applied field such that ionisation activity ahead of, and behind,
the avalanche tip is greatly enhanced. At a critical avalanche size (exp( x) = N
c
),
the space charge field is high enough to generate rapidly moving ionisation fronts
(streamers) which propagate at 10
8
cm/s towards the electrodes. When these bridge
the gap, a highly conducting channel is formed within a few nanoseconds.
For pressures used in technical applications (p > 1 bar), the streamer process is
the accepted breakdown mechanism in SF
6
under relatively uniform field conditions.
The critical avalanche size for streamer formation is found to be that for which
the streamer constant k = nN
c
is approximately 12. The breakdown voltage is then
easily calculated using the linear relationship between /p and E/p:

p
= A
_
E
p
_
B
where A = 27.7 kV
1
and B = 2460 bar
1
cm
1
.
The minimum streamer inception or breakdown level will occur when the critical
avalanche size is achieved at the anode. Thus:
d = AEd Bpd = k
The breakdown voltage V
s
(=Ed) is then:
V
s
=
B
A
(pd) +
k
A
= 88.8 (pd) +0.43 (kV)
where pd is in bar cm.
Note that the breakdown voltage is a function only of the product (pressure x
spacing). This is an example of the similarity relationship (Paschens Law) which
allows gas-insulated equipment to be made more compact by increasing the pressure
above atmospheric.
As indicated above, once the gap is bridged arc formation in SF
6
is extremely
rapid. The voltage collapse time depends on the pressure (p), spacing (d) and geo-
metry, and is typically 10 d/p nanoseconds. In certain situations, this can present
serious problems in GISequipment. Sparking during closure of a disconnector switch,
for example, can generate travelling waves in the GIS bus which have very fast rise
times (up to 100 MV/s). Doubling at open circuits elsewhere in the systemcan result
in insulation being stressed with high amplitude (>2 p.u.) pulses with very short rise
times (<10 ns). There are also problems with grounding and shielding, as very high
fields (500 kV/m) can appear across parts of the grounded enclosure during the pulse
transit.
2.3.2 Quasi-uniform fields (coaxial cylinders)
If the field is varying with position across the gap, the initial avalanche formation will
occur within a critical volume for which ( ) > 0 (i.e. E/p > 88.8 kV/cm bar).
Under these conditions, the streamer inception criterion is:
exp
__
x
c
0
(x) dx
_
= N
where x is the distance fromthe inner electrode along a field line and x
c
is the position
of the boundary of the ionisation region.
For coaxial electrode geometry (inner radius r
0
, outer r
1
) the field distribution is:
E(r) =
V
r
ln
_
r
1
r
0
_
SF
6
Also, at onset, = 0 at position r
c
, so that:
E(r
c
) = E
crit
=
Bp
A
Using these relationships, together with the streamer criterion, it can easily be shown
that the surface field at onset is:
E(r
0
) =
Bp
A
_
1 +
_
k
Bpr
0
_
1/2
_
and that:
x
c
=
_
kr
0
Bp
_
1/2
With the above values of A and B, this yields:
E(r
0
)
p
= 89
_
1 +
_
0.07
pr
0
__
(pr
0
in bar cm)
and
x
c
=
0.07
r
0
p
(cm)
For the large curvature electrodes and high pressures used in GIS, the field at
the inner conductor at onset is therefore very close to the critical reduced field of
89 kV/cm bar.
Note that the streamer forms when the primary avalanche has developed a rela-
tively short distance. For r
0
= 8 cm, p = 4 bar, for example, x
c
will be 1 mm.
The streamer will then propagate until the combination of the space charge field and
the geometric field is unable to sustain further ionisation. In order for breakdown to
occur, it will then be necessary to increase the surface field above the onset level.
In the relatively low divergence field in a (clean) GIS system only a small increase
above the onset voltage is necessary to initiate breakdown.
2.3.3 Effect of surface roughness
Although laboratory measurements using polished coaxial electrodes are in agreement
with the theoretical criterion that the inner surface field at breakdown should be close
to the critical field of 89 kV/cm bar, this value cannot be sustained in large scale
equipment with a practical (machined) surface finish.
One reason for this is the fact that increased ionisation occurs in the vicinity of
microscopic surface protrusions (surface roughness). This results in reduction of the
breakdown field strength by a factor . Figure 2.2 shows calculated values of the
factor as a function of the product ph (pressure protrusion height) for a range of
spheroidal protrusions [2].
It can be seen (a) that the breakdown voltage can be reduced to a low level and
(b) that there is a critical protrusion size for the onset of roughness effects.
h
b
h/b
1
2
10
10
1
0.5
1.0
r
o
u
g
h
n
e
s
s

f
a
c
t
o
r
,

10
2
10
3
ph (bar m)
10
4
Figure 2.2 Roughness factor for uniform field breakdown in SF
6
x
r
E(x)
E
0
Figure 2.3 Hemispherical protrusion on a uniform field electrode
The existence of a threshold value of ph can readily be demonstrated [2] for the
hemispherical protrusion shown in Figure 2.3. For this case, the field at an axial
distance x above the protrusion is given as:
E(x) = E
0
_
1 +
2r
3
(x +r)
3
_
For the protrusion to have no effect, E
0
(the macroscopic field) at onset must be equal
to (E/p)
crit
(=Bp/A).
Also:
(x) = AE(x) Bp = Bp
_
1 +
2r
3
(x +r)
3
_
Bp =
2Bpr
3
(x +r)
3
Breakdown occurs when:
_
x
c
0
(x) dx = k
SF
6
that is
2Bpr
3
_
1
2
(x +r)
2
_
x
c
0
= k
therefore
1r
2
(x
c
+r)
2
=
k
Bpr
therefore
x
c
r
= 1
_
1
1 k/Bpr
_
1/2
With k = 12 and B = 0.246 bar
1
m
1
:
x
c
= r
_
1 (1 49/pr)
1/2
_
For x
c
to be real, pr must be >50 bar m. At a working pressure of 5 bar, surface
roughness would therefore begin to affect the onset level for protrusion heights greater
than 10 m.
Because of surface roughness effects (and other electrode phenomena such as
micro discharges in charged oxide layers, etc.) practical SF
6
-insulated equipment is
designed such that the maximum field is everywhere less than 40 per cent of the
critical value. In a typical GIS, for example, the basic insulation level (BIL) will
correspond to a peak reduced field of only 35 kV/cm bar.
With a good technical surface finish, streamers should not form in a clean coaxial
electrode systemunder these conditions. Further, if a local defect does cause streamer
formation, the streamer should not be able to propagate into the low field region of
the gap. The fact that breakdown can occur, even at the lower reduced field associated
with theACworking stress (15 kV/cmbar) indicates that an additional mechanismis
operative. This is discussed in the following section on non-uniform field breakdown
in SF
6
.
2.4 Non-uniform field breakdown in SF
6
Highly divergent fields can exist in GIS under certain conditions as, for example,
when a needle-like free metallic particle is attracted to the inner conductor or is
deposited on the surface of an insulator. Such defects can result in very lowbreakdown
levels and, with large defects (e.g. particles several mm long), failure can occur
even at the working stress of the equipment. For this reason, there have been many
laboratory studies of the breakdown characteristics of highly non-uniform field gaps
in SF
6
.
These studies have shown that there are two distinct types of breakdown, depend-
ing on the rate at which the voltage is applied to the gap. When the stress is applied
relatively slowly, as with alternating voltage or long rise time switching surges,
corona space charge plays an important part in controlling the field distribution by
200
100
V
(
k
V
)
1 2
pressure (bar)
3 4
breakdown
onset
r
0
=5mm
3mm
2mm
1.5mm
Figure 2.4 AC corona onset and breakdown characteristics for a 40 mm rodplane
gap in SF
6
[4]
the so-called corona stabilisation process [3]. With shorter rise time surges (light-
ning impulse or fast transients), breakdown occurs directly by a stepped leader
mechanism [5]. For both cases, the breakdown voltage is lower when the high field
electrode is positive and most attention has therefore been given to breakdown under
positive point conditions.
2.4.1 Corona stabilised breakdown
Figure 2.4 shows ACvoltagepressure characteristics for point-plane gaps in SF
6
[4].
The shape of these curves is typical of all non-uniform field gaps with slowly varying
voltage applied, in that there is (a) a broad pressure region over which the breakdown
voltage is much higher than the (streamer corona) onset voltage and (b) a critical
pressure at which breakdown occurs directly at onset (i.e. the first streamer leads
directly to breakdown).
The peak in the mid-pressure range is due to the effects of space charge injected
by streamer activity around the point.
For a positive point, for example, the electrons generated by the corona are quickly
removed at the point while the positive ions diffuse relatively slowly into the low
field region. This space charge tends to shield the point and stabilises the field there
to a level close to the onset value.
As the voltage is raised, the space charge density (and the shielding effect) inten-
sifies, and a voltage considerably above onset is required to cause breakdown. The
breakdown usually occurs as a result of filamentary (leader) discharges developing
around the shielding space charge, so that spark channels in the stabilisation region
typically take a very irregular, curved path.
SF
6
As the pressure is increased, the individual streamers become more intense, the
corona regionbecomes confinedtoa smaller regionat the tipof the point andthe stabil-
isation becomes less effective, so that the breakdown voltage is reduced. Eventually,
the shielding effect is lost, and the streamer which forms at onset is able to initi-
ate a discharge which develops completely across the gap at the onset voltage. This
discharge has been shown to be identical to the stepped leader discharge which is
found to occur in non-uniform field gaps under fast pulse conditions.
2.4.2 Leader breakdown
With fast-fronted surges, where the voltage passes rapidly through the theoretical
streamer onset level, the initial streamers can be very intense and may lead to the
formation of a highly ionised leader channel before there is time for space charge
stabilisation of the field at the tip of the protrusion [5].
In addition to the rate-of-rise of voltage, the statistics of appearance of initiatory
electrons may play an important role. For negative-point conditions, electrons are
produced by field emission; with the positive point, however, the trigger electrons
result from detachment from negative ions in the vicinity of the point [6].
Before the stress is applied, the gas contains a negative-ion population of a few
thousand ions per cm
3
. (These are produced by the action of cosmic rays, which
typically generate 10 ion pairs per cm
3
per second in gases at atmospheric pressure.)
For discharge initiation to occur, it is necessary to find one of these ions in the very
small critical volume where ( > ). This critical volume is vanishingly small at the
theoretical onset level and increases with voltage.
For a fast-fronted wave, the field may therefore be well above the minimumonset
level when inception occurs so that the streamer corona is more vigorous than would
be the case for AC or DC stress.
If the streamer corona is large enough, a stepped leader discharge may be
initiated. The mechanism of the stepped leader may be summarised, with reference
to Figure 2.5, as follows.
During the dark period ab which follows the initial corona, charge separation
in the streamer filaments generates a succession of ionising waves which build-up
their conductivity. Eventually, one of the streamer filaments is transformed into a
highly conducting leader channel step; this behaves essentially as an extension to the
point electrode and a new corona burst immediately occurs at its tip b. The range
of this second corona determines the length of the second channel step c. During
each streamers dark period, there are regular reilluminations of the leader channel,
probably associated with the relaxation processes which are occurring in the streamer
filaments.
The leader propagates into the gap in steps typically of a fewmmuntil the streamer
activity is too weak for further channel steps to form. If the voltage is high enough,
the leader can cross the gap, resulting in breakdown. As the interstep interval is
100 ns for p 3 bar, the breakdown formative time lag can be greater than 1 s.
As the field along the leader channel is much lower than that in the streamer
filament, breakdown can occur by the stepped leader process at much lower voltages
time
a b c d
Figure 2.5 Schematic of leader development
than would be required for streamer breakdown. For point-plane gaps, the minimum
leader breakdown voltage is found to be almost independent of pressure and average
breakdown fields of 25 kV/cm are typical of short (2050 mm) gaps [7].
In configurations where the background field in the low field region is falling
less steeply (as, e.g., for the case of a particle fixed to the inner conductor of a GIS),
leader breakdown can occur at average fields of only 15 kV/cm.
Figure 2.6 shows Vp characteristics in SF
6
for corona-stabilised breakdown and
for the minimumbreakdown voltage under impulse conditions. As p
1
is typically only
about 0.5 bar, the minimum breakdown voltage in non-uniform fields at pressures
typical of GIS is determined by the conditions for leader propagation in the absence
of preexisting corona space charge. Models have been developed [8, 9] which allow
the conditions for direct leader inception and propagation to be predicted for a wide
range of geometries.
It must be emphasised that, under surge conditions, the leader propagation field
is the minimum level at which breakdown can occur. Depending on the statistics
of initiation of the primary streamer, there is a probability of corona stabilisation
occurring so that, even for lightning impulse, the 50 per cent probability voltage
pressure characteristic will exhibit a stabilisation peak. It is important therefore to
determine the low probability breakdown level when carrying out surge breakdown
tests in inhomogeneous fields.
2.5 Breakdown in GIS
2.5.1 Streamer-controlled breakdown
The design stresses used in GIS are low enough (<50 per cent E
crit
) that streamer
inception will not occur even at the full rated impulse level. However, scratches or
other small defects on the inner electrode surface may result in streamer formation.
SF
6
streamer
breakdown
leader
breakdown
corona-stabilised
(AC or DC) breakdown
minimum
impulse
breakdown
level
streamer
onset
gas pressure, p
p
1
p
c
v
o
l
t
a
g
e
,
V
Figure 2.6 IdealisedVpcharacteristics for minimumimpulse (direct leader) break-
down and corona-stabilised (ACor DC) breakdown in a point-plane gap
in SF
6
[7]
For a small scratch or protrusion (h 100 m), the field perturbation is very
localisedandthe geometric fieldhas tobe quite highinorder toachieve streamer onset.
Such defects would probably be detected only under impulse-voltage test conditions
at levels close to the BIL.
At suchhighvoltages, the conditions for breakdown(whether bythe leader process
or by a direct streamer channel mechanism) are automatically satisfied at onset. The
breakdown voltage of the GIS may therefore be predicted on the basis of the streamer
criterion:
exp
__
x
0
(AE(x) Bp) dx
_
= N
c
provided that E(x), the spatial distribution of the perturbed field, is known.
As discussed earlier, the statistics of discharge initiation can play an important role
in determining the probability of breakdown in perturbed quasi-homogeneous fields
and this has significant implications for insulation coordination on SF
6
-insulated
equipment. Various models have been developed for calculating surge breakdown
probability on the basis of the negative-ion density distribution and the evolution of
the critically-stressed volume with time during the surge [10, 11].
2.5.2 Leader breakdown
For large defects, such as needle-like particles of several mm length attached to
the high voltage conductor, the onset voltage for streamer corona will be low.
This means that the onset voltage is lower than the leader propagation voltage and
breakdown is preceded by corona. This is the situation discussed in the section on
non-uniform fields, where either corona-stabilised or direct leader breakdown may
occur, depending on the voltage waveform.
2.5.3 Particle-initiated breakdown
Free conducting particles (FCPs) are the most common cause of failure in GIS, and
long, thin particles are most dangerous because of the strong field enhancement
associated with such defects.
If FCPs are present in a coaxial system they become charged by the applied field
and, at a relatively low voltage, lift-off will occur. If the particle is rod shaped, it will
stand up on the outer conductor and corona onset will occur. For DC stress, parti-
cles will cross the gap at the onset voltage. For AC conditions the particles initially
make small hopping excursions at the outer electrode. As the stress is increased, the
excursions become longer and, because of inertial effects and the bouncing action
at the electrode, the bounce interval becomes longer. At each contact with the elec-
trode, a PD occurs and the interval between contacts is an important parameter in the
assessment of the severity of particle-induced PD activity in GIS.
As the voltage is increased, particles may cross the gap to the upper electrode
where they will receive a newcharge, which depends on the polarity and magnitude of
the applied voltage at the instant of contact, and will then move back towards the lower
electrode. The crossing does not necessarily lead to breakdown, and a voltage increase
will usually be necessary for the conditions for leader breakdown to be achieved.
In coaxial electrode systems, breakdown is most likely to occur when a particle
strikes the inner conductor just as the voltage reaches a positive maximum [12].
The particle then behaves like a needle fixed to the HV conductor.
Early studies of the mechanism of AC particle-initiated breakdown in GIS led to
some confusion, as the particle-triggered breakdown voltagepressure characteristics
did not show the strong peak observed with AC stress applied to a coaxial electrode
system having a particle fixed to the inner conductor [13]. This led to the consider-
ation of mechanisms such as density reduction in the wake of the moving particle,
and the possible triggering action of the microdischarge which occurs at the instant
of contact.
The actual reason for the disparity between the fixed and free particle data may be
inferred fromthe earlier discussion on corona-stabilised and direct leader breakdown.
For the fixed particle, there is enough time under ACconditions to guarantee effective
stabilisation at each voltage maximum, resulting in the typical peaked Vp character-
istic. With the free particle, however, the sudden arrival of the particle, together with
the step function increase in the field at the tip as a result of the small spark occur-
ring on contact, means that the behaviour is more similar to that for a fixed particle
subjected to a fast fronted impulse voltage. Under these conditions, the minimum
breakdown voltage is that associated with direct leader breakdown, which is almost
independent of pressure (see Figure 2.6).
SF
6
In practice, the probability of breakdown in GIS at each contact will be a func-
tion of the particle charge, velocity and orientation, the phase and magnitude of the
applied voltage, the statistics of breakdown of the particleelectrode microgap and
the probability of formation of a stabilisation corona. It may therefore be necessary
to wait for a relatively long period (2030 s) to ensure that a test particle will not
trigger breakdown at a given voltage.
Measurements made in a 125/250 mmcoaxial systemwith fixed and free particles
showed good agreement between the (t = 20 s) AC breakdown voltage with free
particles and the minimum breakdown voltage with 1 s rise time impulses applied
to a particle fixed to the inner conductor [14]. As the minimum fixed point impulse
breakdown level corresponds to breakdown by a leader mechanism, it is clear that
models of leader inception and propagation can be used to predict the conditions of
particulate contamination which will result in breakdown in GIS.
Present indications are that, for normal working stresses, particles of length less
than 4 mm should not be able to cause breakdown [9]. However, smaller particles
may be scattered onto the surfaces of insulating barriers or spacers, where they may
cause breakdown under subsequent impulse stresses. It is important, therefore, to
ensure that free conducting particles of significant size (1 mm) are not present in
GIS under working conditions.
2.6 Possible improvements in SF
6
insulation
The dielectric performance of present designs of GIS is probably close to the best
that can be achieved with careful component design and good quality control, using
existing materials and construction methods. Techniques which may offer further
improvements in GIS insulation include.
2.6.1 Use of additives or gas mixtures
The dielectric properties of SF
6
can be significantly improved [15] by using leader-
suppressing additives such as Freon 113, although the use of such additives in GIS
would require confirmation that they have no effect on the solid insulation in the
system. Gas mixtures containing buffer gases such as N
2
in concentrations of up to
80 per cent have dielectric strengths which are not much below that of SF
6
under
clean conditions and may be less susceptible to particulate contamination. SF
6
/N
2
mixtures are particularly attractive for use in gas-insulated transmission lines because
of their lower cost and reduced environmental impact.
2.6.2 Improved spacer formulation and construction
There is continual progress in the development of resins and fillers with improved
properties in terms of mechanical strength, resistance to tracking and compatibility
with SF
6
discharge products. The use of semiconducting surface coatings to prevent
the build-up of surface charge may also be beneficial, and the presence of ribs on the
surface of the spacers may offer significant advantages, particularly in the presence
of particulate contamination [16].
2.6.3 Particle control
Although every effort is made to ensure that FCPs are removed during preassembly
cleaning of GIS, particles can be produced during operation, for example as a result
of abrasion of sliding contacts. Simple slotted trays in the outer conductor make
very effective particle traps and their use in the vicinity of solid spacers can offer
useful protection. Other proposals for particle control have included techniques for
covering FCPs with a sticky insulating coating by post-assembly polymerisation of
an appropriate additive to the SF
6
.
Even if improvements are made in one or more of the above areas, it is probable
that it will never be possible to completely eliminate every defect that may result
in local stress enhancement in GIS. It is important, therefore, to ensure that the
procedures used in factory and/or site testing are able, where possible, to indicate
the presence of particles or other defects and that the tests used do not in themselves
cause damage to the insulation, or create conditions which may cause later failure
(e.g., by scattering particles on to spacer surfaces).
These considerations have led to the development of diagnostic techniques which
allow the presence of defects to be recognised as a result of their partial discharge
activity so that action can be taken to remove them before a failure occurs. The
remainder of this chapter is devoted to a discussion of these techniques.
2.7 Partial discharge diagnostic techniques for GIS
2.7.1 Introduction
Although the reliability of GIS is high [17], any internal breakdown that does occur
invariably causes extensive damage and an outage of several days duration is needed
to effect the repair. During this time the associated circuit may be out of operation
and the consequential losses can be high, especially if the GIS is operating at 420 kV
or above. If in addition the GIS is connecting the output of a nuclear station to the
transmission network and the breakdown leads to a reactor shutdown, the financial
penalties could be most severe.
Modern designs of GIS, such as that shown in Figure 2.7, have benefited from
the experience gained with earlier versions, and they have a high level of reliability.
In the future it appears likely that further improved GIS designs will be supplemented
by quite extensive diagnostic monitoring, which will have two main advantages;
it will allowthe user to adopt condition-based maintenance and reduce revenue expen-
diture, and will lead to the ideal situation of being able to detect a developing fault in
time to prevent an unplanned outage.
Much progress in diagnostic techniques for GIS has been made in recent years,
and they are used increasingly in factory testing, site commissioning and during the
service life of the equipment [1822]. In the UK, diagnostic couplers are specified for
SF
6
Figure 2.7 A modern 400 kV GIS
all newGIS, and in some cases have been fitted retrospectively to existing substations.
The insulation can then be monitored for signs of any incipient weakness, and action
taken to prevent it developing into complete failure. With one technique, this is being
done continuously and remotely, in line with the trend among utilities to operate GIS
unattended.
So far the main objectives of GIS diagnostics have been to detect whether there is
any defect in the GIS, to identify it as a particle, floating shield and so on, and to locate
it so that it can be repaired. Various diagnostic techniques have been demonstrated in
the laboratory, and with some of them several years experience on site, both during
commissioning tests and with the GIS in service, has been gained. The results of the
on-site work have been very promising, and have shown the need for these further
developments:
(i) Complex and often intermittent discharges can be found in GIS, and a better
understanding of the physical processes leading to breakdown is needed to
allow the diagnostic data to be interpreted and its significance assessed.
(ii) Continuously monitoring one or more GIS can produce very large quantities of
data, and it is important not to overburden the engineer with its interpretation.
The discharge data needs to be analysed by an expert system, and the engineer
informed only when some condition arises which needs attention.
(iii) Amonitor installed in a GIS to detect defects in the insulation can in addition be
used to record the condition of circuit breakers, transformers and other plant,
and so provide a complete diagnostic system on which predictive maintenance
of the GIS can be based [20].
It is these aspects which are becoming increasingly important, and where the main
advances can be expected.
2.7.2 The range of diagnostic techniques for PD detection
2.7.2.1 Fundamental processes
The statistics of GIS reliability [23, 24] show that the most common cause of electri-
cal failure is a free metallic particle, which can become attracted to the high voltage
conductor and produce a microdischarge which triggers breakdown. Other causes of
failure are discharges from any stress-raising protrusion, capacitive sparking from an
electrode which is not properly bonded to either the high voltage conductor or earth,
and so on; and the common feature of all these defects is that they generate PD activ-
ity in advance of complete breakdown. With the exception only of the mechanical
noise from a bouncing particle, PD detection is the basis of all dielectric diagnostics
in GIS.
A PD is the localised breakdown of gas over a distance of usually less than
a millimetre. For surface defects such as small protrusions, the discharge takes the
form of corona streamers which give rise to current pulses with very short rise times
(<1 ns). Discharges in voids, and microsparks associated with poor contacts or with
the transport of conducting particles, are also characterised by high rates of change of
current. In all cases, the very short rise time of the PD pulse causes electromagnetic
energy to be radiated into the GIS chamber, and the energy dissipated in the discharge
is replaced through a pulse of current in the EHV supply circuit. In microsparks
and intense coronas, the discharge is followed by rapid expansion of the ionised
gas channel, and an acoustic pressure wave is generated. PD is also accompanied
by the emission of light from excited molecules, and by the creation of chemical
breakdown products. The PD therefore has many effects physical, chemical and
electrical and in principle any of them could be used to reveal the presence of the
discharge.
2.7.2.2 Light output
Detecting the light output from a discharge is probably the most sensitive of all
diagnostic techniques, because a photo-multiplier can detect the emission of even
a single photon. The radiation is primarily in the UV band, and since this is absorbed
strongly both by glass and SF
6
it is necessary to use quartz lenses and a reasonably
short path length. Although this is a powerful laboratory tool for finding the onset
of activity from a known corona point, there are many difficulties in using it to detect
a discharge which might be anywhere in a GIS.
2.7.2.3 Chemical byproducts
This approach initially appears attractive because chemical decomposition is immune
to the electrical interference which is inevitably present in the GIS, and with any
steady discharge the concentration of the diagnostic gas should rise in time to a level
SF
6
where it can be detected (this assumes, of course, that an absorbing reagent is not
used in the chambers).
The main decomposition product of sulphur hexafluoride is sulphur tetrafluoride
(SF
4
), but this is a highly reactive gas. It reacts further, typically with traces of water
vapour, to form the more stable compounds thionyl fluoride (SOF
2
) and sulphuryl
fluoride (SO
2
F
2
). These are the two most common diagnostic gases, and, by using
a gas chromatograph and mass spectrometer, they may be detected with sensitivities
down to 1 p.p.m. As a simpler but less sensitive alternative, chemical detection tubes
can be used.
In small-volume laboratory tests, a reasonably small discharge of 1015 pC can
be detected, typically after some tens of hours. However, in a GIS the diagnostic gases
would be greatly diluted by the large volume of SF
6
in which they occur, and much
longer times would be needed. It therefore appears that the chemical approach is too
insensitive to be considered for PD monitoring.
2.7.2.4 Acoustic emission
Acoustic signals arise both from the pressure waves caused by partial discharges and
from free particles bouncing on the chamber floor. The latter is the only instance
of a diagnostic signal not coming from a PD (although of course the particle gen-
erates a PD as well). The signals in GIS have a broad bandwidth, and travel from
the source to the detector by multiple paths [21, 25]. Those originating at the cham-
ber wall propagate as flexural waves, at velocities which increase with the square
of the signal frequency to a maximum of approximately 3000 m/s. Propagation
through the gas is at the much lower velocity of 150 m/s, and the higher frequen-
cies in the signal are absorbed quite strongly. The alumina- or silica-filled epoxide
barriers which are used to support the inner conductor also attenuate the signal
markedly.
The different propagation velocities of the wave as it passes through various
materials, and the reflections occurring at boundaries between them, give rise to a
complex acoustic waveform. This signal can be picked up by accelerometers or acous-
tic emission sensors attached to the outside of the chamber. The acoustic signal froma
particle bouncing on the chamber floor is characterised by a signal not correlated with
the power frequency cycle. It also has other features, such as the crest factor (ratio of
the peak/r.m.s. value), the impact rate and the ratio of the lift-off/fall-down voltages,
fromwhich the particle shape and its movement pattern can be inferred. Other sources
of discharge may be identified in a similar way from their own characteristics [26].
One advantage of acoustic measurements is that they are made non-intrusively,
using external sensors which may be moved from place to place on the GIS. Because
of the rather high attenuation of the signals, the sensors should preferably be on
the chamber containing the source. This in itself gives the approximate location
of the defect, but a more accurate position can be found to within a few cm using
a second sensor and a time of flight method.
The acoustic technique is not suited to a permanently installed monitor, because
too many sensors would be needed.
2.7.2.5 Electrical methods
There are two approaches to detecting the electrical charge in a partial discharge;
in the external circuit by a conventional PD measurement system, and internally by
detecting the resonances set up in the GIS chambers.
(i) Conventional method The test circuit is that given in IEC Publication 270, and
the charge flowing through a coupling capacitor fitted in parallel with the GIS
is measured using a quadrupole and detector. The PD current pulse at the defect
has a duration of less than 1 ns, and propagates as a travelling wave in each
direction along the chambers. The pulses are attenuated and undergo multiple
reflections, but do not appear immediately in the external circuit. After about
a microsecond or so, the pulses die away and the GIS appears to the external
circuit as a lumped capacitor with a depleted charge. From then, a replacement
charge flows into the GIS, and is measured by the detector.
To obtain the maximum sensitivity of measurement, a completely shielded
test arrangement is required [27], which is possible for a test assembly but
may be inconvenient when testing a complete GIS. Also, the total capacitance
of a GIS is high, and it must be divided into sections for test. In addition,
there is no means of locating the discharge, and since a coupling capacitor is
not normally provided in a GIS the technique cannot be used for in-service
measurements.
(ii) UHF method The current at the PDsite rises in less than a nanosecond, and can
radiate EM waves with energy spectra extending to frequencies of 2000 MHz
or more. This excites the GIS chambers into various modes of electrical reso-
nance, which because of the low losses in the chambers can persist for up to
a microsecond. The resonances are indicative of PD activity, and if they are
picked up by couplers installed in the GIS may be displayed on a spectrum
analyser.
The resonant technique was developed in the UK [22], where over the past
15 years or so much experience has been gained using UHF for PD detection at
frequencies from 300 to 1500 MHz [28].
The UHF technique will later be described in detail, when it will be seen that
its advantages are its high sensitivity; the ability to locate discharges accurately
by time of flight measurements; and that it can readily be used in a continuous
and remotely operated monitoring system.
2.7.3 Comparison of the techniques
It is difficult to compare the various diagnostic techniques from results gained under
different experimental conditions, so a CIGRE Working Group (15-03) arranged for
them to be used simultaneously to detect a range of artificial defects [29]. The tests
were made in a 6 m long section of 420 kV GIS chambers into which was placed one
of the following defects:
a free metallic particle
a particle attached to the surface of a barrier
SF
6
10 15
10
5
0
5
1
50 100 150 200 250
U, kV
300
b
c
a
Figure 2.8 Signal to noise ratio of PDfroma needle on the busbar (taken from[29])
a IEC 270 (pC/pC)
b UHF (dB)
c acoustic (mV/mV)
a corona point at the HV conductor
a similar point on the chamber wall.
The vessel was energised by a 0510 kV metalclad test transformer, the voltage
being increased slowly until breakdown occurred. During this time, diagnostic
measurements were made using the following techniques:
a conventional electrical detection according to IEC Publication 270 with either
a standard detector at 1 MHz, or the phase-resolved partial discharge (PRPD)
evaluation system at 200 kHz [30]
b UHF, using an internal coupler at frequencies up to 1500 MHz
c acoustic, using an external acoustic emission sensor at 34 kHz
d chemical, using detector tubes; this technique proved too insensitive to give
a result over the limited test period.
To illustrate the results reported in Reference 29, those for a needle attached to the
busbar are reproduced as Figure 2.8, in which the data from the various techniques
has been expressed as signal/noise ratios, so that the results can be compared.
The general conclusions of this investigation were that:
the acoustic, IEC 270, and UHF techniques all show good sensitivity
acoustic measurements are non-intrusive and can be made on any GIS, but the
attenuation of the signal across barriers and along the chambers is rather high
conventional PD measurements need an external coupling capacitor, and cannot
be used on GIS in service
the UHF technique is suitable for in-service monitoring.
2.7.4 Overview of UHF technology
AGIS installation consists of a network of coaxial transmission lines which acts as a
waveguiding structure for UHF signals, with an inherently low loss. In the absence
of barriers and discontinuities, the attenuation at 1 GHz in a waveguide of this size
(typically 0.5 m diameter) would be only 35 dB/km. In practice, reflections at
discontinuities within the GIS chamber cause a reduction in signal strength which
has been observed to be in the region of 2 dB/m [28]. These reflections can cause
resonances to appear, such as those set up between dielectric barriers [31].
When coaxial lines are used for signal transmission, the usual mode of signal
propagation is the transverse electromagnetic (TEM) mode, in which the electric
and magnetic field components are transverse to the direction of propagation. The
frequency of operation is kept below the cut-off frequencies at which higher-order
transverse electric (TE) and transverse magnetic (TM) modes begin to be excited,
thus ensuring non-dispersive propagation, a normal requirement for the maintenance
of signal fidelity.
In the case of a GIS, the coaxial structure is a consequence of the need to contain
the gaseous insulation, and its dimensions are accordingly defined by high voltage
requirements. At UHF (3003000 MHz), the GIS dimensions are such that the TE
and TM modes of propagation cannot be neglected [32]. Excitation of a purely TEM
mode signal would require symmetrical excitation of the waveguide, whereas the
location of a PD current pulse is always asymmetrical with respect to the coaxial
cross-section and therefore couples strongly with higher-order modes. These modes
are closely related to those of the hollow cylindrical waveguide, and are therefore
capable of propagating across gaps in the HVbusbar, which would blockTEMsignals.
For these reasons, it is necessary to account for all modes of propagation within the
measurement bandwidth to describe adequately the UHF signal resulting from a PD.
The design of internal couplers for detection of UHF signals in GIS involves
a compromise between the conflicting requirements of minimising the field enhance-
ment while maximising the UHF sensitivity. The coupler must not create an additional
risk of breakdown, and is normally mounted in a region of relatively weak HV field,
at an inspection hatch for example, where it is shielded in a recess in the outer
conductor. A disadvantage is that the UHF fields also tend to be weaker in these
regions, since they are subject to the same boundary conditions as the HV field.
UHF antennas of a form which would be desirable for good sensitivity, such as a
radial monopole, are unacceptable as they would invite breakdown. However, other
forms of broadband planar couplers, such as the spiral, have been investigated [33]
and shown to have good sensitivity. Circular plate couplers have proved useful, and
are more readily accepted in GIS because they are similar to capacitive dividers,
and can be seen not to cause stress enhancement. Circular couplers are themselves
resonant structures at UHF frequencies, and the effect of design parameters has been
investigated [33], showing how their sensitivity can be enhanced within the design
constraints.
A more detailed description of UHF theory and the generation and transmission
of UHF signals will be found in the next section.
SF
6
propagation
PD
source
source parameters:
location of the PD
length of discharge path
shape of current pulse
excitation
2b
2a
UHF
coupler
coupling factors:
UHF field strength
coupler response
noise levels
GIS
chamber
propagation effects:
delay
dispersion
reflections
attenuation
extraction
Figure 2.9 Transfer functions involved in the UHF detection of PD in GIS
2.8 The generation and transmission of UHF signals in GIS
2.8.1 Introduction to UHF theory
Detection of PD by the UHF method involves the stages of energy transfer that are
shown in Figure 2.9.
To take full advantage of the UHF technique, an understanding of the basic
processes involved is important. In the following notes, a represents the radius of the
inner conductor of the GIS, andb represents the radius of the outer conductor. Asystem
of cylindrical coordinates (r, , z) will be used to describe the electromagnetic field
components.
2.8.2 Excitation
The shape of the streamer current pulse i(t ) at the PD source is most important in
determining the characteristics of the UHF signal. The energy radiated in the UHF
range is highly dependent on the rate of change of PD current. However, for a given
pulse shape, the UHF signal amplitude scales linearly with the current flowing at the
defect.
For small defects, the UHF signal amplitude is proportional to the product ql
when the pulse shape is constant [34]. Here, q represents the charge contained in
the PD current and l is the length over which it flows. Because the length of the
streamer itself rarely exceeds 1 mm, l is predominantly a function of the defect size
(e.g., particle or protrusion length).
The UHF signal excited by a PDsource depends on the position of the defect in the
transverse plane. This is because the coupling coefficients to each of the waveguide
modes vary across the coaxial cross-section of the GIS.
10
5
0
5
10
a
m
p
l
i
t
u
d
e
,

m
V
0 50 100 150
time, ns
200 250
Figure 2.10 Typical UHF signal excited by PD in a 400 kV GIS, as measured at the
output of a UHF coupler
2.8.3 Propagation
The electromagnetic waves radiated from the defect region begin to propagate in the
GIS chamber. Different frequency components of the PD pulse propagate at different
velocities, causing dispersion of the pulse. The overall effect of dispersion is to
cause the signal to appear as a long, oscillating waveform with a somewhat random
appearance (Figure 2.10). The UHF signals obtained from a GIS coupler typically
have a duration of 1001000 ns. Some of the signal is rapidly attenuated because it
is below the cut-off frequency of the mode in which it is propagating. The highest
frequency components travel along the coaxial lines with a velocity approaching c.
Propagation through barriers takes place at a lower velocity, c/
r
, where
r
is the
relative permittivity of the insulating material (typically 56). The relative arrival
times of the wavefronts at couplers on either side of the PD source can often be used
to locate the defect.
Any non-uniformities in the GIS will cause partial reflections of the UHF signals.
Most discontinuities inside the GIS have a complicated reflection pattern, because
they do not reflect the signal at a plane, but over a distributed volume (e.g., a conical
gas barrier). The attenuation of the UHF signal along the GIS duct (between one
coupler and the next) is mainly due to the confining of signal energy within the
chambers by the partially reflecting discontinuities. These effects cannot be analysed
theoretically, except in a greatly simplified form[31]. However, numerical techniques
and experimental measurements have resulted in guideline figures for the attenuation
caused by reflecting obstacles in the GIS.
2.8.4 Extraction
Internal UHF couplers are normally mounted at a recess in the outer conductor.
Because of the boundary conditions in this region, the radial component of electric
field is predominant. The intensity of this electric field is therefore the primary factor
affecting the signal level that can be obtained from the coupler. Externally mounted
couplers (e.g., at an inspection window) will be affected by the field patterns in the
structure on which they are mounted. In this case, the mounting arrangement should
SF
6
be considered as part of the coupler, so that the reference plane is still the inner surface
of the outer conductor. The couplers function is to maximise the output voltage for
a given radial component of UHF electric field [35].
The frequency response of the coupler should be suitable for the frequency range
of the UHFsignal. In a 400 kVGIS, the UHFenergy is normally concentrated between
500 and 1500 MHz. UHF signals from low level PD can only be detected if they are
of sufficient amplitude to be distinguished from electrical background noise. Internal
couplers are best from this point of view, as the noise levels are low, especially if
the GIS is cable fed. Where external couplers must be used, they should be screened
from interfering signals [36].
2.8.5 Waveguide modes and UHF propagation
This section introduces the basic field patterns that can exist in a coaxial waveguide,
and describes their relevance to the UHF detection of PD.
2.8.5.1 TEM and higher-order modes
The transverse electromagnetic (TEM) mode is the familiar mode of propagation in
a coaxial line, in which the electric and magnetic fields are directed wholly in the
plane transverse to the direction of propagation. Higher-order modes are classified
as transverse electric (TE) or transverse magnetic (TM) types. TE modes have only
E
r
and E
components to their electric fields, and TM modes have E

r
, E
and
E
z
components. An infinite number of these higher-order modes exists, designated
using subscripts such as TE
nm
and TM
nm
. Each of these modes has a unique cut-off
frequency, below which it does not propagate.
2.8.5.2 Calculating the cut-off frequencies of TE and TM modes
Cut-off frequencies of the higher-order modes are related to the mode eigenvalues
u
nm
and v
nm
by the following equations:
f
TE
nm
=
c
2
u
nm
(2.1)
f
TM
nm
=
c
2
v
nm
(2.2)
The mode eigenvalues are dependant on the radii a and b of the GIS conductors [35]
and are roots of the following equations, which involve the Bessel functions J
n
(x),
Y
n
(x) and their first derivatives J
n
(x), Y
n
(x):
J
n
(ua)Y
n
(ub) Y
n
(ua)J
n
(ub) = 0 (2.3)
J
n
(va)Y
n
(vb) Y
n
(va)J
n
(vb) = 0 (2.4)
For a given value of the integer n, these equations possess an infinite series of roots,
which are numbered m = 1, 2, 3, . . .. Thus, u
nm
is the mth root of Equation (2.3) and
v
nm
is the mth root of Equation (2.4). These values can be determined numerically
using standard mathematical software packages.
TE
11
E
r
TEM
E
r
TM
21
E
r
Figure 2.11 Patterns of the radial electric field over the GIS cross-section for
some coaxial waveguide modes. The rectangular plane represents the
position of zero electric field
2.8.5.3 Electric fields of propagating modes in GIS
In cylindrical coordinates, the electric field of the TEM mode can be expressed in
terms of the voltage V on the inner conductor as:
E
r
=
V
r ln(b/a)
(2.5)
The radial electric field components of TE and TM modes vary in a more complex
manner, as defined by the following equations:
TE modes: E
r
= K
n
r
cos(n)[J
n
(ur)Y
n
(ua) Y
n
(ur)J
n
(ua)] (2.6)
TM modes: E
r
= Kv cos(n)[J
n
(vr)Y
n
(va) Y
n
(vr)J
n
(va)] (2.7)
In these equations, K is an arbitrary constant defining the amplitude of the excitation.
Variation of the radial field components defined by Equations (2.5)(2.7) is illustrated
in Figure 2.11 for a mode of each type.
2.8.5.4 Importance of higher-order modes for PD detection
The total field radiated by a PD current can be represented by the sum of the field
patterns of the waveguide modes that it excites. This is analogous to the representation
of a non-sinusoidal signal by a Fourier series in the time domain. In the case of PD,
at any instant in time the field pattern is represented in the three-dimensional space
by the sum of three-dimensional fields. By its nature, a PD source is usually located
asymmetrically within the coaxial line. During the early stages of propagation, any
excitation of the TEM mode must be associated with considerable contributions from
SF
6
other modes to produce a field that is only non-zero in the region close to the PD
source. Hence, the higher-order modes are excited in greater proportion.
As the duration of the PD pulse decreases, the spatial variation of the radiated
fields must increase because the propagation velocity c is fixed. The number of modes
required to adequately represent the field increases as a result. Conversely, if the PD
pulse is relatively slowly varying compared with the time taken for the field to traverse
the waveguide cross-section, the higher-order modes become less significant in the
representation of the field pattern and the TEM mode is predominant.
APDsource such as a protrusion excites more higher-order modes when it is at the
outer conductor than at the inner conductor [34]. This is because the spatial variation
of the modes is inherently greater at the inner conductor, so fewer are required to
represent the rapidly changing field in this region. In certain circumstances, this
property can be used to discriminate between different PD locations [37].
2.8.5.5 Relative contributions of TEM and higher-order modes
The VHF content of PDsignals in the frequency range up to 300 MHz is usually below
the cut-off frequencies of any higher-order modes in 400 kV class (or smaller) GIS.
PD detection that is carried out at VHF therefore depends upon the level of the TEM
mode signal for its sensitivity. The loss of sensitivity that results from this restriction
can be illustrated by comparing the higher-order and TEM mode contributions to
the electric field at a coupler for two excitation pulses having equal amplitudes but
different pulse widths. The configuration used to make the comparison is that of
a 10 mm PD path (e.g., a protrusion) located at the outer conductor in the centre of
a 3.6 m long 400 kV GIS chamber (a = 0.05 m, b = 0.25 m). The resulting radial
electric field at a distance of 1.2 mfromthe PDsource can be determined by computer
simulation. AGaussian current pulse i(t ) at the PD source was defined as:
i(t ) = I
max
e
(t t
0
)
2
/2
2
(2.8)
For the first example, a 1 pC pulse was defined by I
max
= 2.0 mA, t
0
= 2 ns and
= 200 ps. The contributions to the total field are shown in Figure 2.12a. The
higher-order modes make a much larger contribution to the total field in response to
this pulse, with a peak-to-peak amplitude of 75 mV m
1
compared with 12 mV m
1
for the TEMmode. The power available fromeach signal is proportional to the square
of the field strength, which increases the significance of this amplitude difference.
Also, the higher-order modes do not have the dead time that occurs between reflections
of the TEM mode pulse, during which it does not deliver any energy to the coupler.
The second example, shown in Figure 2.12b, was generated by changing to
500 ps while leaving t
0
and I
max
unchanged in Equation (2.8). This increases the
pulse width at half-amplitude from 0.50 ns to 1.25 ns. The peak amplitude of the
TEMmode field remains unchanged but the width of the reflected pulses has increased
accordingly. In contrast, the higher-order mode field has altered significantly. The
amplitude is nowcomparable to that of the TEMmode, and the high frequency content
has decreased.
0
2
0

m
V

m
1
/
d
i
v
2
0

m
V

m
1
/
d
i
v
5 10 15 20 25 30 35 40
time, ns
0 5 10 15 20 25 30 35 40
TE & TM modes TEM only
time, ns
a
b
Figure 2.12 Simulation results comparing the mode contributions to the radial
electric field E
r
at 1.2 m from a PD source. The PD current has
a path length of 10 mm and the peak current is 2 mA.
a Gaussian PD pulse of half-amplitude width 500 ps
b Gaussian PD pulse of half-amplitude width 1.25 ns
E
r
has been shown in Figure 2.12 rather than the coupler output voltage,
to illustrate the typical electric field strengths to which UHF couplers are sub-
jected. The couplers are normally capacitive, and their sensitivity decreases at lower
frequencies [33]. Although the amplitude of the TEM mode field is unchanged in
Figure 2.12b, the resulting coupler output would be significantly reduced because
the rate of change of the electric field is lower. Making the coupler larger to counter-
act this effect would have the undesirable consequence of increasing the amount of
low frequency noise coupled from the GIS. Measurements of the PD current pulses
generated by small defects such as particles and protrusions in SF
6
have shown typi-
cal values of less than 500 ps for the half-amplitude width. Theoretical studies have
indicated that for the short PD pulses typical of small defects, limiting PD detection
to VHF frequencies (TEM mode only) results in the majority of the available signal
energy being neglected.
2.8.6 Attenuation of UHF signals
The GIS can be considered as a series of loosely coupled chambers in which the
UHF resonances occur. Over the shorter timescale (10100 ns), energy is transferred
SF
6
between adjacent chambers, with those nearest the PD source maintaining the higher
energy levels. Dissipation losses (skin effect) become significant in the longer
timescale (1001000 ns), and are the reason for the ultimate decay of the signal.
2.8.6.1 Losses due to the skin effect
The skin effect causes signal attenuation through dissipation in the surface resistance
of the conductors. The losses tend to increase with increasing frequency. However,
attenuation at UHF in GIS is theoretically quite low (typically 35 dB per km), due
to the high ratio of the cross-sectional area to the surface area in the waveguide.
2.8.6.2 Attenuation caused by barriers and discontinuities
In long gas-insulated lines (GIL), which have relatively fewdiscontinuities (insulating
posts at intervals to support the inner conductor and the occasional gas barrier),
attenuation of UHF signals can be quite low, allowing PD sources to be detected
at distances of >100 m [38]. However, in the more compactly constructed GIS,
the number of bends, junctions, gas barriers and circuit breakers is such that the
empirical figure of 12 dB m
1
is more appropriate.
Various research groups have attempted to assign specific attenuation values to
individual components, but it is now generally accepted that this cannot be done.
The reason is that the overall signal level at a given position in the GIS arises from
interactions between signals reflected from discontinuities. These interactions are
dependent on the distances between the discontinuities, and the signal levels also
depend on the position of the coupler relative to barriers, junctions, etc. [37]. This
can lead to situations where a UHF coupler at a greater distance from a PD source
detects a larger signal than does one closer to it.
As a consequence of these effects, it is not possible to calculate with any accuracy
the attenuation that will be experienced by a PD signal propagating between two
UHF coupling points on a GIS. In order to ensure that a UHF monitoring system
has sufficient sensitivity to detect a 5 pC discharge located midway between a pair
of couplers, a practical approach has been proposed by CIGRE [39]. The two-stage
procedure involves first determining an artificial pulse, which, if injected into one of
the couplers, will radiate a UHF signal equivalent to a real 5 pC PD. An on-site test
is then carried out, injecting the calibrated pulse into each coupler and ensuring that
it can be detected at adjacent couplers by the UHF monitoring system.
2.9 Application of UHF technique to PD detection in GIS
2.9.1 Design and testing of UHF couplers
For radially-directed PD currents, the majority of the energy available to an elec-
tric field sensor (coupler) is in the radial component of the electric field. In the
region close to the outer conductor where couplers are normally mounted, the other
components of the electric field are very small because of the waveguide boundary
conditions. Consequently, the purpose of a coupler is to provide the maximum trans-
fer of energy fromthe incident radial electric field to the 50 input of the monitoring
system. Couplers can be classified according to whether they are mounted internally
or externally:
(i) Internal couplers must be fitted to the GIS during construction or retrofitted
during a planned outage, because degassing of the GIS chambers is necessary.
These couplers often take the form of a metal disc insulated from the GIS
enclosure by a dielectric sheet. The measurement connection is made through
a hermetically sealed coaxial connector that is usually connected to the centre
of the disc.
(ii) External couplers are usually portable sensors that are fitted to an aperture
in the metal cladding such as an inspection window or exposed barrier edge.
These couplers are suitable for periodic insulation testing of GIS for which
a permanently installed monitor is not economically viable or for older GIS that
cannot be retrofittedwithinternal couplers. External couplers are sometimes less
sensitive than their internal counterparts because the UHFsignal is attenuated by
impedance discontinuities at the surfaces of the barrier and window materials.
They may also be more prone to electrical interference signals when they are
not shielded as well as internal couplers. For these reasons, it is preferable to
fit internal couplers during the construction phase, and this is common practice
in the UK.
2.9.1.1 Coupler calibration system
The UHF excitation model developed in Reference 32 allows the electric field at the
outer conductor of the GIS to be determined. This is the radial electric field at the
point where the coupler is to be mounted. In practice, the presence of the coupler and
the mounting arrangement will distort the field pattern. For this reason, the mounting
of the coupler for calibration purposes should duplicate the GIS mounting structure.
Couplers are calibrated by measuring their transfer function in terms of their output
voltage in response to a defined incident electric field. To allow for the many types
of coupler and mounting arrangement that exist, a standard measurement scheme has
been developed which is independent of the GIS in which the coupler is to be used.
The calibration system (see Figure 2.13) measures the frequency response of
the coupler and its mounting arrangement when subjected to a known electric field
normal to the ground plane in which the coupler is mounted. The incident field is first
calibrated using a monopole probe having a known frequency response. The probe
is then replaced by a mounting plate suitable for holding the coupler to be tested.
Adigitiser records the signal from the coupler under test, and a continuously updated
display of the coupler gain is provided by the signal processing unit. Figure 2.14
gives the response of an external window coupler used on a 500 kV GIS. Note that
the coupler sensitivity has units of length and is represented by an effective height h.
This arises because the transfer function relates the output voltage (V) to the incident
field (Vm
1
). This particular coupler is shown, mounted on a window, in Figure 2.15.
SF
6
coupler test port
calibrated field
broadband TEM cell
coupled
signal
digitising
unit
signal
processing
unit
reference
signal
matching
unit
splitter
signal
source
Figure 2.13 Diagram of the UHF coupler calibration system
100
10
1
0.1
0 500 1000
frequency, MHz
1500 2000
s
e
n
s
i
t
i
v
i
t
y
H
e
,

m
m
9.2mm, 100%
Figure 2.14 Frequency response plot for an external, window coupler as measured
by the calibration system
2.9.1.2 Disc couplers design guidelines
Studies of resonance in disc couplers [33] have shown that the UHF signal level can
be increased by making the connection to the disc close to its edge rather than in
the centre. This also improves the bandwidth of the coupler by increasing sensitivity
at lower frequencies. For a given disc size, the UHF signal level increases when
the dielectric height is increased or a material of lower dielectric constant is used.
Both of these changes lead to a reduction in the Q factor of the disc resonances, and
Figure 2.15 An external window coupler on a 500 kV GIS
Table 2.2 Optimising sensitivity of disc coupler
Parameter to be changed To increase sensitivity
size increase diameter
dielectric material reduce permittivity
increase thickness
connection move away from centre
increasing the dielectric height allows the disc to couple with a greater proportion
of the radial UHF electric field inside the GIS. As the radius of the disc coupler is
decreased its resonant frequencies move upwards, away from the frequency range
where most of the UHF energy is concentrated. Generally, the disc should be made
as large as is practical for the size of the GIS in which it is to be installed. This
information is summarised in Table 2.2.
2.9.1.3 External couplers
External couplers, such as the one shown in Figure 2.15, are usually based on antenna
designs and fabricated using normal PCBtechniques. At UHF, the losses in glass fibre
substrates are insignificant and the cost of low loss microwave substrates cannot be
justified. In some cases, external couplers can be calibrated using the systemdescribed
above. However, in some circumstances the coupler may be so far removed from
the GIS chamber (e.g., in a CT housing) that the mounting arrangement cannot be
SF
6
optical converter unit
equipment
cabinets
UHF coupler
Figure 2.16 General arrangement of the PDM system
reproduced reliably. In these cases, the coupler must be optimised in a GIS test rig,
using the pulse injection technique described below.
When the coupler is not in the HVchamber of the GIS, more complicated designs
can be used because there is no concern about field enhancement and the risk of
flashover. Some designs are described in Reference 40, which also describes the test
procedure. Optimisation should take place in the time domain, because it is difficult
to make comparisons based on a spectrum analyser plot of the coupler output in
response to an injected pulse.
2.9.2 Design of a PDM system for GIS
The general arrangement of one particular partial discharge monitor (PDM) is shown
in Figure 2.16.
The monitoring system consists of the following basic parts:
UHF couplers to take the UHF signals from the GIS.
Optical converter units (OCUs) each OCU contains circuitry to detect and
process the UHF signals from one three-phase set of couplers. The UHF data is
then transmitted via an optical fibre link back to the equipment cabinets, which
are located in a central place.
The central data handling, processing, storage and display are carried out within
equipment cabinets located in the relay room. The racks contain the electronics
to receive and handle the streamed data from the OCUs, the PC and control unit
for data storage and display.
100
50
0
10
20
30
40
50
0
90
180
270
360
Figure 2.17 Busbar corona, streamers and leaders
100
50
0
10
20
30
40
50
0
90
180
270
360
Figure 2.18 Free metallic particle
2.9.3 Display and interpretation of PD data
The features of the UHF discharge pulses that are most useful for interpretation
purposes are their amplitude, point on wave and the interval between pulses. These
parameters enable typical defects such as fixed point corona, free metallic particles
and floating electrodes to be identified. Other defects occur less commonly, but have
their own distinctive features.
The UHF data may be displayed in any way which reveals the characteristic
patterns typical of the defects causing them, as, for example, in the three-dimensional
patterns shown in Figures 2.17 and 2.18. Here, the pulses detected in 50 (60) con-
secutive cycles are shown in their correct phase relationships over the cycle. In the
three-dimensional displays, 0
corresponds to the positive-going zero of the power

frequency wave, 90
to the positive peak, and so on.

2.9.3.1 Corona
Three distinct phases in which corona develops from a protrusion are seen as the
voltage is raised:
(i) Inception, where discharges occur first on the half-cycle that makes the
protrusion negative with respect to the other electrode. Inception is therefore
SF
6
on the negative half-cycle when the protrusion is on the busbar, and on the
positive half-cycle when it is on the chamber wall. The initial discharges are
of very low magnitude (less than 1 pC), and are centred on one of the voltage
peaks.
(ii) Streamers start at a slightly higher voltage, and appear as a regular stream of
pulses on the peak of the positive half-cycle. At the same time, the negative
discharges become larger and more erratic. This difference between the positive
and negative discharges reveals whether the protrusion is on the busbar or
chamber wall. Streamer discharges will not lead to breakdown.
(iii) Leaders follow further increases of the applied voltage, and appear every few
cycles as large discharges on the positive half-cycle. They propagate in steps
until either they become extinguished, or reach the other electrode and cause
complete breakdown. Figure 2.17 shows a typical display of PD from a pro-
trusion on the busbar at this stage of the discharge process. Leaders are the
precursors of breakdown, and there is always a risk of failure when they are
present.
2.9.3.2 Free metallic particle
A particle lying on the chamber floor becomes charged by the electric field, and if
the upward force on it exceeds that due to gravity it will stand up and dance along
the floor. This generates a discharge pulse each time contact is made with the floor,
since the particle then assumes a new value of charge. The pulses occur randomly
over the complete power frequency cycle, but their peak amplitudes follow the phase
of the voltage. Atypical PDpattern for a free metallic particle is shown in Figure 2.18.
At higher voltages, the particle will start to jump towards the busbar, and after several
cycles may reach it. Two factors combine to make this an especially serious condition
which often leads to breakdown: (i) as the particle approaches the busbar it discharges
and generates a voltage transient which increases still further the stress at its tip;
(ii) breakdown can occur by leader propagation well before any space charge has had
time to develop and shield the tip.
2.9.3.3 Floating electrode
This arises if the contact to, for example, a stress shield deteriorates and sparks
repetitively during the voltage cycle. The sparking is energetic because the floating
component usually has a high capacitance, and this degrades the contact further.
Metallic particles are produced, and may lead to complete breakdown. The discharges
are concentrated on the leading quadrants of the positive and negative half-cycles,
and their amplitude does not vary with the applied voltage.
Often the gap is asymmetrical, and sparks over at different voltages on the two
half-cycles. Then, a different charge is trapped on the floating component, and
this gives rise to the characteristic wing-shaped patterns shown in the plan view
of Figure 2.19. These provide a positive confirmation that the defect is a floating
component.
0
10
20
c
y
c
l
e

n
o
.
30
40
50
0 90 180
degrees
270 360
Figure 2.19 Floating electrode, plan view
2.9.4 AI diagnostic techniques
PDMsystems can generate large amounts of data, which needs to be interpreted auto-
matically so that the results are immediately available to system operation engineers.
Various artificial intelligence (AI) techniques may be used to classify the type of
defect present, which gives a guide to any action that might need to be taken.
The real need, however, is for an expert system that will give the probability
and time to failure of the GIS. The expert system will need to include additional
information on the severity of the defect, its location in the GIS chamber and the
electric field at that point. This full risk analysis is not possible at present, but much
effort is being devoted to its development.
2.9.4.1 Classification algorithms
PD signals may be classified using analytical techniques such as distance classi-
fiers, artificial neural networks (ANNs) and fuzzy algorithms. In general, features
are extracted from the PD signatures and the techniques trained to recognise them.
Since no single technique appears adequate to recognise all types of defect, the
combined outputs from a number of techniques may be used to give a higher prob-
ability of success in the final classification. All the analytical techniques need to
be trained using extensive databases of PD signals, usually recorded in labora-
tory tests, and with sufficient care the confidence level in the results may exceed
90 per cent.
Aclassification technique that is used with much success is the ANN. This can be
applied to identify the single-cycle patterns recorded by the monitoring system out-
lined above; for example, the particle and floating components shown in Figures 2.18
and 2.19. As shown in Figure 2.20, groups of multiple ANNs are used first to separate
the signal fromnoise, and then to identify the source of each with a known confidence
level. This technique may itself be used in conjunction with others.
SF
6
2D+3D
discharge
data
particle
busbar corona
chamber corona
signal
noise
floating electrode
void
undefined
undefined/
insufficient data
mobile phone
radar noise
motor noise
lights
undefined
Figure 2.20 Classification of PD signals by ANNs
2.9.4.2 Trend analysis and alarms
Often, the PD signals recorded on site are intermittent in nature, and it is necessary
to build-up a history of the activity over a period of time. The historical records are
used to determine the trends in the activity of a PD source, such as its amplitude
and repetition rate. Averaging is applied to eliminate spurious results and reveal any
underlying change in the PD activity, and this can be a further basis for sending an
alarm signal to the operator.
2.9.5 Service experience
The first online UHF systemfor PDmonitoring in GIS was commissioned in Scotland
in 1993 and the technology has since been adopted by other utilities in the UK,
South-East Asia, the Middle East, South America and the USA. In the case of new
substations, PDM systems are often specified for key installations, as, for example,
where the GIS is supplying a major population centre, or is connected to a critical
generation source such as a nuclear power plant.
In many cases, the use of UHF monitoring is also seen as being cost effective
for GIS which have been in service for a number of years, where the benefits are
improved availability and increased plant life. The availability of suitable external
couplers which allowa PDMsystemto be retrofitted without the need for an outage is
an important factor in this situation, although planned maintenance or refurbishment
operations provide the opportunity for internal couplers to be installed where there
are no suitable dielectric apertures at which external sensors can be mounted.
Service experience with UHF monitoring has been excellent. Several utilities
have reported the identification of defects that would have resulted in major failures,
any one of which would have incurred costs considerably in excess of the cost of
the PDM system. Typical examples have included sparking at internal shields, track-
ing on solid insulation and, in one case, failure of a joint between two sections of
busbar.
For the future, it seems inevitable that the use of monitoring in all types of high
voltage equipment will become increasingly common. PD monitoring of the insula-
tion in GIS will merely be one component in integrated systems which will monitor a
range of mechanical, thermal and electrical parameters in all power plant, including
transformers, switchgear, CTs, bushings and cables. There will be a growing need for
the development of sophisticated information management techniques to coordinate
and prioritise the output from such systems in order to provide a basis for effective
remedial action.
2.10 References
1 CHRISTOPHOROU, L.G., and DALE, S.J.: Dielectric gases. Encyclopedia
of Physical Science and Technology, 1987, 4, pp. 246262
2 SOMERVILLE, I.C., TEDFORD, D.J., and CRICHTON, B.H.: Electrode
roughness in SF
6
a generalised approach. Proceedings of XIII international
conference on Phenomena in ionised gases, Berlin, 1977, pp. 429430
3 FARISH, O.: Corona-controlled breakdown in SF
6
and SF
6
mixtures.
Proceedings of XVI international conference on Phenomena in ionised gases,
Dusseldorf, 1983, pp. 187196
4 SANGKASAAD, S.: AC breakdown of point-plane gaps in compressed SF
6
.
Proceedings of 2nd international symposium on HV technology, Zurich, 1975,
pp. 379384
5 CHALMERS, I.D., GALLIMBERTI, I., GIBERT, A., and FARISH, O.: The
development of electrical leader discharges in a point-plane gap in SF
6
, Proc.
R. Soc. Lond. A Math. Phys. Sci., 1987, 412, pp. 285308
6 SOMERVILLE, I.C., FARISH, O., and TEDFORD, D.J.: The influence of
atmospheric negative ions on the statistical time lag to spark breakdown, in
CHRISTOPHOROU, L.G. (Ed.): Gaseous dielectrics IV (Pergamon Press,
New York, 1984), pp. 137145
7 CHALMERS, I. D., FARISH, O., GIBERT, A., and DUPUY J.: Leader devel-
opment in short point-plane gaps in compressed SF
6
, IEE Proc. A, Phys. Sci.
Meas. Instrum. Manage. Educ. Rev., 1984, 131, pp. 159163
8 PINNEKAMP, F., and NIEMEYER, L.: Qualitative model of breakdown in
SF
6
in inhomogeneous gaps, J. Phys. D: Appl. Phys., 1983, 16, 12931302
9 WIEGART, N. et al.: Inhomogeneous field breakdown in GIS the prediction
of breakdown probabilities and voltages, IEEE Trans. Power Deliv., 1988, 3,
pp. 923946
10 KNORR, W., MOLLER, K., and DIEDERICH, K.J.: Voltagetime charac-
teristics of slightly nonuniform arrangements in SF
6
using linearly rising and
oscillating lightning impulse voltages. CIGRE paper 15.05, Paris, 1980
SF
6
11 BOECK, W.: SF
6
-insulation breakdown behaviour under impulse stress, in
Surges and high voltage networks (Plenum Press, New York, 1980)
12 WOOTTON, R.E., COOKSON, A.H., EMERY, F.T., and FARISH, O.:
Investigation of high voltage particle-initiated breakdown in gas insulated
systems. EPRI reports RP7835,1977; EL-1007,1979
13 COOKSON, A.H., and FARISH, O.: Particle-initiated breakdown between
coaxial electrodes in compressed SF
6
, IEEETrans., 1973, PAS-92, pp. 871876
14 FARISH, CHALMERS, I.D., and FENG, X.P.: Particle-initiated breakdown
in a coaxial system in SF
6
/air mixtures. Proceedings of IEEE symposium on
Electrical insulation, Washington DC, 1986, pp. 206209
15 CHALMERS, I.D. et al.: Extended corona stabilisation in SF
6
/Freon mixtures,
J. Phys. D: Appl. Phys., 1985, 18, pp. L107112
16 VANAGIWA, T., ISHIKAWA, T., and ENDO, F.: Particle-initiated breakdown
characteristics on a ribbed spacer surface for SF
6
gas-insulated switchgear,
IEEE Trans. Power Deliv., 1988, 3, pp. 954960
17 CIGREWG23.10: GIS in service experience and recommendations. CIGRE
paper 23-104, Paris, 1994
18 IEEE Substations Committee, Working Group K4 (GIS diagnostic methods):
Partial-discharge testing of gas-insulated substations, IEEE Trans. Power
Deliv., 1992, 7, (2), pp. 499506
19 BARGIGIA, A., KOLTUNOWICZ, W., and PIGINI, A.: Detection of partial
discharge in gas-insulated substations, IEEE Trans. Power Deliv., 1992, 7, (3),
pp. 12391249
20 JONES, C.J., IRWIN, T., HEADLEY, A., and DAKERS, B.: The use of diag-
nostic based predictive maintenance to minimise switchgear life cycle costs.
Proceedings of 9th conference on Electric power supply industry, Hong Kong,
1992, 5, pp. 225234
21 LUNGAARD, L.E., TANGEN, G., SKYBERG, B., and FAUGSTAD, K.:
Acoustic diagnosis of GIS: field experience and development of expert
systems, IEEE Trans. Power Deliv., 1992, 7, (1), pp. 287294
22 HAMPTON, B.F., and MEATS, R.J.: Diagnostic measurements at UHF in gas
insulated substations, IEE Proc. C, Genes. Trans. Distrib., 1988, 135, (2),
pp. 137144
23 FUJIMOTO, N., and FORD, G.L.: Results of recent GIS fault survey. IERE
workshop on Gas insulated substations, Toronto, September 1990
24 KOPEJTKOVA, D., MOLONY, T., KOBAYASHI, S., and WELCH, I.M.:
Atwenty-five year reviewof experience withgas-insulatedsubstations. CIGRE
paper 23-101, Paris, 1992
25 LUNGAARD, L.E., and SKYBERG, B.: Acoustic diagnosis of SF
6
gas
insulated substations, IEEE Trans. Power Deliv., 1990, 5, (4), pp. 17511759
26 SCHLEMPER, H.D.: TE-Messung und TE-Ortung bei der Vor-Ort-Prfung
von SF
6
-Anlagen. Haefely Trench symposium, Stuttgart, March 1995
27 SCHLEMPER, H.D., KURRER, R., and FESER, K.: Sensitivity of on-site
partial discharge detection in GIS. Proceedings of 8th international symposium
on HV engineering, Yokohama, 1993, 3, pp. 157160
28 HAMPTON, B.F., PEARSON, J.S., JONES, C.J., IRWIN, T., WELCH, I.M.,
and PRYOR, B.M.: Experience and progress with UHF diagnostics in GIS.
CIGRE paper 15/23-03, Paris, 1992
29 CIGRE WG 15.03: Diagnostic methods for GIS insulating systems. CIGRE
paper 15/23-01, Paris, 1992
30 FRUTH, B., and NIEMEYER, L.: The importance of statistical characteristics
of partial discharge data, IEEE Trans. Electr. Insul., 1992, 27, (1), pp. 6069
31 KURRER, R., FESER, K., andHERBST, I.: Calculationof resonant frequencies
in GIS for UHF partial discharge detection, in CHRISTOPHORU, L.G., and
JAMES, D.R. (Eds): Gaseous dielectrics (Plenum Press, New York, 1994)
32 JUDD, M.D., FARISH, O., and HAMPTON, B.F.: Excitation of UHF signals
by partial discharges in GIS, IEEE Trans. Dielectr. Electr. Insul., 1996, 3, (2),
pp. 213228
33 JUDD, M.D., FARISH, O., and HAMPTON, B.F.: Broadband couplers for
UHF detection of partial discharge in gas insulated substations, IEE Proc., Sci.
Meas. Technol., 1995, 142, (3), pp. 237243
34 JUDD, M.D., and FARISH, O.: Transfer functions for UHF partial discharge
signals in GIS. Proceedings of 11th international symposium on High voltage
engineering, London, 1999, 5, pp. 7477
35 JUDD, M.D., and FARISH, O.: A pulsed GTEM system for UHF sensor
calibration, IEEE Trans. Instrum. Meas., 1998, 47, (4), pp. 875880
36 JUDD, M.D., FARISH, O., PEARSON, J.S., and HAMPTON, B.F.: Dielectric
windows for UHF partial discharge detection, IEEE Trans. Dielectr. Electr.
Insul., 2001, 8, pp. 953958
37 MEIJER, S.: Partial discharge diagnosis of high-voltage gas-insulated systems
(Optima Grafische Communicatie, Rotterdam, 2001), ISBN: 90-77017-23-2
38 OKUBO, H., YOSHIDA, M., TAKAHASHI, T., and HOSHINO, T.: Partial
discharge measurements in a long distance SF
6
gas insulated transmission line
(GIL), IEEE Trans. Power Deliv., 1998, 13, (3), pp. 683690
39 CIGRE Task Force 15/33.03.05: Partial discharge detection system for GIS:
sensitivity verification for the UHF method and the acoustic method, Electra,
(CIGRE), 1999, (183), pp. 7487
40 WANNINGER, G.: Antennas as coupling devices for UHF diagnostics in GIS.
Proceedings of 9th international symposiumon High voltage engineering, Graz,
1995, 5
Chapter 3
Lightning phenomena and protection systems
R.T. Waters
3.1 From Franklin to Schonland
Benjamin Franklin was born in 1706, and was fascinated in his youth by electrical
phenomena. In 1732 he founded with his collaborators the Library Company of
Philadelphia with the support of the Penn family, and achieved a new insight into
electrical science with his definition of a single electric fire. This is equivalent to
the free-electron concept, and implies the principle of the conservation of charge.
In 1750 he showed that a needle brought near to a charged conductor caused
a spark, but when further away discharged the conductor silently (a mechanism today
known as a glow corona). Franklin wrote that: this power of points may be of some
use to mankind. Foremost in his mind was to achieve protection against lightning,
and in the famous Philadelphia Experiment (performed under Franklins direction by
dAlibard in France in 1752) point discharges were used to prove the electrification
of the thundercloud and to identify correctly the direction of the field formed by
the cloud charges [1]. The public effect of these successes was great, and they were
regarded as the most important work since Newton.
Franklin, in fact, had envisaged from his laboratory tests two complementary
concepts for the function of a lightning rod:
the possible harmless discharge of the cloud
the attraction of the flash and the conduction of the charge safely to ground.
Although the first process is not feasible, the second is very efficient, at least for
negative lightning. Franklin [2] clearly showed the need to earth the down conduc-
tor. During the well known controversy about the optimum design of a lightning
rod (or air termination in present terminology), experiments by Benjamin Wilson
in 1777 indeed suggested that sharp points might be too attractive to the flash, and
he favoured spheres. Consequently, George III had points replaced on government
buildings with some enthusiasm as these had become associated with the American
colonists. In a letter dated October 14, 1777, to a friend, Franklin wrote that: The
King changing his rods to blunt conductors is a matter of small importance to me.
It is only since he thought himself and his family safe from thunder of Heaven that he
dared to use his own thunder in destroying his innocent Subjects. As with so much
concerning the lightning flash, the design of air terminations continues to be a subject
of research [3].
It is noteworthy that the termstriking distance, still in current use to calculate the
range of attractionof the rodterminationtothe lightningflash, was coinedbyFranklin.
In addition, from his experiments with point discharges, he correctly deduced that
the cloud appeared from the ground to be negatively charged [4].
It was over a century later that Preece [5] proposed a quantification of the zone
of protection that is afforded by a vertical mast; he anticipated the modern rolling-
sphere concept, and set the sphere radius equal to the striking distance, which was
itself equated to the mast height. This leads to a protection angle of 45
.
It was a further century and a half later that the electrical structure of the thunder-
cloud, as a dipole with an upper positive charge [6], and the manner of growth of the
lightning flash [7] were clarified.
The pioneering observations of the lightning flash structure achieved by Basil
Schonland in the 1920s and 1930s constituted the biggest advance in the field since
Benjamin Franklins work in the late 18th century [8, 9]. Schonland read mathe-
matics at Cambridge and his research on lightning from1925 at Capetown University
capitalised on South Africas active thunderstorms. He made electric field measure-
ments and, having collected data from some 23 storms, showed the cloud base to be
negatively charged as Franklin had proposed. This resolved more modern disputes,
and also confirmed the dipole nature of the cloud charge, with a high altitude positive
charge in the cloud. He also recorded field changes due to lightning strikes to the
ionosphere; such flashes are now commonly seen from orbiting vehicles [10].
He carried out high speed photography of lightning using, at first, a streak camera
designed for the purpose by Charles Boys, who had not succeeded in obtaining
records. It displayed the multiplicity and duration of the lightning strikes, andAllibone
and Schonland [11] soon demonstrated the similarity of the lightning growth and
streak photographs of sparks later obtained in the laboratory. The spatial growth
of the lightning discharge, particularly as controlled by the physics of the stepped
first leader and the upward discharge from ground, determines the probability of
strikes to grounded structures. The statistics of the lightning current and its rate of
rise, on the other hand, determine the severity of the resultant strike. Our know-
ledge of these statistics was largely established by the formidable work from 1946 to
1974 by Berger at Mt San Salvatore in Italy. These spatial and electrical properties,
which are not independent, together comprise the risk of lightning damage by direct
current injection to power lines and industrial plant, or by coupled surges in electronic
systems.
Recent years have seen an acceleration in all aspects of lightning research, both
experimental and theoretical, and in the techniques of protection, which will be
described here.
Lightning phenomena and protection systems 79
3.2 Phenomenology of lightning
3.2.1 Characterisation of the flash
Data from Anderson and Eriksson, Berger, Gary, Uman and many others have
quantified lightning parameter characteristics in respect of:
1 incidence of ground flashes
2 flash polarity
3 structure height
4 multiple strokes
5 flash duration
6 peak currents in first and subsequent strokes
7 current shapes.
3.2.2 Incidence
Substantial advances have been made since 1989 in the location and tracking of
lightning, using radio direction finding technology. Before these developments, data
on ground flash density N
g
(flashes per square kilometre per annum) was primarily
obtained fromlightning flash counters triggered by lightning radiation fields, particu-
larly with the CIGRE 10 kHz counter which is mostly responsive to cloud-to-ground
flashes and which has an effective range of 20 km. For example, Anderson and
coworkers [12, 13] aimed to relate the measured N
g
to the commonly available data
on T
d
, the mean number of thunderstorm days in a given locality as evidenced by
audible thunder. Some 400 counters over five years in South Africa suggested an
approximate relationship
N
g
= 0.04T
1.25
d
(3.1)
where T
d
varied regionally between 4 and 80, corresponding to a range of ground flash
densities between 0.2 and 10 flashes/km
2
/annum. In tropical areas such as Indonesia,
the average T
d
values are about 270. Lightning activity is infrequent for latitudes
greater than 50
N or S, and recent orbital observations suggest that 90 per cent of

flashes occur over land.
Surveys in the UK based upon observer reports over several decades have
indicated N
g
values in the range of 0.10.7 flashes/km
2
/annum with large seasonal
and geographical variations within that range. Gary et al. [14] reported that the
keraunic level T
d
over France averages 20, but can be as high as 30 in the Alps,
and CIGRE counter studies found N
g
to range geographically between 2 and
4 flashes/km
2
/annum.
The number of ground flashes in a given region can be greatly increased by
the frequency of strikes to tall structures. Eriksson [15] found that the incidence of
upward-initiated flashes increased exponentially with structure height h. Structures
of h > 400 m involved 95 per cent upward flashes (where the lightning flash was
initiated by a leader from the structure). For h < 100 m, upward leaders caused
initiation of only ten per cent of strikes. For 3000 flashes studied, Eriksson [16]
quoted for the number of strikes per annum to a structure of height h:
N = N
g
2.4 10
5
h
2.05
(3.2)
He determined that for structures of height greater than 100 m a significant number
N
u
of the strikes represented by Equation (3.2) were upward flashes:
N
u
= N
g
3 10
9
h
3.53
(3.3)
Many tall structures furthermore are located on elevated terrain, and these equations
then much underestimate the risk. Diendorfer and Schulz [17] found the 100 m tower
at Peissenberg, Austria (elevation 1287 m), to be struck 83 times in two years. This
is likely to be the result of large thunderstorm fields at high locations rather than the
effect of altitude on the flash (section 3.4.4.5), and all flashes were of the upward
type [18].
Dependence on observer reports and lightning counter measurements has become
almost obsolete in many countries since ground flash density data have become
more widely available from real-time lightning location detection using networks
of sensors. The online display of ground flash activity, usually with only seconds
of delay, and with a capacity to record up to 100 flashes per second, provides
economic advantages for the scheduling of alternative power routes and protecting
vital installations in, for example, the telecommunications, petrochemical, defence
and aviation industries. The UK Lightning Flash Location System (EATechnology)
[19] has monitored cloud-to-ground flashes since 1989. The four-loop detection
system installed at six recording stations in the UK and Ireland can discriminate and
locate groundflashes down to 3kAandtowithina claimed100m. The important capa-
bility of discriminating cloud-to-ground strikes from intercloud flashes is achieved
by utilising the low attenuation of the mainly vertically-polarised radiation associ-
ated with the ground strikes. At 1.1 kHz, radiated waves in the earth-to-ionosphere
waveguide are effectively recorded.
Although the data fromsuch stations generally confirmoverall levels (Figure 3.1),
they give much higher localised values of N
g
than had been expected from earlier
observation and detection methods, with up to eight ground flashes/km
2
/annum in
some areas of the UK. One flash in three is found to strike ground. The shape and
amplitude of the signals also give important indications of peak current magnitude
and flash polarity.
The National Lightning Detection Network (Global Atmospherics Inc) [20, 21]
employs 107 stations over the 10
7
km
2
of the USAwith a satellite link to calculate the
time, location, polarity and current magnitude of each flash. Resolution of compo-
nent return strokes is also possible. This systemuses IMPACTsensors which combine
magnetic field direction finding with time-of-arrival (TOA) sensors (these latter com-
prising the LPATS Lightning Position and Tracking System). TOA measurement is
particularly useful at long range. VHF TOA systems also allow three-dimensional
reconstruction of lightning paths. In Japan a SAFIR VHF-DF network detects total
lightning (cloudcloud and cloudground).
Since 1989 an average of 20 million ground strikes per annum, corresponding to
a mean value of N
g
2, have been recorded in the USAwith a location best accuracy
1000
800
600
400
0.04
200
100
200
300
400
200 0
0
200 400 600
1.2
1
.4
1
.0
0
.7
0
.
1
0
.
0
6
National Grid eastings, km
Note 1 This lightning density map was compiled by E.A. Technology Ltd.
from data accumulated over 10 years
Note 2 A linear interpolation should be used to determine the value of the
lightning flash density, N
g
, for a location between two contour lines
0 Grid
0.08
0.06
0
.1
0
.2
0.3
0
.5
0.7
0.08
0
.
0
4
0
.
0
2
1200
N
a
t
i
o
n
a
l

G
r
i
d

n
o
r
t
h
i
n
g
s
,

k
m
200
Figure 3.1 Ground flash density in the UK (BS 6651:1999)
1
of about 500 m. In a recent development, newsensors enabling simultaneous detection
of both cloudcloud and cloudground flashes provide an earlier warning of active
storms. These newradio detection techniques provide not only vital contemporaneous
information, but also the considerable benefits of a growing archive of lightning
statistics for risk analysis.
A recent addition to global lightning observations derives from the development
of the NASA optical transient detector (OTD) [22]. This detects lightning events
from low orbit, sweeping a given surface location three times every two days. The
detection efficiency is about 60 per cent, with no discrimination between cloud flashes
and cloud-to-ground strikes. Astudy using the OTD system in combination with the
US ground sensor network [23] has enabled estimates to be made for the first time
of the ratio Z of cloud flashes and ground strikes. The results give Z = 2.94 1.28
(standard deviation), but local values can be as high as 9. Low Z values of unity or
less are often observed in mountainous regions, where this high probability of ground
strikes is perhaps to be expected. The orbiting OTD records of total lightning flash
density N
c
+N
g
reached 30 flashes/km
2
/annum or higher.
3.2.3 Polarity
Flash polarity is of relevance for two reasons. In the first place, field studies suggest
that the highest peak currents are associated with positive flashes. Second, analogies
with the long laboratory spark, and some generic models of the mechanism of strikes
to grounded structures (section 3.4.4), both indicate that the protection of structures
from positive direct strikes with standard air terminations may sometimes be much
less effective than for the more common negative flash. For the same reason, the
probability of a direct strike to a structure may be lower because of a smaller attractive
area for positive flashes (section 3.4.5). This may also have led in the past to an
underestimate of the ratio of positive to negative flashes.
Before recent lightning tracking networks were developed, the flash polarity could
be deduced, even in the absence of lightning current recording, by means of magnetic
links on transmission lines and structures and from the polarity of transient local
electric field changes. Most of these results suggested that a minority (12 per cent) of
downward flashes is of positive polarity. Over 29 years of observation, Berger [24]
recorded 1466 negative flashes and 222 positive flashes. At Peissenberg, Fuchs [18]
found that 95 per cent were of negative polarity. There is evidence that winter storms
in some regions (e.g. Japan and northeast USA[25, 26]) or lightning to high altitude
locations may more commonly result in positive flashes. In the last decade, lightning
location networks have improved the statistics, with in addition peak current estimates
of 50 per cent accuracy. UK data shows that a surprising 30 per cent of flashes are
positive, with peak currents up to 250 kA. Geographical variations are also apparent.
In the northern and western UK, positive flashes are more frequent than negative;
the opposite is true for southern England [27]. The OTD/ground network studies in
USA [23] show a correlation between the high Z values (indicating a low ground
strike ratio) and a high positive-to-negative flash ratio. The authors suggest that a
high altitude cloud dipole, together with an enhanced lower positive charge centre,
could account for this. The positive strikes were associated with high (275 kA) peak
currents.
It is notable that bipolar lightning, where successive return stroke currents are of
opposite polarity, is not uncommon in both natural and triggered flashes. This proba-
bly indicates the involvement of different cloud charge regions in a multistroke flash.
3.2.4 Flash components
The total flash duration incorporating all the individual strokes of a flash is partic-
ularly important in protection design for power systems, since multiple strokes to
an overhead line can prejudice the overvoltage protection based upon autoreclose
switchgear or surge arresters. Restrikes can be caused in switchgear and excessive
power dissipation in arresters by long duration overvoltages.
Schonland and Collens [7] showed from electric field change measurements that
the flash often comprised multiple successive strokes. The mean strokes/flash value
was 2.3, and the points of strike to ground of individual strokes may be significantly
displaced (section 3.2.8). They found probability values of:
P(single stroke) = 45%
P(> 10 strokes) = 5%
(3.4a)
More recent studies by Rakov et al. [28] find that 80 per cent of negative flashes
contain two or more strokes, but that few positive flashes do so.
In Schonlands work the mean flash duration was estimated to be 200 ms, and:
P(< 64 ms) = 5%
P(> 620 ms) = 9%
(3.4b)
3.2.5 Peak current
The basic knowledge of peak lightning currents i
0
from direct measurements is owed
mainly to Berger [24], Anderson et al. [13, 29], Garbagnati and Piparo [30], Dien-
dorfer et al. [17, 31] and Janischewskyj et al. [32]. Again, the function of modern
lightning tracking systems allows new, more extensive semiquantitative indications
of current magnitudes.
Bergers measurements of the peak current probability distributions for return
strokes preceded by negative stepped and dart leaders, and for positive lightning,
were of log-normal form and ranged from a few kA to above l00 kA with a median
value of about 30 kA. As pointed out by Berger himself, his 70 m towers at 914 m
altitude often initiated negative lightning by upward positive leaders of at least 1 km
in length; positive lightning almost always began with an upward stepped (negative)
leader, so that the peak current values for his recorded strikes might differ fromnormal
downward flashes. In Diendorfers measurements, the median peak currents to the
tower were 12.3 kA, compared with 9.8 kA to the surrounding terrain (both values
obtained from radiated signal analysis). Direct measurement of lightning currents is
also feasible from rocket-triggered flashes. However, such lightning strikes are again
initiated by upward leaders and are thus atypical of natural downward flashes [33, 34].
When structures of height 60 m or less are struck, this is usually the result of the
approachof a downwardleader towithinthe strikingdistance before the upwardleader
is launched. The number of analysable flashes is consequently not large. Anderson
and Eriksson [12] have examined a worldwide sample of 338 flashes, where over a
third of these were obtained by Berger at Mt San Salvatore. The overall median peak
current in the first stroke of a flash was 34 kA, and for high current flashes:
P(> 100 kA) = 2.5% (3.4c)
The current in subsequent strokes of a multiple flash was about 60 per cent lower than
this.
Anderson and Eriksson point out that the statistical distribution of peak currents
below 20 kA can be well represented by a log-normal distribution, with a second,
narrower log-normal distribution for higher currents. The practical merit of this
approach in respect of overhead line protection is that the lower current flashes can be
identified with the risk of shielding failure, whereas high current flashes may cause
backflashovers. In the UK Standard BS 6651, the cumulative probability range is
taken to extend from 1 per cent for currents to exceed 200 kA to 99 per cent for
currents to exceed 3 kA.
Data from the EA Lightning Location System in the UK [35] estimate current
magnitude and polarity. A signal strength scale of 1 to 5 is used, corresponding to a
range of current from less than 10 kAto more than 80 kA. There is a possibility that
the bimodal log-normal distribution derives from separate log-normal distributions
for each polarity. The observed positive flash majority in Scotland has a median peak
current of 80 kA, whereas the more common negative flashes in southern England
have peak currents of 1215 kA[27].
The electric field component of the radiated signal will have a magnitude propor-
tional to the peak current and the return stroke velocity, and inversely proportional
to the distance of the sensor from the flash. Errors in estimated currents will arise
from propagation losses and uncertainties in relation to the return stroke process. The
precision is judged to be about 25 per cent.
Current measurements in USA during 19951997 gave median values of peak
current of between 19 kA and 24 kA, depending upon the geographical region [36].
The network has a low detection efficiency for flashes with a peak current below
5 kA. This has been confirmed by Diendorfer et al. [31], from direct comparisons
of strike currents to their tower and simultaneous records from the ALDIS location
system in Austria.
On the basis of recent data from such systems, Darveniza [37] has suggested that
a median current of about 20 kAis more appropriate than the long accepted values of
31 kA(CIGRE) or 33 kA(IEC).
3.2.6 Current shape
Data on the temporal characteristics of the return stroke current are largely owed
to the work of Berger, with additional estimates of current front duration from the
USA [38]. Bergers risetimes were of median value 5 s for the first stroke current;
for subsequent strokes steeper fronts of up to 100 kA/s were measured. More
recent values from the literature are 965 kA/s (median 24 kA/s) for the first
stroke, and l0l62 kA/s (median 40 kA/s) for subsequent strokes, and it is in the
nature of extreme value statistics that as newresearch becomes more intensive, higher
maximum values are recorded. BS6651 [39] gives (di/dt ) max = 200 kA/s.
The rate of change of current measured by Berger in the first and later strokes of
the flash may be represented by:
di
dt
= 3.9i
0.55
0
[kA/s] (first stroke) (3.5)
di
dt
= 3.8i
0.93
0
[kA/s] (subsequent strokes) (3.6)
Bergers measurements also provide mean values for the current tail duration and its
statistical range:
first stroke: mean 75 s, range 30200 s
later strokes: mean 32 s, range 6.5140 s.
Up to half of all ground strikes also show a follow-through continuous current
component of about 100 Awhich can persist for some hundreds of milliseconds. The
long continuous current in the positive leader initiating an upward flash from a tall
mast can carry several kiloamperes for tens of milliseconds [18].
3.2.7 Electric fields
3.2.7.1 Field below the thundercloud
The electric field at ground level is observed to change from the fair weather value of
about +130V/m(created by the 0.6 MCnegative charge on the earth) to a maximum
15 or 20 kV/m below the bipolar charge of a thundercloud, the negative charge
centre being at a lower altitude (about 5 km) than the main positive charge (10 km).
These are large charges of about 40 C, which create a cloud to ground potential of
the order of 100 MV. There is frequently a smaller lowaltitude positive charge at about
2 km that probably resides on precipitation from the upper cloud and is sometimes
responsible for positive downward leader initiation.
The negative thundercloud field has been shown from balloon measurements
to increase linearly from ground level at a rate of about 0.1 kV/m
2
to reach, for
example, a value of about 65 kV/m at 600 m altitude [40, 41]. This increase arises
because positive corona with a current density of up to 10 nA/m
2
from grounded
objects creates a screening effect from positive space charge and an intensification
with increasing height. An extensive region of charge density of about +1 nC/m
3
is
responsible. The screening effect is thought also to be the cause of very long (10 km)
horizontal leaders before a final leader-to-ground jump. Horizontal leader channels
are commonly observed in Japanese winter storms; the effect is clearly relevant to
the side flash risk.
There has also been interest in the localised strike inhibition effects possibly
associated with multipoint air terminations. This arises from the generation in time t ,
from a termination of radius R
0
, of an expanding, spherical, positive ion cloud of
charge Q, mobility k and an increasing radius:
R(t ) =
_
l +
3kQt
4
0
R
3
0
_
1/3
R
0
(3.7)
The inhibition effect of a positive space charge layer would be subject to wind dissi-
pation, and the retardation of a positive upward leader from a ground structure is not
yet quantified [42, 43].
The increase of field with altitude, together with the concentration of field at tall
structures, accounts for the increased frequency of lightning strikes at such structures,
especially in mountain environments, and to rockets and aircraft. At the St Privat
dAllier research centre of EDF, upward positive lightning leaders were initiated
from a 100 m rocket wire when the ground field value was 10 kV/m. There is some
doubt of course whether measured characteristics well represent natural lightning.
In 1988 tests at Kennedy Space Center showed upward positive leader growth from
the rocket wire for several milliseconds before a downward negative stepped leader
was triggered. Nevertheless, despite these reservations, simultaneous current and
electric and electromagnetic field data can be accumulated which are impossible to
obtain from natural lightning. In Florida electric field changes have been recorded
within tens of metres of the flash [28].
3.2.7.2 Electrostatic and electromagnetic field changes during the flash
Although measurement of radiated electric field is useful for peak current estimates
(section 3.2.2), close-in (within 20 km) electric field transient changes are more
informative for the study of the development of the flash. This is especially so when
combined with time-resolved photography of the flash, as in the work of Schonland
and Berger. These field changes have been measured by rotary field mills, with a
resolution time of 0.11.0 ms, or fast response capacitive probes with buffer ampli-
fiers. The main characteristics of the field changes observed for negative lightning
are shown in Table 3.1.
These field change studies, with synchronised photography, have provided much
detail on spatial growth. The triggered lightning data of Rakov et al. [28] on R and
M changes have allowed physical modelling of conduction in the return stroke.
At the present time, the vulnerability of electronic equipment to LEMP (lightning
electromagnetic pulses) produced close to lightning strikes is of much concern
(section 3.5). Fuchs [18] found the maximum values of fields and their rates of
change at a distance of 200 mfromthe Peissenberg tower to be 30 kV/m, 84 kV/m/s,
Table 3.1 Electrostatic field changes for negative lightning
Nomenclature Source Character
L downward or upward leader of first
or subsequent strokes
close to leader: gradual
negative-going field change
at distance: gradual
positive-going
change charge on leader 320 C
R first and subsequent return strokes positive step field changes
C continuous current component slow positive change
M continuous current component positive pulse transients
J interstroke intervals slow positive change
K interstroke intervals positive pulse transients
35 A/m and 237 A/m/s. Concerning close-in magnetic fields, Schnetzer et al.
[44] found induction by the displacement current density
0
dE/dt to dominate for
distances beyond 50 m from the flash. Close-in electric fields are much amplified by
the radiation from the return stroke. Calculations by Schwab and Cvetic [45] of the
magnetic potential derivative dA/dt in the region of the insulators on a 110 kV
tower show these fields to be 50 to 120 kV/m, larger than the backflashover stresses
arising from return stroke current in the tower surge impedance (section 3.5.5).
3.2.8 Spatial development
The spatial growth of the lightning flash was described by Schonland and his
colleagues [7, 46, 47] (Figure 3.2). The overall behaviour of a negative strike to
ground, as summarised by Berger [24] is that in most cases a down-coming first
negative leader is initiated at an altitude of about 5 to 10 km, advancing apparently dis-
continuously in steps of 10200 mlength (mean value 25 m) at intervals of 10100 s
(mean 50 s), and establishes a path for the ensuing upwardly-directed first return
stroke. The mean downward negative leader velocity is 0.10.8 m/s, but Schonland
could not measure the velocity of the stepped advance, which exceeded 50 m/s.
Figure 3.2 Spatial development of the negative downward flash (courtesy of
Progressive lighting: A comparison of photographic and electrical
studies of the discharge, by B.F.J. Schonland and H. Collens,
Proceedings of the Royal Society. Series A, vol. 143, 1934, pp. 654674)
These apparent (two-dimensional) velocities may be about 30 per cent lower than the
real velocities. Because the transient local electric field changes recorded during the
negative leader growth showed no marked stepping, this suggested to Schonland that
the majority of the negative charge lowered by the leader was transported during the
intervals between the visible leader steps. He postulated that the stepped first leader
was itself preceded by a non-visible pilot leader which progressed continuously rather
than in steps; laboratory evidence for this is discussed in section 3.3.1.6. The down-
ward negative leader visible below the cloud base is undoubtedly accompanied by
an upward-directed leader, from the same negative charge centre of the cloud, that is
not observable from ground level but whose development can be modelled [48].
If the downward leader approaches within striking distance of the position of a
prospective earth termination, an ascending positive leader of a velocity in the range
2060 mm/s is launched from that position. Such low velocities are also found in
triggered upward leaders [49] and in cloud flashes [50], which suggest propagation at
lowelectric field. After the formation of the bright return stroke by the junction of the
descending and ascending leaders, the velocity of the upwardly-directed luminous
front along the leader track increases from 15 to 150 m/s (0.05c to 0.5c, where c is
the light velocity) with increasing peak current. After a delay of about 50 ms, further
component strokes of the flash may occur along the same path for most of its length,
and for these strokes a continuous downward dart leader with a velocity of 120 m/s
precedes its return stroke. Multiple stroke flashes can be of one second duration.
From the point of view of estimating the hazard associated with direct strikes to
earth, the risk factor (section 3.5) is significantly increased in the light of observations
that more than one ground termination may occur in multiple stroke flashes. This
results from the creation of a new leader path from some point along the downward
dart leaders that precede the second and later strokes of a flash. Ishii et al. [50] used a
fast antenna network to locate ground termination points to within 500 m, and found
that for negative flashes the average distance between stroke terminations was 2 km
(interstroke time intervals 15130 ms) and for positive flashes 13 km (30190 ms).
For mountain environments [18], tall structures [51] or airborne objects [52], the
negative flash is frequently initiated by an upward positive leader. In these situations,
the less common positive polarity flash is also observed to begin with an upward
negative leader, which shows all of the stepping characteristics of the negative down-
ward leader [24]. Bergers rare records of a positive downward leader to a lowaltitude
termination showed a continuous growth with no stepping characteristic.
3.3 Physics of lightning
3.3.1 Long sparks in the laboratory
3.3.1.1 Leader mechanisms
The physical similarities between the small-scale electric spark and natural light-
ning inspired Franklins exploration of these phenomena and led to his innovative
safety measures. Modern research continues to use controlled laboratory tests for
both physical studies and improvements in the design and protection of electronic and
power plant against both natural and internally generated voltage and current surges
[53, 54]. These advances have indicated answers to three basic questions about the
lightning flash:
(i) How can a lightning leader develop in electric fields as low as 50 kV/m?
(ii) What is the mechanism of the stepped negative leader?
(iii) Why does an upward leader initiate from a grounded structure when a given
striking distance is achieved?
3.3.1.2 Spark leaders
The process that establishes a long discharge path in the atmosphere by the longitu-
dinal growth of a leader channel was revealed first in the lightning flash. The work of
Allibone and Meek [55] made a detailed study of the leader channel that also preceded
the laboratory spark, using a moving film camera in order to obtain time-resolved
streak photographs analogous to the rotating lens Boys camera lightning records. By
the use of resistive control of the impulse generator, a step-like progression of the
spark leader could be created. This current limitation of the bright full sparkover path
also enabled the fast, lowluminosity leader phase to be discerned on the streak picture.
Waters and Jones [56, 57] avoided the difficulty of the masking effect of the bright
phase by employing a high aperture (f/1.0) reflector lens, equipped with a rotating
mirror synchronised with the impulse generator, to photograph arrested leader strokes
in which the impulse voltage was chopped before full sparkover. With no circuit limi-
tation of current, the leader progressed continuously until sparkover or its prior arrest.
These laboratorytests were usuallymade withpositive impulses, since the positive
leader is more easily initiated than the negative. The minimum breakdown voltage
for a 2 m gap with a positive impulse is about 800 kV. However, very long gaps of up
to 32 m show a rate of increase of breakdown voltage with gap length of only 55 kV
per additional metre [58]. For gaps of some metres the negative polarity breakdown
voltage is significantly greater than the positive value, but recent tests by Mrazek
[59] for gaps of up to 50 m showed that the polarity difference disappears in long gap
sparkover.
It is the ratio of electric field to air density within the leader channel that is the
important parameter to maintain ionisation, so that the field of 500 kV/m in the
positive streamers ahead of the leader can fall to the 55 kV/m in the long leader if
the gaseous heating of the leader channel can result in such a density reduction. The
leader channel is distinguished from the cold plasma in the streamer filaments that
precede it by its temperature elevation to 10003000 K. Leader growth in long gaps
can then resemble closely the continuous growth in lightning. In 50 m gaps Mrazek
observed the negative leader to progress in 3.6 m steps at intervals of 24 s (mean
velocity 0.15 m/s).
3.3.1.3 Leader gradient
The electrical gradient within the leader channel is influential on the stability of
leader propagation. Successful growth of the spark leader requires that its tip potential
remains high enough to support corona ionisation ahead of itself, and that the conduc-
tivity of the established leader channel is maintained. Waters and Jones [56] showed
that the positive leader (without external current limitation) is accompanied by an
extensive corona discharge at the leader tip. A deceleration of the leader to about
10 mm/s occurs that might result in its extinction if the applied voltage is insuffi-
cient. Otherwise an acceleration to 100 mm/s or more bridges the gap and leads to
the arc phase that is analogous to the lightning return stroke.
The work of the Les Renardieres Group [6062] in 10 m gaps using electric field
probes has enabled the mean gradient g over the length of the leader to be deduced.
It is clear that as the length of the leader grows, the mean gradient within the channel
decreases. For example, in a 10 m gap at critical sparkover (1.8 MV), the mean
gradient along the leader is 0.5 MV/m when the leader is 2 m long, and this falls to
0.18 MV/m when it has grown to 5 m. This decreasing gradient characteristic means
that the potential at the tip of the leader varies little during its development, which
is consistent with the uniform velocity and current (1520 mm/s, 0.61 A) that is
observed in this case. It is also, of course, the underlying cause of the non-linear
breakdown voltage and the falling breakdown-gradient/gap-length characteristic of
long sparks.
All the recent theories of the leader channel attempt to account for this negative
gradient characteristic, which is also the key to understanding howthe lightning leader
of tens of kilometres length is able to maintain its conductivity. These theories need a
quantitative knowledge of the leader diameter as a function of current and time. As in
lightning, measurement of this diameter by direct photography is difficult because of
the limited resolving power of this method. This problem in long sparks has largely
been solved by the use of strioscopy (Schlieren photography), combined with high
speed image recording. Ross [63] found that the leader channel thermal boundary
was so clearly defined by this technique that the diameter could be measured with a
precision of 0.1 mm (Figure 3.3). The diameter of a given section of the channel
was found to increase with time. No strong shock wave was observed under these
conditions, and the rate of radial expansion was less than 100 m/s in all cases. Later
work [61] shows that a pressure wave is generated from the leader tip, but confirms
that the subsequent radial expansion of the channel is subsonic. Figure 3.4 shows,
via a 0.66 mm slit, part of a 2.2 m gap. Asweep-mode image converter time-resolves
the image in the axial direction, showing the birth of the leader with an associated
pressure wave, the channel diameter growth at 18 m/s and the instant of sparkover
with a strong shock wave resulting from the arc channel expansion. This indicates
that the gas pressure within the leader channel remains constant at the ambient value
throughout its temperature rise. The leader diameter typically increased in 10 m gaps
from an initial value of about 1 mm or less to about 5 mm for a spark leader of 6 m
length just before sparkover.
The expansion of the leader channel and the decreasing axial gradient character-
istic are related parameters. The absence of radial shock waves and the nature of the
strioscopic records of the channel show that the neutral-particle density within the
channel is a decreasing function of time, since a constant mass of gas is involved. This
reduced density is the basic explanation of howthe leader channel conductivity can be
(a) (b)
(c) (d)
Figure 3.3 Leader channel expansion in the long spark [46]. Schlieren photographs
of exposure time 1 s per frame. Field of view 100 mm diameter in a
1.5 m rodplane spark gap. Successive frames at
a 5.2 s
b 8.1 s
c 15.2 s
d 36.5 s
sustained at lowoverall electric fields. At atmospheric density a field of about 3 MV/m
is necessary to maintain direct-impact ionisation by electrons, but an expansion of the
leader radius by a factor of four will reduce the required field to below 0.19 MV/m.
This approaches the average field available in long gap sparkover (the critical impulse
sparkover voltage for a 10 m rodplane gap is 1.8 MV).
3.3.1.4 Energy dissipation and storage in the leader growth
The observed current shows that at critical breakdown the mean power input to the
leader growth is about 1 MW. Some of the energy supplied, amounting under these
130/3000 s
T
B
=60.2s
Figure 3.4 Shock wave generation at the leader head (Gibert). Schlieren record of
section of vertical spark channel viewed through a horizontal slit. Image
swept downwards by an electronic streak camera. Conical image shows
leader expansion to time of sparkover at 60 s. Radial shock waves are
visible
conditions to about 40 J/m extension of the leader, is stored in the redistributed
electrostatic field, and most of the remainder is utilised in discharge processes. This
division of energy was studied by examination of the space charge in the gap using
a rotating fluxmeter [61], which indicated an approximate equipartition of energy.
Much of the expended energy causes heating and expansion of the gas; calibrated
photomultiplier measurements show that very little energy is lost by radiation.
3.3.1.5 Physical theories of the leader channel
Kline and Denes [64] considered the influence of the gas density on the electric
gradient at which ionisation and electron attachment rates become equal. Extrapola-
tion of their estimates to a hot gas column at 5000 K gives a value for the product
ET of this critical electric gradient and temperature of 5 10
8
VK/m. Two separate
theories suggest that the value of ET at each point in the expanding leader remains
approximately constant. In the approach of Ross et al. [63], it is noted that the constant
current condition that is observed in the leader channel at critical breakdown implies:
_
n
e
n
_
v
e
= constant (3.8)
where n
e
, n are the electron and neutral particle volume densities at any instant
during leader growth, and v
e
is the electron drift velocity. This equation supposes
that the mass of gas within the expanding leader channel remains constant. Now the
equilibrium electron density in the leader, with a positive ion density of n
i
, is given
by the condition:
k
1
nn
e
k
2
n
2
n
e
k
3
n
i
n
e
= 0 (3.9)
where k
1
, k
2
andk
3
are, respectively, the rate coefficients for ionisation, the three-body
collisions that control the electron loss process of attachment to form negative ions
and electronion recombination. Since the leader channel can certainly be regarded
as a quasi-neutral plasma with n
e
n
i
, we have:
n
e
n
=
k
1
k
2
n
k
3
(3.10)
If the ratio of electric gradient to neutral gas density in the leader E/n were to increase
as n decreases because of channel expansion, then the ratio n
e
/n would also increase
strongly. This is because with the channel expansion k
1
is much increased, and k
2
and
k
3
are decreased. Furthermore, the electron drift velocity v
e
increases approximately
linearly with E/n in the range of interest. An increasing value of n
e
v
e
/n would
imply an increasing leader current with time, which is contrary to Equation (3.8).
The constant current condition therefore indicates a constancy of E/n and ET. If we
define a leader constant K of value:
ET = K = 5 10
8
VK/m (3.11)
and assume a constant specific heat, then the rate of rise of temperature in any element
of the leader is:
dT
dt
= BEi
L
(3.12)
where B is also constant and i
L
is the leader current. Combining these equations:
i
L
=
K/BE
3
dE/dt
(3.13)
Under critical breakdown conditions, where the leader current and velocity v
L
=
dL/dt are constant, the velocity is:
v
L
=
K/BqE
3
dE/dt
(3.14)
where q is the injected charge for unit leader length.
If the initial gradient is takenas 0.5MV/m(correspondingtothe initial conditionin
the positive streamer corona) and an initial leader radius of 1 mm, then by integration
the electric gradient E can be calculated at the position x = 1 m from the leader
origin. Table 3.2 shows the variation in both the gradient and radius a during the
leader growth from a length L = 1 to 7 m.
Table 3.2 Leader gradient E with increasing leader
length L at a position 1 m along channel
L (m) 1 2 3 4 5 6 7
E (MV/m) 0.39 0.28 0.23 0.10 0.18 0.17 0.15
a (mm) 1.0 3.8 4.6 5.0 5.4 5.7 6.0
Gallimberti [61] has described a hydrodynamic/thermodynamic analysis of the
expansion of the leader channel, again based upon the expansion of a constant mass of
gas within the channel. It quantifies such physical considerations as the conservation
of neutral particles, heat transport at the boundary, energy transfer fromcharge carriers
to neutrals, ambipolar diffusion and a radial distribution of gas temperature. It is
possible to obtain, as a function of time, the rate of the channel expansion, the
increase in its temperature and the weak shock generation at the leader boundary.
The value of ET within the channel again remains constant, and the same value of
5 10
8
VK/m emerges.
This relationship between electric gradient and channel temperature is a useful
concept for lightning modelling, but gas temperature estimates for the lightning
leader channel are obviously tentative. Orville [65] quotes the work of Maecker, who
derived from spectrosopic analysis a temperature of about 1500 K for this phase and
concluded also that the leader was at atmospheric pressure. This would be consistent
with an electric gradient E = 0.3 MV/mwhich is comparable with long spark leaders
(section 3.3.1.2).
3.3.1.6 Space leaders in the negative leader channel
The negative leader channel is easily observed in air gaps such as rodrod electrodes,
where the positive leader fromthe anode preionises the gap with an active high current
corona and promotes the formation and growth of a negative leader from the cathode.
The negative leader can be initiated without the assistance of an associated positive
discharge only for gaps over 3 m in length, and has been studied for gaps up to 7 m at
Les Renardieres [62]. The negative leader, like its lightning counterpart, progresses
in steps. The detailed structure of this progress in the negative polarity rodplane gap,
which is almost certainly present but is inaccessible in lightning photographs because
of its low luminosity, throws much light on the probable stepping mechanism of the
negative lightning leader:
(i) Anegative long spark leader grows continuously with a velocity of 10 mm/s
and a current of about 1 A, accompanied by regular fast bright elongations,
current pulses of up to 100 Aand large negative corona bursts. These are seen
in time-resolved streak mode (Figure 3.5a) and framing mode (Figure 3.5b)
photographs, interpreted in Figure 3.5c.
(ii) Between these elongation events, one or more space leaders formwell ahead of
the negative leader tip, connected to the main leader only by corona streamers
(as in the static photograph of Figure 3.5d). As represented schematically
in Figure 3.5e, the extension of each successive space leader takes place at
1
2
3
4
10 20 30
z, m
1
2
3
4
z, m
t, s
a
b
1
5
4 9 8
4
1 8
5
2
3
7
6
2
6
c
Figure 3.5 Negative space leaders in the 7 metre spark (Les Renardieres)
a Image-converter streak display; b Image-converter frame display
(Courtesy of IEE Proceedings, Vol. 125, 1978, p. 1160);
c Streak display interpretation (Courtesy of IEE Proceedings,
Vol. 125, 1978, p. 1160): 1 retrograde positive corona; 2 negative
corona; 3 upward positive leader; 4 downward negative leader;
5 space stem; 6 space leader elongation; 7 final jump duration;
8 leader corona region
1m
0.6m[
impulse shape:
60/3000 s
polarity: negative
withstand:
Ucr =2.86 MV
(U
50
=2.96 MV)
charge injected:
Q
total
= 430C
leader lengths:
L
right
=0.32 m
L
left
=0.24 m
L
space
=0.05 m
d
a b c d e
e
Figure 3.5 Continued.
d still photograph of space leaders
e visualisation of negative leader growth in lightning
both of its ends, with velocities of 30 mm/s towards the downcoming negative
leader and 10 mm/s towards ground. Bright elongations of the whole length of
the negative leader occur whenever it connects with the space leader(s) ahead
of its tip.
(iii) New space leaders then form in the new corona burst from the tip of this elon-
gated channel. The leader current does not show any significant discontinuities
during this stepped leader growth.
This unanticipated space leader mechanism was first reported by Stekolnikov
and colleagues [66, 67], and is easily overlooked experimentally, since the sudden
elongation (i) is a much brighter phenomenon than the space leaders (ii). The local gas
heating which triggers the formation of space leaders within the negative corona is a
stochastic process, and their existence is clearly relevant to any model of the negative
leader in both the long spark and lightning. The transformation of the corona discharge
ahead of the down-coming leader into a space leader will determine channel tortuosity,
branching and the path taken in the later stages of the lightning strike to earth. It is
clearly possible to draw a parallel between the pilot leader, postulated by Schonland
to precede the lightning stepped leader in order to account for the absence of large
step field changes, and the space leaders ahead of the laboratory negative leader.
3.3.2 Lightning leader propagation
3.3.2.1 Correlation of leader channel charge, current and velocity
In the same way that the electric gradient and temperature in the long spark leader
informs physical studies of the lightning leader channel, so the mode of propagation
can be expected to develop from similar ionisation processes ahead of the leader in
both cases. The current in the lightning leader is several orders of magnitude smaller
than the tens of kiloamperes in the return stroke that neutralises the leader charge.
Petrov and Waters [68] showed that a comparison of the charge q per unit length of
the leader and the leader current is possible by considering the ionisation at the head
of the high temperature leader channel itself. By representing the head of the leader to
have a simple circular face, then within its radius r
0
a charge density is maintained
by the leader current i
L
, so that:
i
L
= r
2
0
v
L
(3.15)
where v
L
is the leader velocity, and for current continuity:
i
L
= qv
L
(3.16)
The velocity v
L
depends upon ionisation processes at the surface r
0
, and for
propagation along the x axis:
v
L
=
_
dx
dt
_
r
0
=
[dn
e
/dt ]
r
0
[dn
e
/dx]
r
0
(3.17)
If n
0
and v
0
are the values of electron density n
e
and drift velocity v
e
at the leader tip
surface r
0
, the current density at that surface is:
j
0
= en
0
v
0
(3.18)
The continuity condition at r
0
, taking account of the ionisation coefficient
(photoionisation neglected), is:
_
dn
e
dt
_
r
0
=
_
1
e
__
dj
0
dx
_
+n
0
v
0

n
0
v
0
r
0
+n
0
v
0
(3.19)
for the approximation dj
0
/dx j
0
/r
0
.
If the electron density behind the leader head is considerably smaller than n
0
, then
we can also approximate:
_
dn
e
dx
_
r
0
=
n
0
r
0
(3.20)
Then the leader velocity is:
v
L
= v
0
+v
0
r
0
(3.21)
In the intense field at the leader head, can become very large and the leader velocity
v
L
can greatly exceed the electron drift velocity. This allows the further simplification:
v
L
v
0
r
0
=
i
r
0
(3.22)
where
i
is the ionisation rate coefficient. These approximations lead to unique basic
relationships between current i
L
, velocity v
L
and linear charge density q for the
lightning leader:
i
L
=
v
3
L
2
i
(3.23)
and
q =
_
2
i
_
1/3
i
2/3
L
(3.24)
Here is the charge density in the leader head that is necessary to maintain its
propagation [68]. This will be approximately constant during the leader propagation,
so that it is possible to deduce the important proportionalities:
i
L
q
3/2
v
3
L
(3.25)
Plausible values can be obtained from the model on the basis of laboratory measure-
ments. The ionisation frequency is typically
i
= 5.610
6
per second and the leader
head charge density has been estimated as 1 C/m
3
. These values give:
q 46 10
6
i
2/3
L
10
13
v
2
L
[C/m, A, m/s] (3.26)
Table 3.3 gives numerical examples of this leader propagation model, together with
return stroke calculations (section 3.3.2.2).
Adownward leader length of say 10 kmwould lower a total charge of 3.9 Cduring
a stroke at the median current of 31 kA. This agrees well with lightning observations.
The leader velocity of 0.062 m/s is lowcompared with the slowest observed velocity
of 0.1 m/s. This would be consistent with the formation of the leader step by the
simultaneous growth of two or more space leaders in series ahead of it (so doubling
or more the velocity). The median leader current of 25 Ain Table 3.3 compares with
the 40 to 50 Arange estimated by Mazur and Ruhnke [48].
Equation (3.26) also scales down effectively to long laboratory sparks. During the
stage of stable leader development in a 10 mgap, the measured leader current is 0.8A,
the propagation velocity 18 mm/s and the charge flow 45 C/m. Equation (3.26)
Table 3.3 Leader current, velocity and charge
with associated return stroke current and
velocity
i
0
(kA) 5 31 100 300
v
r
0.14c 0.26c 0.39c 0.55c
q (mC/m) 0.12 0.39 0.86 1.80
i
L
(A) 4.1 25 82 245
v
L
(m/s) 0.035 0.062 0.092 0.13
for i
L
= 0.8 Agives 20 mm/s and 40 C/m. This accord is important to counter the
frequently expressed doubt that the laboratory discharge can represent adequately the
lightning event.
3.3.2.2 Return stroke current
Although the properties of the leader channel can be used to model the strike process
to ground structures (section 3.4), the consequential damage caused by lightning is
strongly current dependent via the so-called action-integral
_
i
2
0
dt of the return stroke.
For this reason, the probability density of the lightning peak current distribution has
been extensively studied in the field. The connection between the properties of the
leader development and the prospective peak current i
0
in the lightning stroke is
not known; however, it is possible to relate the much smaller current in the leader
phase i
L
to the return stroke current i
0
by noting that the return stroke results in the
neutralisation of the leader channel charge q. Much of this charge certainly resides in
the ionised region surrounding the channel, as a result of the corona discharge ahead
of the leader tip propagation [69, 70]. Using the simplest of assumptions that both
charge density and velocity are constant along the channel, then:
q =
i
0
v
r
(3.27)
where i
0
is the return stroke current and v
r
is the effective return stroke velocity.
Lightning field studies of the return stroke [71] indicate i
0
v
3
r
, similar to the
relationship in Equation (3.25) for leader current and velocity, which suggests that
i
0
i
L
. A precise cubic relationship between i
0
and v
r
can also be deduced from
recent return stroke models [72].
It is recalled in section 3.2.8 that lightning observations show that return stroke
velocities lie within a range from0.05c to 0.5c, and return stroke currents mainly from
5 to 300 kA. If we correlate these values, then the proportionality i
0
v
3
r
becomes:
i
0
= 1.75 10
6
c
3
v
3
r
(3.28)
As a consequence of Equations (3.26)(3.28) we can deduce i
0
= 1220i
L
. Table 3.3
includes numerical examples of return stroke current and velocity using these
relationships.
It is also noteworthy that the charge density variation with peak current is now
obtained from Equations (3.26) as:
q = 40i
2/3
0
[C/m, kA] (3.29a)
This may be compared with the empirical equation derived from Bergers field data
by Dellera and Garbagnati [73]:
q = 38i
0.68
0
(3.29b)
3.4 Lightning termination at ground
3.4.1 Striking distance
The probability of a downward lightning flash terminating on a structure can be
calculated by empirical or by physical methods. At a critical point in the development
of a downward leader in the vicinity, an upward leader may be launched from the
grounded structure, and this will determine the location of the strike unless another
competing upward leader precedes it to make the first connection to an adjacent
structure. The distance at this moment of launch between the tip of the downward
leader and the origin of the successful upward leader is defined as the striking distance.
Calculation of the striking distance enables both a risk assessment to be made
of the probability of a flash to a structure and also of the efficacy of the protection
afforded by a grounded air termination, overhead ground wire or Faraday cage.
It is useful to classify into three types the striking distance calculations that are
presently used:
(i) Geometric models, in which the striking distance is assumed to be independent
of both the prospective peak current i
0
in the return stroke and the geometri-
cal contours of the ground structures. This simplest of models is nevertheless
the basis of international standards for lightning protection because of its
convenience and utility.
(ii) Electrogeometric models, where the striking distance is represented by a func-
tion only of i
0
, and is again supposed independent of the local geometry of
the ground structures. Because of the prime influence of i
0
in the prospective
magnitude of injected lightning overvoltages, this method was developed for
and is widely used in insulation coordination of overhead line power systems.
(iii) Generic models, which take account more realistically that both i
0
and the
structure geometry will together determine the striking distance. With these
more refined concepts, such factors as flash polarity and ground elevation
can be incorporated, and calculations can employ electric field and statistical
packages. These are recent advances but, except in specialised risk assessment
requirements, generic models have so far remained in the scientific rather than
the engineering domain. There is a case for their increased use as data on
lightning parameters improve, and an extensive review is included here.
3.4.2 Geometric models and lightning standards
As envisaged in Preeces early work, the height h of a structure can be used to estimate
its exposure risk (or the efficiency of an air termination). Suppose that a mast (or a
long horizontal conductor) of height h is approached by a downward leader, whose
tip has a horizontal distance component r from the mast axis (or conductor position).
If the leader tip reaches its striking distance r
s
at the instant when it is equidistant
from the mast top (or the conductor) and the ground surface, then this horizontal
component will define the attraction radius r = r
a
:
r
a
= h
_
2r
s
h
1
_
1/2
(3.30)
In present international standards this concept is simplified to the proportionality
r
a
= kh, where the choice of k = 1 or 3 may be used (section 3.5.2), with the unstated
implication from Equation (3.30) that r
s
= h or 5h.
In the BS 6651:1999 Code of practice for the protection of structures against
lightning [39] are three fundamental recommendations each linked to the pheno-
menology of the flash:
(i) The conventional 30
or 45
cones of attraction are retained only for structures

less than 20 min height. The risk of side flash is estimated by the rolling sphere
method (Figures 3.6a and b) [74].
(ii) The increased vulnerability of electronic equipment (for telecommunications,
telemetry, computing, control and instrumentation, and power electronic instal-
lations) to transient fields is recognised by a recommended increase in the
number of downconductors (which may be structural steel members).
(iii) The risk factor for direct strikes is quantified.
These practical measures are discussed in section 3.5, but fromthe point of viewof
the physics of the lightning flash, developments (i) and (iii) are particularly relevant.
The adoption of the rolling sphere criterion has arisen from the pragmatic desire for a
straightforward alternative geometrical construction to the cone that will better predict
the side flash risk. It also offers a generally useful application to other problems such
as attachment points to aircraft. The discovery of space leaders in the recent work
on long gap negative breakdown suggests that the rolling sphere is indeed a better
physical analogy than the cone. As stated in section 3.2.8, the original concept of
Schonland that a non-visible pilot leader develops from the previous tip to establish
the next step can be invoked. It can be postulated that, as occurs in the laboratory long
spark, the negative step is preceded by a space leader developing in both forward and
retrograde directions froman origin that lies ahead of the leader tip. This bidirectional
growth will thus be well simulated by a spherical region; in this interpretation, the
centre of the rolling sphere can be regarded as the origin of the space leader ahead of
the previous step rather than the tip of the step itself, as envisaged in the Standard.
Since step lengths of 10 to 200 m are observed, a radius for the rolling sphere of 5 to
100 m seems supportable. This compares with the recommended radii of 20 or 60 m
in the Standard. The smaller choice of rolling sphere radius is the more onerous for
60 m protected
volume
additional
air termination
air termination
air
termination
air termination
a
b
last leader step occurs
from centre of circle
R
S
= striking distance =
radius of sphere
zone of protection for objects or people
R
S
Figure 3.6 Rolling-sphere representations of the ground strike (BS 6651)
the protection system design. The US NFPAStandard 780 [75] gives the values 46 m
and 30 m, and the IEC Standard 61662 [76] specifies the four radii 60, 45, 30 and
20 m as options.
This flexibility in the choice of rolling sphere radius is intended both to recognise
the expected effects of structure height on the strike process and to allow a choice
of security rating for the structure. In addition, it implicitly admits that a purely
geometric model does not incorporate the reality that the striking distance will be
smaller, and protection more difficult, for smaller prospective peak currents [77].
In the long laboratory negative spark, several space leaders can develop simulta-
neously in series ahead of the leader tip, all subsequently causing bright leader channel
elongations and stepping. So it is possible to envisage, as a geometrical representation
of this process, not merely a single rolling sphere, but a string of small radius spher-
ical regions ahead of a descending leader. This method would imply a higher rate of
predicted shielding failures (Figure 3.7).
3.4.3 Electrogeometric models
3.4.3.1 Effect of peak current
A simplistic geometrical model has obvious limitations in not quantifying the initia-
tion of the upward leader in terms of the field enhancement associated with the electric
Figure 3.7 Spherical-string representation of the ground strike
charge on the downward leader. Quantitative estimates must rely on modelling,
and the first authors to develop the now generally used relationship:
r
s
= ki
n
0
(3.31)
for a given structure were Armstrong and Whitehead [78, 79]. They based this
relationship upon physical concepts due to Wagner [71], and
r
s
= 9.4i
2/3
0
(3.32)
is commonly used [80, 81].
The striking distance as a function of current was calculated by these authors on
the basis of the assumptions that a 1 C leader charge was equivalent to a 20 kApeak
current. Lightning observations by Berger, Anderson and Kroninger [82] suggest
a leader charge of 4.5 C for a median lightning current of 30 kA. Armstrong and
Whitehead [58] also needed to specify the mean electric field to complete the lightning
channel over the striking distance. They used values of 5 kV/cm(for negative flashes)
and 3 kV/cm (for positive flashes).
The electrogeometric model developed by Armstrong and Whitehead [58] pro-
vides a more realistic representation of the protection principle than a simple shielding
angle. It is in extensive use for the design of lightning protection not only for power
lines and substations, but other important structures such as chemical plant and
aviation and rocket installations.
3.4.3.2 Electrogeometric boundary for strikes to a vertical mast
The electrogeometric model makes the same important simplifying assumption as
before that the striking distance is independent of the geometry of the grounded
y
r
a
r
s
x
s
h/2
h/2
r
s
Figure 3.8 Electrogeometric model for a vertical mast
structure. The value of r
s
at a given peak current is assumed to be the same for a
strike to the mast as for one to the surrounding terrain, so the calculations are as for
Equation (3.30). For example, for a vertical mast of height h, and for large striking
distances where r
s
> h/2, the limitingboundaryfor strikes tothe mast top(Figure 3.8)
is defined by the parabola:
y =
x
2
+h
2
2h
(3.33)
This parabola is the locus of the point S which defines, for each value of prospective
peak current i
0
, the limiting condition for a strike to the top of the mast. Here y =
r
s
= f (i
0
), since r
s
is assumed to be the striking distance to both mast and ground.
From the above parabolic equation,we can define the attraction radius of the mast for
a flash with this current:
r
a
(i
0
, h) = x(r
s
) = h
_
2r
s
h
1
_
1/2
(3.34)
and the attraction area of the mast, i.e. the protected area at ground level:
a(i
0
, h) = x(r
s
)
2
= h
2
_
2r
s
h
1
_
(3.35)
The attraction distance and attraction area are termed collection distance and area in
lightning standards (section 3.5.2). If the mast is a lightning rod (or air termination)
then the terms protection distance and area are appropriate. It should be noted that
the attraction distance is smaller than the striking distance except for r
a
= h = r
s
.
The shielding angle concept can be retained where:
s
(i
0
, h) = tan
1
_
x(r
s
)
h
_
= tan
1
_
2r
s
h
1 (3.36)
Calculation of the attraction area, together with a probability distribution for the
lightning currents, enables the risk factor for the mast to be found (section 3.5.1,
Equation (3.58)).
Table 3.4 Electrogeometric model: striking distance, attrac-
tion area and shielding angle for a vertical mast
i
0
(kA) 5 31 100 300
(median value)
h (m) 20 60 20 60 20 60 20 60
r
s
(m) 27.5 93 203 422
a (m
2
10
3
) 2.2 2.4 10 24 24 64 52 150
s
53
25
71
55
77
67
81
75
For r
s
< h/2, the protected area is simply:
a(min) = x(r
s
)
2
= r
2
s
(3.37)
Examples of these calculations are shown in Table 3.4. These also illustrate the
inherent implication of the electrogeometric approach that elevated structures are
more selectively struck by higher current flashes. It is based on representation of
a downward flash and takes no account that for structures over 100 m tall or at
high elevation the local geometrical enhancement of the thundercloud field by the
structure may be sufficient for the lightning flash to be triggered by an upward leader
even before any downward leader is observed.
3.4.3.3 Shielding calculations for an overhead line
The same simplifying assumption of striking distance to be a function only of peak
lightning current is the basis of a well known approach to the calculation of the
shielding of the phase conductors that is provided by an overhead ground wire
(or wires).
The probability of a shielding failure, defined by a direct strike to a phase
conductor, is in the case of a single ground wire a function of the volume of the
prospective strike zone bounded by (Figure 3.9a):
(i) the parabola defining equidistance between the phase conductor and the ground
plane
(ii) the linear locus PRequidistant betweenthe groundwire andthe phase conductor
(iii) the circular locus centred on the phase conductor, defining the maximum safe
striking distance r
s
(maximum) = r
m
.
This last value is determined by the basic insulation level (BIL) of the line, since
this defines a maximum allowable peak current for such a direct strike to a phase
conductor. From the unprotected volume, the risk factor for a direct strike to a phase
conductor (section 3.5) can be calculated using the N
g
value for the region.
Complete shielding is rarely justified economically. For the line of Figure 3.9,
complete shielding is achieved, with a critical shielding angle
c
, for QR = PR.
The procedure is then to specify, according to the system insulation coordination
requirements, the maximum allowable overvoltage, the anticipated peak lightning
H
h
r
s
0
s
R
O
Q
P
prospective
strike
zone
a
b
h
r
s
r
s
0
c
H
R
O
PQ
complete shielding for 0
s
=0
c
[

Figure 3.9 Electrogeometric modelling for overhead line shielding
a Strike zone for line with shield wire
b Evaluation of critical shielding angle
current and corresponding striking distance. Then, using Figure 3.9b, the associated
critical shielding angle is calculated from:
c
=
_
2

_
= sin
1
_
1
h
r
s
_
sin
1
_
H h
2r
s
cos
c
_
(3.38)
where h and H are the mean heights of the phase conductor and the ground wire.
3.4.4 Generic models
Since the late 1980s, models of the striking distance, based to a large extent upon
the improved knowledge of the physics of long sparks, have taken account of both
prospective peak lightning current and the geometry of the grounded structure. Some
examples follow.
3.4.4.1 Eriksson model
Following a review of field data, Eriksson [16] developed a quantitative model to
calculate the attraction radius for a vertical mast on flat terrain. The criterion for
launch of the upward leader is based upon the critical radius concept developed from
long spark studies [83]. Effectively, this represents the apex of the mast as having a
radius of curvature of 0.35 m, since positive leader inception in the laboratory varies
little for radii smaller than this. An induced field of 30 kV/cm at the surface of this
notional anode is required for upward leader inception. The magnitude of the induced
field depends upon the charge on the downward leader and the prospective peak
current, and for a downward leader directly above the mast a strike is then inevitable.
For a vertical downward leader displaced horizontally from the mast, the possibility
of completion of a strike to the mast is calculated from the downward (negative) and
upward (positive) leader velocities. The model does not directly utilise the physics of
leader and leader corona growth, but is a powerful and simple approach for which a
regression analysis yields the relationship between attraction radius r
a
(m), lightning
return current i
0
(kA) and mast height h (m):
r
a
= 0.84i
0.74
0
h
0.6
(3.39)
The resulting attraction radius is calculated for different lightning return stroke
currents i
0
and mast heights h, as shown in Table 3.5, where the values deduced
for the electrogeometric model are also shown.
3.4.4.2 Dellera and Garbagnati model
This also uses the critical radius approach to determine the upward leader inception
[73, 84], and can be applied to both downward and upward lightning. The model
computes the spatialtemporal development of both downward and upward leaders
numerically, using a charge simulation program to calculate the electric field distri-
bution, and simulates incrementally their directional and velocity development in the
direction of maximumfield. The inclusion of the streamer zone between the leader tips
and the use of field computation techniques facilitate predictive engineering appli-
cations such as a realistic representation of the local ground topology at the mast or
transmission line. Examples of the output of this model are shown in Table 3.5. They
generally indicate a smaller attraction radius than the Eriksson model and a stronger
variation with mast height. The charge simulation software also allows the cloud
charges to be included, and the probability of upward-directed flashes is computed
as a function of structure height. Equal probability of upward or downward flashes is
estimated for a height of 230 m.
Table 3.5 Generic models of attraction radius for a single mast
(downward negative flash)
i
0
(kA) 5 31 100 300
h (m) 20 60 20 60 20 60 20 60
r
a
(m) electrogeo. model 26 28 56 87 87 143 129 218
r
a
(m) Eriksson 17 32 64 125 147 296 345 668
r
a
(m) Dellera 20 30 35 95 90 240 170 535
r
a
(m) Rizk 22 45 95 150 180 280 315 560
r
a
(m) Petrov 17 28 57 95 124 206 258 429
3.4.4.3 Rizk model
Here, a critical potential criterion determines inception of the upward leader, and then
the completion of the lightning strike is determined from two aspects of long spark
studies [61]. The first aspect sets a critical streamer gradient between the downward
and upward leaders of 5 kV/cm at normal air density. Second, the voltage gradient
along the leaders, which is known fromthe work of the Les Renardieres Group [43] to
decrease temporally from5 kVcm1 to 0.5 kV/cmor less, is also represented, together
with the effect of reduced air density at high altitudes. An analytical evaluation of the
electric field allows simple ground topologies (based upon a semiellipsoidal terrain)
to be simulated. The predictions for a mast on flat terrain are shown in Table 3.5,
where the attraction radii are similar to those deduced from the Eriksson model.
On elevated terrain, the effect of the thundercloud field is predicted to increase the
attractive radius of tall masts by over 100 per cent for lowpeak currents. A50 per cent
probability of an upward flash is expected for h = 160 m.
3.4.4.4 Petrov and Waters model
This aims to be a flexible model, based upon the physics of the leader channel
described in section 3.3.2, which is adaptable for stroke polarity and terrain altitude.
The procedures which will be described here lead to the values of attraction radius
for negative flashes at sea level that are shown for comparison in Table 3.5. A mast
height of 190 m is calculated for 50 per cent probability of upward flashes (Table 3.8,
section 3.5.4). This model agrees with values deduced from the electrogeometric
approach for currents up to the median, but generic models predict significantly
greater attraction radii for the larger peak currents.
(i) Upward leader criterion
Aphysical condition for the launch of a successful upward leader that will complete
the junction with the down-coming negative leader [68] is not in this case based on a
critical radius for leader inception, but upon a critical interaction between a putative
upward discharge and the downward leader. In Figure 3.10, the leader channel is
represented by a vertical linear charge of length L with a charge per unit length q and
leader tip charge Q. The mast is shown as an ellipsoid of height h and half width b.
Q
E
L
q
h
b
E
min
E
*
cr
H
S
x
0
Figure 3.10 Analytical modelling of a downward leader above a ground mast [53]
At large distances S between the lightning channel and the mast, the range of the
electric field intensification above the top of the mast may be insufficiently extensive
to support a successful upward leader, although there may well be corona streamer
activity and weak leader growth as the downward leader approaches. The streamer
corona from the top of the mast propagates to a distance where the electric field falls
to the minimum streamer gradient E
s
, at a distance x
0
from the top of the mast. For
standard sea level conditions, the electric field E is equal to about 5 kV/cm for a
streamer of positive polarity and about 10 kV/cm for negative polarity. The criterion
for the lightning strike to the mast is thus a critical upward streamer length so that an
upward leader can be successfully developed.
Evaluation of the critical streamer length is made from long spark studies: if an
upward positive leader of charge q
u
per unit length grows simultaneously with a hemi-
spherical upward leader corona region of radius L
u
, the charge per unit radius within
the leader corona will be equal to that in the leader. Thus the charge within the upward
leader corona zone is given by q
u
L
u
. This charge produces at the hemispherical
surface of the upward leader corona a field:
E
S
=
q
u
2
0
L
u
(3.40)
The critical streamer length L
u
is associated with a critical value of q
u
. In the case
of a positive upward leader, long spark studies show that the minimum charge for
positive leader inception is about 2040 C/m (section 3.3.1.1), so enabling the
minimum L
u
to be found. It is known also from optical and electronic measurements
that the minimum length of the streamer zone of the positive leader in long air gaps
is about 0.7 m [53]. This corresponds to a critical streamer charge q
u
of 20 C/m.
(ii) Termination to a plane ground
For specific cases of gaps with simple geometry, analytical expressions for the poten-
tial and electric field may be obtained. In particular, the axial electric field distribution
created by the vertical downward leader channel represented in Figure 3.10 at a height
H above an earth plane surface has the form:
E(x, H, L) =
_
Q
4
0
__
1
x
2
+
1
(2H x)
2
_
+
_
q
4
0
__
1
x
1
x +L
+
1
2H x
1
2H x +L
_
(3.41)
In the absence of a structure the field at ground level (x = H) is, for H L:
E
g
=
_
1
2
0
__
Q
H
2
+
q
H
_
(3.42)
If the leader is sufficiently close to ground so that this field is equal to the critical
streamer propagation field E
s
, then the leader will complete the strike to ground
without any upward discharge growth.
Although the streamer is rarely visible in lightning photographs, it can be inferred
from long gap studies to be essential for the propagation of the leader. For analytical
purposes, we may represent the net charge in the streamer zone as the charge Q
(Figure 3.10). Petrov and Waters [53] showed in the following way the correspon-
dence between the streamer charge Q and the linear charge density q consequently
established upon the leader channel. If the streamer zone of the downward leader is
represented as a hemisphere, with a radius equal to the streamer zone length L
S
, then
we can write:
Q = qL
S
(3.43)
But the field at the head of the streamer zone is:
E
S
=
Q
2
0
L
S
(3.44)
The important relationship between the charges Q and q is therefore from
Equations (3.43) and (3.44):
Q =
q
2
2
0
E
S
(3.45)
As a result, the charge Q of the leader head and the leader channel charge density
q are both determined by the streamer zone length. For example, the charge q on
the channel is 0.39 mC/m for a median current stroke (Table 3.3). Equation (3.45)
gives the charge Q on the negative streamer system as 2.7 mC and the streamer zone
length L
S
= 7 m. So the charge Q can be expressed in terms of q, and in the case of
E
g
= E
S
the height H represents the striking distance r
s
. From Equation (3.42):
E
S
=
_
1
2
0
__
q
2
2
0
E
s
r
2
s
+
q
r
s
_
(3.46)
which yields a value of striking distance:
r
s
=
_
q
4
0
E
S
_
[1 +
5] (3.47)
Using now the relationship between q and i
0
obtained from Equations (3.26) and
(3.29a), and a value of E
S
= 10 kV/cm for the propagation field of the downward
negative leaders, we get:
r
s
= 1.16i
2/3
0
[m, kA] (3.48)
These striking distances shown in Figure 3.11a must be regarded as a lower limit for
the striking distance to a plane ground, since no upward positive discharge has been
assumed. In practice, such upward leaders are observed from local asperities on level
terrain, and r
s
may consequently be larger than in Equation (3.48).
(iii) Strikes to a mast
When a vertical mast is approximated by a semiellipsoid of half width b as in
Figure 3.10, the electric field between it and a vertical coaxial downward leader can
also be represented analytically. The striking distance is then calculated by employing
the criterion for a critical upward streamer length from (i) earlier. Petrov and Waters
[68] showed that for a negative flash the striking distance is:
r
s
= 0.8[(h +15)i
0
]
2/3
[m, kA] (3.49)
the factor 0.8 (h+15)
2/3
giving a convenient numerical representation (for b = 1 m)
of the analytically calculated field enhancement in the leader-to-mast space for the
attainment of the strike condition. Striking distances calculated from Equation (3.49)
are shown in Figure 3.11b, together with those for the electrogeometric model. These
give comparable values for a 20 m mast but the PetrovWaters model suggests that
an electrogeometric approach significantly underestimates the striking distance for
a 60 m mast. Most importantly, both the electrogeometric and generic modelling
show that the minimum rolling sphere radius of 20 m recommended in international
standards is optimistically larger (for reliable protection) than the striking distance
for low current strikes of 3 kAand below.
3.4.4.5 Effect of altitude on striking distance
Phelps and Griffiths [86] studied the effect of air density and humidity on positive
streamer growth. Fromthis work a relationship between the streamer gradient E
S
and
the relative air density and absolute humidity was obtained by Eriksson et al. [87]:
E
S
= 425
1.5
+(4 +5) [kV/m] (3.50)
where the relative air density with respect to standard sea level values of 760 torr
and 293 K is:
=
293p
760T
(3.51)
0 20 40 60 80 100 120
peak current, kA
a
5
10
15
20
25
0
30
s
t
r
i
k
i
n
g

d
i
s
t
a
n
c
e
,

m
0 50 100 150 200 250 300 350
a
b
c
0
100
200
300
400
500
600
700
s
t
r
i
k
i
n
g

d
i
s
t
a
n
c
e
,

m
a peak current, kA
h = 60m
i
0
= 100 kA
20m 100 kA
60m 31kA
20m 31kA
60 m 5 kA
20 m 5kA
0
100
200
300
400
500
600
s
t
r
i
k
i
n
g

d
i
s
t
a
n
c
e
,

m
0 2 4 6 8 10 12
c altitude, km
Figure 3.11 PetrovWaters model
a Striking distance calculations for a negative flash to plane ground
b Mast at sea level. Curve a: electrogeometric model; curve b:
generic model (60 m mast); curve c: generic model (20 m
mast) [53]
c Mast at altitude [53]
Pressure, temperature and humidity change with increasing altitude z and, for the
range 0 < z < 10 km, Petrov and Waters [68, 88] represented the meteorological
data by the equations
p(z) = p(0) exp
_
z
z
0
_
(3.52)
where p(0) is the sea level pressure and z
0
= 8 km:
T (z) = T (0) kz (3.53)
where T (0) is the sea level temperature and k = 6 K/km:
(z) = (0) exp
_
z
z
H
_
(3.54)
where (0) is the sea level absolute air humidity and z
H
= 3 km. The standard value
for (0) in high voltage testing is 11 g/m
3
.
At an elevation of z km, the reduction of the critical field E
S
implies a significant
increase of striking distance. At the San Salvatore measuring station altitude of 914 m,
for example, the critical positive streamer field is 4.4 kV/cm, which partly contributes
to Bergers observations of flashes of either polarity resulting from upward first
leaders.
Incorporation of altitude effects into Equation (3.49) for a vertical mast gives for
an altitude of z km:
r
s
= 0.8[(h +15)i
0
]
2/3
_
1 +z
2
h +80
_
(3.55)
This greater striking distance (Figure 3.11c) represents a significant increase of risk
factor in high mountainous regions and in aviation. At z = 5 km, the critical streamer
propagation field is predicted to be 2 kV/cm compared with 5 kV/cm at sea level.
3.4.4.6 Air termination geometry
In recent years, there has been considerable interest in early streamer emission air
terminals for lightning protection [89, 90]. These generate a locally triggered streamer
discharge fromthe terminal with a shorter delay than a standard device. However, the
efficiency of such systems is not proved experimentally [43, 9195] or by field obser-
vations. On the contrary, there are physical grounds to believe that early streamer
initiation (also claimed to be encouraged by air terminations incorporating radioac-
tive sources) would make more difficult a successful upward leader, that is to say
one which would propagate over the distance required to intercept the downward
negative leader [96]. The influence of the geometry of the air terminal on the striking
probability has been investigated experimentally by several authors [68, 85, 97100].
These investigations showed a higher striking probability to a grounded rod with a
blunt top which probably results from a delayed streamer onset. The determination of
a possible optimumradius of curvature of the top of a lightning conductor is clearly of
practical interest. Petrov and Waters [101, 102] have calculated the striking distance
Table 3.6 Optimum half width b(opt) and radius of curvature for
semiellipsoid mast: negative flash striking distances for mast
(b = 1 m), sphere and horizontal cylinder
i
0
(kA)
h
(m)
b(opt)
(m)
(m) Striking distance r
s
(m)
b = b(opt) b = 1 m sphere cylinder
r = h r = h
10 20 3.5 0.6 42 40 17 13
60 4.8 0.4 75 66 16 12
31 20 3.5 0.6 100 85 41 29
60 4.8 0.4 174 141 37 26
100 20 3.5 0.6 210 185 99 74
60 4.8 0.4 384 307 87 62
for a semiellipsoid mast with a half width b. In Table 3.6, for various lightning cur-
rents and mast heights, numerical examples are given of optimum half widths b(opt)
and radii of curvature = b(opt)
2
/h at the ellipsoidal mast top that result in the
maximum striking distances. Additional calculations for a rod lightning conductor
with a sphere on the top showed also that the optimum radius of the sphere for which
the striking distance is a maximum is of the order of 0.7 m. This approach has a
similar outcome to the equivalent radius concept [83].
Additional results are included for grounded hemispheres and long semicylinders,
with their bases lying on a plane earth, as a guide to the behaviour of basic building
shapes. These calculations show sensitivity to the structure geometry for negative
lightning. There is little influence on the striking distance r
s
in the case of positive
lightning (section 3.4.5).
3.4.5 Positive lightning
3.4.5.1 Striking mechanisms for positive flashes
The lightning detection techniques of the 1990s indicate that positive polarity flashes
may occur more frequently than had been supposed previously. Although undoubtedly
less common than the negative flash, positive lightning is significant because such
strikes often carry particularly damaging peak currents exceeding 100 kA. Observa-
tions of positive lightning by Berger [20] (section 3.2.8) showed that, at an elevation
of 914 m, the initiation was by a stepped upward negative leader that could be of
great length. However, positive downward leaders initiated from the cloud have been
recorded by Berger and others, in which case the determination of the striking dis-
tance based solely upon a criterion for upward leader inception cannot always be used.
Laboratory experiments with high positive impulse voltages show that the negative
upward leader may not propagate until a very late stage, in which case the striking
distance is determined instead by the streamer zone length of a positive downward
leader at the final jump phase preceding sparkover. The efficiency of lightning rods
against positive polarity downward leaders can thus be expected in some cases to be
substantially lower than against negative polarity leaders. The necessity of further
clarification of the mechanism of the positive lightning flash to determine striking
distances was emphasized also by Golde [103].
The following calculations by Petrov and Waters [101, 102] indicate that a sig-
nificantly lower sensitivity of striking distance r
s
(+) to the structure height can be
expected for positive polarity downward lightning. At the same time, the variation of
striking distance with prospective peak current can be expected to be the same (r
s

i
2/3
0
) as shownfor negative lightninginsections 3.3.2and3.4.4.4, since the charge q on
the channel and the peak return current will be similarly related as in the negative flash.
Two mechanisms of lightning strike for downward flashes are envisaged for this
polarity, depending upon the criterion that succeeds to establish the striking distance.
(i) Positive streamer criterion
For positive lightning, the downward leader may not need to initiate a negative
upward leader before the striking distance has already been achieved by the successful
bridging of the gap to the grounded structure by the positive streamers. In this case,
the striking distance r
s
(+) is determined by the streamer zone length of the down-
ward leader in the transition to what in long spark observations is known as the final
jump phase. To calculate r
s
(+), it is necessary again to calculate the electric field
distribution between the lightning channel and the grounded structure. Streamers of
the downward positive leader will propagate to a distance where the electric field
falls to the minimum streamer gradient E
S
5 kV/cm at sea level. At the approach
of the lightning channel to the mast, the electric field intensity and the range of its
intensification are increased, and this will extend the streamer zone length of the
positive leader. When the minimum value of the electric field between the downward
leader and the grounded structure becomes equal to E
S
, then positive streamers can
successfully propagate to the structure. The instant at which this condition is reached
will define the striking distance r
s
(+). Of course, the ongoing development of the
leader towards the structure will almost inevitably initiate a late upward negative
leader before the ultimate establishment of the return stroke.
(ii) Negative streamer criterion
For tall structures or high lightning currents, the conditions for negative upward
leader initiation may well be fulfilled before the minimum electric field between a
downward positive leader and the grounded structure becomes equal to E
S
(+). Here,
the striking distance r
s
(+) will instead be defined by that upward leader initiation
condition, in the same way as for the negative flash. As already discussed, negative
leaders in very long spark discharges are also known to propagate by steps, which
are themselves initiated by space leaders ahead of the main negative channel. The
length of the steps in long air gaps is practically independent of both the gap length
and the shape of the high voltage impulse, and takes values of between 2.5 and 3.6 m
[59, 62, 66, 104, 105]. The critical range of field intensification in the vicinity of the
grounded structure is then important, and the intensified field must, at an adequate
distance x
0
from the structure, reach a critical value E
S
(). Measured values for this
electric field in the negative streamer zone have been variously reported as 11 kV/cm
by Volkova and Koriavin [67] and as 1016 kV/cm by the Les Renardieres Group
[62]. For a minimum value of 10 kV/cm, and a corresponding value of critical charge
per unit length of q = 160 C/m, the value of x
0
for a vertical mast is 2.9 m. This
represents physically the minimum length of the upward negative streamer zone and
the negative leader step length.
The mast height and prospective lightning current will determine which of these
criteria (i) or (ii) is achieved the earlier. The discrimination boundary calculated for
mechanisms (i) and (ii) is shown in Figure 3.12a.
3.4.5.2 Striking distance for positive lightning as a function of peak current
and mast height
Figure 3.12b shows the striking distance dependence on the lightning peak current
and mast height resulting from these calculations. The solid lines correspond to the
application of the positive streamer criterion (i), and the broken lines are obtained
from the negative streamer criterion (ii).
The striking distances indicated by criterion (i) show a weak dependence on mast
height h. The calculations can be approximated by the numerical relationship:
r
s
(+) = 1.08i
2/3
0
ln
_
h
15
+10
_
[m, kA] (3.56)
As already noted the increase of striking distance with return stroke current is again
of the form r
s
i
2/3
0
since it is proportional to the charge on the leader for either
polarity.
It is seen from Figure 3.12b that for lightning currents less than a median value
of 31 kA, the first criterion (i) is fulfilled earlier (at a larger striking distance) than
criterion (ii). However, for tall structures and high current flashes, large striking
distances will frequently follow the prior achievement of criterion (ii). The striking
distances indicated by criterion (ii) may be approximated by the relationship:
r
s
(+) = 0.103[(h +30)i
0
]
2/3
[m, kA] (3.57)
These calculations for downward flashes from positive cloud charges show that the
striking distance is merely 1020 per cent of that associated with negative lightning.
However, as Berger [20] has shown, for tall masts at altitude the upward leader is
usually the initiation mechanism for flashes of either polarity. This is discussed in
section 3.5.4.4.
3.5 Risk factors and protection
3.5.1 Risk assessment
As in the case of striking distance calculations, the assessment of lightning risk can
be made at three levels of refinement. The probability of a strike is usually expressed
(ii)
strike polarity: positive
strike polarity: positive
(i)
50
100
150
200
250
300
a
b
50 100 150 200
100kA
60kA
31kA
10kA
10
20
30
40
50
60
70
80
p
e
a
k

c
u
r
r
e
n
t
,

k
A
0
350
50 100 150 200 250 300 350
mast height, m
0 400
s
t
r
i
k
i
n
g

d
i
s
t
a
n
c
e
,

m
0
90
0 250
mast height, m
Figure 3.12 Positive flash modelling
a criterion boundary for upward negative leader from a mast.
Region (i): striking distance achieved before upward negative
leader inception; region (ii): striking distance achieved by upward
negative leader inception
b striking distance to a mast [53]. Solid curves: striking distance
from criterion (i); broken curves: striking distance from criterion
(ii)
as the risk factor, which is quantified as the estimated number of strikes per annum
to the structure. This can be calculated:
(i) independently of the prospective lightning current magnitude, and taking into
account only the collection area of the structure; this is the procedure used in
international standards
(ii) noting that the risk factor will increase with increasing prospective lightning
current; in this case, an attraction area a(i
0
) is calculated from r
s
(i
0
) by
geometrical relationships, and the annual number of lightning strokes to the
structure for all peak currents then gives the risk factor:
R = N
g
A
c
= N
g
_
0
a(i
0
)p(i
0
) di
0
(3.58)
where p(i
0
) is the normalised probability density of the lightning current distri-
bution, the integration gives the collection area A
c
, and N
g
is the local density
of lightning strikes to earth
(iii) using a generic model of a criterion for a successful upward leader to find a(i
0
).
Amore refinedassessment of riskgoes further thanthe riskfactor R. This approach
estimates the probability that the strike is sufficiently severe to cause an electrical,
thermal, electromagnetic or mechanical shock that is prejudicial to the structure. An
example of these risk statistics is the calculation of the risk of flashover R
F
of an
insulated electrical system as a result of a lightning overvoltage [106]. The per-unit
probability density p(V) defines the range and magnitudes of such overvoltages. The
shape of the p(V) curve depends both on the statistics of natural lightning and on the
structural and other characteristics of the system, including the effect of protective
measures. The cumulative probability function P(V), on the other hand, is system
dependent and defines the increasing risk of flashover of the system insulation with
increasing voltage V for a voltage shape that is representative of lightning overvolt-
ages (usually a 1.2/50 impulse but sometimes a chopped impulse or a non-standard
shape). The product p(V)P(V) is the probability of an overvoltage V arising which
results in a system flashover. The total risk of flashover per annum in a region of
ground flash density N
g
is:
R
F
= N
g
_
0
p(V)P(V) dV (3.59)
3.5.2 Standard procedure for the calculation of risk factor
Using the simple geometric method, the risk-of-strike assessment is determined in
the Standard BS 6651:1999 [39] by the calculation:
R = (ABCDE)A
c
N
g
10
6
(3.60)
where the important weighting factors A to E (in the range 0.32.0) concern the
structure and its contents and location. The collection area A
c
(m
2
) is calculated
in BS 6651 by adding an attraction radius r
a
= h (structure height) to the plan
dimensions. IEC 61662 [76] adds an attraction radius r
a
= 3h, so increasing the risk
factor by almost an order of magnitude.
As far as collection area is concerned, however, a much greater impact on BS 6651
arises from its advice for the protection of electronic equipment. This reasonably
suggests that a lightning hazard to vulnerable equipment is caused by strikes to
surrounding ground, associated structures and incoming or outgoing mains services
and signal lines. On the basis of a collection distance d(m) equal numerically to the
earth resistivity (m), say typically l00 m (100 m), the values of A
c
would be
increased by factors of 10 to 1000 above those calculated by use of the attraction radius
r
a
. Consequently, high risk factors in the range 0.050.1 for the UK are then found.
The risk of lightning damage often justifies expenditure on the protection of power
supplies, signal lines and telephone cables. The Standard BS 6651 classifies four
types of structure, with a consequential loss rating that ranges from one for domestic
dwellings to four for major industrial infrastructure. The acceptability of a given risk
factor will depend on the individual structure. Avalue of 10
5
is sometimes quoted as
a guide, which corresponds to a very cautious single annual failure in 100 millennia.
3.5.3 Electrogeometric calculation of risk factor
Armstrong and Whitehead [78] used their electrogeometric model to take account of
the structure of an overhead transmission line to design its lightning shielding. They
deduce for the flash rate to an overhead line of effective width w and height h:
R = 0.1N
g
(2h +w) (3.61)
which implies an attraction distance r
a
= h and a 10
3
weighting factor. Whitehead
[107] was able to test the utility of the electrogeometric model by an eight-year
study involving 51 cases of shielding failure and 52 cases of backflashover. Table 3.7
shows results from this study giving specific tripout rates (lightning outages per
100 km years) at thunderstorm day values of T
d
= 40 (hence Whiteheads STR-40
nomenclature).
Whitehead [108] importantly noted that application of the electrogeometric model
should take account of the variability of both:
a striking distance: statistically low values for a given peak current will increase
the risk of shielding failure; one standard deviation allowance for this purpose
indicates that r
s
(statistical) = 0.9r
s
(i
0
) should be used
b terrain: the mean height of the phase conductor can be much higher in exceptional
terrain such as deep valleys.
Induced overvoltages occur on phase conductors from nearby ground strikes.
Whitehead concluded that such overvoltages are generally harmless in respect of
Table 3.7 Line outage statistics [86]
Line voltage Impulse flashover voltage U
50
Shielding angle, STR-40
(kV) (kV)
S
345 1600 31
5.7
22
3.44
15
0.19
500 1800 20
0.23
q
Q z
0
L
H
h y
b
r
s
A
Figure 3.13 Modelling of a laterally displaced leader [53]
outages on UHV lines, but are significant with respect to a BIL of less than 60 kV at
distribution voltages.
3.5.4 Generic modelling of risk factor for a negative flash
3.5.4.1 Strikes to the top of a mast
From the attraction radii calculated in Table 3.5, risk factors are found from generic
models using Equation (3.58). For the determination of an attraction area a(i
0
) that
takes account of both lightning current and structure geometry, it is first necessary to
calculate the striking distance as a function of lateral displacement of the downward
lightning channel (Figure 3.13) [109].
As before for the PetrovWaters model, the criterion for determination of the
striking distance uses analytical field formulae to find the range of field intensifi-
cation by the grounded structure that is equal to a critical upward streamer length.
Figure 3.14 shows, for a mast height of 60 m, the striking distance calculated for
various angles between the vertical axis and the lightning channel. For the range
0 < < 90
, the striking distance dependence on the mast height and lightning

current can be approximated for slender structures, in extension of Equation (3.49),
by the relationship:
r
s
(i
0
, h, ) = [(h +15)i
0
]
2/3
[0.8 cos +0.24 sin ] [m, kA] (3.62)
There is a maximum collection angle
m
, above which lightning will not strike the
mast. The maximum lateral displacement
m
is the attraction radius r
a
and the
collection area of the mast is:
a(i
0
) =
2
m
(3.63)
Since
= r
s
sin (3.64)
the differentiation with respect to of Equations (3.62) and (3.64) gives:
r
a
=
m
= 0.54[(h +15)i
0
]
2/3
(3.65)
100 kA
31 kA
10 kA
50
100
150
200
250
300
s
t
r
i
k
i
n
g

d
i
s
t
a
n
c
e
,

m
0
350
10 20 30 40 50 60 70 80 90
angular displacement, degrees
0 100
Figure 3.14 Strikingdistance dependence onangular displacement (negative flash).
Mast height 60 m [53]
and
m
= 53
(3.66)
The values of attraction radius are shown in Table 3.5 for comparison with other
generic models. Although some differences are seen between these models, they
nevertheless allow significant conclusions to be drawn on the estimation of risk
factor. Present day standards give acceptably conservative estimates of risk factor
for low current flashes, but greatly underestimate strike frequency for high current
flashes. Even the electrogeometric model, which recognises the increased risk for
high current flashes, is shown to underestimate the risk factor for a mast structure by
about four-fold.
For a Franklin rod air terminal, a large striking distance gives efficient protection
and a large sheilding angle. The value
m
determines also the conventionally used
shielding angle of the mast:
s
= arctan

m
h
= arctan
_
0.54[(h +15)i
0
]
2/3
h
_
(3.67)
Figure 3.15 shows the shielding angle dependence on the lightning current and
mast height. Comparison with the angle deduced from an electrogeometric model
(fromTable 3.4) suggests that the electrogeometric approachmayoverestimate shield-
ing efficiency for low current flashes and underestimate it for high current flashes.
By incorporating the downward leader velocity into this model, Petrov and
DAlessandro [100] have calculated the time interval between the upward leader
inception from the mast top and its stable propagation phase. For h = 60 m and
i
0
= 100 kA, this interval is almost 200 s. They conclude that, contrary to the early
peak current, kA
a 60m
a 20m
b 20m
b 60m
10
20
30
40
50
60
70
80
90
s
h
i
e
l
d
i
n
g

a
n
g
l
e
,

d
e
g
r
e
e
s
0
100
0 50 100 150 200 250 300 350
Figure 3.15 Shielding angle variation with peak current for mast height 20 and 60 m
(negative flash)
curves a: electrogeometric model
curves b: generic model [53]
streamer emission concept (section 3.4.4.6), a delayed upward leader inception may
be more effective for lightning protection.
3.5.4.2 Side flashes to a mast
The above calculations are for strikes to the top of a mast. However, a side flash
below the top of tall structures can arise, especially as a result of negative downward
leaders that have a horizontal path in the vicinity of a high tower [110]. Petrov and
Waters [109] have shown that the striking distance for such side flashes is:
r
ss
2.4i
2/3
0
[m, kA] (3.68)
The striking distance for side flashes is significantly less than that for downward
lightning to the mast top because of the low field intensification near the side of the
mast. The value of r
s
= 50 m, corresponding to the rolling sphere concept of the
standards, is suitable only for a peak current i
0
= 95 kAaccording to this model.
3.5.4.3 Inverted cone strike zones
Because the maximum collection angle of 53
(Equation 3.66) does not depend on

the mast height or the lightning current, the collection region of the mast will be an
inverted cone of this angle whose apex is at the mast top (Figure 3.16). As Table 3.5
shows few major numerical differences between generic models, the maximum col-
lection angle concept would appear to have good validity. A recent Fractal model
of leader tortuosity [111] has shown that 95 per cent of strikes will occur within a
half-angle of 60
.
64
53
strike zone
to mast top
Side-flash
strike zone
ground flash
zone
Figure 3.16 Collection cones for apex strikes and side flashes in negative lightning
a attraction zone boundary for vertical downward flashes
b side flash boundary for horizontally oriented flashes
Noting also that the side flash striking distance is larger than that to the plane
earth surface, which is (section 3.4.4.4):
r
s
= 1.16i
2/3
0
[m, kA] (3.69)
we can by means of Equation (3.68) define the side flash zone boundary, for all
values of peak current, as an inverted cone of half angle 64
whose apex is at the mast

base. This together with the collection angle derived in the treatment of strikes to the
mast top gives an easily applied current-dependent concept that is simpler to apply to
both side flash and mast top risk calculations than a fixed radius current-independent
rolling sphere locus (Figure 3.16).
3.5.4.4 Lightning to tall structures
Only generic models can provide an estimation of the risk of lightning initiation
from tall structures. The critical ambient thundercloud field necessary for upward
leader inception decreases with the increase of structure height h because of the
intensification of the local field. For structures with height h > 200 m, the critical
ambient field becomes lower than 20 kV/m. The electric field E
cl
created by a storm
cloud at the earths surface before a lightning flash can approach this critical value
(section 3.2.7). Because of space charge effects, the electric field increases with
height above ground (section 3.2.7.1) and the field intensity before a lightning flash
can be simultaneously E
cl
= 5 kV/m at the earths surface and E
cr
= 65 kV/m at
600 m. The electric field created by the stormcloud may be substantially less than the
critical electric field needed for upward leader inception in the case of structures with
height h < 100 m, but for higher structures an upward leader may form and initiate a
strike without a preceding downward negative leader. To incorporate this effect in the
PetrovWaters model for a downward flash requires the addition of the electric field
E
cl
created by the thunderstorm cloud to the electric field created by the downward
negative leader (if present). Such a consideration allows an estimate to be made of
the proportion of downward and upward flashes occurring to very high structures.
Application of the PetrovWaters model then gives a maximum lateral displace-
ment for a strike to the mast top that is approximated by the relationship:
max
0.54[(h +15)i
0
]
2/3
exp(h) (3.70)
where, for example, for a storm cloud field of E
cl
= 4.2 kV/m the mast height
enhancement coefficient = 0.0021. The coefficient will of course increase with
the intensity of the storm cloud field E
cl
.
Since the risk factor of lightning strikes to a structure calculated from
Equation (3.58) is:
R =
_
0
2
max
(i
0
, h)p(i
0
) di (3.71)
where p(i
0
) is the probability density function for the current amplitude distribution,
then in the case when the lightning current amplitude is, for example, log-normally
distributed this integral may be calculated analytically. Substituting the values for
lateral displacement and calculating the integral we have for the collection area:
A
c
= 153(h +15)
4/3
exp(0.0042h) [m
2
, m] (3.72)
Here, the values i
0
= 31 kA and = 0.7368 for the mean and standard deviation
of the peak current statistical distribution have been used [109]. The risk factor for
a structure with height h can then be expressed as:
R = 153 10
6
N
g
(h +15)
4/3
exp(0.0042h) (3.73)
In Table 3.8, the risk factor is presented for a ground flash density
N
g
= 1 km
2
year
1
. It is seen from the Table that the influence of the storm cloud
field becomes important for structures with heights h > 100 m, indicating that the
contribution of upward, structure-initiated flashes increases with the structure height.
The results describe well the observed data of Petrov and DAlessandro [100, 112],
including the probability of strikes to very high structures. It is known [66], that
the average number of strikes per year to the 540 m television tower in Moscow
Table 3.8 Risk factor for a negative flash to a tall mast
( for N
g
= 1)
h (m) 20 50 100 200 300 400 500
R (downward flashes) 0.02 0.04 0.09 0.2 0.3 0.5 0.6
R (all flashes) 0.02 0.05 0.13 0.45 1.2 2.5 5.1
is equal to R 30, using a value of ground flash density for the Moscow area of
N
g

= 34/km
2
/annum.
3.5.5 Protection of overhead power lines
The vulnerability of power systems to lightning is illustrated by considering that a
direct lightning strike of median current (31 kA) to an 11 kV line (surge impedance
260 ) could produce a peak voltage of 4 MV. Modern lightning location techniques
show that even in Northern Europe there were 540 000 ground strikes in 1992. On
the UK 132 kVsystem, the lightning fault rate is between 0.31.5 faults/l00 km/year,
with damage limited to 510 per cent of cases because of a high BILand fault current
interruption. 11 kV fault rates are lower, but the much larger system and lower BIL
results in many more incidents and a higher consequential damage in 40 per cent of
cases on lines and 100 per cent on cables and equipment.
Power supply plant such as pole-mounted transformers or substation equipment
have to be protected against high voltage surges resulting from shielding failures
and backflashovers. Lightning overvoltages can be predicted using electromagnetic
transient programs for which the main parameters are the peak current, earth resis-
tivity, line geometry, arrester characteristics and cable type. The return stroke current
source is often modelled as a transmission line. Even SF
6
-pressurised GIS are vulner-
able to very fast transients. The last two decades have seen advances in applications
of metal-oxide surge arresters, solid-state overcurrent and distance protection, line
sectionalisers and automatic reclose circuit breaker technology. ZnOgapless arresters
in polymeric housings are effective provided that they are properly specified and are
positioned close to the protected equipment. This will limit the overvoltage at the
equipment terminals to V
a
+2ST , where V
a
is the residual voltage at the arrester, S
is the surge steepness and T the transit time from arrester to the equipment terminals
(see chapter 5, section (i)). The IECStandard [113] gives important recommendations
on the specification of arrester rated voltage, maximumcontinuous operating voltage,
discharge current, energy rating and residual voltage.
The availability of online lightning tracking data offers transmission systemengi-
neers new possibilities for improved plant asset management and security of supply
[21]. Early warning of severe storms is being improved, and the high precision of
ground strike location within 500 m is often useful in fault location after a lightning
event. Archival data of lightning activity also allows realistic estimates of risk factor
and seasonal and geographical variations.
Where a direct strike is prevented successfully by shielding, overvoltages from
backflashover to the phase conductor will be proportional to the peak current and
are determined by the surge impedances of the tower and shield wire and the footing
resistances [114]. The classical lattice diagrammethod remains useful to take account
of factors causing variability in backflashover voltages, such as the steepness of
the lightning current front, the point-on-wave of the supply voltage and the line
shielding by the upward leader. A useful and substantial reduction of steepness and
amplitude results from corona attenuation during the surge propagation along the
line. Until recently, this effect has often been misinterpreted as either a reduction in
surge propagation velocity or a change in the surge impedance of the line. In fact,
the corona discharge affects neither of these, but absorbs energy during the travelling
voltage front because of the ionisation of the air around the conductor. Al-Tai et al.
[115] have shown how the corona attenuation factor can be predicted in terms of the
line conductor geometry.
When transient analysis of the network provides an estimate of the overvoltage
probability density p(V), the level of protection required can be chosen on the basis of
the riskfactor R(section 3.5.3) and the riskof flashover R
F
(section3.5.1). Close-inor
direct strikes tosubstations must be preventedbecause the magnitude andrate of rise of
the unattenuated overvoltage would disable protection by gaps or autoreclosure. Gary
et al. [14] recommend a rolling sphere design of the shielding. Complete shielding
will be achieved, even for the minimumobserved lightning peak current of 2 kA, with
a choice of sphere radius R = 15 m. Alternatively, a design current of 5 kAor greater
would include 97 per cent of flashes, and is less onerous with its value R = 27 m.
For 0.2 flashes/year to the substation, this would be equivalent to one failure in
150 years.
The dependence of flashover voltages on the complex overvoltage shapes arising
on power systems [116], and the consequent critical values of lightning peak current
have been considered by Darveniza et al. [117] who deduced empirical equations for
time to flashover for partly chopped impulses, and Hutzler and Gibert [118] who both
modelled and tested backflashover impulse shapes. In more recent work, Haddad
et al. [119] have quantified the flashover voltage for air gaps in parallel with ZnO
surge arresters. Another factor is that following fault protection operation, an air
gap can suffer significant arc erosion of the electrodes which will alter its volttime
breakdown curve.
3.5.6 Protection of electronic equipment
3.5.6.1 Strategy
Lightning electromagnetic pulses (LEMP) are a particular hazard for electronic
systems. The indirect surges arising from inductive coupling and resistive voltage
drops, and to a lesser degree capacitive and radiative coupling, are sufficient to cause
severe damage; a pulse energy of less than 1 microjoule can easily destroy an inte-
grated circuit. As in power equipment the requirements for the protection are speed
and reliability in overvoltage control, survivability and system compatibility in nor-
mal operation. The strategy of the protection is the same in both cases: where power
engineers speak of main and backup protection, electronic engineers specify primary
and secondary protection. Arecent comprehensive reviewof the standards for the pro-
tection of low voltage systems against lightning and other surges has been published
by Hasse [120].
Three stages are generallynecessarytoachieve immunitytolightningproblems:
(i) transient amplitudes must be controlled by means of the building design and
equipment layout
(ii) the equipment must meet electromagnetic compatibility standards
(iii) surge protector devices should be used to minimise let-through voltages.
3.5.6.2 Mitigation of surges
IEEE C62.41 [121] reviews field experience for transient overvoltages and overcur-
rents recorded on internal power supplies. This and its companion Standard IEEE
C62.45 [122] are presently being updated [123]. Transient magnitudes were found
to be as high as 6 kV and 10 kA. Resistive coupling is an important cause of such
transients when multiple earth points are present. Single-point grounding is ideal, but
because a number of down conductors are usually involved, bonding of all conducting
paths that may share lightning current is recommended.
Annexe Cof BS6651(General advice onprotectionagainst lightningof electronic
equipment within a structure) [39] gives practical advice on transient control. As
specified in BS 6651, the effective earth resistance should if possible not exceed
10 . However, with good bonding between N down-conductors, a resistance to
earth of 10N per conductor should give a satisfactory earth. Removable links are
advisable to enable testing of individual conductors. For structures taller than 20 m,
down conductor spacing around the periphery of the structure should be no greater
than 10 m because of the risk of side flashes. Horizontal bonding conductors should
be present every 20 m of elevation, and on all ridges, eaves and parapets. Large flat
roofs on high risk structures justify a conductor mesh of 5 m10 m. Other standards
differ in detail from these specifications, and are sometimes more stringent. All such
conductors can be structural members of good conductivity, reinforcement rods and
stanchions, or lightning conductor strip or rod. Sharp bends should be avoided if
possible, and low resistance joints and connections are vital.
Whether internal installations such as heating systems should be bonded to the
lightning earth will depend on the risk of sparkover between the installations and
earth. BS 6651 defines the minimum clearance distances between down conductors
and internal metalwork above which bonding is optional rather than mandatory. It is
also important to limit to a safe level the step voltage to which persons who are in the
vicinity during a direct strike are exposed. If necessary, electrical insulation should
be added to the grounding system.
Inductive coupling to electronic systems, arising from the steep current fronts in
down conductors and by induction fromnearby flashes, can be minimised by suitable
routing of cabling and the avoidance of loops. Incoming and outgoing power feeds,
LAN connections and signal and telephone lines can be subject to both resistive and
inductive transients, and common entry points and single point earthing are important.
For some applications optical isolation is worthwhile.
3.5.6.3 Electromagnetic compatibility
Much engineering effort has aimed to identify the types of disturbance that can be
experienced from both near and distant storms and to specify the performance tests
that must be carried out to meet these satisfactorily. In the last decade, EMCStandards
61000 and 61312 [124, 125] have defined these hazards and test techniques. In par-
ticular, multiple screening of structures, interior bonding, grounding requirements,
deployment of surge protection devices (SPD) and comprehensive testing of systems
and devices are justified.
3.5.6.4 Protector devices
Lightning transients can be effectively limited by protection so as to avoid data and
software corruption, partial discharges and irreversible damage to hardware. Such
protection is arranged so that the transient control level (TCL) limits overvoltages to
below the equipment transient design level (ETDL).
The type of protector device will relate to its location in the electronic system.
Locations are defined in the standards as falling into three categories, from a low
exposure category A(e.g. plug-in equipment that is expected to experience only low
transient levels) to category C which is appropriate for major exposure to transients,
such as an incoming mains supply. High vulnerability data, signal or telephone lines
are also included in category Cbecause transient attenuation is weak in such networks.
Mains power protectors are tested during manufacture with 1.2/50 impulses of up
to 20 kV peak and 8/20 impulse currents of up to 10 kA. Data system protectors are
tested up to 5 kV and with 10/700 125 Aimpulses.
In order to achieve the required TCL, the let-through voltage of the protector
must be specified [126]. For mains supplies, the primary series fuse, residual current
and miniature circuit breaker boards have secondary shunt varistor protectors together
with Land Cfilters. Filter design, whether high or lowpass, is not straightforward for
fast transients because of the reduction of inductance by magnetic saturation effects
and the effect of series and shunt capacitance.
For electronic devices, two-tier protectors using gas discharge tubes (GDT)
followed by metal oxide varistors (MOV) provide transient control. For steep
front transients, the let-through voltage of the GDT can be as high as 1 kV before the
tube fires and reduces the follow-through arc voltage to about 20 V.
Fortunately, the solid-state technology which has increased the vulnerability of
electronic equipment has also offered new protection techniques. The metal oxide
varistor with its almost ideal non-ohmic currentvoltage characteristic:
I = AV
n
(3.74)
where n is 25 to 30 has found widespread application after its development in the
1970s. Its high capacitance and limited speed remain a limitation for high band-
widths and bit rates, and ageing and condition monitoring are also problems. Recent
developments in thyristor technology offer crowbar protection with a fast response
as an alternative to the GDT for robust primary protection. These devices can now be
manufactured with a drain current of less than 10 A, a response time below 50 ns
and a peak current of 750 Awith good resealing.
The efficiency of fortress screening, combined with conventional hybrid
GDT/diode and GDT/MOV protection of electronic components even against full
triggered lightning currents of up to 52 kAwas proven in a test programme of triggered
lightning in Japan at a 930 m altitude site [127].
3.5.7 Strikes to aircraft and space vehicles
Lightning strikes to airborne structures are essentially triggered events, and are thus
common, especially since altitude effects are also present. On average, civilian
airliners receive one stroke per annum [128], 90 per cent of which arise from positive
leaders triggered from the aircraft. The standards for the protection of aircraft against
lightning strikes define zone 1 as the initial lightning attachment region. This zone is
usually taken as being restricted to aircraft extremities where electric field enhance-
ment will tend to favour attachments to such sites, which are defined as the extremity
plus a region 0.5 m aft or inboard of it. Severe lightning strikes with currents and
action integrals greater than l00 kA and 0.25 10
6
A
2
s, respectively, should be
avoided outside zone 1 regions. Reported flight experience, however, indicates that
occasionally very severe strikes do occur outside this zone. Aknown hazard to aircraft
arises from the nose to tail sweeping of the lightning attachment points due to the
aircraft motion. This has led to new schemes for determining the initial attachment
zone such as the rolling sphere method and the swept leader method. The definition
of aircraft attachment zones has been improved by the work of the FULMAN pro-
gram [129]. Methods for determining the initial attachment zone are electrical field
analysis or test methods such as arc attachment to scale models.
Aircraft flight safety is sometimes discussed in terms of the probability of a catas-
trophic incident per flying hour. As an example, to illustrate aircraft zoning, it is
essential [130] to assume a vanishingly small probability of a hazardous lightning
strike, for example less than one in 10
9
flying hours, assuming:
one strike every 1000 flying hours
that the most severely damaging lightning strikes are associated with cloud to
ground strikes and there is only one strike to ground involved in every ten aircraft
strikes
only one in ten of ground strikes has severe parameters exceeding the protection
levels appropriate to a swept stroke zone (zone 2).
These assumptions give a lightning strike to aircraft exceeding zone 2 protection
requirements every 10
5
flying hours. Hence, in order to achieve the required prob-
ability of a potentially hazardous strike to a zone 2 region, zone 1 has to contain
99.99 per cent of all cloud to ground strikes.
To model the electric field of a real aircraft, the boundary element method derives
a solution to the Laplace equation over the surface. The method has the advantage
that only the surface of the aircraft has to be meshed. The far-field boundary must
also be included as a second two-dimensional surface, but this outer mesh can be
remote and simple.
Radomes to protect aircraft radar systems from the elements do not prevent
occasional but costly damage from lightning strikes, even after being tested to
industry-agreed standards [130]. Radomes usually form the nose cones of aircraft,
a zone 1 location which makes them susceptible to damaging strikes. To conduct
these lightning currents and prevent damage to the insulating surface of the radome,
while maintaining effective transparency to radar signals, thin metal diverter strips
consisting of a large number of gaps in series may be fitted.
The new lightweight composite materials which are now being used for radomes,
and the introduction of a new generation of airborne weather radar with forward-
looking wind shear detection, require increased radar transparency from the radome.
Since 1993 all aircraft carrying more than 30 passengers are required to have
windshear detection capabilities. Other non-conductive aircraft parts include a new
generation of satellite communication and antenna fairings installed on the exterior
of aircraft.
Strikes that prejudice air worthiness are fortunately rare, but a Boeing 707 was
lost to a lightning strike in the USA in 1963. A survey of USAAF aircraft loss and
damage between 1977 and 1981 revealed a financial cost of M$10. Boulay [52]
has described inflight data obtained from strikes, both triggered and intercepted,
to suitably instrumented aircraft. These strikes confirmed that at an altitude above
3 km these usually involved cloud discharges. This suggests that both low altitude
military sorties and helicopter flights are most at risk from high current cloud-to-
ground lightning. Uhlig et al. [131] express the need for an agreed test waveform for
aeronautical equipment, since the mechanism of the flow of return stroke current in
mid channel is unclear.
Lightning research directly associated with the NASAspace shuttle activity, both
at the launch pad and in space using the optical transient detector (section 3.2.2),
followed serious lightning incidents in Apollo 12 and the total destruction of an
Atlas-Centaur rocket in 1987. Thayer and colleagues [132] analysed the considerable
distortion of the ambient electric field by a vertical space vehicle at launch, and the
consequent risk of triggered lightning. An interesting example of the use of the generic
Rizk model to the design of lightning protection for a satellite launch pad is given by
Joseph and Kumar [133].
3.6 Note
1
Figure fromBritish Standards reproduced with the permission of BSI under licence
number 2003SK/0157. British Standards can be obtained from BSI Customer Ser-
vices, 389 Chiswick High Road, London W4 4AL (Tel +44(0)20 8996 9001).
3.7 References
1 COHEN, I.B.: Benjamin Franklins experiments (Harvard University Press,
1941)
2 FRANKLIN, B.: Letter to the Royal Society, London, December 18th, 1751
3 MOORE, C.B., RISON, W., MATHIS, J., and AULICH, G.: Lightning rod
improvement studies, J. Appl. Meteorol., 2000, 39, pp. 593609
4 AUSTIN, B.: Schonland: scientist and soldier(Institute of Physics Publishing,
London, 2001)
5 PREECE, W.H.: On the space protected by a lightning conductor, Phil. Mag.,
1880, 9, pp. 427430
6 WILSON, C.T.R.: On some determinations of the sign and magnitude of
electric discharges in lightning flashes, Proc. R. Soc. A, 1916, 92, pp. 555574
7 SCHONLAND, B.F.J., and COLLENS, H.: Progressive lightning, Nature,
1933, 132, p. 407
8 SCHONLAND, B.F.J.: Benjamin Franklin: natural philosopher, Proc. R.
Soc. A, 1956, 235, pp. 433444
9 SCHONLAND, B.F.J.: The lightning discharge, Handbuch der Physik, 1956,
22, pp. 576628
10 BOECK, W.L.: Lightningtothe upper atmosphere as seenbythe space shuttle.
International conference on Lightning and stat elec, Florida, 1991, paper 98
11 ALLIBONE, T.E., and SCHONLAND, B.: Development of the spark
discharge, Nature, 1934, 134, pp. 735736
12 ANDERSON, R.B., and ERIKSSON, A.J.: Lightning parameters for
engineering applications, Electra, 1980, 69, pp. 65102
13 ANDERSON, R.B., ERIKSSON, A.J., KRONINGER, H., MEAL, D.V., and
SMITH, M.A.: Lightning and thunderstormparameters, IEEConf. Publ. 236,
1984, pp. 5761
14 GARY, C., LE ROY, G., HUTZLER, B., LALOT, J., and DUBANTON, C.:
Les proprietes dielectriques de lair et les tres hautes tensions (Editions
Eyrolles, Paris, 1984)
15 ERIKSSON, A.J.: Lightning and tall structures, Trans. South Afr. Inst. Electr.
Eng., 1978, 69, pp. 238252
16 ERIKSSON, A.J.: The incidence of lightning strikes to power lines, IEEE
Trans., 1987, PWRD-2, (3), pp. 859870
17 DIENDORFER, G., and SCHULZ, W.: Lightning incidence to elevated
objects on mountains. International conference on Lightning protection,
Birmingham, 1998, pp. 173175
18 FUCHS, F.: Lightning current and LEMP properties of upward discharges
measured at the Peissenberg tower. International conference on Lightning
protection, Birmingham, 1998, pp. 1722
19 LEES, M.I.: Measurement of lightning ground strikes in the UK, in Lightning
protection of buildings, structures and electronic equipment, (ERA, London,
1992), pp. 2.1.12.1.10
20 KRIDER, E.P.: Wideband lightning detection and mapping system, in
Lightning protection of buildings, structures and electronic equipment
(ERA, London, 1992), pp. 2.2.12.2.11
21 CUMMINS, K.L., KRIDER, E.P., and MALONE, M.D.: The US National
Lightning Detection Network and applications of cloud-to-ground lightning
data by electric power utilities, IEEE Trans. Electromagn. Compat., 1998,
40, (4)
22 CHRISTIAN, H.J., DRISCOLL, K.T., GOODMAN, S.J., BLAKESLEE, R.J.,
MACH, D.A., and BULCHLER, D.E.: The optical transient detector. 10th
international conference on Atmos elec, Osaka, 1996, pp. 368371
23 BOCCIPIO, D.J., CUMMINS, K.L., CHRISTIAN, H.J., andGOODMAN, S.J.:
Combined satellite and surface based estimation of the intra-cloud-cloud-to-
ground lightning ratio over the continental United States, Mon. Weather Rev.,
2001, 129, pp. 108122
24 BERGER, K.: in GOLDE, R.H. (Ed.): Lightning (Academic Press, 1977),
chap. 5
25 ISHII, M., SHINDO, T., HONMA, N., and MIYAKE, Y.: Lightning location
systems in Japan. International conference on Lightning protection, Rhodes,
2000, pp. 161165
26 RAKOV, V.A.: A review of positive and bipolar lightning discharges, Bull.
Amer. Met. Soc., June 2003, pp. 767776
27 WAREING, B.: The effects of lightning on overhead lines. IEE seminar on
Lightning protection for overhead line systems, London, 2000, paper 1
28 RAKOV, V.A., UMAN, M.A., WANG, D., RAMBO, K.J., CRAWFORD, D.E.,
and SCHNETZER, G.H.: Lightning properties fromtriggered-lightning exper-
iments at Camp Blanding, Florida (19971999). International conference on
Lightning protection, Rhodes, 2000, pp. 5459
29 ERIKSSON, A.J., and MEAL, D.V.: The incidence of direct lightning strikes
to structures and overhead lines, IEE Conf. Publ. 236, 1984, pp. 6771
30 GARBAGNATI, E., and PIPARO, G.B.L.: Parameter von Blitzstromen,
Electrotech. Z, 1982, a-103, pp. 6165
31 DIENDORFER, G., MAIR, M., SCHULZ, W., and HADRIAN, W.: Light-
ning current measurements in Austria. International conference on Lightning
protection, Rhodes, 2000, pp. 4447
32 JANISCHEWSKYJ, W., HUSSEIN, A.M., SHOSTAK, V., RUSAN, I.,
LI, J.X., and CHANG, J-S.: Statistics of lightning strikes to the Toronto
Canadian National Tower, IEEE Trans., 1997, PD-12, pp. 12101221
33 FIEUX, R.P., GARY, C.H., and HUTZLER, B.P., et al.: Research on
artificially triggered lightning in France, IEEE Trans., 1978, PAS-97,
pp. 725733
34 LAROCHE, P.: Lightning flashes triggered at altitude by the rocket and wire
technique. International conference on Lightning and stat elec, Bath, 1989,
paper 2A.3
35 EATECHNOLOGY PLC: www.eatechnology.co.uk
36 GLOBALATMOSPHERICS INC: www.LightningStorm.com
37 DARVENIZA, M.: Some lightning parameters revisited. International
conference on Lightning protection, Rhodes, 2000, pp. 881886
38 UMAN, M.A., and KRIDER, E.P.: Natural and artificially initiated lightning,
Science, 1989, 246, pp. 457464
39 BRITISH STANDARD 6651: Code of practice for protection of structures
against lightning, 1999
40 CHAUZY, S., MEDALE, J.C., PRIEUR, S., and SOULA, S.: Multilevel mea-
surement of the electric field underneath a thundercloud: 1. A new system
and the associated data processing, J. Geophys. Res., 1991, 96, (D12),
pp. 2231922236
41 SOULA, S.: Transfer of electrical space charge from corona between ground
and thundercloud: measurements and modelling, J. Geophys. Res., 1994, 99,
pp. 1075910765
42 HAMELIN, K.J., HUBERT, P., STARK, W.B., and WATERS, R.T.: Panneau
a pointes multiples pour la protection contre la foudre. CNET research report
81/230, 1981
43 UMAN, M.A., and RAKOV, V.A.: A critical review of nonconventional
approaches to lightning protection, Bull. Amer. Met. Soc., December 2002,
pp. 18091820
44 SCHNETZER, G.H., FISHER, R.J., RAKOV, V.A., and UMAN, M.: The
magnetic field environment of nearby lightning. International conference on
Lightning protection, Birmingham, 1998, pp. 346349
45 SCHWAB, A.J., and CVETIC, J.M.: Radiated field stresses in the environ-
ment of lightning current. International conference on Lightning protection,
Birmingham, 1998, pp. 341345
46 SCHONLAND, B.F.J., and COLLENS, H.: Progressive lightning, Proc. R.
Soc., 1934, 143, pp. 654674
47 SCHONLAND, B.F.J., MALAN, D.J., and COLLENS, H.: Progressive
lightning II, Proc. R. Soc., 1935, 152, pp. 595625
48 MAZUR, V., and RUHNKE, L.H.: Model of electric charges in thunderstorms
and associated lightning, J. Geophys. Res., 1998, 103, pp. 2320023308
49 HUBERT, P.: Triggered lightning in France and New Mexico. Endeavour,
J. Geophys. Res., 1984, 8, pp. 8589
50 ISHII, M., SHIMIZU, K., HOJO, J., andSHINJO, K.: Terminationof multiple-
stroke flashes observed by electromagnetic field. International conference on
Lightning protection, Birmingham, 1998, pp. 1116
51 McEACHRON, K.B.: Lightning to the Empire State Building, J. Franklin
Inst., 1939, 227, pp. 149217
52 BOULAY, J-L.: Triggered and intercepted lightning arcs on aircraft. Interna-
tional conference onLightningandstat elec, Williamsburg, 1995, pp. 22.122.8
53 WATERS, R.T.: Breakdown in non-uniform fields, IEE Proc., 1981, 128,
pp. 319325
54 WATERS, R.T.: Lightning phenomena and protection systems: developments
in the last decade, in Lightning protection of buildings, structures and
electronic equipment (ERA, London, 1992), pp. 1.2.11.2.8
55 ALLIBONE, T.E., and MEEK, J.M.: The development of the spark discharge,
Proc. R. Soc. A, 1938, 166, pp. 97126
56 WATERS, R.T., and JONES, R.E.: The impulse breakdown voltage and time-
lag characteristics of long gaps in air I: the positive discharge, Philos. Trans.
R. Soc. Lond. A, Math. Phys. Sci., 1964, 256, pp. 185212
57 WATERS, R.T., and JONES, R.E.: The impulse breakdown voltage and time-
lag characteristics of long gaps in air II: the negative discharge, Philos. Trans.
R. Soc. Lond. A, Math. Phys. Sci., 1964, 256, pp. 213234
58 PIGINI, A., RIZZI, G., BRAMBILLA, R., and GARBAGNETI, E.: Switch-
ing impulse strength of very large air gaps. International symposium on High
voltage eng., Milan, 1979, paper 52.15
59 MRAZEK, J.: The thermalization of the positive and negative lightning chan-
nel. International conference on Lightning protection, Birmingham, 1998,
pp. 510
60 LES RENARDIERES GROUP: Long air gaps at Les Renardieres: 1973
results, Electra, 1974, 35, pp. 49156
61 LES RENARDIERES GROUP: Positive discharges in long air gaps, Electra,
1977, 53, pp. 31132
62 LESRENARDIERESGROUP: Negative discharges in long air gaps, Electra,
1981, 74, pp. 67216
63 ROSS, J.N.: The diameter of the leader channel using Schlieren photography,
Electra, 1977, 53, pp. 7173
64 KLINE, L.E., and DENES, L.J.: Prediction of the limiting breakdown strength
in air from basic data. International symposium on High voltage eng., 1979,
paper 51.02
65 ORVILLE, R.E.: in GOLDE, R.H. (Ed.): Lightning (Academic Press, 1977),
chap. 8
66 GORIN, B.N., and SHKILEV, A.V.: Electrical discharge development in long
rodplane gaps in the presence of negative impulse voltage, Elektrichestvo,
1976, 6, pp. 3139
67 BAZELYAN, E.M., and RAIZOV, Y.P.: Lightning physics and lightning
protection (Institute of Physics Publishing, London, 2000)
68 PETROV, N.I., and WATERS, R.T.: Determination of the striking distance of
lightning to earthed structures, Proc. R. Soc. Lond. A, Math. Phys. Sci., 1995,
450, pp. 589601
69 DIENDORFER, G., and UMAN, M.A.: An improved return stroke model with
specified channel base current, J. Geophys. Res., 1990, 95, pp. 1362113664
70 COORAY, V.: Amodel for subsequent return strokes, J. Electrost., 1993, 30,
pp. 343354
71 WAGNER, C.F.: The relationship between stroke current and the velocity of
the return stroke, IEEE Trans., 1963, PAS-68, pp. 609617
72 COORAY, V.: The Lightning Flash (The IEE, London, 2003)
73 DELLERA, L., andGARBAGNATI, E.: Lightningstroke simulationbymeans
of the leader progression model, IEEE Trans. Power Deliv., 1990, PWRD-5,
pp. 20092029
74 JONES, C.C.R.: The rolling sphere as a maximum stress predictor for
lightning attachment zones. International conference on Lightning and stat
elec, Bath, 1989, paper 6B.l
75 NFPASTANDARD 780: Installation of lightning protection systems, 1997
76 IEC STANDARD 61662: Assessment of the risk of damage due to lightning,
1995
77 GRZYBOWSKI, S., and GAO, G.: Protection zone of Franklin rod.
International symposium on High voltage engineering, Bangalore, 2001
78 ARMSTRONG, H.R., and WHITEHEAD E.R.: Field and analytical studies
of transmission line shielding, IEEE Trans., 1968, PAS-87, pp. 270281
79 WHITEHEAD, E.R.: in GOLDE, R.H. (Ed.): Lightning (Academic Press,
1977), chap. 22
80 GOLDE, R.H.: Lightning and tall structures, IEE Proc., 1978, 25, (4),
pp. 347351
81 GILMAN, D.W., and WHITEHEAD, E.R.: The mechanism of lightning
flashover on high-voltage and extra-high-voltage transmission lines, Electra,
1973, 27, pp. 6989
82 BERGER, K., ANDERSON, R.B., and KRONINGER, H.: Parameters of
lightning flashes, Electra, 1975, 41, pp. 2327
83 CARRARA, G., and THIONE, L.: Switching surge strength of large air gaps:
a physical approach, IEEE Trans., 1976, PAS-95, pp. 512520
84 DELLERA, L., and GARBAGNATI, E.: Shielding failure evaluation: appli-
cation of the leader progression model, IEE Conf. Publ. 236, 1984,
pp. 3136
85 RIZK, F.A.M.: Modelling of lightning incidence to tall structures, IEEE
Trans. Power Deliv., 1994, PWRD-9, pp. 162178
86 PHELPS, C.T., and GRIFFITHS, R.F.: Dependence of positive corona
streamer propagation on air pressure and water vapor content, J. Appl. Phys.,
1976, 47, pp. 19291934
87 ERIKSSON, A.J., ROUX, B.C., GELDENHUYS, H.J., and MEAL, V.: Study
of air gap breakdown characteristics under ambient conditions of reduced air
density, IEE Proc. A, Phys. Sci. Meas. Instrum. Manage. Educ. Rev., 1986,
133, pp. 485492
88 PETROV, N.I., and WATERS, R.T.: Conductor height and altitude: effect on
striking distance. International conference on Lightning and mountains (SEE),
Chamonix Mont-Blanc, 1994, pp. 5257
89 ALEKSANDROV, G.N., BERGER, G., and GARY, C.: CIGRE paper
no. 23/13-14, 1994
90 VAN BRUNT, R.J., NELSON, T.L., and STRICKLE, H.K.L.: Early streamer
emission lightning protection systems: an overview, IEEE Electr. Insul. Mag.,
2000, 16, (1), pp. 524
91 ALLEN, N., HUANG, C.F., CORNICK, K.J., and GREAVES, D.A.: Sparkover
in the rodplane gap under combined direct and impulse voltages, IEE Proc.
Sci. Meas. Technol., 1998, 145, pp. 207214
92 ALLEN, N.L., CORNICK, K.J., FAIRCLOTH, D.C., and KOUZIS, C.M.:
Tests of the early streamer emission principle for protection against
lightning, IEE Proc. Sci. Meas. Technol., 1998, 145, pp. 200206
93 ALLEN, N.L., and EVANS, J.C.: New investigations of the early streamer
emission principle, IEE Proc. Sci. Meas. Technol., 2000, 147, pp. 243248
94 CHALMERS, I.D., EVANS, J.C., and SIEW, W.H.: Considerations for the
assessment of early streamer emission lightning protection, IEE Proc. Sci.
Meas. Technol., 1999, 146, pp. 5763
95 HARTONO, Z.A., and ROBIAH, I.: A study of non-conventional air ter-
minals. International conference on Lightning protection, Rhodes, 2000,
pp. 357360
96 MACKERRAS, D., DARVENIZA, M., and LIEW, A.C.: Review of claimed
enhanced protection of buildings by early streamer emission air terminals, IEE
Proc. Sci. Meas. Technol., 1997, 144, pp. 110
97 ALEKSANDROV, G.N., and KADZOV, G.D.: On increasing of efficiency of
lightning protection, Elektrichestvo, 1987, (2), pp. 5760
98 RISON, W., MOORE, C.B., MATHIS, J., and AULICH, G.D.: Comparative
tests of sharp and blunt lightning rods. International conference on Lightning
protection, Birmingham, 1998, pp. 436441
99 DALESSANDRO, F., and BERGER, G.: Laboratory studies of corona cur-
rent emissions from blunt, sharp and multipointed air terminals. International
conference on Lightning protection, Birmingham, 1998, pp. 418423
100 PETROV, N.I., and DALESSANDRO, F.: Theoretical analysis of the pro-
cesses involved in lightning attachment to earthed structures, J. Phys. D, Appl.
Phys., 2002, 35, pp. 18
101 PETROV, N.I., and WATERS, R.T.: Striking distance of lightning to earthed
structures: effect of stroke polarity. International symposium on High voltage
eng., London, 1999, pp. 220223
102 PETROV, N.I., and WATERS, R.T.: Striking distance of lightning to earthed
structures: effect of structure geometry. International symposium on High
voltage eng., London, 1999, pp. 393396
103 GOLDE, R.H.: Lightning (Academic Press, 1977), chap. 17
104 LUPEIKO, A.V., and SYSSOEV, V.S.: Dependence of probability of strikes
to aircraft models on the parameters of spark discharges, Proceedings MEI
(in Russian), 1990, (231)
105 GAYVORONSKY, A.S., and OVSYANNIKOV, A.G.: New possibilities of
physical modelling of orientation process and the influence of lightning leader
on the protected object. International conference on Lightning protection,
Florence, 1996, pp. 440443
106 JONES, B., and WATERS, R.T.: Air insulation at large spacings, IEE Proc.,
125, pp. 11521176
107 WHITEHEAD, E.R.: Edison Electric Institute Report Pathfinder Project, 1972
108 WHITEHEAD, E.R.: CIGRE survey of the lightning performance of EHV
transmission lines, Electra, 1974, 33, pp. 6389
109 PETROV, N.I., PETROVA, G.N., and WATERS, R.T.: Determination of
attractive area and collection volume of earthed structures. International
110 YUMOTO, M.: Technology of electrical discharges ranging from nanometer
scale to megameter scale, IEEJ Trans. Fundamentals and Materials, 2004,
124, pp. 1314
111 PETROV, N.I., PETROVA, G.N., and DALESSANRO, F.: Quantification of
the possibility of lightning strikes to structures using a fractal approach, IEEE
Trans. Diel. Elec. Insul., 2003, 10, pp. 641654
112 DALESSANDRO, F., and PETROV, N.I.: Assessment of protection system
positioning and models using observations of lightning strikes to structures,
Proc. R. Soc. Lond. A, 2002, 458, pp. 723742
113 IEC STANDARD 60099-5: Surge arresters part 5 selection and application
recommendations, 1996
114 IEEE WORKING GROUP: A simplified method for estimating lightning
performance of transmission lines, IEEE Trans. Power Appar. Syst., 1985,
104, pp. 919932
115 AL-TAI, M., GERMAN, D.M., and WATERS, R.T.: The simulation of surge
corona on transmission lines, IEEE Trans., 1989, PD-4, pp. 13601368
116 LES RENARDIERES GROUP: Double impulse tests of long air gaps, IEE
Proc. A, Phys. Sci. Meas. Instrum. Manage. Educ., 1986, 133, pp. 395479
117 DARVENIZA, L.D., POPOLANSKY, F., and WHITEHEAD, E.R.: Lightning
protection of UHV lines, Electra, 1975, 41, pp. 3969
118 HUTZLER, B., and GIBERT, J.: Breakdown characteristics of air insula-
tions exposed to short-tailed lightning impulses. IEE Conf. Publ. 236, 1984,
pp. 158162
119 HADDAD, A., GERMAN, D.M., WATERS, R.T., and ABDUL-MALEK, Z.:
Coordination of spark gap protection with zinc oxide surge arresters, IEE
Proc. Gener. Transm. Distrib., 2001, 148, pp. 2128
120 HASSE, P.: Overvoltage protection of low voltage systems (IEE Publishing
London, 2000)
121 IEEE STANDARD C62.41: Recommended practice on surge voltages in
low-voltage AC power circuits, 1991
122 IEEE STANDARD C62.45: Guide on surge testing for equipment connected
to low-voltage AC power circuits, 1992
123 MARTZLOFF, F.: The trilogy update of IEEEStandard C62.41. International
124 IEC STANDARD 61000: Electromagnetic compatibility (EMC), 1995
125 IEC STANDARD 61312-1: Protection against lightning electromagnetic
pulses, 1995 (parts 2&3 in draft)
126 HASBROUCK, R.T.: Performance of transient limiters under laboratory
simulated and rocket-triggered lightning conditions. International conference
on Lightning and stat elec, Bath, 1989, paper 12B1
127 NAKAMURA, K., WADA, A., and HORII, K.: Long gap discharge to an EHV
transmission tower by a rocket triggered lightning experiment. International
conference on Lightning and stat elec, Florida, 1991, paper 62-1
128 LARSSON, A., LALANDE, P., BONDIOU-CLERGERIE, and
DELANNOY, A.: The lightning swept stroke along an aircraft in flight
phenomenology and numerical simulations. International conference on
129 ZAGLAUER, H., WULBRAND, W., and DOUAY, A.: Definition of lightning
strike zones in aircraft and helicopters. International conference on Lightning
stat elec, Toulouse, 1999
130 HARDWICK, C.J.: Review of the 1999 Joint Radome Programme.
International conference on Lightning and stat elec, Toulouse, 1999,
paper 01-2322
131 UHLIG, F., GONDOT, P., LAROCHE, P., LALANDE, P., and
HARDWICK, J.: A basis for a new improved definition of the lightning
external environment in the aeronautical field. International conference on
132 THAYER, J.S., NANAVICZ, J.E., and GIORI, K.L.: Triggering of lightning
by launch vehicles. International conference on Lightning stat elec, Bath, 1989
133 JOSEPH, N.T., and KUMAR, U.: Evaluation of the protective action of LPS
to Indian satellite launch pad. International symposium on High voltage eng.,
Bangalore, 2001
Chapter 4
Partial discharges and their measurement
I.J. Kemp
4.1 Introduction
Partial discharges are localised gaseous breakdowns which can occur within any plant
systemprovided the electric stress conditions are appropriate. Because the breakdown
is only local, failing to result in a following current flow, it is described as partial.
Why are partial discharges (PD) of importance to high voltage engineers? Partial
discharge activity is both a symptomof degradation in the insulating systems of power
plant irrespective of the causative stress and a stress mechanismin itself. Wherever
degradation occurs in an insulating system, be it due to electrical, mechanical, thermal
or chemical/environmental conditions, it is generally accompanied by the generation
of partial discharges. Once present, these then tend to dominate as the stress degra-
dation mechanism. Irrespective of whether the insulating system is gaseous, liquid,
solid or a combination of these, partial discharge activity will cause degradation.
It can therefore be appreciated why understanding the processes by which partial
discharges cause degradation is so important to the development of new insulating
systems capable of withstanding this stress mechanism. In addition, it can also be
appreciated why understanding the correlations among the measurable parameters
of discharge activity and the nature, form and extent of degradation present is so
important to the engineer responsible for the maintenance and asset management of
existing plant systems.
4.2 Partial discharge degradation mechanisms
The electrons, ions, atoms, radicals andexcitedmolecular species producedina partial
discharge move under the influence of the following forces variously:
thermal excitation
the electric field
electrostatic forces
the electric wind, generated by the collision of the ionic species, moving under
the influence of the electric field, with the molecules of the surrounding gas.
The distribution of the reactive species within the gas discharge, and their resulting
impact at the discharge surfaces, will be complex. The following sections discuss
this complete interaction by considering the different stress mechanisms likely to be
prevalent.
4.2.1 Particle impact stress
As has been explained earlier in this book, a gas discharge consists generally of
electrons, positive and negative ions and photons. In relation to partial discharges,
when these particles impact on a surface at the ends of the discharging channel,
they may cause degradation at that surface. Any of these particle types may con-
tain sufficient energy to cause bond scission, often with an associated electron
release.
An impacting ion at an insulating surface may result in local molecular changes
as a result of either an electronic interaction between the incoming charged particle
and the shell electrons of the molecules of the insulating material or through a tight
interaction between the ionising ion and one (or more) ions of the surface lattice.
As detailed by Hepburn [1], the interaction which occurs when an electron collides
with a molecular surface will depend upon the structure and energy state of the
impacted species and upon the electron energy.
An energetic electron impacting upon an uncharged molecular species can interact
in four ways:
(i) an electron, with velocity v
1
, colliding with a molecule, M, can lose part of its
kinetic energy to the molecule without becoming attached:
e
(v
1
) +M e
(v
2
) +M
the electron continues on with a lower velocity, v

2
, and the excited molecule
M
either emits the extra energy as a photon:

M
h +M
or loses the energy by collision with a second molecule:
M
+M
1
M+M
1
(ii) an electron can collide with a molecule and become attached:
e
+M M
forming a negative ion

(iii) an electron can be energetic enough to detach an electron from a molecule:
e
+M e
+e
+M
+
increasing the number of free electrons in the systemand creating a positive ion
Partial discharges and their measurement 141
(iv) an electron can become attached to a molecule and cause a division into charged
and neutral subspecies:
e
+M M
1
+M
2
In a similar manner to the possible reactions between electrons and molecules, there
are a variety of interactions possible between an electron and an ion:
(i) an electron colliding with a negative molecular ion can cause a reaction similar
to that given above [3] but the resultant molecule is neutral:
e
+M
+e
+M
(ii) for a number of molecular species, it is possible for the molecule to become
doubly negatively charged:
e
+M
M
2
(iii) when an electron interacts with a positive ion the electron can become attached
to the molecule and any excess energy may be released as a photon:
e
+M
+
M+h
As with electron impact, the transfer of energy froma photon to a molecule will cause
changes in a number of ways:
(i) the energy transfer can cause photoionisation: where the ionisation energy is
less than that of the photon, the excess can be released as a less energetic photon
or as increased kinetic energy in the electron:
h
1
+M M
+
+e
(v
1
) +h
2
h
1
+M M
+
+e
(v
2
)
(ii) a molecule can be split into ionic subspecies:
h +M M
+
1
+M
2
(iii) a molecule can be split into ionic and neutral species and an electron:
h +M M
1
+M
+
2
+e
(iv) a molecule can be divided into free radical species, which are highly reactive
h +M M
1
+M
2
In relation to photon impact with an ion, the energy transferred from a photon to an
ionic molecular species can cause changes to occur in the following manner:
(i) release of an electron from a negative ion:
h +M
M+e

(ii) release of an electron from a positive ion:
h +M
+
M
2+
+e
(iii) splitting a negative ion into neutral and charged species:

h +M
M
+
1
+M
2
+2e
(iv) splitting a positive ion into neutral and charged species:

h +M
+
M
1
+M
+
2
The photon energy and the energy state of the molecular or ionic species involved in
the reactions given above will determine which of the reactions will occur.
Particle impact has been attributed variously as the mechanismof partial discharge
degradation [2, 3]. Interestingly, in polyethylene, Mayoux [4] has shown that degra-
dation due to ion bombardment occurs at a significant rate only if the charge density
of ions exceeds 1.5 10
2
cm
2
. Further, he has demonstrated that although, in
theory, electrons may produce degradation, electron energies in excess of 500 eV are
required to cause substantial damage. However, once again, the synergetic nature of
discharge degradation mechanisms cannot be overstressed; that one form of stressing
does not in itself result in degradation does not mean that it can be dismissed as a
contributory mechanism to degradation.
In respect of the potential for particle impact damage at a given insulating
surface, it is useful to consult one of the many excellent texts on gas breakdown,
e.g. Reference 5, to determine, for a given gaseous situation, the nature of the par-
ticles involved and for a given set of conditions (partial pressure etc.) the statistical
distribution of energies associated with those particles. Having obtained this infor-
mation, consultation of materials texts will provide details on molecular structure
and bond energies. Comparison of the two sets of information should provide some
guidance of the likelihood of bond scission etc. and where it might occur. An example
of the latter form of data is provided in Table 4.1.
Table 4.1 Chemical bond energies associ-
ated with epoxy resin
Bond type Group structure Energy (kJ/mol)
CH aromatic 435
CH methyl 410
CH methylene 400
CO ether 331
C=O ketone 729
CC aliphaticaliphatic 335
CC aromaticaliphatic 347
4.2.2 Thermal stress
The energy injected into the gaseous environment by the discharge will increase the
temperature of the gas in the local vicinity. In turn, this thermal differential will cause
the gas molecules in the hotter region to migrate to the cooler regions. Although
attempts have been made to ascertain the gas temperature under partial discharging
[6], in general these results must be treated with caution. In addition, the combinations
of particle input energy transfer, chemical bond restructuring and other potentially
exothermic reactions will result in temperature increases at a discharging surface.
The authors personal experience has suggested that the thermal stress created by a
partial discharge may be sufficient to cause damage to polymeric materials but not to
other forms of solid insulating material. Even in the case of polymers, the degradation
sustained due to thermal stressing alone is likely to be insignificant compared with
other stresses present. However, since the degradationis due toa synergetic interaction
of a number of stresses, it must still be considered where partial discharges are present.
The reaction of polymeric materials to a purely thermal stress is dependent on the
structure of the material [1].
For example, simple structures, such as polyethylenes, degrade by random chain
scission:
CH
2
CH
2
CH
2
CH
2
CH
2
CH
2
CH
2
CH
2
CH
2
CH
2
CH
2
+

CH
2
the left-hand side of the reaction continues:

CH
CH
2
CH
2
CH
2
CH
3
CH
+CH
2
=CH
2
CH
2
CH
3
thus producing subgroups of length determined by the fold back length of the chain
transferring the radical element.
Other simple polymers, e.g. polyvinyl chloride, degrade not by chain scission but
by loss of side constituents:
CH
2
CHClCH
2
CHClCH
2
CHCl CH
2
CH=CHCH
2
CHCl
+HCl
thus although the polymer chain remains the same length it loses stability by
developing unsaturated sites in the chain.
Depolymerisation will occur in situations where the polymer molecule contains
no easily abstracted atoms or groups. Polymethylmethacrylate is an example of this
type of reaction:
[CH
2
C(CH
3
)(COOCH
3
)][CH
2
C(CH
3
)(COOCH
3
)]
[CH
2
C(CH
3
)(COOCH
3
)]
CH
2
C(CH
3
)(COOCH
3
)CH
2
C
(CH
3
)(COOCH
3
)
CH
2
C(CH
3
)(COOCH
3
) +CH
2
=C(CH
3
)(COOCH
3
)
here, it can be seen that the polymer chain unzips, i.e. each segment of the polymer
will return to the prepolymer state.
The presence of aromatic rings, i.e. benzene-type structures, in a polymer stiffens
the chain and raises the glass transition temperature, i.e. the temperature at which
the structure changes from a glassy to a plastic state. This can be seen in the work
by Black [7] who has determined that the glass transition temperature of polymer (1)
below, which has aliphatic rings in the chain, is 80
C whereas polymer (2), which

has aromatic rings, has a glass transition temperature of 380
C:
[NH(CH
2
)
6
NH(C=O)(CH
2
)
6
(C=O)]
n
(1)
[NH(CH)
6
NH(C=O)(CH)
6
(C=O)]
n
(2)
The reaction of a polymer to thermal stress in air will also result in thermal oxidation
of the material. Main chain scission or side group removal will create a radical species
which reacts readily with oxygen to form peroxy radical species:
CH
2
CH
2
CH
2
CH
2
CH
2
+O
2
CH
2
CH
2
O
2
The peroxy radical can abstract hydrogen from a polymer group in the vicinity to
form a hydroperoxide and a second radical species:
CH
2
CH
2
O
2
+RH CH
2
CH
2
O
2
H +R
The hydroperoxide species can also decompose due to the application of heat to form
radical species:
ROOH RO +OH
From the preceding discussion, it can be seen that the application of heat to a poly-
meric insulating system will produce a number of reactive sites and species given
the application of sufficient heat.
These are just some of the reactions which can occur due to the thermal effects.
However, as indicated earlier, all effects must be considered to occur in a synergetic
manner.
4.2.3 Mechanical stress
Avibrational mechanical stress will be set up in a solid insulating material subject to
partial discharge stressingunder normal ACoperatingconditions due tothe interaction
of trapped charge in the solid matrix from the discharge interacting with the applied
AC electric stress field. In addition, trapped charged particles of similar polarity will
be repelled from each other and dissimilar charges attracted to each other, again
resulting in a local mechanical stress within the solid matrix. There will also be
a mechanical stress resulting from the impact of larger particles at the discharging
surface depending on the mass number and collision velocities of these particles. It is
unlikely that this will have sufficient energy to cause fracture, as the shock wave is
likely only to have energies of the order of 10
12
J [8]. However, once again the
synergetic effects of such a process cannot be dismissed.
Particle impact from a partial discharge can also result in bond breakage and the
production of ionic and radical species, as indicated earlier. These species may react
with the gas but may also react with the solid to produce, for example, in the case
of polymers, extra crosslinks. These extra crosslinks may produce a stiffer section in
the polymer making it less resistant to shear, tensile and compressive forces induced
in the polymer by the electric stress/trapped charge effects. Arbab, Auckland and
Varlow [9] have been long proponents of mechanical stress damage to materials via
partial discharge/AC electric field stressing, and their various papers on this subject
are extremely illuminating.
4.2.4 Chemical stress
As indicatedearlier, particle impact, thermal stressingandmechanical stressingcanall
result in changes to the chemical structure of a solid insulating systemsubject to partial
discharge stressing. In addition, the species generated in the gaseous environment of
the discharge may also interact chemically with the solid material when they impinge
at its surface. Given the range of potential interactions (on the basis of the range
of gases, liquids and solids plus contaminants) involved, it is impossible within the
confines of this chapter, to detail all possible effects. However, to give the reader
some sense of the issues involved, some examples are presented.
Air is the most common atmospheric medium through which partial discharges
propagate and, as such, it is worthy of consideration from a chemical viewpoint.
Air is a complex mixture of gases, of which the major components are nitrogen and
oxygen with minor concentrations of argon, water vapour and oxides of carbon. The
molecular species, ions, etc., generated by discharges in air [1014] are, therefore,
most likely to be combinations of nitrogen, oxygen, carbon oxides and hydrogen
(from breakdown in atmospheric moisture). The gaseous species produced will be,
for anACstress situation, of both positive and negative polarity e.g. O
and O
2
, N
+
and O
+
2
. In addition, in a discharge atmosphere the polar nature of water molecules
causes them to be attracted to charged species in the discharge.
The ionic species formed in positive and negative DC corona in air are shown in
Table 4.2 to illustrate these differences.
Column A shows the principal species formed during discharges where the high
voltage point electrode is negatively charged. The species listed are all hydrated,
i.e. had attached (H
2
O)
n
groups. The principal species generated when the point is
positively charged are found in columns B and C; the species listed in column B are
not hydrated, those in column C are hydrated.
The difference in the ionic species generated in the discharges is significant in
that the character of the chemical reactions which will occur on the material surface,
due to the impact of the species produced in discharges from positive and negative
points, will be different. In the case of an AC field stress condition, in which both
potential discharge surfaces may act to form discharges, all species may be present.
The production of reactive oxygen species and oxides of nitrogen in the dis-
charge atmosphere is particularly important when considering degradation processes.
The triplet form of oxygen, ozone (O
3
), is a strong oxidising agent and oxides of
Table 4.2 Ionic species formed
in DC coronas
Column A Column B Column C
CO
2
O
+
H
+
CO
3
O
+
2
N
+
O
2
N
+
NO
+
O
3
NO
+
N
+
2
NO
NO
3
NO
+
2
NO
+
NO
nitrogen are known to react in air to form nitric acid, also strongly degrading. This
phenomenon and, indeed, the importance of chemical degradation in partial dis-
charge stressing in general is exemplified by the work of Shields and the present
author [15, 16] in comparing discharge degradation of mica in an air and a nitrogen
environment. Under similar experimental conditions, including discharge repetition
rate and magnitude, it was found that degradation was much more severe over a given
time period in an air environment. Given the physical similarities between the two
gases, it would be expected that the only differences between degradation in the two
atmospheres would be attributable to their chemical differences. In this respect, the
formation of nitric acid (HNO
3
) at the mica surface in air appears to offer the most
likely explanation of the variation in degradation. Although surface reactions are pos-
sible with active nitrogen, no nitric acid will be produced in a nitrogen atmosphere.
The active nitrogen will transfer energy to the mica structure in order to return to the
ground state. In air, however, where oxygen and water are also present the following
reactions are likely to occur:
3O
2(g)
+h 2O
3(g)
and
N
2(g)
+O
2(g)
+h 2NO
(g)
NO
(g)
+
1
2
O
2(g)
NO
2
and finally
2NO
2(g)
+H
2
O
(g,l)
HNO
3(g,l)
+HNO
2
It is therefore suggested that the increased erosion of mica in air can best be explained
by nitric acid, formed in the discharge environment, causing surface erosion by an
acid reaction mechanism on the mica. This mechanism would also account for the
presence of metallic elements fromthe electrode at the degraded surface, found during
the experimental programme, as the electrode too would suffer similar erosion.
As damage was observed in both gaseous environments, however, a second degra-
dation mechanism had to be postulated to account for degradation in nitrogen. Given
the stress conditions prevalent, chemical, bulk thermal and surface/bulk field effects
could be rejected, and, on this basis, the most likely source of degradation was consid-
ered to be energetic particle bombardment, as discussed earlier. Bearing in mind that
mica consists of a lattice of SiO
4
units, this mechanism would involve the energetic
particles within the gas transferring their energy to the mica surface causing either
direct bond scission:
M
+SiOSi SiO +Si +M

where M
is the energetic particle, or bond scission by cumulative localised heating:

NM
T +NM
where N is the number of energetic particles and T is the increase in
temperature, then:
SiOSi + SiO +Si
where is the heat applied.
As air contains a large proportion of nitrogen, and the presence of other molecules
does not preclude this mechanism, it was assumedthat a similar reactionwas occurring
in air, concomitantly with, but secondary to, the acid degradation mechanism.
Another example of the importance of chemical stressing under partial discharge
stressing is exemplified by the work of Hepburn et al. [17] but this time in an organic
polymeric material, i.e. epoxy resin, as opposed to the inorganic, crystalline mica
structure.
Examination of the epoxy resin degraded surfaces following partial discharge
stressing in air indicated the presence of various nitrogen compounds, carbonaceous
anhydrides, acids and peracids and led to the following reactions being proposed for
the resin degradation.
Nitric acid is known to react with organic compounds as follows:
nitric acid breaks down into a nitrous oxide and a hydroxyl radical
HNO
3
OH +NO
2
an organic radical is formed when hydrogen is extracted by the hydroxyl
radical
RH +OH R +H
2
O
the organic radical then reacts with nitrogen oxide to form nitrate
R +NO
2
RNO
2
nitrite rather than nitrate may be formed
R +NO
2
RONO
The reactions described would account for the presence of nitrogen compounds but
do not explain the other reactions taking place.
It is known [10, 11, 18] that reactive carbonaceous compounds are present in air
discharges. Formation of reactive carbon species and possible routes to production
of anhydrides are thought to rely on either:
a activated oxygen attack on the methyl group:
RCH
3
+O
2
RCH
2
+OOH
or
b hydroxyl radical attack on the methyl group:
RCH
3
+OH RCH
2
+H
2
O
both of these initiating reactions produce a methylene radical on the resin chain.
The radical reacting with oxygen produces an aldehyde:
RCH
2
+O
2
RCH
2
O
2
R(C=O)H +OH
the aldehyde reacts with oxygen as follows:
R(C=O)H +O
2
RC=O +OOH
Carbonyl and hydroxyl radicals interact as follows:
RC=O +OH R(C=O)OH
Asecond interaction with a hydroxyl radical produces another radical as shown:
R(C=O)OH +OH R(C=O)O +H
2
O
Interaction of the two radicals highlighted will produce a linear anhydride, as detected
on the epoxy resin surface after electrical stressing:
R(C=O)O +RC=O R(C=O)O(C=O)R
Given that activated oxygen species are less prevalent in a moist atmosphere [11]
and that anhydrides are widespread following stress in a moist atmosphere but less
so in a dry atmosphere, reaction b was considered the more probable initiating step.
The production of nitrated species following normal air discharges is explained by
the higher levels of nitrogen oxides in a moist atmosphere [11].
Interactions between the radical species involved in development of anhydrides
can also be used to postulate reactions for the production of acids and peracids detected
on the stressed resin surface.
The production of radical species, R, on the bisphenol chain (by removal of the
methyl group) was postulated earlier. Interaction with oxygen and carbon species
allows the following reaction:
R +O
2
+RH RO
2
+RH RO
2
H +R
Peroxides can thus be formed on the resin surface.
Where the radical species, RO
2
and R, are formed in close proximity, additional
crosslinks can be formed in the resin matrix by the oxygen molecule:
R +RO
2
ROOR +h
Once again the importance of chemical stressing is exemplified.
A careful trawl through the literature (e.g. Goldman, Mayoux, Bartnikas,
Wertheimer and the present author plus appropriate general chemical texts, variously)
will provide much information and data from which potential reactions can be
hypothesised based on the specific materials and gaseous environment in use.
4.2.5 Electrical stress
The superposition of an electric field due to charge deposition froma partial discharge
at a solid insulating surface will result in both local microscopic and macroscopic
effects which may cause degradation. Electric fields can be responsible for dis-
sociation and transport of ionised and ionisable byproducts resulting in increased
losses and local stress enhancements. Charge trap filling and other forms of charge
capture will result in local field effects. In turn, these may result in local elec-
tronic breakdowns around the stress enhanced site. For further details of this form of
degradation/breakdown, the reader is referred to the excellent book by Dissado and
Fothergill [19].
4.2.6 Synergetic interaction of stresses
As has been emphasised throughout this section, although, for the sake of conve-
nience, the various forms of stress which apply during partial discharge stressing
can be compartmentalised, it is the synergetic interaction of these stresses which
results in degradation. Different materials, gaseous atmospheres, contamination
levels, discharge magnitudes and orientation all will result in a unique combination
of stress effects at a discharging surface.
This synergy is epitomised by the work variously by Mayoux [4, 20, 21] for
polyethylene, as noted earlier. Using an electron gun, an ion gun and an ultra-violet
radiation source, he attempted to separate the different particle effects using infra-
red and ultra-violet spectroscopy and scanning electron microscopy. Unfortunately,
although on a quantitative basis the damage observed using each of the above tech-
niques individually yielded damage similar to that observed under discharge action,
the energies required to produce similar damage, under similar conditions, were well
in excess of those found in a gas discharge. For example, under electron bombard-
ment, polyethylene was found to exhibit negligible degradation below 500 eV and
with ion bombardment, under 100 eV. Mayoux was obliged, somewhat inevitably, to
conclude that the structural transformations found when polyethylene is subjected to
partial discharge stressing cannot be considered as the superposition of the effects
due to the individual components of the discharge acting independently.
The reader interested in finding out more about the mechanisms and phenomena
associated with partial discharge degradation would do worse than read variously the
early papers of Mason and Garton and, more recently, the papers involving Densley,
Bartnikas, Mayoux, Wertheimer, Kemp and, for underpinning mechanisms associated
with the materials (in particular polymers), the book by Dissado and Fothergill [19].
4.3 Partial discharge measurement
The various techniques that can be applied, either directly or indirectly, to deter-
mine the presence of, and characterise, partial discharge activity are described in
this section.
4.3.1 Electrical detection
The electrical detection of partial discharge activity falls within three distinct
approaches:
(i) measurement of each individual discharge pulse
(ii) measurement of the total, integrated loss in the insulating system due to
discharge activity
(iii) measurement of electromagnetic field effects associated with discharge activity
using antennae and capacitor probes.
4.3.1.1 Individual discharge pulse measurement
There are two broad approaches to making this kind of measurement, i.e. connecting
a clampon current transformer (CT) to the neutral strap of the plant itemand taking the
output to an oscilloscope or similar recording instrument or connecting a transducer
(typically a capacitor divider type assembly) to the high voltage terminals of the plant
item and measuring the output in a similar way to the CT approach (see Figure 4.1).
Each has its own advantages and disadvantages.
The CT approach is extremely cheap, simple and safe to use, utilising, as it does,
the neutral strap on the plant item (see Figure 4.1a). No disconnections need to be
made since the CTis simply clamped around the neutral and is supplied with a suitable
output connector compatible for coaxial cable. It suffers, unfortunately, three major
disadvantages:
it cannot be effectively calibrated to determine the magnitude of any discharges
present
it is prone to interference from external sources such as pulses from power
electronics circuitry and, indeed, corona discharging fromelsewhere in the system
it does not provide effective phase information on the location of discharges on
the AC voltage power cycle.
In addition, there may not be a neutral available. However, despite these disadvan-
tages, it is used, particularly for motors and, to a lesser extent, for transformers, as a
first pass technique by some engineers. In the USA, the technique that is used employs
a frequency spectrum analyser rather than an oscilloscope and is described as a radio
to oscilloscope or
frequency spectrum analyser (USA)
R
Y
B
N
high voltage, discharge-free,
capacitor
B
Y
N
R
Z to oscilloscope
low voltage
unit
a
b
Figure 4.1 PD pulse measurement on motors (Red/Yellow/Blue phase notation)
a CT connected to motor neutral
b Capacitor coupler connected to high voltage phase terminal (each
measured in turn)
interference (RI) measurement [22]. In this form, it is used to assess spectra on a com-
parative basis among spectra measured at different time intervals e.g. annually, on
the presumption that changes in the spectrum may be indicative of discharge activity.
In use, care should be taken to make a reference measurement with the CTdiscon-
nected from the neutral prior to making the actual measurement to ensure as far as is
practicable that interference is not compromising the measurement. However, it must
be remembered that the neutral will often act to pick up extraneous signals especially
in a noisy environment, and this nullifies the validity of this approach. Only with the
CT connected to the neutral, with the plant deenergised, can a true comparison of this
type be made.
The second approach is to connect the discharge transducers to the high voltage
terminals of the plant item, e.g. the individual phase terminals of a motor, in turn
(see Figure 4.1b). Typically, this discharge transducer consists of a discharge-free
high voltage capacitor connected to a low voltage impedance circuit (RC or RLC)
which in turn is connected to an oscilloscope or similar instrument. By careful choice
of component values, the high voltage is reduced to a safe level at the low voltage
impedance (typically 1000 : 1 ratio) and individual pulses from discharges can be
displayed superimposed on theACpower cycle voltage. This systemcan be calibrated
by injecting a discharge-simulating pulse, of known magnitude, into the detector
circuit. All commercial instrumentation carry such a calibrator on board. It should be
noted that for transformers with bushings containing tapping points, the bushing can
act as the high voltage capacitor.
The second method utilising the high voltage phase terminals is the Rogowski
coil. Essentially a form of CT, but not to be confused with the earlier type used at
the neutral, its design concentrates the magnetic flux more effectively than in the
standard CT. Its principle of operation is based on Amperes Law. An air-cored coil is
connected around the conductor in a toroidal fashion. The current flowing through the
conductor produces an alternating magnetic field around the conductor resulting in
a voltage being induced in the coil. The rate of change of this voltage is proportional
to the rate of change of current. To complete the transducer, this voltage is integrated
electronically to provide an output which reproduces the current waveform.
It is light, flexible and easy to connect to terminals since the coil can come in a
formwhich can be opened and closed. In general, it is less sensitive than the capacitive
coupler approach, but, broadly, there is little to choose between them.
Individual discharge pulse measurement, either by capacitive coupler or
Rogowski coil, can be applied to most items of plant. Generally, it is applied as an
online technique (although it can be used offline) with the exception of cables and, in
this mode, care must be taken to ensure that any discharges detected are coming from
the item of plant under investigation and not from some other item further away with
the discharges coupling electrically through the conductors to the terminals where
the transducer system is located. In this context, Rogowski coils have the advantage
over capacitor couplers, providing an indication of pulse direction in reaching them.
The discharge pattern produced by individual discharge events is generally recog-
nised as comprising the amplitude of individual discharge events, the number of
discharge pulses per power cycle and the distribution of these pulses within the power
cycle, i.e., their phase relationship. In addition to the discharge pattern produced by
these parameters at a given time and under a given applied electrical stress, one would
normally also be interested in any changes that occur in the pattern as a function of
the magnitude of applied electrical stress and time of application. Armed with this
information, it may be possible to evaluate the nature of the degradation sites and
thereby provide an assessment of insulation integrity. However, although the pre-
ceding paragraphs might imply that this process can be relatively straightforward, in
practice a variety of factors conspire to make interpretations based on such patterns
a highly complex affair.
Of these factors, the major culprits are external interference and the complexity
of the discharging insulating system. External interference, in general, can take the
following forms [23]:
PD and corona from the power system which can be coupled directly to the
apparatus under test (in an online test) or radiatively coupled (in online or offline
tests)
arcing between adjacent metallic components in an electric field where some of
the components are poorly bonded to ground or high voltage
arcing from poor metallic contacts which are carrying high currents
arcing from slip ring and shaft grounding brushes in rotating machinery
arc welding
power line carrier communication systems
thyristor switching
radio transmissions.
It can generally be minimised through the use of inline filters and some form of
discrimination circuit if the problem is in the high voltage line. (An example of this
is the PDA system developed for turbine generators [23], utilising two couplers in
each phase.) In addition, as indicated earlier, Rogowski coils should discriminate,
on the basis of polarity, the direction of a pulse reaching them. If the problem is
airborne interference, e.g. from rectifiers, it can be more difficult to eliminate but an
assessment of noise activity and its characterisation can be made prior to the high
voltage insulation test. In relation to the complexity of the discharging insulating
system, this is generally beyond ones control. The problem, in this case, lies in
the potentially vast number of discharging sites and their variety. This combination
tends to swamp out the characteristic patterns associated with specific discharge
site conditions. Having indicated these caveats, however, there is no doubt that the
information contained in a discharge pattern can be extremely useful in assessing
insulation integrity.
Another aspect of pattern interpretation that should be noted at this stage is the
importance of regular measurements on a given insulating system. Ideally, patterns
should be obtained at regular intervals throughout the life of the insulating system
since there is no doubt that the trend in the discharge pattern of a given insulating
system with time provides far more useful information on insulation integrity than
any measurement at only one point in time.
Although it is always possible to cite examples where a single measurement
can be extremely beneficial, there are many situations where it provides relatively
little information. As a diagnostician (be it of plant or human beings), once one
has established that the absolute levels of the vital characteristic parameters do not
indicate the imminent death of the patient, one is primarily interested in the rate at
which these characteristic parameters are changing through life in comparison with
similar systems (or humans) of comparable design and stress history.
That said, in relation to the interpretation of the discharge pattern, the starting
point would be the magnitudes of the discharges being detected and their repetition
rate. Broadly, the larger the discharge magnitude, the larger the site of degradation
with which it is associated and the greater the likely rate of degradation. Similarly,
the greater the repetition rate or number of discharges per cycle or unit time, the
greater the number of discharging sites. Once these issues have been considered, the
next aspect for interpretation would be the location of the discharges on the power
cycle waveform. Cavity-type discharge sites generally yield a discharge pattern within
which most pulses are in advance of the voltage peaks, i.e., 0
90
and 180
270

of the AC power cycle. In contrast, discharge sites containing a sharp metal surface
generally produce a pattern with pulses symmetrically spaced on both sides of the
voltage peak(s).
The next consideration would be the relative magnitudes of discharges on the
positive and negative half cycles of applied voltage. Broadly, similar magnitudes on
both half cycles imply that both ends of the discharge are in contact with similar
physical surfaces. This contrasts with the situation which is prevalent if the pattern
indicates (say) different magnitudes of pulses on the two half cycles, i.e. quadrants one
and three. In this instance, it would be likely to assume that the discharge was active
between two insulating surfaces but perhaps of different surface topology resulting in
varying stress enhancements, thus changing the supply of initiatory electrons when
the necessary cross-gap stress is reached to produce a spark and the time which an
insulator takes to dissipate surface charge (potential) build-up.
Having decided whether the discharge site(s) are insulating-bound, insulating-
metallic or metallic (corona), one would next consider the variation in discharge
magnitude with test voltage and then with the time of application of voltage should
an offline measurement be a possibility. If the discharge magnitudes remain constant
with increasing test voltage, this tends to imply that one is observing discharge activity
within fixed cavity dimensions. This suggests either a most unusual condition where
all the cavities are similarly dimensioned or, far more probably, one is observing a
single cavity situation possibly of quite large surface dimensions. This contrasts
with the condition in which the discharge magnitudes of the pattern rise with test
voltage. In this instance, one is generally observing cavities, or gaps between insu-
lating surfaces, of differing size. As the test voltage is increased, in addition to those
cavities of low inception voltage discharging more often per cycle, cavities of higher
inception voltage are also starting to discharge. In this case, one would be thinking of
internal discharges in a number of insulation-bound cavities of different size, exter-
nal discharge across a varying length gas gap (between, for example, two touching
insulated radial conductors) or perhaps even surface discharges in an area of high
tangential stress.
Having decided that one is dealing with (say) a cavity-type degradation condition
and that the cavity surfaces are either insulation-bound or metallic-insulation bound,
and having determined something of the cavity size/number distribution, observing
the pattern at a fixed voltage over a period of time may yield still more information.
For example, if the activity tends to lessen with time, then, depending on the insu-
lation system under investigation, one might be observing a pressure increase in the
cavities (caused by gaseous byproduct formation and resulting in a higher breakdown
voltage for given cavity dimensions see Paschens Law), a build-up of surface
charge within the cavity structure (thus inhibiting the realisation of sufficient poten-
tial drop across the cavity to produce further discharges following discharge activity
and potential equalisation) or perhaps a build-up of water or acids within the cavity
structure (increasing surface conductivity and allowing charge to leak away).
The preceding discussion has largely been related to cavity-type degradation
sites identified by the location of discharges primarily in advance of the voltage
peaks however, the same approach holds true for gaps with a sharp metallic
boundary. For example, depending on the numbers of discharges, their spacing and
magnitude on one half cycle compared with the other, one should be able to say
something about the relative sharpness of any points. This relates to the much higher
electric stress realised around a sharp point for a given applied voltage, rather than
that at a plane surface, and the related spark activity when the point and then the plane
are alternately the cathode within a cycle of applied voltage.
There are many more conditions that could be discussed within this section relat-
ing discharge magnitude, phase distribution, test voltage and time of application to
specific conditions. However, it is not the purpose of this chapter to provide a com-
prehensive guide of this type. Other papers within the literature are available for this
purpose [24]. Rather, it is to indicate to those unfamiliar with the interpretation of
partial discharge patterns an effective and efficient way forward in their use. Although
it may be sufficient to either memorise or have available a look-up table of specific
discharge pattern characteristics from which one can make an interpretation, one will
have a much greater possibility of making an accurate interpretation if one under-
stands the physical and chemical processes that can produce a given discharge pattern
(or result in a change in discharge pattern) and can relate these to the specific form
and design of the insulation system under investigation.
Again, the importance of establishing a trend in the discharge pattern for a given
insulating system over its lifetime cannot be stressed too strongly. In general, the par-
ticular degradation characteristics are often developed relatively early in the life of an
insulating system. Thereafter, the parameters of primary importance are the absolute
discharge magnitude and, more importantly, its rate of rise over given times both
relative to the values obtained from other insulating systems of similar design and
history.
Modern instruments acquire, store and process this information digitally and pro-
vide, in general, a three-dimensional plot of this data on screen for analysis purposes.
The three axes are phase (0
360
), discharge magnitude (usually in pC) and number

of discharges (or frequency) resulting in the so-called qn plot.
The x axis and y axis, which represent the phase angle of the PD pulse occur-
rence and the magnitude of the PDpulse, respectively, formthe floor of the histogram,
see Figure 4.2. The phase notation is conventionally that of a positive sine wave
as opposed to positive or negative cosine the 0
is the zero crossing point of

the rising edge (of a signal with no DC bias). The z axis, n, represents the num-
ber of pulses per unit time that have occurred within the specific phase-magnitude
window.
Figure 4.3 demonstrates how a qn pattern is constructed. The x axis of the
PDtime domain signal is divided into a number of time windows, each corresponding
to a number of degrees, or a phase window, of the power cycle. Likewise, the ampli-
tude or y axis is divided into discrete windows. The phase and amplitude windows
correspond to the row and column indices of a two-dimensional matrix. The value at
each index in the matrix corresponds to the number of PD occurring in that phase and
amplitude window, per unit time. In Figure 4.3, it can be seen how the PD a, b, c, d
and e are fitted into their particular phase amplitude window and the H
n
(, q) graph
modified accordingly.
0
0
397
794
1191
1588
0
a
p
p
a
r
e
n
t

c
h
a
r
g
e
,

p
c
n
u
m
b
e
r

o
f

d
i
s
c
h
a
r
g
e
s
p
h
a
s
e

w
i
n
d
o
w
s
,

90
180
270
360
10
20
30
Figure 4.2 qn plot of discharge activity
The distribution is then the accumulation of every PD, in its respective
phase/charge window. If the PD pulses occur statistically, then the distribution is
a probability distribution. The value at each index can be correlated to the probability
of a PD occurring within that particular phase/apparent charge window. The more
cycles making up the distribution the more (statistically) accurate the distribution
will be, with the one caveat that the PD pattern must be relatively stable over the
sampling period.
Clearly, the ability of an instrument of this type to accurately portray the statistical
pattern of PD activity depends on its sampling rate and memory size.
It should be noted, however, that this type of pattern constitutes only the primary
data of PD activity, i.e. magnitude, phase, and number. Changes in the shape of
the pattern with time are now also being recognised as potential indicators of the
nature, form and extent of PD activity. Changes in the statistical moments of the
distribution mean, variance, skewness and kurtosis, when considered across both
the x and the y axes are beginning to show correlations with specific forms of
degradation associated with PD activity.
It should also be noted that this is not the only form of PD data under consider-
ation. The power of digital systems to acquire, store and process the primary data is
being utilised to generate new, and potentially more useful, information. Parameters
generated in this way include [25]:
the discharge energy (p) where p = q V for any discharge, V being the
instantaneous voltage at which the discharge (q) occurs
the discharge phase inception voltage, V
i
, where the discharge pulse sequence
starts
the discharge phase extinction voltage, V
e
, where the discharge pulse
sequence ends

phase angle, [
180 0 360
0
PD apparent
charge, q
phase angle, [
180
360
0
0
a
c d
e
b
P
D

a
p
p
a
r
e
n
t

c
h
a
r
g
e
,

q
n
u
m
b
e
r

o
f

P
D

p
u
l
s
e
s
,

n
Figure 4.3 Build-up of qn distribution
the discharge current, I = l/T |q
i
|, where T is the duration of the power
frequency half cycle and i is the number of discharges observed during T
the discharge power, P = l/T |q
i
V
i
|
the discharge intensity N, the total number of discharges as observed during
time T
the quadratic rate D = l/T |q
2
i
|.
All these quantities can be analysed either as a function of time or phase angle.
However, there has, as yet, been insufficient empirical data gathered to enable cor-
relation to be made unambiguously between any of these parameters and the nature
and extent of PD activity.
Those readers interested in learning more about this form of partial discharge
measurement and its interpretation might read the following papers and their citations
[2630].
It should be noted at this stage that the standard (IEC60270) is not immune to
inherent errors. For example, Zaengl [3133] has analysed the effects of detection
circuit integration error and sensitivity for various integration filter bandwidths and
rise times of pulses. Zaengl also introduced the concept of a parasitic inductance,
distinct from the integration circuit, capable of producing additional variations in
the detector response circuit. The standard makes no reference to such a parasitic
inductance. This phenomenon has been carried forward through simulation studies
[34], which show that the calibration of any measurement PD system depends on
the parasitic inductance, that it influences the measurement sensitivity according to
rise time and filter bandwidth and that, if not considered, it may therefore result in
erroneous measurements.
4.3.1.2 Pulse sequence analysis
In addition to providing statistical patterns of activity from acquired PD data, e.g.
qn patterns, there has been increasing interest in recent years in the nature of
the sequence of pulse events. This approach explores the deterministic nature of PD
events. Generally, a given PD event terminates when the local voltage drops below
some critical value. Although this can be purely a function of the external macroscopic
electric field, it is far more likely to be due to a combination of this macroscopic field
and the local field in the vicinity of the PDevent due to space or surface charge effects.
Factors such as the distribution of (say) surface charge and its rate of decay will, in
such circumstances, have a strong influence over the occurrence of a further PDevent.
On this basis, it is not difficult to see why a given PD sequence may be deterministic
and not random in nature. Equally, over relatively long time periods, changes in the
nature of PD activity may be evident if this approach is taken. Such changes would
not be apparent on the basis of statistical patterns of activity. For example, bursts of
PD activity followed by relatively quiet periods were noted in treeing experiments
on resin [35]. The underlying mechanisms responsible for degradation under PD
stressing may be elucidated more readily using the pulse sequence approach, it can
be argued, since, from the deterministic nature of the results produced, it provides a
much greater insight into the underlying mechanisms than does the statistical, random,
approach.
Rainer Patsch of the University of Seigen, often in collaboration with Martin
Hoof when he too worked there, has been a key advocate of this approach since the
early 1990s. Their work [3639] has demonstrated that the sequence of discharges
and, in particular, the change of the external voltage between consecutive discharges,
provides important information from which the nature and form of the degradation
can be inferred. In addition to aiding in the classification of PDdefects, they have also
demonstrated that the technique can be utilised to separate several sites discharging
contemporaneously.
The increasing speed and memory capacity of digital signal acquisition, storage
and processing systems is making the pulse-sequence-analysis approach more readily
available than previously.
4.3.1.3 Calibration a word of caution
It might strike anyone investigating, for the first time, the range of instruments avail-
able commercially for the electrical detection of individual partial discharge events,
as somewhat incongruous that there are almost as many different bandwidths, centre
frequencies etc. as there are detectors. Being essentially an impulse characteristic,
a partial discharge pulse will contain frequencies across the spectrum from almost
DC through to GHz. On this basis, any bandwidth of detector is likely to detect some
energies containedwithinthe discharge pulse. However, as canbe appreciated, a broad
bandwidth detector will detect far more frequencies than a narrowband detector, i.e. it
will be more sensitive. How are these differences resolved to ensure that, irrespective
of detector characteristics, an accurate measurement of PD activity is achieved? This
is the role of the calibration process.
Essentially, to calibrate an individual event partial discharge detector, a discharge-
simulating pulse of known magnitude (the calibration pulse), is injected into the
detector at the point where the real discharge pulse enters the detector. By utilising
this strategy, irrespective of the detectors individual characteristics, however those
characteristics modify the PD pulse they will also modify the calibration pulse in
exactly the same way and comparability will be maintained. For example, if a
given detector modifies a PD pulse to be of only half the magnitude it would have
been if detected by a very broadband detector, the detector circuitry will modify the
calibration pulse to also be of only half the magnitude, and the correct magnitude of
discharge will be inferred.
Unfortunately, as has been demonstrated, this calibration strategy does not
necessarily work for generators [40], motors [41] or transformers [42]. In the case
of these items of plant, by injecting a discharge-simulating pulse of known mag-
nitude at various points through their windings, and detecting the response at the
phase terminals with detectors of different bandwidth, it has been demonstrated that
the detected magnitude is a function of both the location of the injected pulse and
of the bandwidth of the detector utilised. To make matters worse, it has been shown
that no direct relationship exists between the magnitude that one detector will suggest
and that of others. Indeed, there is not even an indirect relationship with (say) one
detector consistently suggesting a larger discharge magnitude than another for the
same injected discharge-simulating pulse. It is all a function of location.
What is going wrong and, specifically, why is the calibration strategy failing to
ensure that detectors with different characteristics produce the same result? Quite
simply, there is a fallacy within the calibration strategy. The strategy is predicated
on the notion that the injected discharge-simulating pulse emulates the real PD pulse
as it enters the detector. This is indeed true when, in general terms, the discharge
site is close to the detector terminals. However, in the case of generators, motors and
transformers, this may not be true. In these circumstances, the partial discharge pulse
may have to propagate tens of metres to reach the detector. This propagation path may
be complex both electrically and literally! It is composed of inductive, capacitive
and resistive components distributed, in all probability, in a non-linear fashion. Being
complex, the impedance paths will be dependent on frequency and will vary as a
function of the individual frequencies which make up the discharge pulse. As the PD
pulse propagates, frequencies will be lost (especially high frequencies), there will
be resonance effects etc. In short, having propagated any significant distance either
electrically or physically to the detector, the PDpulse will, in all likelihood, no longer
resemble the pulse that set out from the discharge site. Therein lies the fallacy in the
calibration process. The calibration pulse does indeed emulate the PD pulse as it
occurred at its site of origin but after it has propagated any significant distance this
equality may be lost and the detected discharge magnitude will become a function of
location of discharge site (controlling the frequencies of the PD pulse which reach
the detector) and, of course, the detector bandwidth [41, 43].
Figure 4.4 illustrates the issue. The calibration pulse contains essentially similar
frequencies across the measured spectrum. Note that the discharge pulse, at its site of
origin, would also appear similarly. However, as can be seen, following propagation,
the frequency characteristics of the PD pulse may look quite different, as illustrated.
Although it is simplifying the argument greatly, by inspection of Figure 4.4, it can be
appreciated that in this case a narrowband detector detecting around (say) 800 kHz
would, relative to the detected calibration pulse, suggest an extremely small PDpulse
whereas a detector utilising a narrowband detection system around 260 kHz would
suggest that the PD pulse is large!
For those unfamiliar with PD measurements of this type, it might appear that
the obvious solution is to use only wideband detectors. Unfortunately, although this
would decrease errors introduced due to this phenomenon and result in an increased
detector sensitivity, it would prove impractical in field measurements. The greater the
bandwidth of detector, the greater the likelihood of detecting pulses which do not orig-
inate fromdischarges. Given the broad range of frequencies available fromPDpulses,
a balance must be sought between detector sensitivity and erroneous information.
m
a
g
n
i
t
u
d
ePD pulse
calibration pulse
500 750 250 1000
frequency, kHz
Figure 4.4 Calibration pulse characteristic compared with PDpulse characteristic
following propagation
The field of partial discharge measurements has had a number of salutary experi-
ences where noise has been interpreted as PD activity and plant has been removed
from service unnecessarily with serious economic consequences.
Now, if the situation as described in the preceding paragraphs were very prevalent,
no discharge measurements made on generators, motors or transformers could be
trusted and, fortunately, that is not the case! The reasons for this lie, in part, in
the fact that much of the degradation suffered in such plant (and hence associated
PD activity) occurs near the high voltage terminals. The PD pulses thus have only a
relatively short distance to propagate to the detector, resulting in only minor errors.
Also, most diagnosticians would rely on trend analysis at least as much as any absolute
measure of PD magnitude and, with a range of caveats, this is probably unaffected
provided the same detector is used consistently.
This is merely a word of caution for those instances when, in comparing two items
of plant, possibly of different design and manufacture, one is tempted to suggest that
one itemis in poorer condition than the other because it is displaying higher discharge
levels. The odds are in favour of this being true but it may not be!
4.3.1.4 Noise and wavelet analysis
As indicated earlier, the electrical detection of PD can be affected sometimes
seriously by various forms of noise [23], with measurements in the field rendered
ineffectual or severely compromised.
Over the years, methods employing discrimination circuits [23], traditional filter-
ing techniques [44], neural networks [45] etc. have been designed to suppress noise
with limited success. However, a new, more powerful tool is now being applied to
this problem wavelet analysis. It is likely that this technique will bring radical
improvement to noise reduction and suppression.
Since its introduction in practical applications in the mid-1980s, as a powerful
tool for signal analysis and processing, wavelet analysis has been increasingly applied
to solve engineering problems [4648]. Essentially, the technique allows the user to
obtain two-dimensional information on PD pulses in both the time and frequency
domain, and to extract features of PD pulses in measurement data. By careful choice
of wavelet transform, ideally coupled to a clear understanding of the PD detector
characteristics, sophisticated feature extraction is possible for PD pulses [49, 50].
The required properties for a wavelet in this application include compactness, limited
duration, orthogonality and asymmetry for analysis of fast transient irregular pulses,
e.g. the Daubechies wavelet family. This, in turn, makes it possible to extract PD
pulses from extraneous noise in a way which would be quite impossible with tradi-
tional filtering techniques [51, 52]. Indeed, it is possible to extract PD pulses from
measurement data where the PD pulse is embedded within the noise, i.e. below the
noise plane. This is illustrated in Figure 4.5, which shows raw measurement data and
processed data following the application of the wavelet transform. As can be seen,
there are three PD pulses within the measurement timescale, with one of these below
the noise plane. Time taken for analysis is short, making the technique attractive for
field analysis.
4
time, s
a
2
0
60 40 20 20 40 60 80 100 120 0
60 40 20 0
time, s b
20 40 60 80 100 120
2
4
2
0
2
4
6
m
a
g
n
i
t
u
d
e
,

m
V
m
a
g
n
i
t
u
d
e
,

m
V
Figure 4.5 PD measurement in presence of noise
a Raw data, PD activity and noise superimposed
b Processed data to show PD activity
Another benefit of the technique is its ability to compress PD measurement data
[53]. Because only limited coefficient data related to PD events need to be used,
following wavelet analysis, to reconstruct precisely the actual PD signal extracted,
the amount of data storage space can be greatly reduced, i.e. five per cent of the
original data stream. The technique may, when fully developed to this application,
revolutionise PD measurements in the field.
Electrical PD detection as discussed in the preceding section has clear advan-
tages in terms of sophistication over most other approaches, but is generally viewed
as difficult in measurement and, in particular, in interpretation. It is also relatively
expensive and is normally only applied as a front line approach for motors and gen-
erators where most other techniques, with the exception of tan (see section 4.3.1.5),
cannot be applied. For motors/generators, many users have chosen to install perma-
nent high voltage capacitors and low voltage impedance units at the phase terminals
of their plant, thus removing the need to deenergise the plant to make the high voltage
connections. All that is required is to connect the measuring instrument to the low
voltage connections. Capacitors for this application are relatively inexpensive.
4.3.1.5 Loss measurements associated with discharge activity
This technique operates on the principle that the current loss in an insulating system
will increase markedly in the presence of discharges. By monitoring the tan of the
system ( being the angular quantity indicative of the relative values of resistive to
capacitive current), relative discharge activity can be inferred.
The insulation system should, ideally, behave as a perfect capacitor, i.e., there
should only be capacitive current through the system. However, inevitable losses
in the system mean that there is also a small component of resistive current. As
losses increase, the resistive current becomes larger and the angle created between
the capacitive current and the resultant vector-summated current () increases, as does
the tan . In the presence of partial discharge activity, the losses increase enormously
and dominate as the loss mechanism.
Traditionally, tan measurements would be made offline, allowing the insulation
system to be energised progressively to full rated voltage (generally in 0.2 voltage
steps to full working voltage). This has enabled a plot to be made of tan versus
applied voltage similar to the ones shown in Figure 4.6.
motor 2
slow rate of change
good bulk properties
high start value
contaminated winding
motor 1
motor 4
motor 3
low start value
clean winding
high rate of change
poor bulk
t
a
n
,
o
0.2 0.4 0.6 0.8
1 (V line) V, applied
Figure 4.6 Tan versus applied voltage
Any sudden change in tan (at the so-called knee point or tip-up point on the
tan plot) would be considered indicative of PDinception and the rate of change with
increasing voltage would be indicative of the relative severity of the PD. In addition,
with experience, the starting value of tan informs on the level of contamination
present on the system usually on stator end windings since it is in motors and
generators that the technique has found greatest favour although it has also been
applied to transformers and bushings [54]. On this basis, by considering the plots
shown in Figure 4.6, motor 1 plot would be indicative of a clean winding (low start
tan , and minimal rate of change of tan with voltage), motor 2 a clean winding
but with serious discharge problems (low start tan , low inception voltage (knee
point) and high rate of change of tan with voltage), motor 3 a contaminated winding
(high start tan ) but one which has low PD activity (low rate of change of tan with
voltage) and motor 4 a contaminated winding which also suffers from serious PD
activity (even more serious than motor 1).
As can be seen, tan informs on PDactivity and thereby on the appropriate action
to be taken. Traditionally, the measurement employed a Schering Bridge type circuit
being connected at the high voltage terminals of the plant but, latterly, instruments
have made direct measurements of the different current components.
The technique has the advantages of being simple in measurement and clear
in interpretation. Its major disadvantage lies in it being an integrated measure of
degradation in the insulating system. In motors and generators, quite large levels
of discharge activity can be sustained both in terms of magnitude and number/cycle
without fear for the insulation integrity. Afew specific sites of degradation which are
much worse than anywhere else will provide much larger magnitudes of discharge
but these may be lost in the integration process with all other activity. That said, most
instruments incorporate a peak magnitude detector to alleviate this problem.
In recent years, there has been an increased pressure to only apply online tech-
niques for economic reasons. In relation to tan , this means that only a single value
can be obtained. As can be seen from Figure 4.7, this would make it impossible, by
this measurement alone, to distinguish between a contaminated but good winding
with low PD activity and a clean but poor winding with high PD activity (see point X
on Figure 4.7). However, trending the value with time and comparing results among
different machines, backed-up by visual inspection, should make the situation much
clearer.
4.3.1.6 Antenna techniques
The radio interference noise generated by partial discharges has been recognised
since the 1920s. Using a range of different antennae, attempts have been made to
quantify and characterise the noise produced by partial discharges in particular
corona associated with overhead lines and insulators [5560].
Although the results from these studies would suggest that antenna measure-
ments of partial discharges should enable characterisation of the nature, form and
extent of discharge activity, no such rigorous study has yet been undertaken. This
is disappointing given the attractiveness of a technique which is both non-intrusive
0.8 0.6 0.4 0.2
V, applied
clean surface/
poor bulk
contaminated surface/
good bulk
X
single measurement
at V line (online)
which is it?
V line
t
a
n
,
o

Figure 4.7 Tan measured only at V line problem of online measurement
and requires no connections to plant under test. Recent measurements by the present
author, as yet unpublished, on rotating machines and on air-cooled transformers
have proved extremely successful in this respect. Using several specifically designed
antennae coupled to a high bandwidth oscilloscope, not only could the presence of
discharges be detected but pulse shapes obtained could be correlated with specific
types of discharge. In addition, an estimate of discharge magnitudes in terms of pic-
ocoulombs could be obtained via an indirect laboratory calibration of the antennae.
This technique will be much more widely used in the future.
4.3.1.7 Capacitive probe techniques
When a discharge occurs, an electromagnetic wave is produced which propagates
away fromthe PDsite. Where the plant is metal clad, the wave will propagate towards
the earthed metal enclosure. Provided there is a gap somewhere in the enclosure, e.g.
a gap in the gasket or gap at the busbar chamber cover in the case of metal clad
switchgear, the electromagnetic wave is free to travel to the atmosphere outside
the switchboard. The action of the wave connecting with the earthed metalwork
produces a transient earth voltage (TEV) which can be detected by a capacitive probe
if positioned at the gap. The amplitude of the detected signal is normally in the
millivolts to volts range, and this is generally translated into dB for measurement.
The principle is shown in Figure 4.8.
Although this method of detection will give some indication of severity, there is
no detail of exact location. When more than one capacitive probe is used, however,
it is possible to gain some knowledge of location using the time of flight principle, as
EM wave
seal/gasket
probe
case
high voltage component
discharge producing
electromagnetic emissions
Figure 4.8 Transient earth voltage (courtesy of EA Technology Limited)
the probe nearest the source should detect the discharge first. This may indicate the
panel from which the source is emanating.
An advantage of this type of testing is that the non-intrusive nature of the mea-
surements allows for no disruption or outage to the plant under test. All items of
plant within a substation can be monitored, such as circuit breakers, busbars, current
and voltage transformers and cable end joints. Installing the test equipment while
the substation is online has obvious benefits, and savings are also made due to the
removal of outage costs and the relevant manpower reduction. The advantages and
disadvantages of this technique are detailed in the work of Brown [61].
It is possible that external electromagnetic noise will interfere with the readings
taken, and it is therefore essential that a background survey be completed prior to
commencement. As with all types of PD measurement, if the interference is too
severe, it may not be possible to complete the test, as results may not be sufficiently
analysed. Problems are encountered when these noise sources produce voltages on
other metallic surfaces within the substation using the TEVprinciple. The background
reading should be taken froma metallic surface that is not attached to the switchboard
such as a battery charger or doorframe. If a reading above a certain level is detected in
this circumstance, the probe is unsuitable for monitoring the plant as it is not possible
to differentiate between discharges from the plant or from external sources.
This type of probe can be purchased as a light, portable handheld unit or as
a system with a number of probes connected to (typically) an event counter.
Clearly, the handheld unit is convenient, easy to use and relatively cheap.
However, it will only detect discharges at the moment when the test is conducted.
Although this is useful, it does not include any discharges that may present themselves
at some other time due to load changes, humidity or temperature changes etc. Partial
discharges are often intermittent and therefore a more thorough test regime may be
required. In this event, the multiprobe system can be utilised.
Typically, such a systemmight have eight to ten capacitive probes and these would
be connected to the plant using magnetic clamps. The probes are normally threaded
and screw into the clamps until flush with the earthed enclosure. In turn, the probes
would be connected by separate channels to (typically) an event recorder. Additional
channels should be kept free for antennae. The purpose of the antennae is to detect
any external electromagnetic noise that may filter into the plant environment, be
detected by one or more of the probes and interpreted as partial discharge activity.
By subtracting the events detected by these antennae from those detected by the
probes, interference effects are reduced. These antennae should be positioned in the
corners of the substation and extended vertically at the same height as the plant
under test.
4.3.2 Acoustic detection
Partial discharges produce acoustic noise, as anyone who has listened to the crackling
noise in electrical substations will confirm. Although directional microphone systems
have been used to detect partial discharges from their airborne acoustic emissions,
their application has been largely linked to external busbars, connectors and insulator
assemblies. Acoustic detection has found much greater success and wider application
through the use of piezoelectric sensors.
Piezoelectric polymers, such as PVDF (polyvinylidene fluoride), when com-
pressed, result in the production of an external voltage proportional to the force
applied to the polymer. As such, when built into the head of a handheld probe or
when fixed, with a suitable paste or clamping mechanism, to an appropriate enclo-
sure, they offer a simple means of detecting acoustic signals. Typically, handheld
acoustic probes are coupled to an analogue voltmeter. Unfortunately, although it has
been shown that for simple geometries the resultant voltage is proportional to the
size of the discharge, due to the complex acoustic impedances associated with the
propagation of an acoustic pulse to the probe, no effective calibration is possible
for high voltage plant. The intensity of the emitted acoustic waves is proportional
to the energy released in the discharge. On this basis, the amplitude of the wave is
proportional to the square root of the energy of the discharge and, since energy may
be taken as proportional to the charge squared, there should be a linear relationship
between discharge magnitude and acoustic signal. However, acoustic measurements
are more about detecting the presence of discharges, irrespective of magnitude, and
locating these within the plant item. Typical applications for a handheld probe would
be distribution circuit breaker boxes and small transformers. As the probe is moved
around the enclosure, the larger the voltage detected, the closer the probe is to the
source of discharge activity. On a much larger scale, and using more sophisticated
acquisition instrumentation but the same sensor technology, with a minimum of three
probes fixed to the earthed tank of a large transformer, and a reference signal, it is
possible to determine both the presence and the accurate location of any discharges
present. By measuring the relative times of arrival of the pulse(s) fromthe discharging
site at the three probes, and by assuming a constant velocity of acoustic propagation
through the transformer structure, the relative distance from each of the probes to the
discharging site can be computed and triangulated in three dimensions. Commercial
software is available to do this and can be readily utilised with a laptop computer for
portability.
Knowing the times of arrival of t
1
, t
2
and t
3
from the different sensors, and
assuming a given velocity of propagation, since distance = velocity/time, three
distances can be computed. Knowing these distances, and the location of the probes,
a three-dimensional plot can be made; where these spheres intersect is the discharge
source. The approach is illustrated in Figures 4.9a and 4.9b.
Lundgaard [62, 63] has produced a useful review of acoustic detection of partial
discharges and, as he points out, changes in both signal amplitude and shape occur
as the acoustic signal propagates to any sensors. He cites reduced signal amplitude
as a result of:
geometric spreading of the wave
division of the wave down multiple pathways
transmission losses in propagating from one medium to another and at disconti-
nuities within a given medium
absorption in materials.
sensor 2
sensor 3
sensor 1
PD
t
1
t
2
t
3
discharge
emanating acoustic
waves in 3D
Sensor 2
sensor 1
sensor 3
tank/enclosure
t
1
t
2
t
3
a
b
Figure 4.9
a Partial discharge acoustic emissions arriving at different times at
each sensor according to distance from source
b Arrival of pulses at detector at different times according to distance
travelled
In relation to changes in signal shape, he cites:
frequency-dependent velocity effects resulting in different frequency components
of a given signal arriving at the transducers at different times
frequency-dependent propagation paths, again resulting in different wave
components arriving at the transducer at different times
absorption in materials removing high frequency components preferentially.
In addition, there are significant differences in the acoustic velocities in typical types
of media encountered in (say) a transformer, i.e.:
transformer oil at 25
C, 1415 ms
1
core steel, >5000 ms
1
impregnated pressboard, 1950 ms
1
When one considers the complexity in both structure and materials in (say) a large oil-
filled paper insulated transformer, the factors which can affect the signal propagation,
and the variation in signal velocities with medium, it is impressive that this technique
can be applied to complex structure, large plant.
Typically, in a quiet factory-type environment, discharges in a large power trans-
former can be located to within a volume the size of a football within a couple of hours
and to within the size of a fist within a working day. That said, there is some variation
in the estimated success rate using this technique according to manufacturer. Some
claim total success whereas others are more circumspect admitting that they would
be unlikely to detect a source, using this technique, embedded deep in a winding.
In the field, utilities also report a useful success rate (typically 50 per cent) using
this technique with large power transformers. Indeed, a good example of this is
presented in the work by Jones [64]. He reports that faults, using this approach, can
be categorised as, for example, coming from:
bushing connection stress shields
windings
winding jacking screws
core bolts
winding lead clamps
tapchanger components.
The key issue here is that, although the absolute magnitude of discharge events cannot
be determined from this technique, knowing the location of the source may be just as,
if not more, important. For example, a source identified as coming from a winding
might give serious cause for concern whereas, if the source is corona from a core
bolt, it might not. Asource within the winding will erode the paper and could lead to
a catastrophic failure in time. Acore bolt suffering corona will result in change to the
dissolved gas levels but will not age the transformer in any significant way.
The single probe approach can also be used for capacitors and bushings but great
care must be taken to ensure that placement of the probe on the capacitor or bushing
surface does not distort the electric field resulting in a flashover to the probe.
Finally, it is worth noting that, in the case of distribution circuit breaker boxes,
discharges are often intermittent and a single measurement over a short period of
time may not be sufficient to ensure an accurate assessment. In these situations, an
alternative to the hand held approach is to fix a sensor to the box and leave it on site for
a number of days coupled to an event counter (in the same fashion as per capacitive
probes). In a substation, a number of probes can be used simultaneously in this way,
one to each box, with some form of reference probe also in place, to ensure that any
detected events are coming from discharges within one or more of the distribution
boxes and not from some external source.
In summary, acoustic techniques are relatively cheap and simple to apply, are
utilised online and can detect the presence and location of discharges in the various
items of plant discussed. Their disadvantages include their inability to be applied
to intrinsically noisy plant, e.g. motors/generators, the need for the sensor to be
relatively close acoustically to the discharge source (so the technology cannot, for
example, be applied to cables) and their inability to be calibrated in terms of voltage
output versus size of discharge.
4.3.3 Thermography and other camera techniques
Given that partial discharges are generally hotter than their surrounding media (see
earlier), it might reasonably be thought that the thermal imaging camera could be
applied in their detection and measurement. Unfortunately, most partial discharges
are enclosed in some way, e.g. within solid insulation or within metal clad enclosures.
Given the relatively low temperatures of partial discharges and the high thermal
impedances likely to be present between such discharges and the imaging camera, its
use is very limited in this application.
Certainly where the discharges are external to associated plant, e.g. on overhead
lines/busbars or post/string-type insulators, such techniques can be used but not in any
quantitative way. However, the desire to detect PD from these structures is somewhat
limited compared with other items of plant.
Interestingly, in a similar vein, a daylight corona camera has recently been devel-
oped [65]. This incorporates independent UV video and visible cameras to capture
separate video images of discharges and of associated high voltage plant. The system
detects corona in the 240280 nm region. Corona discharges emit in air mainly in the
230405 nm range of the UV spectrum. Although the corona emission lines between
240 and 280 nmare not as strong as in the 290400 nmrange, this region is also called
the UV solar blind band, i.e. there is no background radiation in this region. Despite
the weaker intensity, the UVsolar blind imager is able to provide high contrast images
due to the complete absence of background radiation.
4.3.4 Chemical detection
Chemical techniques rely on the measurement of byproducts associated with PD
activity and thereby fromwhich PDactivity can be inferred. This necessarily requires
that these byproducts can be detected in some way. Clearly, PD activity associated
Table 4.3 Typical gases absorbed in the oil
under the action of PD
H
2
hydrogen (HH)
CH
4
methane (CH
3
H)
C
2
H
6
ethane (CH
3
CH
3
)
C
2
H
4
ethylene (CH
2
=CH
2
)
C
2
H
2
acetylene (CHCH)
with a closed, unvented void within a section of solid insulation would not lend itself
to detection by this method.
Although chemical detection has been applied to various items of plant involving
gas circulation over the years, e.g. hydrogen-cooled generators and, with relative
success, gas insulated substations, it is primarily in oil-filled equipment that chemical
detection has found favour.
Under the action of partial discharges (and, indeed, other fault conditions), oil
will degrade through bond scission to form characteristic gases absorbed in the oil.
Typical gases produced are given in Table 4.3.
The quantity and mix of gases produced depends on the nature of the fault,
its severity and the associated temperature. The weakest CH bond can be broken
with relatively little energy, i.e. ionisation reactions, with hydrogen being the main
recombination gas. As the strength of the molecular bond increases, more energy
and/or higher temperature is required to create scission of the CC bonds and the
resulting recombination into gases which have either a single CC, double C=C or
triple CC bond.
Being essentially a lowenergy type fault, partial discharge activity tends to favour
the breaking of the weakest CH bond with the production of hydrogen. Carbon
monoxide and carbon dioxide will also be present if the discharge occurs in the
presence of cellulose, i.e. paper insulation (as is generally the case in large power
transformers).
As little as 50 ml of oil suffices for analyses to be performed. This is important
since, although dissolved gas analysis (DGA) has primarily been used for screening
of large, oil-filled transformers (due to the capital involved in such assets) and where
the loss of sampled oil would be insignificant, the increasing use of the technique
with small oil volume plant such as bushings, CVTs etc., has made the volume of oil
to remove critical.
4.3.4.1 Dissolved gas extraction and measurement
Once the sample has been obtained, it can be sent to one of the commercial laboratories
which performs such analyses. It should be stressed at this stage that this is a cheap
method of screening for faults and this too makes it extremely attractive to end users
of oil-filled plant.
The gas can be extracted by a range of methods including the use of Toepler
pump (vacuum extraction), partial degassing, stripping using argon, direct injection
and head space analysis.
Measurements are made via the use of a gas chromatograph, infra-red spectrom-
eter or, indeed, semiconductor sensors or miniature fuel cells.
For field measurements, a range of commercially available fault gas detectors is
available, designed for use with large oil-filled transformers. These include the use of
selectively gas permeable membranes with a miniature fuel cell or Fourier transform
infra-red spectrometer as the gas detector, with portable gas chromatographs. Afault
gas probe is also available, designed for instrument transformers where it is difficult
to obtain an oil sample due to their low oil volume and location. The probe is best
factory fitted and can be used with other forms of low oil volume plant.
4.3.4.2 Interpretation strategies
Several interpretation techniques have been developed and are used in the interpreta-
tion of dissolved gas in oil. These tend to be based on a combination of the quantity
of individual gases present (in parts per million by volume, p.p.m.v.) and the ratios
of these characteristic gases.
Although the presence of partial discharges in oil-filled plant can be inferred from
the absolute levels of different dissolved gases measured, it is primarily through the
ratios of these gases that PD is indicated.
Gas ratios have been in use since 1970 when Dornenburg utilised them to differ-
entiate between fault types. The use of ratios had the advantage that oil volume did
not affect the ratio and hence the diagnosis. Dornenburg first used pairs of gases to
form ratios, to differentiate between electrical and thermal faults. In his first ratios,
an electrical fault was indicated if the ratio of ethylene to acetylene exceeded unity,
and the ratio of methane to hydrogen indicated a thermal fault if greater than 0.1 or
a corona discharge if less than 0.1.
The ratios were developed further and significant levels, for each gas, known as
L1 limits, were introduced. The technique was only to be applied if one of the gas
levels exceeded its L1 limit (see Figure 4.10). As can be seen in Figure 4.10, the
various ratios are indicative of, and capable of differentiating between, low intensity
and high intensity partial discharges.
The development of a thermodynamic model [66] indicated that different
temperatures favoured certain fault gases. The order of gas evolution with increasing
temperature was found to be hydrogen, methane, ethane, ethylene and acetylene,
respectively.
Rogers [67], used the order of gas evolution to form the gas ratios
methane/hydrogen, ethane/methane, ethylene/ethane and acetylene/ethylene. Adiag-
nosis table was created based on nearly ten thousand DGA results, together with
examination of units with suspected faults and failed units. The table went through
several evolutions and was produced in two formats. In the first, Figure 4.11, diagno-
sis was based on codes generated by ratios, and in the second, Figure 4.12, diagnosis
was based on the value of the ratio. As can be seen, different types of partial discharge
can be identified from the gas ratios.
L1 limit (significant value)
Gas H
2
CH
4
CO C
2
H
2
C
2
H
4
C
2
H
6
L1 limit 100 120 350 35 50 65
Ratio Diagnosis
CH
4
H
2
C
2
H
2
C
2
H
4
C
2
H
2
CH
4
C
2
H
6
C
2
H
2
>1.0 <0.75 <0.3 >0.4 thermal decomposition
<1.0 not significant <0.3 >0.4 corona (low intensity partial
discharge)
>0.1 and <1.0 >0.75 <0.3 >0.4 arcing (high intensity partial
discharge)
Figure 4.10 Dornenburgs gas ratios (based on Joseph B. DiGiorgios Dissolved
gas analysis of mineral oil insulating fluids, published by NTT
19962002)
In 1975, CIGRE working group 15.01 assessed over 100 sets of DGA results
from faulty transformers using interpretation schemes that were in use in Europe.
As a result of this assessment, the IEC issued Code 599 [68] in 1978. This code was
based on the Rogers ratios, however, the ethane/methane ratio was omitted as it only
covered a limited temperature range of decomposition but did not assist in further
identification of the fault. The diagnosis table is code based with a rating from 0 to 2
being given depending on the value of the ratio (Figure 4.13).
In use, the code generated by the ratios of dissolved gas data being collected in
the field at times did not match any given in the Standards diagnosis table. This led
various bodies such as utilities, transformer manufacturers, testing laboratories and
consultants to develop their own interpretation techniques based on the IECStandard;
they were widely used to complement the IEC Code rather than replace it. A short-
coming of these techniques was that although they were developed to give more
individual diagnosis for equipment, the diagnosis obtained was often not comparable
between techniques.
Following a review of IEC 599, IEC Code 60599 [69] was issued in 1999. The
code is based on a diagnosis table, Figure 4.14, which retains the original gas ratios
althoughthe diagnosis is nolonger code based. As canbe seen, once againthe presence
of partial discharges of various types can be inferred from the fault gas ratios.
4.3.4.3 Graphical techniques
As a consequence of IECCode 599 being inconsistent in producing reliable diagnoses,
additional schemes were introduced to complement the Code.
CH
4
/H
2
C
2
H
6
/CH
4
C
2
H
4
/C
2
H
6
C
2
H
2
/C
2
H
4
Range Code Range Code Range Code Range Code
0.1 5 <1 0 <1 0 <0.5 0
> 0.1 < 1 0 1 1 1 < 3 1 0.5 < 3 1
1 < 3 1 3 2 3 2
3 2
Code Diagnosis
CH
4
H
2
C
2
H
6
CH
4
C
2
H
4
C
2
H
6
C
2
H
2
C
2
H
4
0 0 0 0 normal
5 0 0 0 partial discharge
1, 2 0 0 0 slight overheating <150
C
1, 2 1 0 0 slight overheating 150200
C
0 1 0 0 slight overheating 200300
C
0 0 1 0 general conductor overheating
1 0 1 0 winding circulating currents
1 0 2 0 core and tank circulating currents, overheated
joints
0 0 0 1 flashover, no power follow through
0 0 1, 2 1, 2 arc, with power follow through
0 0 2 2 continuous sparking to floating potential
5 0 0 1, 2 partial discharge with tracking (note CO)
CO
2
/CO > 11 higher than normal temperature in insulation
Figure 4.11 Rogers fault gas ratios (code basedonJosephB. DiGiorgios Dissolved
gas analysis of mineral oil insulating fluids, published by NTT
19962002)
Duval [70] developed a triangle, Figure 4.15, based on the relative percentage
of methane, ethylene and acetylene gas. The triangle is divided into six regions
representing high energy arcing, low energy arcing, corona discharge and hot spots.
The triangle has the advantage that a diagnosis is always given but, of course, this
will always imply a fault! Hence it must only be used in conjunction with individual
levels of gases which imply the possibility of a fault.
Other graphical techniques include the Church Nomograph Method, based on
data published by Dornenburg and Strittmatter. The data are plotted on sliding loga-
rithmic scales, with each scale representing a different gas. Data points are then joined
together and the slope of the line between adjacent scales is indicative of the type
of fault.
Code Diagnosis
CH
4
H
2
C
2
H
6
CH
4
C
2
H
4
C
2
H
6
C
2
H
2
C
2
H
4
>0.1
<1.0
< 1.0 <1.0 <0.5 normal
0.1 <1.0 <1.0 <0.5 partial discharge corona
0.1 <1.0 <1.0
0.5
<3.0
or >3.0 partial discharge corona
with tracking
>0.1
<0.1
<1.0 3.0 3.0 continuous discharge
>1.0
<0.1
<1.0
1.0
<3.0
or >3.0
0.5
<3.0
or >3.0 arc with power follow
through
>1.0
<0.1
<1.0 <1.0
0.5
<3.0
arc no power follow
through
1.0
<3.0
or >3.0 < 1.0 <1.0 <0.5 slight overheating
to 150
C
1.0
<3.0
or >3.0 1.0 <1.0 <0.5 overheating 150200
C
>0.1
<1.0
1.0 <1.0 <0.5 overheating 200300
C
>0.1
<1.0
<1.0
1.0
<3.0
<0.5 general conductor
overheating
>1.0
<3.0
<1.0
1.0
<3.0
<0.5 circulating currents in
windings
>1.0
<3.0
<1.0 3.0 <0.5 circulating currents core
and tank; overload joints
Note: several simultaneously occurring faults can cause ambiguity in analysis
Figure 4.12 Rogers fault gas ratios (ratio value based) (based on Joseph B.
DiGiorgios Dissolved gas analysis of mineral oil insulating fluids,
published by NTT, 19962002)
As can be imagined, artificial intelligence in the form of expert systems, artificial
neural networks and fuzzy-logic-based systems are finding increasing application in
this area.
Before leaving chemical detection of partial discharges, it is appropriate to note
that for oil/paper insulated plant, the detection of degradation byproducts of the paper
C
2
H
2
/C
2
H
4
CH
4
/H
2
C
2
H
4
/C
2
H
6
Range Code
<0.1 0 1 0
0.11 1 0 0
13 1 2 1
>3 2 2 2
Code Diagnosis Typical examples
C
2
H
2
C
2
H
4
CH
4
H
2
C
2
H
4
C
2
H
6
0 0 0 no fault normal ageing
0 1 0 partial discharges of discharges in gas filled cavities
low energy density resulting from incomplete
impregnation, or supersaturation
or high humidity
1 1 0 partial discharges of as above, but leading to tracking
high energy density or perforation of solid insulation
1 2 0 1 2 discharges of continuous sparking in oil
low energy between bad connections of
different potential or to floating
potential; breakdown of oil
between solid materials
1 0 2 discharges of discharges with power follow
high energy through; arcing breakdown
of oil between windings or coils
or between coils to earth;
selector breaking current
0 0 1 hot spots general insulated
T < 150
C conductor overheating
0 2 0 hot spots
150
C < T < 300
C
local overheating of the core
due to concentrations of flux;
increasing hot spot temperatures;
varying from small hot spots in
core, shorting links in core,
overheating of copper due to
eddy currents, bad contacts/joints
(pyrolitic carbon formation) upto
core and tank circulating currents
0 2 1 hot spots
300
C < T < 700
C
0 2 2 hot spots
T > 700
C
Figure 4.13 IEC 599 1978 fault gas ratios
1
Case Characteristic fault
C
2
H
2
C
2
H
4
CH
4
H
2
C
2
H
4
C
2
H
6
PD partial discharges NS
1
<0.1 <0.2
(see notes 3 and 4)
D1 discharges of low energy >1 0.10.5 >1
D2 discharges of high energy 0.62.5 0.11 >2
T1 thermal fault T < 300
C NS
1
>1 but NS
1
<1
T2 thermal fault <0.1 >1 14
300
C < T < 700
C
T3 thermal fault T > 700
C <0.2
2
>1 > 4
note 1 in some countries, the ratio C
2
H
2
/C
2
H
6
is used, rather than the ratio CH
4
/ H
2
, also in some
countries, slightly different ratio limits are used
note 2 the above ratios are significant and should be calculated only if at least one of the gases is at
a concentration and a rate of gas increase above typical values
note 3 CH
4
/H
2
< 0.2 for partial discharges in instrument transformers; CH
4
/H
2
< 0.07 for partial
discharges in bushings
note 4 gas decomposition patterns similar to partial discharges have been reported as a result of the
decomposition of thin oil film between overheated core laminates at temperatures of
140
C and above
1
NS = non-significant whatever the value
2
an increasing value of the amount of C
2
H
2
/C
2
H
6
may indicate that the hot spot temperature is
higher than 1000
C
Figure 4.14 IEC 60599 1999 edition fault gas ratios interpretation
1
in the oil may also imply the presence of partial discharges. Paper degrades to form
several furans and these can be detected using various techniques of oil analysis.
Although absolute levels of these byproducts are important, as with DGA, as the
ratios of the different byproducts may prove more important in the longer term. It has
been argued that, from these ratios, the temperature resulting in the paper degrading
can be inferred and, in turn fromthis, the integrity of the paper [7174]. Given that the
temperature of the fault can be inferred, as with DGA, it should be possible to infer the
presence (and, indeed, type) of partial discharges. However, this form of monitoring
is not nearly so well established as DGA; the relationship to the detection of partial
discharges has yet to be made. However, it is a potential technique for the future and
most people would now analyse an oil sample for both gas and furan content.
4.3.5 Comparison among different PD measurement techniques relative to
type of plant under investigation
Fromthe foregoing it can be seen that, in general terms, most types of measurement of
PD can be made on most types of plant with some obvious exceptions as mentioned
T1
T2
D2
T3
D1
D+T
% CH
4
% C
2
H
4
% C
2
H
2
80
80 60 40 20
60
40
20
80
60
40
PD
20
Key:
partial discharges
discharges of low energy
discharges of high energy
thermal fault, T < 300C
thermal fault, 300 C < T < 700C
thermal fault, T > 700C
PD
D1
D2
T1
T2
T3
Limits of zones
PD 98% CH
4
D1 23% C
2
H
4
13% C
2
H
2
D2 23% C
2
H
4
13% C
2
H
2
38% C
2
H
4
29% C
2
H
2
T1 4% C
2
H
2
10% C
2
H
4
T2 4% C
2
H
2
10% C
2
H
4
50% C
2
H
4
T3 15% C
2
H
2
50% C
2
H
4
triangle coordinates:
%C
2
H
4
=
100x
x +y +z
%C
2
H
4
=
100y
x +y +z
%CH
4
=
100z
x +y +z
where x = C
2
H
2
y = C
2
H
4
z = CH
4
in p.p.m.
Figure 4.15 IEC 60599 1999 edition Duvals triangle
1
PD measurement technique
Type of IEPD tan C.T. capacitive antenna acoustic chemical thermography,
plant probes etc.
Generators
Circuit
breaker boxes
Transformers
Motors
Cable end boxes
Bushings
Capacitors
CVTs
Overhead
busbars,
insulators, etc.
provided the plant item is air-cooled (i.e. has access vents)
provided the plant item is oil filled

Figure 4.16 A comparison of different PD measurement techniques in relation to
various items of plant
in the text. The decision on which to choose in a given situation becomes a balanc-
ing act among the costs of plant failure (including health and safety considerations,
replacement/repair costs and outage losses), the cost in making a given measurement
(including the frequency of measurement dictated over given time periods and the
cost of potential outages dictated by the need to make solid connections with some
measurement systems) and the quality of information provided by the technique. This
type of decision is unique in each situation. However, for those embarking on the
measurement of PD for their plant, the table shown in Figure 4.16 may prove helpful.
4.3.6 Other items of plant
In addition to the various types of plant highlighted in the foregoing sections, there are
two other plant systems which, because of their specialist nature, have had evolved for
them special techniques to measure PD. This is not to say that they cannot utilise the
techniques described earlier, merely that better techniques are available specifically
related to them. The two items of plant are cables and gas-insulated substations (GIS).
4.3.6.1 Cables
Although other techniques can be applied to cables with limited success in PD mea-
surements, the technique described as cable mapping [75] has evolved specifically
for use with cables where not only is the magnitude and nature of PDrequired but also
their location. Without location, refurbishment of a plant item hidden underground
and extending potentially over many kilometres would be extremely daunting.
The approach is as follows. First, the cable must be disconnected at both ends.
One end is then used for measurement purposes and the other is open-circuited. The
choice is arbitrary. The measurement end also acts as the input for the high voltageAC
supply since, having been disconnected from its inherent supply, an external supply
must be connected. This supply must be sufficiently portable to be carried in a van
to the test site. Due to the impedance of a cable at 50 Hz, this would not be possible
and so measurement on cables is made at between 0.01 Hz and 0.1 Hz.
I =
V
Z
where
Z =
1
2
f C
therefore
I = V2f C
The detector is similar to that described previously for electrical detection of discharge
activity at the high voltage terminals of plant, i.e. a discharge-free capacitor coupled
to a low voltage detector impedance but, in this case, the low voltage impedance is
connected to instrumentation which can record both the magnitude and time of arrival
of discharge pulses. When a degradation site discharges, part of the discharge pulse
propagates towards the detector end and part to the open circuit end. When the latter
encounters the open circuit, it is reflected and propagates back towards the detector
end. In consequence, at the detector, a pulse arrives representing that part of the
discharge which originally propagated from the discharge site followed by a second
pulse representing that part of the pulse which propagated in the opposite direction,
was reflected at the open circuit, and then propagated back along the cable to the
detector (see Figure 4.17).
Knowing the relative times of arrival for the cable as a whole (obtained by
injecting a calibration pulse at the detector end and detecting its time of arrival
back to that point from the open-circuited end), the percentage location of the
degradation site along the cable can be ascertained as can its length, knowing the
velocity of propagation (Figures 4.17a and b) i.e. percentage location of PD along
cable = (1 (t
1
t
2
)/t cable) 100, where t
1
and t
2
are the times of
arrival of the two parts of the pulse and t cable is the total transit time (obtained by
the calibration pulse injection) for the cable. Note, in this equation, t
1
and t
2
are
taken as negative values.
What is produced is a map of the cable with the cable length on the horizontal
axis and any discharge magnitudes on the vertical axis. By inspection, the size of the
degradation site, its presence in only one or in more phases and its location in the
cable or in a connector, can be obtained (Figure 4.18).
In addition, for each degradation site, a qn plot can be obtained from which
the nature of the degradation can be inferred as indicated previously.
This approach can be used to measure PD activity on cables from around 4 m to
3 km in length (the latter figure arising from the attenuation of discharge pulses in
m
a
g
n
i
t
u
d
e

o
f

p
u
l
s
e

signal direct
to detector
t
2 reflected signal
time b
t
1
open-circuited
end of cable
discharge site
signal reflected at o/c
end of cable, arriving
at detector at t
2
a
to detector
VLF
supply
C
Z
signal direct to
detector from discharge
site arriving at t
1
Figure 4.17
a Partial discharge cable mapping
b Signals time of arrival used for location
paper-insulated cables). For cables with attenuation constants substantially different
from paper cables, different figures may apply. It should be noted that a double-
ended measurement is also possible which effectively doubles the cable length which
can be tested. Errors associated with this approach have been addressed by Kreuger
et al. [76].
4.3.6.2 Gas-insulated substations (GIS)
The criticality of GIS to electrical power generation/transmission has made it imper-
ative that PD activity within their enclosures is identified (see Chapter 2). Chemical
byproducts of SF
6
(as used in GIS) are particularly active chemically and will readily
result in degradation within the system.
Due to their nature, and in particular the very high electric fields within their
assemblies, GIS suffer specific and relatively unique problems. These issues are
neatly summarised in a two-part series of papers [77, 78]. As indicated there, the
length of cable
0%
100%
3 phases
of cable
R
Y
B
d
i
s
c
h
a
r
g
e

m
a
g
n
i
t
u
d
e

Figure 4.18 Cable map: PDas afunctionof locationandmagnitude for three phases
of cable
range of possible PD inducing defects include:
moving particles
electrode protrusions
fixed particles on insulating surfaces
floating electrodes
loose, non-floating electrodes
voids in solid insulation of spacers.
To detect PDassociated with these various defects, three approaches are possible,
namely chemical, acoustic and electrical detection.
In relation to chemical detection, as indicated above, under PD stressing SF
6
will
degrade to produce characteristic byproducts [79]. By gas sampling, it is possible to
detect these and thereby infer the presence and, indeed, the nature of PD activity.
Unfortunately, the technique is not quantitative and its effectiveness can be com-
promised by desiccants in the GIS chambers which absorb water but also these
byproducts.
Acoustic detection can be effective [8082] but, due to the high attenuation of
acoustic signals at flanges etc., each section must be tested in turn.
In relation to electrical detection, within the factory environment, a standard
IEC60270 type approach can be used as described earlier in this chapter. Alter-
natively, rather than using large coupling capacitors, electric field sensors can be
incorporated at a standard flange cover/inspection hatch within the GIS. Essentially,
the gas gap capacitance between the busbar in the GIS and the chamber wall acts
as the coupling/high voltage capacitor. The low voltage impedance coupled to the
sensor is essentially as described previously.
These methods have all been described previously, although it is interesting to
see how they are applied in GIS. However, a novel method of electrical detection has
been developed for field measurements on GIS and it is this which is of particular
interest in this section.
The principle of this UHF technique (further details can be found in Chapter 2) is
that the current pulse which forms the partial discharge has a very fast risetime and
this excites the GIS chambers into multiple resonances at frequencies up to 1.5 GHz
or so (for more details see Chapter 3). These resonances can last for several microsec-
onds (compared with the nanosecond(ish) duration of the originating current pulse).
Once again, suitable UHF couplers can be fitted at flange covers etc. (thus permitting
retrofitting). The PD signals can then be amplified and processed to produce (typi-
cally) the qn type patterns described earlier. The patterns produced are indicative
of the nature of the PD-inducing defect. The work of Pearson et al. [83] and Judd
et al. [84] is particularly illuminating in this context [85].
4.4 Concluding remarks
The study of partial discharges is not an academic pursuit. It is driven by a very
practical desire, a desire to understand a phenomenon which can be utilised to infer
the level of integrity within the insulation systems of high voltage plant and which, in
itself, constitutes a serious stress degradation mechanism. The role of partial discharge
studies remains twofold: to enable the development of new insulating systems which
are resistant to partial discharge stressing and to predict remnant life for existing
plant systems. To this end, studies in partial discharges involve a balance a balance
between understanding the phenomenon and being able to measure it. On one side
lies the understanding of how partial discharges cause degradation, and which are
the key parameters of activity to be measured. On the other side lies the ability to
make the measurement, often under very adverse conditions.
Partial discharge study is a minor, but important, field of research. The continuing
desire to find new, cheaper materials or subject existing materials to ever higher stress
levels in a quest to cut costs will keep it so. The 1950s and 1960s saw the emergence
and development of PD measurement techniques culminating in correlations being
identified among PD activity and specific forms of degradation. These decades also
witnessed a developing understanding, based on strong physical principles, of the
mechanisms of partial discharge degradation and breakdown. In many senses this
was the golden age of partial discharge study. The 1970s saw steady progress in
electrical and acoustic measuring techniques and progress in chemical techniques for
oil analysis. However, in degradation studies, the optimism of the 1950s and 1960s
was waning. With a few conspicuous exceptions, many researchers were moving
to empirical rate studies to breakdown. Accelerated frequency ageing of materials
to breakdown under partial discharge stressing was replacing the more fundamental
studies designed to elucidate the physical and chemical mechanisms of degradation.
There appeared to be a recognition that an understanding of partial discharge degra-
dation mechanisms was more complex and more unattainable than had been
thought in earlier years. A more pragmatic, if limited, approach had emerged. The
1980s and 1990s saw another major shift in activity. Fuelled by the advent of digital
microprocessor technologies, the ability to capture, store and process data on partial
discharges to measure partial discharge activity made an enormous leap forward.
The excitement of these new technologies and how they could be applied to partial
discharge measurement far outpaced the research on degradation studies again with
a few, conspicuous, exceptions relating principally to the understanding of chemical
stress reactions at a discharging surface.
Now, as we move into the new millennium, our ability to measure, store, process
etc. data related to partial discharge activity far outstrips our understanding of what
to do with these data irrespective of whether the data was acquired by electrical,
acoustic or chemical means.
Despite our progress, the synergetic nature of interactions present under par-
tial discharge stressing continues to ensure that a clear understanding of discharge
degradation remains elusive. Without that understanding, the quest for new materials
capable of suppressing partial discharge activity will remain unfulfilled. Equally,
without that understanding, the ability of database techniques and associated
analytical/statistical tools to permit an accurate and objective assessment of insu-
lation integrity and remnant life for existing plant insulating systems will remain
flawed. It is perhaps time to recognise that a conscious shift to more fundamental
degradation studies, to improve our understanding of the mechanisms involved, is
the most efficient and effective way forward.
And finally to those readers who are perhaps embarking on a study of partial
discharges for the first time a thought. As with most fields of endeavour, partial
discharge research is complex. Understanding the mechanisms of partial discharge
degradation, parameters of partial discharge activity and thereby inferring the rate,
form and extent of degradation to predict remnant life is extraordinarily difficult. In a
careers researchof partial discharges, most of us canonlyhope tocontribute one small
piece to the puzzle of this phenomenon. However, once in every few generations, a
researcher of true stature emerges. This person is capable of taking the many and
varied pieces of the puzzle and bringing them together to form an intelligible picture.
That has yet to happen for partial discharges. The most that each of us can hope to
do, until then, is to ensure that the small piece which we contribute to the study of
partial discharges has clarity, objectivity and accuracy.
The major breakthrough in partial discharge research the solution to the puz-
zle must not be hindered by errors in the pieces of research that have gone before.
Contribute your piece but be open to the thought that maybe, just maybe, you are the
one to solve the puzzle!
4.5 Note
1
The author thanks the International Electrotechnical Commission (IEC) for permis-
sion to reproduce: the table Fault Gas Ratios Interpretation and the illustration of
Duvals Triangle from its International Standard IEC 60599 2nd edition 199903
Mineral oil-impregnated electrical equipment in service Guide to the interpretation
of dissolved and free gases analysis. All such extracts are copyright of IEC, Geneva,
Switzerland. All rights reserved. Further information on the IEC is available from
www.iec.ch. IEC has no responsibility for the placement and context in which the
extracts and contents are reproduced by the author; nor is IEC in any way responsible
for the other content or accuracy therein.
4.6 References
1 HEPBURN, D.M.: The chemical degradation of epoxy resin by partial
discharges. PhD thesis, Glasgow Caledonian University, October 1994
2 TANAKA, T.: Internal partial discharge and material degradation, IEEETrans.
Electr. Insul., 1986, EI-21, pp. 899905
3 GAMEZ-GARCIA, M., BARTNIKAS, R., and WERTHEIMER, M.R.:
Synthesis reactions involving XLPE subjected to partial discharges, IEEE
Trans. Electr. Insul., 1987, EI-22, pp. 199205
4 MAYOUX, C.J.: Partial discharge phenomena and the effect of their con-
stituents on PE, IEEE Trans. Electr. Insul., 1976, EI-11, pp. 139148
5 MEEK, J.M., and CRAGGS, J.D. (Eds.): Electrical breakdown of gases(Wiley,
Chichester, New York, 1978)
6 BOULLOUD, A. and CHARRIER, I.: Current density on the plane electrode of
a positive point-to-plane gap. 6th international conference on Gas discharges
and applications, Edinburgh, UK, September 811, 1980, pp. 110113
7 BLACK, W.B., in BOVEY, F.A., and WINSLOW, F.H. (Eds.): Macro-
molecules: an introduction to polymer science (Academic Press, 1978),
Chap. 7
8 FUJITA, H.: An analysis of mechanical stress in solid dielectrics caused by
discharge in voids, IEEE Trans. Electr. Insul., 1987, EI-22, pp. 277285
9 ARBAB, M.N., AUCKLAND, D.W., and VARLOW, B.R.: Mechanical
processes in the long term degradation of solid dielectrics. Proceedings of
IEE 5th international conference on Diel. Mats., Meas. and Appl., Canterbury,
UK, June 1988, pp. 109111
10 PEYROUS, R., COXON, P., and MORUZZI, J.: Mass spectra of ionic species
created by dc corona discharges in air. Proceedings of IEE 7th international
conference on Gas discharges, London, UK, September 1982
11 PEYROUS, R., and MILLOT, R.M.: Gaseous products created by dc corona
discharges in air or oxygen fed point to plane gap. Proceedings of IEE 7th
international conference on Gas discharges, London, UK, September 1982
12 ALLEN, N.L., COXEN, P., PEYROUS, R., and TEISSEYRE, Y.: Anote on the
creation of condensation nuclei by negative discharges in air at low pressure,
J. Phys. D, Appl. Phys., 1981, 14, pp. L20790
13 SONE, M., MITSUI, H., and TAKAOKA, K.: Clustering of water molecules
in several kinds of gases. Conference on Electrical insulation and discharge
phenomena, Pocono Manor, Penn., USA, October 2831, 1990, pp. 696701
14 YAGI, S., and TANAKA, M.: Mechanism of ozone generation in air-fed
ozonisers, J. Phys. D, Appl. Phys., 1979, 12, pp. 15091520
15 SHIELDS, A.J., and KEMP, I.J.: Degradation and breakdown of mica under
partial discharge stressing, IEE Proc., Sci. Meas. Technol., 2000, 147, (3),
pp. 105109
16 SHIELDS, A.J., and KEMP, I.J.: Degradation and breakdown of mica under
partial discharge stressing: transverse discharges, IEE Proc., Sci. Meas.
Technol., 2000, 147, (5), pp. 256260
17 HEPBURN, D.M., KEMP, I.J., SHIELDS, A.J., and COOPER, J.: Degradation
of epoxy resin by partial discharges, IEE Proc., Sci. Meas. Technol., 2000, 147,
(3), pp. 97104
18 HEPBURN, D.M., KEMP, I.J., RICHARDSON, R.T.R., and SHIELDS, A.J.:
Role of electrode material in partial discharge chemistry. Proceedings of 5th
international conference on Cond. and breakdown in solid dielectrics, Leicester,
UK, July 1995
19 DISSADO, L.A., and FOTHERGILL, J.C.: Electrical degradation and
breakdown in polymers, in STEVENS, G.C. (Ed.): (Peter Peregrinus Ltd,
1992)
20 MAYOUX, C.J.: Influence of lowenergyions inthe degradationof polyethylene
by partial discharges, European Polym. J., 1973, 9, pp. 10691075
21 MAYOUX, C.J.: Corona discharge and the ageing process of an insulator,
IEEE Trans. Electr. Insul., 1977, EI-12, pp. 153158
22 HARROLD, R.T., EMERY, F.T., MURPHY, F.J., and DRINKUT, S.A.: Radio
frequency sensing of incipient arcing faults within large turbine generators,
IEEE Trans. PAS, 1979, 98, (3), pp. 11671173
23 STONE, G.: Measuring partial discharge in operating equipment, IEEE Electr.
Insul. Mag., 1991, 7, (4), pp. 920
24 KREUGER, F.H.: Recognition of discharges, Electra, 1970, 11, pp. 6991
25 GULSKI, E.: Digital analysis of partial discharges, IEEE Trans. Dielectr.
Electr. Insul., 1995, 2, (5), pp. 822837
26 BOGGS, S.A., and STONE, G.C.: Fundamental limitations in the measurement
of corona and PD, IEEE Trans. Electr. Insul., 1982, 17, (2), pp. 143150
27 BARTNIKAS, R.: A commentary on pd measurement and detection, IEEE
Trans. Electr. Insul., 1987, 22, (5), pp. 629653
28 BOGGS, S.A.: Partial discharge: overviewand signal generation, IEEE Electr.
Insul. Mag., 1990, 6, (4), pp. 3342
29 KREUGER, F.H., GULSKI, E., and KRIVDA, A.: Classification of pd, IEEE
Trans. Electr. Insul., 1993, EI-28, (6), pp. 917931
30 GULSKI, E., and KRIVDA, A.: Influence of ageing on the classification of
partial discharges in high voltage components, IEEE Trans. Dielectr. Electr.
Insul., 1995, 2, pp. 676684
31 ZAENGL, W.S., OSVATH, P., and WEBER, H.J.: Correlation between band-
width of pd detectors and its inherent integration error. Proceedings of IEEE
international symposium on Electr. Insul., Washington DC, USA, June 1986
32 ZAENGL, W.S., LEHMANN, K., and ALBEIZ, M.: Conventional pd
measurement techniques used for complex hv apparatus, IEEE Trans. Electr.
Insul., 1992, EI-27, (1)
33 ZAENGL, W.S., and LEHMANN, K.: A critique of present calibration
procedures of partial discharge measurements, IEEETrans. Electr. Insul., 1993,
EI-28, (6), pp. 10431049
34 TWIEL, M.M., STEWART, B.G., and KEMP, I.J.: An investigation into the
effects of parasitic circuit inductance onpartial discharge detection. Proceedings
29th EIC/EMCW Expo., Cincinnati, USA, October 2001
35 BRYDEN, J., KEMP, I.J., NESBITT, A., CHAMPION, J.V., DODD, S., and
RICHARDSON, Z.: Correlations amongtree growthandthe measurable param-
eters of partial discharge activity. Proceedings of 11th international symposium
on HV Eng., 4, (79.524), London, UK, August 1999
36 PATSCH, R., HOOF, M., and REUTER, C.: Pulse-sequence-analysis, a promis-
ing diagnostic tool. Proceedings of 8th international symposium on High
Voltage Eng., Yokohama, Japan, August 1993
37 PATSCH, R., and HOOF, M.: The influence of space charge and gas pressure
during electrical tree initiation and growth. Proceedings of 4th international
conference on Props. and Appl, of Diel. Mats., Brisbane, Australia, July 1994
38 HOOF, M., and PATSCH, R.: Pulse-sequence-analysis: a new method for
investigating the physics of pd-induced ageing, IEE. Proc., Sci. Meas. Technol.,
1995, 142, (1), pp. 95101
39 HOOF, M., and PATSCH, R.: Voltage-difference analysis: a tool for partial
discharge source identification. Proceedings of IEEE international symposium
on Electr. Insul., Montreal, Canada, June 1996
40 WILSON, A., JACKSON, R.J., and WANG, N.: Discharge detecting tech-
niques for stator windings, IEE Proc. B, Electr. Power Appl., 1985, 132, (5),
pp. 234244
41 KEMP, I.J., GUPTA, B.K., and STONE, G.C.: Calibration difficulties asso-
ciated with partial discharge detectors in rotating machine applications.
Proceedings of EEIC/ICWAconference on Electr. Insul., Chicago, USA, 1987
42 VAILLANCOURT, G.H., MALEWSKI, R., and TRAIN, D.: Comparison
of three techniques of partial discharge measurements in power transformers,
IEEE Trans. Power Appar. Syst., 1985, PAS-104, (4), pp. 900909
43 WOOD, J.W., SEDDING, H.G., HOGG, W.K., KEMP, I.J., and ZHU,: Par-
tial discharges in HV machines; initial considerations for a PD specification,
IEE Proc. A, Sci. Meas. Technol., 1993, 140, (5), pp. 409416
44 KOPF, U., and FESER, K.: Rejection of narrow-band noise and repetitive
pulses in on-site pd measurements, IEEE Trans. Dielectr. Electr. Insul., 1995,
2, (6), pp. 11801191
45 BORSI, H., GOCKENBACH, E., and WENZEL, D.: Separation of partial
discharge from pulse-shaped noise signals with the help of neural networks,
IEE Proc., Sci. Meas. Technol., 1995, 142, (1), pp. 6974
46 LOUIS, A.K., MAASS, P., and RIEDER, A.: Wavelets: theory and applications
(John Wiley & Sons, 1997)
47 DAUBECHIES, I.: The wavelet transform, time-frequency localization and
signal analysis, IEEE Trans. Inf. Theory, 1990, 36, (5), pp. 9611005
48 MALLAT, S.G.: Atheory for multiresolution signal decomposition: the wavelet
representation, IEEE Trans. Pattern Anal. Mach. Intell., 1989, 11, (7),
pp. 674693
49 ANGRISANI, L., DAPONTE, P., LUPO, G., PETRARCA, C., and
VITELLI, M.: Analysis of ultrawide-band detected partial discharges by means
of a multiresolution digital signal-processing method, Measurement, 2000,
27, pp. 207221
50 MA, X., ZHOU, C., and KEMP, I.J.: DSP based partial discharge character-
isation by wavelet analysis. Proceedings of 19th international symposium on
Discharges and electrical insulation in vacuum (ISDEIV2000), Xian, China,
1822 September 2000
51 MA, X., ZHOU, C., and KEMP, I.J.: Investigation into the use of wavelet
theory for partial discharge pulse extraction in electrically noisy environments.
Proceedings of IEE 8th international conference on Dielectr. Mats. Meas. and
Appl., Edinburgh, UK, 1721 September 2000
52 SHIM, SORAGHAN, J.J., and SIEW, W.H.: Detection of pd utilising digital
signal processing methods, part 3: open loop noise reduction, IEEE Electr.
Insul. Mag., 2001, 17, (1), pp. 613
53 MA, X., ZHOU, C., and KEMP, I.J.: Wavelet for pd data compression. IEEE
conference on Electrical insulation and dielectric phenomena (CEIDP 2001),
Kitchener, Canada, October 2001
54 SIMONS, J.S.: Diagnostic testing of high-voltage machine insulation a review
of ten years experience in the field, IEE Proc. B, Electr. Power Appl., 1980,
127, (3), pp. 139154
55 PAKALA, W.E., and CHARTIER, V.L.: Radio noise measurement on overhead
power lines from2.4 to 800kV, IEEETrans. Power Appar. Syst., 1971, PAS-90,
pp. 11551165
56 SAWADA, Y., FUKUSHIMA, M., YASUI, M., KIMOTO, I., and NAITO,
K.: A laboratory study of RI, TIV, and AN of insulators strings under
contaminated condition, IEEE Trans. Power Appar. Syst., 1974, PAS-93,
pp. 712719
57 BERADELLI, P.D., CORTINA, R., and SFORZINI, M.: Laboratory investiga-
tion of insulators in different ambient conditions, IEEE Trans. Power Appar.
Syst., 1973, PAS-92, pp. 1424
58 IEEE RADIO NOISE AND CORONASUBCOMMITTEE REPORT, RI limits
Task Force, Working Group No. 3: Reviewof technical considerations on limits
to interference from power lines and stations, IEEE Trans. Power Appar. Syst.,
1980, 99, (1), pp. 35379
59 KANNUS, K., LEHTIO, A., and LAKERVI, E.: Radio and TV interference
caused by public 2.4 kV distribution networks, IEEE Trans. Power Deliv.,
1991, 6, (4), pp. 18561861
60 STEWART, B.G., HEPBURN, D.M., KEMP, I.J., NESBITT, A., and
WATSON, J.: Detection and characterisation of partial discharge activity on
outdoor high voltage insulating structures by RF antenna measurement tech-
niques. Proceedings of international symposium on High-voltage engineering,
London, UK, August 1999
61 BROWN, P.: Non-intrusive partial discharge measurements in high voltage
switchgear. IEE Colloquiumon Monitors and condition assessment equipment,
December 1996
62 LUNDGAARD, L.E.: Partial discharge, Pt XIII: acoustic partial discharge
detection fundamental consideration, IEEE Electr. Insul. Mag., 1992, 8, (4),
pp. 2531
63 LUNDGAARD, L.E.: Partial discharge, Pt XIV: acoustic partial discharge
detection practical application, IEEE Electr. Insul. Mag., 1992, 8, (5),
pp. 3443
64 JONES, S.L.: The detection of partial discharge activity within power trans-
formers using computer aided acoustic emission techniques. Proceedings of
IEEE international symposium on Electr. Insul., 1990, Toronto, Canada
65 LINDNER, M., ELSTEIN, S., LINDNER, P., and TOPAZ, J.M.: Daylight
corona discharge imager. Proceedings of international symposium on High
voltage Eng., 4.3494.352, London, UK, August 1999
66 HALSTEAD, W.D.: Athermodynamic assessment of the formation of gaseous
hydrocarbons in faulty transformers, J. Inst. Petroleum, 1973, 59
67 ROGERS, R.R.: Concepts used in the development of the IEEE and IEC codes
for the interpretation of incipient faults in power transformers by dissolved gas
in oil analysis. IEEE Winter Meeting, 1978
68 IEC 599: Guide for the interpretation of the analysis of gases in transformers
and other oil-filled electrical equipment in service, 1978
69 IEC 60599: Mineral oil-impregnated electrical equipment in service Guide to
the interpretation of dissolved and free gas analysis, second edn., 1999
70 DUVAL, M.: Dissolved gas analysis: it can save your transformer, IEEEElectr.
Insul. Mag., 1989, 8, (6), pp. 2227
71 SHROFF, D.H., and STANNETT, A.W.: A review of paper ageing in
power transformers. IEE Proc. C, Gener. Transm. Distrib., 1985, 132, (6),
pp. 312319
72 UNSWORTH, J., and MITCHELL, F.: Degradation of electrical insulating
paper monitored with high performance liquid chromatography, IEEE Trans.
Electr. Insul., 1990, EI-25, (4), pp. 737746
73 EMSLEY, A.M., XIAO, X., HEYWOOD, R.J., and ALI, M.: Degradation of
cellulosic insulation in power transformers. Part 2: Formation of furan products
in insulating oil, IEE Proc., Sci. Meas. Technol., 2000, 147, (3), pp. 110114
74 CARBALLIERA, M.: HPLCcontribution to transformer survey during service
or heat run tests. IEE Coll., Publ. 184, 1991
75 MACKINLAY, R.: Managing an ageing underground high-voltage cable
network, Power Eng. J., 1990, 4, pp. 271277
76 KREUGER, F.H., WEZELENBURG, M.G., WIENER, A.G., and
SONNEVELD, W.A.: Partial discharge Pr XVIII: Errors in the location of
partial discharges in high voltage solid dielectric cables, IEEE Electr. Insul.
Mag., 1993, 9, (6), pp. 1523
77 BAUMGARTNER, R., FRUTH, B., LANZ, W., and PETTERSON, K.: Partial
discharge, part IX: PD in gas-insulated substations 1, IEEE Electr. Insul.
Mag., 1991, 7, (6), pp. 513
78 BAUMGARTNER, R., FRUTH, B., LANZ, W., and PETTERSON, K.: Partial
discharge Part X: PDin gas-insulated sub stations measurement and practical
considerations, IEEE Electr. Insul. Mag., 1992, 8, (1), pp. 1627
79 CHU, F.Y.: SF
6
decomposition in gas insulated equipment, IEEETrans. Electr.
Insul., 1986, EI-21, pp. 693726
80 GRAYBILL, H.Q., CRONIN, J.C., and FIELD, E.J.: Testing of gas insulated
substitutions and transmission systems, IEEE Trans. Power Appar. Syst., 1974,
PAS-93, pp. 404413
81 LUNDGAARD, L.E., RUNDE, M., and SKYBERG, B.: Acoustic diagnoses
of gas insulated substations a theoretical and experimental basis, IEEE Trans.
Power Deliv., 1990, PD-S, (4), pp. 17511759
82 LUNDGAARD, L.E., TANGEN, G., SKYBERG, B., and FAGUSTAD, K.:
Acoustic diagnoses of GIS: field experience and expert systems, IEEE Trans.
Power Deliv., 1992, PD-7, (1), pp. 287294
83 PEARSON, J.S., FARISH, O., and HAMPTON, B.F. et al.: Partial discharge
diagnostics for gas insulated substations, IEEE Trans. Dielectr. Electr. Insul.,
1995, 2, (5), pp. 893905
84 JUDD, M.D., FARISH, O., and HAMPTON, B.F.: Excitation of UHF signals
by partial discharges in GIS, IEEE Trans. Dielectr. and Electr. Insul., 1996, 3,
(2), pp. 213228
85 CIGRE Task Force 15/33.03.05: Partial discharge detection system for GIS:
sensitivity verification for the UHF method and the acoustic method, Electra,
1999, (183), pp. 7487
Chapter 5
ZnO surge arresters
A. Haddad
5.1 Introduction
High voltage systems are often subject to transient overvoltages of internal or external
origin. The resultant surges travel along the transmission line and can cause damage to
unprotected terminal equipment. Corona losses and the earth return path can attenuate
and distort the surges, but the magnitude of the surge may still exceed the insulation
level of the equipment. Surge arresters provide a limitation of the overvoltage to a
chosen protective level. The superiority of the recently developed zinc oxide (ZnO)
material over earlier silicon carbide (SiC) renewed interest and boosted the use of
surge arrester protection.
The ideal surge arrester would be one that would start conduction at a voltage
level at some margin above its rated voltage, hold that voltage level with as little
variation as possible for the duration of the overvoltage surge and cease conduction
as soon as the voltage across the arrester returns to a value close to the rated voltage.
Such an arrester would therefore conduct only that current required to reduce the surge
voltage to the arrester protective level, and absorb the energy that is associated with
the overvoltage. The basic non-linear formula that relates voltage V and current I in
a surge arrester is given by:
I = kV
(5.1)
where k is a constant and is the coefficient of non-linearity.
Surge arresters are used for protection of power system equipment against surge
overvoltages because they offer low protection levels and permit the reduction
of insulation levels, which has a substantial effect on the cost of high voltage
equipment. Early overvoltage protective devices were simple spark gaps capable
of supporting essentially no voltage during conduction. Major development steps in
the evolution of the arrester technology, which are in the direction of the optimum
arrester characteristic, have been made since then.
5.2 Evolution of overvoltage protection practice
The evolution of surge arrester technology has been characterised by both the grad-
ual improvement of the various arrester components, and more importantly by four
successive major steps: the simple spark gap, the valve-type arrester, the introduction
of active gaps and the gapless metal oxide arrester. The latter is associated with ZnO
varistors, and the two former arresters were made with SiCresistors. The introduction
of each arrester had an important impact on protection levels and cost of the power
system equipment as a whole.
5.2.1 Simple spark gaps
During the first half of the twentieth century, protection of apparatus in electrical
power systems was provided by rod gaps (also referred to as coordinating gaps) and
a very high withstand voltage for the insulation [1].
The advantage of rod gaps lies in their simplicity and cost. However, they cannot
support any voltage during their operation and cannot clear the power frequency
follow current, which means that after sparking over, a permanent fault occurs which
leads to interruption of supply. Moreover, the time lag to sparkover and the depen-
dence of sparkover on many factors may result in failure of the protective system. The
combination of a high protective level with spark gaps still finds application today in
lowvoltage applications, but at high service voltages increasing the withstand voltage
of insulation has a marked influence on costs.
At 33 kVsystems and lower, it is not recommended to use simple rod gaps because
they may be bridged by birds; instead duplex gaps and/or triggered gaps of expulsion
type may be employed.
5.2.2 Valve-type arresters
The discovery of the non-linear properties of silicon carbide (SiC) around 1930 made
feasible the introduction of valve-type surge arresters to protect power systems against
atmospheric discharges. The general design of a conventional arrester consists of
plate-type gaps spaced by insulating rings with series non-linear SiC resistors also
known as thyrite. The spark gap performs the switching function and the SiC resistor
limits the follow current and enables the arrester to reseal. By subdividing the spark
gap, it is possible to reseal at higher voltages eliminating steady state energy dissipa-
tion. Compared with spark gaps, valve-type arresters have a number of advantages
but the protective level remains relatively high. The SiC elements have severe speci-
fication standards because most of the voltage is supported by the non-linear resistors
and the arc voltage is comparatively negligible; also most of the energy associated
with the discharge of a transmission line/cable is absorbed by the SiC non-linear
resistors.
ZnO surge arresters 193
5.2.3 Surge arresters with active gaps
A significant improvement of arresters to alleviate problems of utilisation was
achieved by the introduction of active gaps (also called current-limiting gaps) [2].
These are characterised by the presence of blast coils that produce a strong magnetic
field during the passage of arrester followcurrent. The arcs across the individual gaps
are then elongated and blown towards the edge of the chamber, thus producing an
increased voltage drop during the followcurrent period. The power frequency voltage
across the series non-linear resistors is therefore reduced with the consequent reduc-
tion of the follow current. The arc voltage opposes the follow current and interrupts
it before the working voltage reaches zero. Afurther component of the active design
is the grading system which ensures that the voltage is distributed uniformly between
the series spark gaps.
Compared with simple gap arresters, active gap arresters have the following
advantages:
(i) arc voltage is of the order of the voltage drop across the non-linear SiC resistor
(ii) protective level is substantially lowered
(iii) the energy absorbed by the resistors decreases; some of it is absorbed by the
elongated arc and the other part remains in the transmission line because the
follow current is interrupted before voltage zero
(iv) constant voltage during flow of arrester current
(v) the roots of the arc move along the electrode which reduces the change in arcing
found in simple plate gap arresters.
5.2.4 Metal oxide surge arresters
The relatively high protective levels provided by SiC arresters became more and
more an economical disadvantage with the increase of maximum system voltages. In
order to reduce the insulation levels of the apparatus it was, therefore, necessary to
try to reduce the protection levels of the surge arresters. Furthermore, the increase
of transmission line lengths resulted in an increase of the energy that the arrester
had to absorb in case of a line discharge through it. The search for new materials
to obtain superior non-linear VI characteristics led to the discovery of zinc oxide
(ZnO) varistors in the late 1960s. The impedance of ZnO varistors at voltages below
the rated voltage is so high that the resulting current is in the milliampere range. The
direct consequence of this lowcurrent consumption was the possibility of constructing
surge arresters with no series gaps. The first power system gapless metal oxide surge
arresters were completed in the mid 1970s [3, 4]. The absence of the gaps and the
extreme non-linearity of the voltagecurrent characteristic of the material resulted in
the following additional changes in the main features of arrester protection:
(i) elimination of grading resistors/capacitors, which reduced further the number
of parts used for the arrester construction
(ii) energy absorbed by the non-linear resistor represents only a fraction of the
discharge energy of the transmission line because there is no follow current,
andparallel connections of varistors toincrease the energyabsorptioncapability
of the arrester are now possible [5]
(iii) lower discharge voltage (residual voltage which appears across the arrester
terminals when a discharge current is flowing through it [6, 7])
(iv) lower protective level (highest discharge voltage that appears between arrester
terminals during specific conditions of a discharge operation [6, 7])
(v) faster switching capability compared with a spark gap time response.
5.2.5 Existing applications of ZnO surge arresters
5.2.5.1 Applications
In addition to the extensive use in low voltage suppressors used in electronic equip-
ment anddomestic appliances [8], ZnOmaterial has nowlargelyreplacedSiCmaterial
in modern surge arresters. Because of their reliability and superiority, ZnO surge
arresters are used in various overvoltage protection schemes and in coordination with
overcurrent protection [9]. Among these schemes are standard protection of overhead
distribution systems [1013], underground distribution systems [1417], GISsystems
[1820], HVDC systems and cables [18, 19], EHV series capacitors [2123], EHV
shunt capacitors [24], EHVbreakers [25] and rotating electrical machines [26]. Surge
arresters, which limit surges entering underground power cable and provide adequate
protection at open end of feeders [15, 27], known as riser-pole and open-end arresters,
are needed in order to maintain acceptable protective margins on cable systems.
Surge arresters are also used at bonding points of the cable system [16, 17].
They are referred to as sheath voltage limiter (SVL) arresters. The voltage rating
is, however, much smaller than that for phase-conductor arresters, for example: 3 and
6 kV arresters are acceptable on 138 kV systems, 9 kV arresters will be needed only
for higher fault current or very long sections of cable. The rating of SVL arresters
for single bonded systems is usually higher than that for cross bonded systems [16]
because of the induced voltage involved. The main duties for SVLarresters are to:
(i) withstand the sheath induced voltage; this is not onerous because the magnitude
is a few tens of volts
(ii) dissipate short circuit current in case of a line fault
(iii) be water and humidity proof.
5.2.5.2 Design improvements
Field experience from these various applications has led to some important improve-
ments in design features which were suggested to overcome problems encountered
on surge arresters:
(i) use of shunt gap module to increase tolerance of ZnO arresters to temporary
system overvoltage, particularly useful for riser-pole, elbow and oil-immersed
arresters [28]
(ii) use of special thermal shields and ventings for pressure relief to avoid
fragmentation which occurs due to thermal shocks following fault currents
[29, 30]
(iii) soldering techniques for mechanical strength [18]
(iv) use of ceramic grading capacitors to control voltage distribution along the
arrester column [19]
(v) better EJ characteristic of the material to reduce size of the arrester and
improve voltage distribution [20]
(vi) special requirements for the working environment and conditions [31, 32].
5.3 Basic properties of ZnO material
5.3.1 Composition and effect of additives
The semiconducting properties of ZnO were very well known [33] well before Mat-
suoka et al. [34] investigated the non-linear voltagecurrent characteristic of ZnO
doped with alkali earth metal oxides. It was found that with appropriate amounts
(0.5 mol%) of additives having an ionic radius larger than that of Zn
+2
(which would
not dissolve in the ZnO lattice and would thus form a segregation layer at the ZnO
grain boundary), the ceramic exhibits a high coefficient of non-linearity on its VI
characteristic. Furthermore, it was shown that the new material was exceptionally
superior compared with conventionally used SiC varistors [35]. The non-linearity
coefficient, , of early ZnO varistors was in the range 2550. In contrast, that of SiC
varistors was between 2 and 7 [36].
In 1971, Matsuoka [37] disclosed a large number of compositions which produce
useful ZnOvaristors. Acoefficient = 50 was obtained when 99.9 per cent pure ZnO
is doped with 1 mol%Sb
2
O
3
and 0.5 mol%of Bi
2
O
3
, CoO, MnOand Cr
2
O
3
[3842].
Adding more dopants has allowed the fabrication of ZnO varistors with large grain
size and the reduction of the breakdown voltage of a single grain boundary [4347].
Paraseodymium oxide rich composition, which does not contain any bismuth, was
used to manufacture ZnO varistors with a two-phase microstructure and improve its
electrical properties such as energy absorption capability, residual voltage and life
performance [48, 49].
It is postulated that the transition metal oxides are involved in the formation of
interfacial states and deep bulk traps, both of which contribute to the highly non-
linear resistivity property. The grain size, hence the breakdown voltage per each
grain boundary, could be controlled with additives such as TiO
2
, Sb
2
O
3
, BeO, Al
2
O
3
,
K
2
O and SiC. The electrical stability has been improved by adding ZrO
2
, Cr
2
O
3
or
Nb
2
O
5
to bismuth oxide rich compositions. Table 5.1 summarises the effect of various
additives on ZnO varistor properties reported in the presently reviewed literature.
5.3.2 Fabrication process
Figure 5.1 shows a diagram of the ZnO surge arrester fabrication process. The
basic material used to manufacture metal oxide varistors is pulverised, very finely
grained and highly pure (99.9 per cent) ZnO with a particle size of 110 m. Several
doping elements and specially prepared powders [43, 50] are added in the form of
fine oxide powders. The concentration of the individual components range from
Table 5.1 Effect of additives on ZnO material electrical properties
Additives Effect on ZnO properties Reference
Bi
2
O
3
+alkali metals: non-linearity appears 34, 37, 40
CaO, CoO, BaO, SrO, MnO highest with BaO
K
2
O inhibits grain growth 52
3 mol% MeO minimum for good non-linearity 41
seed grains: large grain size 43
ZnO +0.5% BaCo
3
breakdown voltage reduced
reduced
Ga
2
O
3
increases 47, 102
large leakage current
ZrO
2
better energy absorption capability 47
SiO
2
low leakage current 47
Pr
6
O
11
substituted to Bi
2
O
3
2-phase microstructure (no spinel) 48, 49
improved VI characteristic
higher energy absorption capability
better life performance
lower residual voltage
Al
2
O
3
lower residual voltage 44, 48
grain growth enhancer
large leakage current
high
Frit glass: 63% PbO, lower leakage current 50
25% Bi
2
O
3
, 12% SiO
2
same
TiO
2
, Sb
2
O
3
grain growth enhancer
lower potential barrier height
high 44, 45, 102
Cr
2
O
3
enhances thermal stability 50
Nb
2
O
5
increased to 60 46
leakage current decreased
higher energy absorption capability
doubles life performance
p.p.m. to per cent [51], and the composition and proportion of additives differ from
manufacturer to manufacturer. The resulting powder is ball mixed in deionised water
[46, 52] or alcohol [50] for 1024 hours, after which a homogeneous aqueous slurry
with small particles is obtained. The homogeneity of the mixture is of immense
importance for the quality of the end product. The slurry is then calcinated at 700
C
[37, 41] in a spray drier for a duration of 30 min [43] to 2 h [37]. Homogeneous
spheroidal granulates of 100 m in diameter are obtained [51] and are compacted, in
the next production stage, under a pressure of 340400 kg/cm. The pressed material
ZnO crystals + selected additives
wet mixing in
a ball mill
homogeneous
aqueous slurry
calcinating in a
spray drier
homogeneity maintained
in granulated form
disc shaping
of resistors
sintering
dense ceramic bodies
addition of electrodes
and protective coating
electrical testing
and checking
assembly
of resistors
Figure 5.1 Simplified diagram of the manufacturing process of ZnO material
is moulded into disc shaped blocks having a diameter of 3100 mm and a thickness
of 130 mm.
The blocks are then sintered in an electric furnace in air or oxygen for one or two
hours under constant temperature between 10001400
C. This heat treatment has

the effect of further densifying the compacted powder into a solid ceramic body. The
temperature and duration of the process are very critical for the characteristics of ZnO
varistors. It is reported [51] that in this process the adjacent powder particles are united
by means of diffusion, and subsequently grow into large grains. In the next stage, the
dense ceramic bodies are furnace cooled at a rate of 50200
C/h at the end of which

they are coated with a collar material (glass, epoxy) on the peripheral surface to
protect them from the environment and prevent flashover. Metallic electrodes (silver,
gold, copper, aluminium) are also deposited on each face of the blocks in the last
production stage. Finally, before assembling, routine checks are performed to verify
the resistor electrical properties
5.3.3 Microstructure
The grain size, hence the breakdown voltage per each grain boundary, could
be controlled with additives such as TiO
2
, Sb
2
O
3
, BeO, Al
2
O
3
, K
2
O and SiC.
intergranular
layer
ZnO grains
a
b
10
Sp
ZnO
Figure 5.2
a Schematic representation of ZnO material microstructure
b Typical SEM micrograph of ZnO material. (Courtesy of
GREUTER, F., PERKINS, R., ROSSINELLI, M., and
SCHMUCKLE, F.: The metal-oxide resistor at the heart of modern
surge arresters, ABB Rev. Surge Arresters, 1989, 1)
The fabrication process results in a polycrystalline material composed of semicon-
ducting ZnOgrains surrounded by a thin layer of intergranular materials (Figure 5.2a).
The microstructure has been examined by means of several optical techniques such
as x-ray diffraction, electron microprobe, scanning electron microscopy (SEM) and
transmission electron microscopy (TEM) [41, 42, 45, 47, 48, 50, 52]. The resulting
photomicrographs as reproduced on Figure 5.2b reveal that the microstructure of
ZnOvaristors is made up of minute ZnOcrystals or grains, approximately 5 to 20 m
in size. These grains are separated physically from each other by thin layers of
intergranular materials of approximately 0.2 mm in thickness. The nature and com-
position of these layers depends on the additives used for doping ZnO. Very sensitive
measurements [40, 42] showed that the intergranular layer can be as small as 2 .
Spinel particles of Zn
7
Sb
2
O
12
which are thought to play a passive electrical role,
were also observed [39, 41, 48] in the intergranular phase and their origin was traced
back to the presence of Sb
2
O
3
additive in the ZnO composition during the sintering
process. Inadditiontothe spinel crystallites, a pyrochlore phase (Bi
2
{Zn
4/3
Sb
2/3
}O
6
)
was identified in the intergranular layer and a bismuth rich phase at the triple point
which is formed with three or more ZnO grains [47].
5.3.4 Conduction mechanism in ZnO varistors
There are two distinct regions of conduction: the prebreakdown region where
the varistor behaves almost in a linear fashion (the current is thermally activated
[39, 47, 53] which is described by the Schottky emission over the potential barrier
[54]) and the breakdown region where high non-linearity appears and more complex
conduction processes take place. An overview of these studies is given below.
5.3.4.1 Progressive evolution of the breakdown conduction theory
Since the early studies [55], the origin of the non-linearity of conductivity has been
attributed to the regions situated in the ZnO grain boundaries. This was soon con-
firmed by thin film experiments [52] and photoconductivity measurements [39].
Several models describing the conduction phenomenon have been developed to give
an account of the non-linear electrical properties of zinc oxide varistors (Table 5.2).
On the other hand, there are other known mechanisms which have been ruled out due
to incompatibility with ZnO properties. The diffusion theory has been excluded [55]
since the current in ZnO is strongly voltage dependent. The avalanche breakdown
mechanism has also been rejected [27, 39, 53, 55] because it results in a positive
temperature coefficient; ZnO is known to exhibit a strong negative temperature
coefficient.
The first model of the ZnO conduction mechanism was based on space charge
limited current (SCLC) [55]. Although yielding high coefficients of non-linearity
[56], this model is inappropriate to ZnO varistors because of the passive role of
the intergranular layer in determining the breakdown voltage [39, 53]. Subsequently,
a simple tunnelling process [39, 47] and a two-step transport mechanism [52] that
consisted of steps of electron hopping [57] and electron tunnelling were suggested.
None of these models could account for the high observed in ZnO varistors.
Most of the accepted models assume a common description of the band structure in
the vicinity of the intergranular layer region. The potential barrier is assimilated to an
assembly of two back-to-back Schottky junctions. The Schottky emission model [27]
was extended to include thin intergranular layers, electron traps in the intergranular
layer [41] and to consider the depletion regions formed in the ZnOgrains [58, 59]. The
parallel currents in the bismuth rich intergranular layer are important when the varistor
Table 5.2 The various models considered to explain the conduction mechanism in
ZnO varistors
Model Feature Reference
Diffusion theory ZnO current strongly dependent 55
on voltage (rejected)
Avalanche theory ZnO temperature coefficient 39, 55
is negative (rejected)
SCLC theory high but intergranular layer 53, 55, 56
is electrically inactive (rejected)
Simple tunnelling ZnO capacitance is function of 41, 43, 47, 52
voltage and low (rejected)
Electron hopping and tunnelling low (rejected) 52, 57
Schottky emission linear conduction: suitable for 39, 53, 54
prebreakdown regime
Double Schottky barrier (DSB) linear conduction 27
without intergranular layer
Parallel currents important in low current regime 60
DSB with thin layer low 41
DSB with thin layer and better 41
electrons traps
Depletion layer and holes and high but does not account 53, 59, 61
tunnelling for negative capacitance
Depletion layer and holes high 56, 61, 62, 63
Depletion layer and high and improved results 58
recombination of holes
is operated in the lowconduction regime [60]. Afurther improvement of the depletion
layer model is achieved by introducing a process of creation, by hot electrons [56]
and impact ionisation [58, 59, 61], of holes [62, 63] in the depletion layers which are
inversely polarised compared with the double Schottky barrier. Under the effect of
the field, the created holes are collected at the interface and narrow the potential hill
at the intergranular interface and favour tunnelling [53]. However, tunnelling was
rejected and the presence of holes was associated at the interface with lowering of
the potential barrier height [60, 64]. As a result, a sharp increase of current for a very
small increase in applied voltage is obtained. Afurther step in the formulation of the
conduction mechanism takes into account recombination of the created holes [58].
5.3.4.2 Description of the high conduction model
Consider the situation where two ZnO grains meet to form a grain boundary. The
thickness of the intergranular layer is assumed very thin, so that it can be neglected.
Gap states are then formed at the interface and trap a negative charge. This charge
is screened by the ionised donors ZnO. As a result, depletion layers appear on both
sides of the barrier. The resulting double Schottky potential barrier has a height of the
order 13.2 eV and the total spatial extent of the potential barrier is about 0.2 m,
representing the depletion layer width (X
L
+X
R
), which is very large compared with
the thickness of the intergranular layer (2 ). The formof the potential barrier depends
on the interface density of states, the donor density and the applied voltage V.
The main features of this model are:
(i) parabolically decaying potentials on both sides of the interface
(ii) asymmetric distortion of the barrier under an applied voltage V
(iii) decay of the barrier height with voltage
(iv) at large applied voltages, a very large potential drop (5 eV) within very short
distances (0.1 mm); therefore, very large electric fields (1 MV/cm) are
reached in such conditions.
The last feature is of particular importance since under such magnitudes of electric
field, generation of hot electrons process takes place [56, 58]. Under a very high field,
electrons are injected over the potential barrier into the positively biased depletion
region. The strong electric field easily accelerates the particles to attain high kinetic
energy. Minority carriers (holes) are then created by impact ionisation and are swept
towards the interface. As a result, the barrier height is lowered which increases the
current sharply. Further increase of current is obtained if recombination of the swept
holes with trapped electrons at the interface is considered.
5.4 Thermal performance of ZnO surge arresters
5.4.1 Background
The performance of surge arresters in power systems is determined on the one hand
by the electrical and thermal properties of the varistors, and on the other by the design
and installation of the arrester.
Compared with SiC gapped arresters, ZnO gapless metal oxide arresters offer
a protection closer to the ideal. However, because these arresters contain no gaps,
a leakage current flows through the material at working voltages which causes power
losses and heating of the ZnO elements. This can be dangerous to the stability of
the arrester, particularly in the low conduction regime where the VI characteristic
of ZnO material is very sensitive to temperature. Badly dimensioned arresters are
exposed to the risk of thermal runaway. Afurther practical problem relating to power
system usage concerns pollution of arresters, which can cause abnormal voltage and
current distribution in the arrester in both steady state and transient conditions.
5.4.2 Heat dissipation capability and thermal stability of ZnO surge
arresters
In surge arresters, the ZnO valve elements are generally located in an environment
consisting of gas or solids which will limit cooling of the valve elements when
submitted to their continuous operating voltage. As can be seen on Figure 5.3, for the
30 C
ambient temp.
40 C
50 C
60 C
70 C
25
20
15
10
5
0
1 10 100 1,000
resistive current, A peak
v
o
l
t
a
g
e
,

k
V

p
e
a
k
Figure 5.3 Temperature dependence of ZnO voltage current characteristic
same applied voltage level, any temperature rise would increase the current because
of the high sensitivity to temperature of the VI characteristic in the low conduction
regime. Asort of positive feedback will occur and may cause damage to the arrester
if the material heat is not dissipated. For this reason, thermal stability is one of the
most important application criteria for metal oxide surge arresters. It is frequently
analysed with the help of a heat loss input heat balance diagram [6574].
The thermal stability of ZnO surge arresters is affected by ambient tempera-
ture and heat dissipation capability [65, 66], impulse degradation [67, 68] and ageing
[6971]. To obtain thermal stability, the electrical power dissipation in the ele-
ment must be balanced against heat output to the environment. Near the thermal
equilibrium, it is possible to express the thermal dissipation capacity Q of a surge
arrester as:
Q = C
T
(T T
a
) (5.2)
where T is the temperature of ZnO valve elements, T
a
the ambient temperature and
C
T
the thermal dissipation factor. The heat generation, P, which is voltage and
material composition dependent, may be approximated by:
P = Ae
(W
c
/kT )
(5.3)
where W
c
is the activation energy, k = 0.86 10 eV/K (Boltzmann constant), T is
the temperature of the material and A depends on the applied voltage level and the
physical dimensions of the valve elements.
The above curves are shown schematically in Figure 5.4. For an ambient temper-
ature T
a
and an applied voltage V, the two curves intersect at two points Xand Y. The
lower point X is at a stable operating temperature T
x
and is referred to as the lower
stability point. The upper point Y is also at a stable operating temperature T
y
and is
called the upper limit stability point. At these temperatures, the power input equals
the power output.
Y
X
T
a
T
x
T
f
temperature T
y
P
P
Q
Q
P
p
o
w
e
r
,

W
Figure 5.4 Thermal stability diagram for ZnO arresters
Q heat loss at ambient temperature T
a
Q
heat loss at ambient temperature T

f
P power (heat) dissipation at voltage V
1
P
power (heat) dissipation at voltage V

2
with (V
2
> V
1
)
P
power (heat) dissipation at voltage V

3
with (V
3
< V
1
)
The valve element temperature always settles at the stable operating point,
as long as the initial valve temperature does not exceed the instability threshold.
Following any temperature rise up to T
y
, produced for example by the absorption of
an overvoltage energy, the valve element state will revert to the lower stability point X.
For temperatures higher than T
y
, the valve element would have an ever increasing
power dissipation which would be much higher than the heat dissipation capacity of
the arrester. Therefore, when the arrester is operated above point Y, the arrester is no
longer able to dissipate the heat introduced into the ZnO valve elements. As a conse-
quence, the temperature will increase and thermal runaway will occur. The arrester is
designed for steady state operation at temperature T
x
. The energy difference between
points X and Y is called the allowable surge energy. It can be seen, in Figure 5.4,
that there is a maximum temperature range (T
y
T
x
) for the ambient temperature T
a
,
beyond which the valve element would be unstable. It is noted that this range is not
a unique value, but is highly dependent upon the ambient temperature and on the
applied voltage. This is a simplified representation which does not include various
limiting factors, such as transient changes in the rate at which the energy is delivered
and dissipated.
For the ambient temperature T
a
, the maximum energy discharge capacity E
d
(T
a
)
is given by [69, 70]:
E
d
(T
a
) = mC
T m
(T
y
T
x
) (5.4)
where C
T m
is the heat capacity per unit mass 0.54 J/g/
C and m is the mass of the

valve element.
It should be noted, however, that the energy capacity of an arrester may be lower
in practice due to thermal shock or puncture. The larger the difference between T
y
and T
x
, the larger the capacity of the arrester. If the surge energy does not exceed the
allowable value, then the arrester temperature will return to T
x
. However, the time
taken to return to stable conditions is important. Up to one hour may be needed for
a surge arrester to recover its initial conditions after absorption of two rectangular
surges of 600 A amplitude and 2 ms duration [75]. In addition, the upper stability
point is the steady state and not the transient limit, which is higher [71, 72]. For
one ZnO product, it was found [76] that the allowable temperature rise due to surge
current is 110
C if the arrester is operated at an alternating voltage which produces

1 mAcurrent through the valve elements. Consequently, Figure 5.4 can only be used
to evaluate thermal instability arising from steady state conditions, where there has
been a sufficient amount of time allowed to stabilise both valve element and arrester
housing temperature, and is of limited use when thermal instability due to sudden
energy input, switching or lightning surges or even TOV is considered [71].
5.4.3 Thermal runaway
The term thermal runaway as defined by both existing IEC [77] and JEC standards
[78] is used to describe a situation culminating in failure if the ZnO arrester is
operated above the upper stability point Y. However, there are a number of factors
which influence the position of point Y.
5.4.3.1 The ambient temperature
Increasing the ambient temperature lowers the dissipation curve Q and hence
decreases the maximum energy discharge capacity [66, 70]. For an ambient
temperature of T
f
, the upper and lower stability points X and Y coincide with each
other and there is zero energy capacity.
5.4.3.2 Ageing and degradation of the arrester
An ageing characteristic that leads to increased power dissipation also decreases
the maximum energy capacity [65, 66, 70]. However, an ageing characteristic,
which shows decreased power dissipation, increases the maximum energy discharge
capacity [70]. In the first case, which is more realistic, the heat generation P is
expressed as [67, 69]:
P = P
0
(1 +h
t ) (5.5)
where P
0
is the heat generation at the initial stage, h is constant and t is time.
The degradation in metal oxide surge arrester elements increases the leakage
current and hence the heat generation P. The effect of ageing is to shift the stabil-
ity point X to higher temperatures, therefore, degrading it more. Consequently, the
thermal runaway threshold is shifted towards lower temperatures [69, 70], and the
margin T
y
T
x
becomes smaller. The ageing subsequent to the application of voltage
will make the heat generation curve P move to P
which has a new operating point.

This evolution of the operating point on the straight line continues until the stability
points X and Y are at the same position.
5.4.3.3 Applied voltage level
Thermal runaway of metal oxide elements at high AC stress is caused by a positive
feedback process due to the temperature dependence of the resistive component of
leakage current. Moreover, the powervoltage characteristic is highly non-linear so
that a small increase in voltage will result in a considerably higher power. The effect
of this is to shift curve P of Figure 5.4 upwards. This displacement will be of similar
effect as that described for the ageing effect.
5.4.3.4 Heat dissipation capability of the housing
From the heat dissipation point of view, a small unit is better than a complete arrester.
The main reason for this is related to the heat flow along an arrester column. Heat
dissipation is reduced at the centre of a long column [68]. As a consequence, the tem-
perature of the middle section of an arrester is higher than those at the ends [65, 79, 80].
Up to six times improvement can be achieved if cooling metal fins are used [66].
Improving the heat dissipation capability of the arrester means that the curve Q
on Figure 5.4 will be lifted to higher values, hence the thermal property of the arrester
is improved. For this reason, conducting tests on pro-rated units or model units must
ensure that the unit reproduces the heat dissipation and thermal characteristics of the
complete arrester to which the tests are to be related [68].
5.4.4 Thermal runaway critical condition
The critical condition of thermal runaway is evaluated analytically based on the
temperature dependence and the heat loss of the arrester [65]. As shown above,
thermal runaway occurs when the lower stability point X moves to the upper stability
point Y, i.e. the heat dissipation curve Q is tangential to the generation curve P.
At that instant, we can write the critical conditions,
P = Q (5.6)
and
P
T
=
Q
T
(5.7)
Using Equations (5.2) and (5.3), the critical values of temperature T
cr
and power P
cr
[65] are:
T
cr
= T
a
_
1 +
kT
a
W
2
c
(W
c
+2kT
a
)
_
(5.8)
and
P = 4
C
d
D
(T
cr
T
a
) (5.9)
where C
d
is the heat dissipation coefficient W/cm/K and D is the diameter of
elements.
5.4.5 Dynamic stability of ZnO surge arresters
The energy capacity of a surge arrester may also be affected by thermal stability
limitations, because the surge amplitude, duration and energy have critical values
above which thermal runaway is triggered. Nishikawa et al. [76] quoted an allowable
surge energy density of 350 J/cmfor an ambient temperature of 40
Cand a maximum
2 ms duration for a 1.25 p.u. temporary overvoltage.
5.4.6 Simulation of thermal characteristics of ZnO surge arresters
Thermal stability tests are very expensive and need careful measurements with highly
sensitive equipment [74]. It is therefore desirable to have a simple circuit model that
will reproduce the thermal properties of the arrester. Lat [71] proposed an electrical
analogue model of the thermodynamic behaviour of the arrester assembly. The elec-
trical equivalence is based on representing power flow as a current and temperature
as a voltage. The circuit as shown in Figure 5.5 takes account of the electrical power
T
E
T
H
C
H
C
E
W
E
C
E
, C
H
: thermal capacities of the valve element
and adjacent housing, respectively
R
EH
: thermal resistance from element to external
surface of the housing
R
HAO
: thermal resistance from housing to ambient,
radiation and natural convection components
R
HAF
: thermal resistance from housing to ambient,
forced convection temperature
T
E
: valve element temperature
T
H
: housing temperature
T
A
: ambient temperature
W
E
: electrical power input to valve element
W
S
: heat input due to solar radiation
W
S
T
A
R
HAF
R
HAO
R
EH
Figure 5.5 Thermal model of a ZnO arrester. (Courtesy of LAT, M.V.: Thermal
properties of metal oxide surge arresters, IEEE Trans. PAS, 1983,
PAS-102, (7), pp. 21942202, 2003 IEEE)
input, the solar radiation and the ambient temperature. It is claimed that this circuit
is valid for steady state and transient behaviour, giving results which agree closely
with experimental data [71, 81]. Another circuit [82], which considers the same input
components, is a ladder network whose element characteristics are determined exper-
imentally [83]. An important factor, the heat dissipation transfer resistance can be
studied using the three heat transfer mechanisms involving radiation, conduction and
convection [82].
5.5 Degradation and ageing of ZnO surge arresters
The terms varistor degradation and varistor stability are often used to describe the
electrical state of a varistor relative to its past or future state when under the influence
of external stress. The degree of degradation is a good indication of varistor reliability
and is usually used for predicting the life span of ZnO surge arresters.
Severe electrical degradation causes a large increase in leakage current at low
voltages. The degradation process affects mainly the varistor leakage and prebreak-
down conduction and not the behaviour of the varistor at voltages higher than the
breakdown voltage.
Most degradation studies have investigated the changes in the voltage level which
cause a current of 1 mA to flow through the ZnO elements at ambient temperature,
after removing the applied degrading stress. Others [84, 85] have used the thermally
simulated currents (TSC) method. The TSC currents are known to be associated
with stored charge within the varistor which is released at a constant heating rate
e.g. 4
C/min, after a biasing direct voltage has been removed. Measurement of the
TSC spectrum gives useful information on the amount of trapped carriers, the polari-
sation of dipoles and space charge polarisation caused by ion migration which are all
dependent on the degree of degradation [64, 84, 86].
Another method [86] uses changes incurred to the potential barrier height. It is now
established [64, 86] that the barrier height, which determines the amplitude of current,
is decreased after each ageing cycle (continuous applied voltage at high temperature)
and after applying a fast high current surge (4/10 s, 250 kA). The consequence of a
decreasing barrier height is the increase in leakage current. The degree of degradation
has also been evaluated through partial discharge detection [87].
5.5.1 Differences between degradation and thermal runaway
It is useful to note two differences between normal degradation and the thermal
runaway condition [66]:
(i) In degradation mode the gradual increase of current with time is different from
the thermal runaway increase. It has a slower rate of change and shows little
dependence upon temperature. The rate of change of current due to degradation
for most products was found [65, 69, 88] proportional to the square root of time:
i
t (5.10)
whereas this change is in exponential form during the thermal runaway
process [65]:
i e
t
(5.11)
(ii) The change in the VI characteristic, which is caused by degradation, is pre-
served even after removing the stress. In thermal runaway, however, recovery
may be possible after cooling down.
5.5.2 Factors affecting rate of degradation
Anumber of factors have been identified to affect the rate of ZnO degradation.
5.5.2.1 Composition and fabrication process
The physical explanation of degradation and ageing is related to electromigration of
oxygen ions perpendicular to the grain boundaries which lowers the potential barrier
height [86]. Reference 64 gives a comprehensive summary of the various physical
explanations of the degradation process. The amount of additives [84, 89] and the heat
treatment [85], with which ZnO is produced, affect the non-linearity coefficient of
the material. Products with high coefficients degrade more because of the relatively
high currents which flow in the material. Furthermore, it is observed that the long
term leakage current characteristic depends on the composition and manufacturing
process of the elements even if they have similar VI characteristics at the initial
stage.
5.5.2.2 Homogeneity
Evidence of non-uniformity of ZnO material has been shown by means of dot
electrodes [90, 91] and an infra-red radiation thermocamera [91]. Heterogeneous
conductivity leads to inhomogeneous energy dissipation. The resulting local over-
heating may then lead to an irreversible change of varistor properties after surge
stress, because of localised paths of current.
5.5.2.3 Ambient temperature
Since the current is thermally activated in the low conduction region, there is a strong
dependence of the VI characteristic on the temperature, so that high temperatures
cause high currents to flow, and this process is known to accelerate the ageing of the
sample. As an example [81], continuous operation under 135
C for one day degrades

the ZnO arrester by a decrease of 12 per cent in the voltage level, V
1mA
, which
produces a current of 1 mAthrough the arrester.
5.5.2.4 Working voltage
The highly non-linear characteristic of ZnO has permitted the reduction of the pro-
tective level. However, continuous application of the working voltage progressively
degrades the ZnO elements and may end up in thermal runaway after some operating
time [66]. The rate of degradation is largely affected by the magnitude [68, 92] and
type (direct or alternating) [89] of the applied voltage. Testing ZnOat higher voltages
and temperatures results in accelerated ageing.
5.5.2.5 Design and physical arrangements of ZnO surge arresters
The main features that could affect the degradation rate are the heat dissipation capa-
bility of the complete housing [66, 68, 93] and the voltage distribution along the
arrester column. As will be discussed later, the voltage distribution along an arrester
column is distorted by stray capacitances, and may enhance the non-uniform tem-
perature distribution, which in turn will accelerate the degradation of the arrester.
Besides, extensive laboratory measurements [69, 70] showed that degradation could
affect the collar material (rim) severely, which may overshadow the ageing of the
ZnO body. New organic materials are, however, more efficient as far as corona [94]
and long term stress [70] are concerned.
5.5.2.6 Pollution of surge arresters
The effect of pollution is related to that of the dry bands which cause partial discharges
and corona effects. These types of electrical stress cause surface erosion and localised
damage to the collar material protecting the ZnO blocks [94]. The partial discharges
cause a drop in oxygen content of the gas surrounding ZnO blocks together with
simultaneous formation of corrosive gases and this was found to accelerate ZnO
degradation [95]. Moreover, a white deposit of silicone, whose origin is not explained,
is observed [96] in contaminated arresters where corona activity is present.
5.5.2.7 Environment
The working environment of ZnOprotective devices, which may be gaseous (air, SF
6
)
or liquid (transformer oils) and solid (sidewall), plays an important role in the life
span of the material. ZnO elements in SF
6
are degraded more than those in air [69].
However, adding to SF
6
a small amount (3 per cent) of oxygen, the presence of which
seems to play an important role in the chemical process, ensures less degradation [95].
Water vapour and moisture presence also cause an increased leakage current at low
temperatures [95, 97]. The relativelyhighworkingtemperature of ZnOvalve elements
in transformer oils (95
C) degrades these arresters at a faster rate. For an applied

voltage V = 0.7 V
1mA
, the arrester life has been estimated to be as short as 0.36
years [98].
5.5.2.8 Surge absorption capability
In general, just after the application of a surge, there is a measurable degradation
of ZnO electrical properties. Three types of overvoltages are of interest; switching
surges, temporary overvoltages and lightning surges:
(i) Switching surges Following the absorption of a high energy surge, the cooling
conditions of the arrester have very little effect [90]. Experiments [4, 66, 68]
showed that degradation is more severe under high current surge than high
energy surge. Degradation resulting from energy surge absorption is there-
fore negligible in practice and has almost no effect on the life of the elements.
However, high energy surge applied to ZnOelements exhibiting localised spots
of conduction could melt the material [66, 99]. Switching surge degradation is
pronounced under unidirectional surges but can be reduced to some extent by
applying surges of opposite polarity [67]. Only 1 per cent degradation was
observed [68] on the direct voltage which produces 1 mA (V
1mA
), after the
discharge of positive and negative switching surges. After absorbing two rect-
angular surges of 600 A and 2 ms duration followed by power frequency, less
than 4 per cent maximum variation in voltage, V
1mA
, was observed [4], and
the ZnO element could return to its initial working conditions of voltage and
temperature after just one hour. As can be seen on Figure 5.6a, V
1 mA
undergoes
as little as 5 per cent change when the arrester is subjected to 1000 switching
surges of 600 Amagnitude and 2 ms duration [4].
(ii) Temporary overvoltage (TOV) No change is measured after TOV absorption
[68]. Durability of MOV against TOV depends mainly on global structural
uniformity [100]. Temperature effects, however, may lead to thermal run-
away if the surge parameters are higher than specific allowable values given in
manufacturers specifications (Figure 5.6b) [76].
(iii) Lightning discharge Degradation due to discharge of the surge energy largely
depends on the peak value of the discharge current rather than the discharge
energy. The presence of AC stress does not enhance nor restore the surge
degradation. Afew high current surges degrade ZnO arresters more than many
high energy surges of lower current; two (8/20 s) surges of 65 kAamplitude
caused the leakage current to increase by 50 per cent [68]. Figure 5.7a shows the
changes in V
1mA
as a function of the number of applied surges for two current
amplitudes (40 kA and 100 kA) of a fast surge (4/10 s) [4]. The allowable
number of surges decreases when the amplitude of the surge increases; e.g. a
life of 50 years allows 130 surges of 50 kA or only 30 surges of 100 kA [89].
Figure 5.7b shows the overall effect of fast surge absorption on the VI curve.
As may be observed in the figure, the relative degradation is more severe for
the lower voltages of the VI characteristic.
5.5.3 Destruction mechanism
Ultrasonic methods that are sensitive for detection of physical defects such as small
voids or cracks, have been used to check the physical condition of overstressed
ZnO elements [91]. There was no microstructure difference which could be revealed
between newand degraded (aged) blocks when scanning electron micrography (SEM)
was used [69]. However, when a high energy such as those of excessive direct or
alternating voltage levels or a very long surge overvoltage (>100 s) is absorbed, a
puncture will occur in the ZnO element [99]. Such failure may be caused by cur-
rent concentration in non-uniform elements. At 820
C, the main additive Bi

2
O
3
melts. ZnO grains along the localised current path may then melt as a result of high
temperature.
0
2
4
6
1 10 100 1000
number of discharges
10000
(2ms, 400A)
(2ms, 600A)
1.4
0.9
0.02 0.2 2
2
8
r
e
l
a
t
i
v
e

c
h
a
n
g
e

i
n

V
1
m
A
,

%
V
(
p
e
a
k
)
/
V
1
m
A
,

%

1.9
0.4
a
allowable duration, s b
Figure 5.6
a Degradation of ZnO surge arresters after absorption of switching
surges. (Courtesy of NISHIKAWA, H., SUKEHARA, M., and
YOSHIDA, T., et al.: Thyristor valve gapless arrester element
lifetime performance and thermal stability, IEEE Trans. PAS,
1985, PAS-104, (7), pp. 18831888, 2003 IEEE)
b Degradation of ZnO surge arresters following temporary over-
voltages. (Courtesy of NISHIKAWA, H., SUKEHARA, M., and
1985, PAS-104, (7), pp. 18831888, 2003 IEEE)
In the case of high current fast surges (time to crest less than 50 s), a cracking
has been observed. The heat, which is generated within a very short time, may not
transfer sufficiently quickly to other parts of the element. This thermal expansion
stress is transformed into a mechanical one and produces the crack. This localised
destruction may be healed by cutting the material around the defect [100].
2
0
2
4
6
8
0.1 1 1000 100 10
number of discharges
(410s, 100kA)
(4 10s, 40kA)
100
90
80
70
60
50
0
10
2
10
3
10
4
20C
(0)
10
5
(12)
(24)
amperes
r
e
l
a
t
i
v
e

c
h
a
n
g
e

i
n

V
1
m
A
,

%
a
p
p
l
i
e
d

v
o
l
t
a
g
e

r
a
t
i
o
,

A
V
R
%
a
b
Figure 5.7
a Degradation of ZnO surge arresters caused by lightning impulses.
(Courtesy of HAHN, E.E.: Some electrical properties of zinc oxide
semiconductor, J. Appl. Phys., 1951, 22, (7), pp. 855863)
b Changes in VI characteristic following surge absorption.
(Courtesy of NISHIKAWA, H., SUKEHARA, M., and
1985, PAS-104, (7), pp. 18831888, 2003 IEEE)
5.6 Life estimation of ZnO surge arresters
5.6.1 Long term accelerated ageing tests
As described in the previous sections, the continuous applied voltage on ZnO
surge arresters during their service duty causes a permanent flow of current which
affects the arrester qualities in the long term. As a consequence of this degra-
dation, the life span during which the arrester will provide efficient protection
will be limited. In order to assess this life span, long term accelerated ageing
tests have been performed [4, 65, 66, 69, 70, 75, 76, 84, 88, 89, 92, 96]. These tests
are normally run for long periods (approx. 1500 hours) under a constant elevated
voltage level ( knee of conduction) and at a constant temperature which is con-
siderably higher than that anticipated in actual use (70150
C). The current I or

power loss P is monitored versus time t until a given value is reached. This
value is often referred to as the end of life criterion. Various criteria have been
adopted:
(i) when power loss reaches a critical value such as 1 W/cm of length [96]
(ii) when the leakage current reaches a fixed value, e.g. 100 mA[76]
(iii) when power loss reaches a value that is double the initial value measured at the
beginning of the test [70, 84, 88].
During these long term tests, it has been observed that the Pt and It curves go
through three stages: (i) a stable period of working life where the arrester does not
show any change in its characteristics; then comes (ii) a degradation period where
P and I increase with time; and after this degradation stage thermal runaway is
preeminent; (iii) a sudden sharp increase of I and P takes place.
It was suggested [66] that the end of life criterion should be chosen at the begin-
ning of the degradation period. Figure 5.8 shows schematically the various shapes
measured during long term tests.
The difference in the long term accelerated ageing curves is essentially due to
the difference in composition and manufacturing process of the tested ZnO products
[70, 89]. Dissimilar shapes have been recorded for ZnO products showing identical
initial VI characteristic but having different compositions [89]. Another important
factor, which could play a key role in the curve shape, is the nature of the collar
material. It is known that the collar material can overshadow the characteristics of
the ZnO material when it is of unsuitable nature [69, 70].
A
B
C
D
E
m
A

o
r

W
Figure 5.8 Measured typical shapes of current/power versus time curves during
long term accelerated ageing tests
5.6.2 DakinArrhenius plots of life span
The theory developed by Dakin [101] has been widely used for ZnO surge arrester
life prediction [4, 65, 66, 69, 70, 76]. On the basis that the ageing process of ZnO is
a chemical reaction, its duration may be described by the Arrhenius equation:
log =
A
T
+B (5.12)
where is the life span, A, B are constants depending on the material and stress and
T is the absolute temperature.
The time taken for power or current to reach the end of life criterion is a measure
of the tested ZnO element lifetime, , at the test applied voltage and temperature.
If lifetimes obtained in this way at various temperatures are plotted on a log scale
versus the reciprocal of the test absolute temperature, DakinArrhenius life plots
can be obtained [65, 69]. Extrapolation to lower temperatures allows us to obtain the
life span of the ZnO surge arresters under normal working conditions. Figure 5.9
shows a typical example of life plots. Most studies [75, 96, 97] have indicated that
the measured current and power dissipation data agree with the DakinArrhenius
equation. However, there is a third component to consider in P and I equations; that
is the number of surges that the arrester absorbs during its working service life [67].
Subsequently, however, it has been clearly shown [70] that the above equation could
be validated for ZnO only within a limited range of temperature (115175
C) since
10
7
100 YR
20 YR
10 YR
1 YR
ambient temperature, C
10
3
10
4
10
3
10
4
l
i
f
e

h
o
u
r
s
70%
70
75%
80%
85%
90%
Figure 5.9 Typical DakinArrhenius life plots. (Courtesy of BRONIKOWSKI,
R.J., and DuPONT, J.P.: Development and testing of MOVE arrester
elements, IEEETrans. PAS, 1982, PAS-101, (6), pp. 16381643, 2003
IEEE)
the data for 70
C were significantly lower than those expected by extrapolation of

the data from higher temperatures.
5.6.3 Alternative methods of life estimation
The extrapolation to lower temperatures has always been treated with caution [84],
and other methods [65, 67, 69, 70, 102] of estimating life have been adopted:
(i) Kirkby et al. [70] proposed that the energy capacity determined from the
thermal stability diagram would be of more value to the user than a predicted
uncertain life.
(ii) Tominaga et al. [65] and Vicaud [69] suggested an analytical model in which it
is assumed that the power P and current I are proportional to the square root
of time:
P = P
0
(a
t +b) (5.13)
I = I
0
(c
t +d) (5.14)
where P, I are power and current at time t , P
0
, I
0
are power and current at start
of test, b, d are constants and a, c are determined fromthe DakinArrhenius law:
a = a
0
e
(W
c
/kT )
(5.15)
c = c
0
e
(W
c
/kT )
(5.16)
with k Boltzmann constant =0.8610
4
eV/K, W
c
activation energy 0.5 eV
for ZnO and a
0
and c
0
constants.
If the end of life criterion is defined, the combination of the above equations easily
yields the life span. However, experience showed that the development of Pt or It
could deviate from a square root law and exhibit a more complex behaviour.
5.7 Test procedures for the characterisation of ZnO arrester
Modelling of ZnO surge arresters uses experimental data acquired over a wide range
of current and voltage amplitudes and types. Such tests include accurate recording of
the voltage and current traces as well as the measurement of the VI characteristic.
There are two distinct regions of conduction in ZnO arresters
5.7.1 Prebreakdown regime of conduction: AC and DC tests
The low current region of the VI curve is determined from direct or alternating
voltage tests. Current magnitudes up to 10 mAcan be used without damage to the ZnO
material. However, currents above 20 mAcan cause too much heat dissipation which
can lead to irreversible degradation of the arrester. Alaboratory circuit for testing ZnO
surge arresters withalternatinganddirect voltages incorporates suitable power sources
and adequate transducers, with simple dividers and shunt resistors giving sufficient
accuracy for the determination of the VI curve. For 50 Hz alternating voltages, the
20
15
10
5
0
5
10
15
20
v
o
l
t
a
g
e
,

k
V
0.25
0.2
0.15
0.1
0.05
0
0.05
0.1
0.15
0.2
0.25
c
u
r
r
e
n
t
,

m
A
15
10
5
0
5
10
15
0 10 20 30 40
v
o
l
t
a
g
e
,

k
V
0.15
0.1
0.05
0
0.05
0.1
0.15
c
u
r
r
e
n
t
,

m
A
I
V
V
I
0 10 20 30 40
time, ms
time, ms
Figure 5.10 Typical ACrecords of voltage andcurrent obtainedinthe prebreakdown
region
current waveshape changes significantly around the knee of conduction as illustrated
in Figures 5.10 and 5.11 and in order to compare DC and AC characteristics, only the
peak resistive component of the AC current is used in the AC VI curve.
5.7.2 Breakdown regime of conduction and up-turn region: impulse tests
In constructing the VI curve, switching impulses of up to 2.5 kAcan be used since
they are within the energy requirements of the arresters. In the up-turn region with high
magnitude currents up to 100 kA, fast or lightning impulses are commonly applied
depending on arrester class and type. These fast impulses have lower energy content
compared with switching impulses, hence their suitability for high current amplitude
testing. Figure 5.12 shows a typical laboratory arrangement for testing ZnO arresters
with impulse currents. With suitable setting of the current impulse generator, the test
loop should be as short as possible to minimise stray inductance effects. The voltage
measurement requires transducers with a fast response time (usually less than 20 ns)
30
20
10
0
10
20
30
0 10 20 30 40
time, ms
v
o
l
t
a
g
e
,

k
V
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
c
u
r
r
e
n
t
,

m
A
V
I
Figure 5.11 Typical AC records of voltage and current obtained in the breakdown
region (around the knee of conduction)
impulse
generator
ZnO
current
shunt
voltage
divider
v i
Figure 5.12 Simple layout for impulse testing of ZnO surge arresters
andnoovershoot ontheir transfer characteristic. The transducer shouldbe locatedvery
close to the arrester to avoid induced spikes [103]. Current measurement also requires
specialised tubular current shunts or fast response current transformers. Figure 5.13
shows some typical fast impulse voltage and current records obtained on a 15 kV
rated surge arrester.
Figure 5.14 shows a typical measured VI curve for polymeric surge arresters
using the methods described above. As can be seen, the VI characteristic is highly
non-linear in the breakdown region above 10 mA.
5.7.3 Voltage distribution along arrester columns
5.7.3.1 Effect of stray capacitances
At working power frequency voltages, the resistive component of current is very
small compared with the capacitive component. The stray capacitances then play an
important role in determining the voltage distribution [104]. The voltage distribution
along an arrester is not uniform because of the effect of the stray capacitances to both
the high and lowvoltage electrodes. This non-uniformity should be kept to a minimum
in order to avoid damage to the upper parts of the arresters which are inevitably more
50
40
30
20
10
0
10
0.5
0.4
0.3
0.2
0.1
0
0.1
time, s
r
e
s
i
d
u
a
l

v
o
l
t
a
g
e
,

k
V
d
i
s
c
h
a
r
g
e

c
u
r
r
e
n
t
,

k
A
0 1 2 3 4 5
V
I
50
40
30
20
10
0
10
2.5
2.0
1.5
1.0
0.5
0
0.5
time, s
r
e
s
i
d
u
a
l

v
o
l
t
a
g
e
,

k
V
d
i
s
c
h
a
r
g
e

c
u
r
r
e
n
t
,

k
A
0 1 2 3 4 5
V
I
60
50
40
30
20
10
0
10
6
5
4
3
2
1
0
1
time, s
r
e
s
i
d
u
a
l

v
o
l
t
a
g
e
,

k
V
d
i
s
c
h
a
r
g
e

c
u
r
r
e
n
t
,

k
A
0 1 2 3 4 5
V
I
50
40
30
20
10
0
10
5
4
3
2
1
0
1
time, s
r
e
s
i
d
u
a
l

v
o
l
t
a
g
e
,

k
V
d
i
s
c
h
a
r
g
e

c
u
r
r
e
n
t
,

k
A
0 1 2 3 4 5
V
I
a b
c d
Figure 5.13 Typical voltage and current traces obtained with a fast impulse
generator
stressed (Figure 5.15). In contrast, in the high conduction regime, the effect of stray
capacitances is secondary and the distribution becomes more uniform because of the
low ZnO resistance.
5.7.3.2 Effect of pollution
In addition to the stray capacitances, pollution of the arrester housing is a key factor
in the performance of the arrester. The electric field uniformity along the arrester may
be improved by means of grading rings at the top of the arrester and at the flanges
for tall column arresters which are used in EHV. The effect of stray capacitances is
further reduced if the base of the arrester (low voltage side) is raised above ground
[105109]. Additional grading resistors and/or capacitors are also used in particular
applications but the cost and size of the arrester are increased. Besides, the effective-
ness of the protection offered by the arrester will depend on the grading elements
reliability [20].
5.7.3.3 Voltage distribution measurements methods
Measurement of the voltage distribution along an arrester can be achieved in the same
way as for insulator strings. Anumber of techniques have been used such as:
(i) sparking gaps (rods, spheres) in which the voltage is determined from a known
breakdown voltage of the gaps
AC curve DC curve
impulse rated voltage
60
50
40
30
20
10
0
r
e
s
i
d
u
a
l

v
o
l
t
a
g
e

a
t

p
e
a
k

c
u
r
r
e
n
t
,

k
V
peak discharge current, A
1
E
0
6
1
E
0
5
1
E
0
4
1
E
0
3
1
E
0
2
1
E
0
1
1
E
+
0
0
1
E
+
0
1
1
E
+
0
2
1
E
+
0
3
1
E
+
0
4
Figure 5.14 Measured VI curve over the whole range of current for 15 kV rated
surge arrester
(ii) voltage divider methods which use potentiometers or microcapacitances
(iii) electrostatic probes; for which the voltage is calculated from the calibration
curve of the probe
(iv) the neontube methodwhere the voltage is determinedfromthe knowndischarge
voltage of the neon tube [110].
All the above cited methods use metallic links to the point where the voltage is to
be measured. This results in the disturbance of the equipotential distribution in the
vicinity of the measuring point. Therefore, regardless of the method accuracy, the
Figure 5.15 Computed equipotential distribution on a polymeric surge arrester
true voltage that would exist without the presence of the measuring device, is slightly
different from the measured value.
Recently, two new optical methods have been developed [111, 112], which are
characterised by the advantage of having an electrically insulated input and out-
put, and hence no extra distortion of the field distribution is introduced. Moreover,
these methods cause a minimumelectromagnetic interference and consume very little
energy.
(i) One method [105, 112], which allows measurement of current at any point along
the arrester column, is an optoelectronic technique in which a sensor comprising
a light emitting diode (LED) is inserted between two adjacent ZnO elements so
that when current flows through the elements, the LED is also subjected to this
current.
(ii) The other method [112, 113] measures the field at any point between the high
and low voltage electrodes. It uses the Pockels effect principle. A light beam,
which can be generated either by a laser or an LED and sent through a light
waveguide (LWG), is linearly polarised before it is goes through the Pockels
device. The refraction index of Pockels devices is electric field sensitive. As
a result of this property, the two orthogonal components of the light electro-
magnetic wave travel with different velocities inside the Pockels device. The
output light beam is then elliptically polarised proportional to the field strength.
An analyser detects the phase shift before the beam is transmitted to the photo
detector which converts the light into electricity. From the calibration curve, a
very accurate field measurement can be achieved without field distortion. This
method can be used for field strengths up to 40 kV/cm [113].
0.01 0.1
current density, mA/cm
2
1 10
1000
100
10
1
0.1
SiC ZnO
f
i
e
l
d
,

V
/
m
m
Figure 5.16 EJ characteristics of ZnO and SiC materials
5.8 Characteristics of ZnO surge arresters
5.8.1 Background
Since 1968 when Matsushita Electric Co. Ltd announced the new technology of ZnO
varistors [34], many investigators have been involved in studying the properties of
ZnO in order to understand and formulate the complex process of ZnO conduction.
The importance of ZnO arresters is a consequence of their voltagecurrent (VI)
characteristic, which is far closer to the ideal than that of SiC, the gapped arresters
which they have largely replaced. Figure 5.16 shows the measured field current
density (EJ) characteristics of both ZnO and SiC materials.
It is not surprising, in view of the extreme non-linearity of its conduction process,
that the ZnO characteristic is a function of temperature and depends on the type of
the applied voltage, especially in the low conduction region. Furthermore, a large
dependence on frequency is observed. These features represent a departure from the
ideal, and may have important implications for ageing of arrester elements and change
of arrester performance.
5.8.2 Frequency response of ZnO material
Early measurements [39, 52, 114] on the frequency response of ZnO material ceram-
ics were achieved using a General Radio capacitance bridge, and because of the very
small size of the tested samples, the applied voltage was merely 1 V r.m.s. In the
frequency range 1010
5
Hz, the calculated relative permittivity from capacitance
measurements is unexpectedly high (10001600) considering that the relative per-
mittivity for ZnO is in the range
r
= 8 to 10, and for the additives is in the range
r
= 16 to 25. This can be partially accounted for by the microstructure of ZnOmate-
rial. As used in surge arresters, the material consists of ZnO grains (approximately
25 m) of high conductivity separated by a very thin layer (<100 ) of additives
which controls the non-linear characteristics. Therefore, the true dielectric thickness
1400
1200
1000
800
600
10
1
10
2
10
3
10
4
10
5
frequency f, Hz
10
6
10
7
10
8
d
i
e
l
e
c
t
r
i
c

c
o
n
s
t
a
n
t
r
Figure 5.17 Permittivity dependence on frequency in ZnO material. (Courtesy

of LEVINSON, L.M., and PHILIPP, H.R.: Long time polarization
currents in metal oxide varistors, J. Appl. Phys., 1976, 47, (7),
pp. 31773181)
of a ZnO block is very small compared with the physical thickness. Microscopic
photography measurements showed that the ratio of sample thickness to dielectric
thickness was about 1000. The direct effect of this reduced thickness would be the
increase of the observed capacitance. Consequently, the permittivity of the material
would not necessarily be as high as the calculated values from the experimental data.
The nature of the additives is clearly shown to influence the measured values of both
capacitance and dissipation factor [50, 115]. A more plausible explanation is based
on the existence of depletion layers in the ZnO grains adjacent to the intergranular
layer and the trapping of electrons at the interface. The net effect of the former is the
rise of the material capacitance [38].
Extensive characterisation work [38, 42, 50, 115, 116] used bridges, Q-meters and
transmission line techniques over a wide range of frequency (3010
8
Hz). In order
to validate the application of dielectric theory, the measurements were performed at
very low stresses, well below the knee of conduction, where the conduction in ZnO
material is linear. A decreasing dielectric constant with increasing frequency was
measured (Figure 5.17). Measurements at higher frequencies (1 MHz1 GHz) on a
sample 0.162 cm thick and 0.305 cm in diameter confirmed a relative permittivity
for the material of about 900, and microwave transmission line techniques revealed
an inductance value of about 0.35 nH associated with the body of the ZnO sample.
The dielectric constant measured with polarisation current techniques [116] shows a
decrease with increasing frequency and a strong temperature dependence at very low
frequencies below 100 Hz.
The dissipation factor, however, was found first to decrease with increasing fre-
quency, to a minimumbetween 110 kHz, and then increased to reach a peak at around
300 kHz. Finally, it decreased for higher frequencies. The observed peak is reminis-
cent of a typical broadened Debye resonance [117]. In this range of frequency, tan
10 10
2
10
3
10
4
10
5
10
6
10
7
10
8
10
2
10
3
10
4
10
5
10
6
10
7
10
8
10
9
10
10
p
a
r
a
l
l
e
l

r
e
s
i
s
t
i
v
i
t
y

j
p
,
c
m
frequency f, Hz
Figure 5.18 Dependence of ZnO resistivity on frequency. (Courtesy of
LEVINSON, L.M., and PHILIPP, H.R.: Long time polarization
currents in metal oxide varistors, J. Appl. Phys., 1976, 47, (7),
pp. 31773181)
varied only by a factor of ten. The immediate consequence of this behaviour would
be a strong frequency dependence of the resistance in a parallel RC circuit repre-
sentation. Accordingly, the equivalent parallel resistivity decreases with increasing
frequency (Figure 5.18). At low frequencies, the resistivity was mainly attributed to
the resistance of the intergranular layer, which is very high compared with the grains
resistance. At higher frequencies, however, it was supposed that the intergranular
resistance fell to the low limiting value representing the grains resistance.
Apeak in loss angle accompanied by a fall in permittivity is a common dielectric
behaviour but the MaxwellWagner model [118], which is used to explain the dielec-
tric behaviour of inhomogeneous solids and polycrystalline semiconductors, fails to
account for the decreasing parallel resistivity with increasing frequency. The highly
disordered intergranular layer and the existence of interface states and electron traps
are thought to be the cause of the model failure. The loss angle peak can be interpreted
as being caused by electron trapping [119].
Although of great importance, the above-published data were obtained from
experiments performed in order to examine the basic physics of these materi-
als in which the samples studied were of very small size (thickness = 2 mm,
diameter = 0.32 cm) and the voltages were very low (up to 10 V). Consequently,
fewer problems were encountered in generating the voltages and measuring the phys-
ical characteristics. The dangers of extrapolating and scaling the properties of such
non-linear materials are clear, since different phenomena may appear in large samples
6
5
4
3
2
1
0
0.0001 0.001 0.01
k
V
0.1 1.0 10.0
mA
DC 20
50 150 Hz
Figure 5.19 Frequency dependence of ZnO VI curve
and with very high voltages which would be more characteristic of elements used in
surge arresters in power systems.
The first high voltage variable frequency test results on ZnO elements [120, 121],
which used frequencies between 30 and 10
6
Hz, showed a rapidly falling dissipation
factor up to 1 kHz and a decreasing capacitance with increasing frequency in the high
frequency range above 10 kHz. Subsequent research [132] showed the dependence
of the VI curve (Figure 5.19) and power consumption in ZnO surge arrester blocks.
Such data yielded decreasing material resistance, capacitance and dissipation factor
with increasing frequency.
5.8.3 Impulse response
There is now an extensive published literature on the response of the ZnO material
and complete surge arresters to impulse currents of different magnitudes and shapes.
Unlike in the prebreakdown region, the resistive current dominates for impulse
currents in the kiloampere range.
5.8.3.1 Initial voltage overshoot
One peculiar observation in the ZnO fast transient response, which was observed by
many investigators, is the initial spike/overshoot measured on the residual voltage
(Figure 5.20). Careful studies linked this phenomenon to circuit inductance [120], the
impulse rate of rise [122, 123], the impulse amplitude [124], the nature and amount
of additives [125], the preinjection of carriers (using a double impulse technique)
[125, 126], the value of the non-linearity coefficient and the difference in the rate of
charge accumulation at the electron and hole traps located at the interface of the grain
and intergranular layer [127]. This type of overshoot is, however, not observable on
SiCmaterial [127, 128]. This leads to the attribution of the overshoot to ZnOmaterial,
50
40
30
20
10
0
10
1.5
1.2
0.9
0.6
0.3
0
0.3
0 1 2 3
time, s
d
i
s
c
h
a
r
g
e

c
u
r
r
e
n
t
,

k
A
r
e
s
i
d
u
a
l

v
o
l
t
a
g
e

k
V
4
V
I
Figure 5.20 Measured spike on ZnO surge arrester residual voltage
its limited capability to switch fast suggesting that it takes a finite time to turn on the
material from a low to a high conduction regime.
It should be emphasised, however, that this time to turn on is not conclusively
established. Subnanosecond tests [129] on small ZnO varistors showed no detectable
time delay in the conduction process, and high voltage tests [128] claimed a time
delay of 3 ns. One major difference between these two sets of conflicting results is
the size of test and measurement loops. Minimisation of these loops and adoption of
coaxial field-free techniques established that the overshoot; hence the delay to turn
on is caused by circuit arrangements [130]. Figure 5.21 illustrates how the voltage
spike recorded with a parallel divider is avoided with the coaxial measurement.
5.8.3.2 Effect of discharge current impulse shape and magnitude
Laboratory tests have shown that faster discharge current impulses produce higher
peaks of residual voltage but the VI curve constructed from the voltage at the
instant of peak current versus peak current is not affected. This may indicate that
the resistance of the material is not greatly influenced by the current rate of rise.
The higher magnitudes of current, however, have a faster rate of rise for a given test
circuit. It is suggested that discrete current paths form through the material using the
lowest potential barriers. With increasing current magnitude, the number of paths
increases with some of them getting shorter. Parallel branches having resistances and
inductances in series can represent such a process [131].
5.8.4 Combined stress response
The improved understanding of the material behaviour and complete arrester perfor-
mance has allowedbetter characterisationof ZnOsurge arresters andmore widespread
application. Various types of voltage including direct, alternating, variable frequency,
impulse voltages and their combinations have been used to test ZnO arresters. Each
set of tests has revealed an important property of the material: frequency dependence
25
20
15
10
5
0
5
0 1 2 3
time, s
V
p
,

k
V
V
s
,

V
,
I
,
k
A
4 5
25
2
1.5
1
0.5
0
0.5
I
V
p
V
s
40
30
20
10
0
10
4
3
2
1
0
1
0 1 2 3
time, s
V
p
,

k
V
V
s
,

V
,

I
,
k
A
4 5
I
V
p
V
s
50
40
30
20
10
0
10
5
4
3
2
1
0
1
0 1 2 3
time, s
V
p
,

k
V
V
s
,

V
,
I
,
k
A
4 5
I
V
p
V
s
50
40
30
20
10
0
10
5
4
3
2
1
0
1
0 1 2 3
time, s
V
p
,

k
V
V
s
,

V
,
I
,
k
A
4 5
I
V
p
V
s
Figure 5.21 Initial voltage overshoot (V
p
) on ZnOarresters caused by measurement
errors. No overshoot on coaxial measurement (V
s
)
of the VI curve [132]; crossover of the different VI characteristics [133]; over-
shoot on the residual voltage [134, 120, 122127]; finite time to turn-on [128]. Mixed
(AC plus DC) voltage tests provide an insight into three important aspects of ZnO
characteristics.
5.8.4.1 Mixed (AC+DC) characteristics
ZnO surge arresters are used in HVDC systems [121, 135] where, because of the con-
verters used in such systems, high frequency voltages are generated. As a result,
a mixed voltage in which an alternating voltage is superimposed upon a direct
voltage appears on the network. A comparison of both the currentvoltage and
power-dissipationvoltage characteristic curves for direct and alternating voltages,
respectively, show a crossover point near the knee of high conduction (Figure 5.22).
This crossover is also frequency dependent, and its physical cause has not been
explained, although it appears to be associated with the dynamic response of the
material:
(i) Lowvoltage capacitance measurements Although the above introduction estab-
lishes engineering interest in mixed voltage tests, hitherto such a procedure has
been used mainly in physical studies, where it is a widely used laboratory
method for measuring capacitance [42, 44, 115, 49, 136]. Here, a very small
alternating voltage is superimposed on a larger direct voltage. The AC level is
K=0 (DC)
K=1 (AC)
K=0.5
6
5
4
3
2
1
0
0.01 0.1 1
current, mA (peak)
v
o
l
t
a
g
e
,

k
V

(
p
e
a
k
)
10 100
Figure 5.22 Crossover of DC and AC characteristics
then kept constant while the direct voltage is varied [44, 136, 137]. The impor-
tance of these CV
dc
curves lies in the close link between the capacitance and
the physical behaviour of the material. CV
dc
curves are far more sensitive to
microstructural fluctuations than IV curves [44], and have been used to inves-
tigate many of the physical properties of ZnO material, such as the density of
interface states [42, 52], donor concentration [49, 115] and change in depletion
layer.
(ii) Crossover of DC and AC characteristics The crossover was first attributed to a
significant contribution of dielectric loss under alternating applied voltage in the
prebreakdown (low current) region [48, 58, 70, 121, 132, 133, 139], and a time
lag which was observable in the current in the breakdown region. However,
according to the measured data, less than 200 ms are needed for the current
to reach its steady state value, so that this explanation involving such a short
response time wouldnot account for anycrossover at 50Hz. Another explanation
relates to polarisation currents which are much higher than conduction currents,
in contrast to the breakdown region where conduction prevails [116, 140, 141].
It was predicted that lower conduction currents would be obtained at higher
frequencies for the same applied voltage, because the time lag would prevent
the current from reaching its static value. However, the polarisation currents
are more important at very low temperature and the order of magnitude of the
ratio of polarisation currents to DC equilibrium current is less than that of the
ratio of AC to DC values. Polarisation phenomena can thus give only a partial
explanation of the crossover.
45
40
p
e
a
k

r
e
s
i
d
u
a
l

v
o
l
t
a
g
e
,

k
V
35
30
25
0.01 0.1
peak discharge current, kA
V
dc
=0 kV V
dc
= 15kV V
dc
=+15kV
1 10
Figure 5.23 Typical effect of DC prestress (trapped charge) on the protective
characteristic of a 15 kV-rated ZnO surge arrester
5.8.4.2 Mixed (DC+impulse) characteristics and effect of trapped charge
In AC power systems, under normal service conditions, surge arresters on isolated
lines and cables can experience a direct voltage stress arising from trapped charge.
If followed by a switching surge of reverse polarity, severe requirements will be
imposed on the surge arresters. Combinations of direct and surge voltages can also
occur in HVDC systems and due to circuit breaker restrikes during switching of
capacitor banks or unloaded transmission lines where trapped charges are present.
Circuit breaker restrikes can cause high rate-of-rise overvoltages of peak values of 2
to 3 p.u. following rapid polarity reversal, and may cause up to 8 p.u. especially for
some types of oil circuit breakers [142]. This condition may be particularly onerous
where the surge is opposite in polarity to the preexisting working voltage [143146].
For a given level of peak discharge current through the arrester, a higher level of
residual voltage of upto ten per cent is found to appear across an arrester when the
polarity of the applied impulse opposes that of the direct voltage (Figure 5.23). This
effect is especially marked in the peak residual voltage which is affected by the
increased di/dt . Experimental tests have shown [147] that the effect increases with
decreasing arrester protection level.
5.8.4.3 Mixed (AC+impulse) characteristics and effect of working
voltage level
Transient high amplitude surges commonly occur in high voltage power systems,
due to either external or internal influences, such as lightning strikes or switching
operations. For most systems, these conditions are likely to occur when the system is
already operating at its normal alternating working voltage. Laboratory experiments
[148] revealed that for an impulse voltage superimposed on the power frequency
40
35
30
25
20
0.01 0.1
peak discharge current, kA
firing angle
r
e
s
i
d
u
a
l

v
o
l
t
a
g
e

a
t

c
u
r
r
e
n
t

p
e
a
k
,

k
V
1 10
0 90 270
Figure 5.24 Typical effect of AC prestress (working voltage) on the protective
characteristic of a 15 kV-rated ZnO surge arrester
voltage, it was found that firing at peaks with opposite polarity impulses produces
the highest residual voltage levels (Figure 5.24).
In the conditioning test recommended by IEC 99-4:1991 [77] for the operat-
ing duty test, it is specified that four groups of five lightning nominal current
impulses (8/20 s) shall be applied at intervals of 5060 s between impulses and
25 to 30 minutes between groups. These impulses should be applied superimposed
on an elevated system voltage determined by ageing tests. The firing angle is speci-
fied to be 15 to 45 electrical degrees from the zero crossing. The American Standard
ANSI/IEEE C62-11: 1987 [149] specifies a firing angle of 60
before the crest volt-

age. It is recommended that the impulse should be of same polarity as the alternating
voltage at the instant of firing. This IEC recommendation regarding the firing angle
in the duty tests was originally specified for gapped surge arresters and was to ensure
satisfactory firing of the series gap. However, since modern ZnO surge arresters do
not have any series spark gaps, it may be more appropriate to include a firing angle
that produces higher residual voltages.
5.8.4.4 Multiple impulse response
Between 60 and 70 per cent of all lightning ground flashes contain more than
one stroke. A typical flash will consist of three or four strikes with time inter-
vals between strokes of 20 to 200 ms. In addition, induced lightning surges on
transmission lines are known to produce a transient overvoltage with two peaks of
reverse polarity. Arapid series of transients may also occur during certain switching
transients.
Extensive laboratory tests have shown that the cumulative nature of multiple
impulses forms a significant part of the ZnO material degradation process, especially
the insulating walls of the ZnO elements [150, 151]. Such deterioration was related
to energy dissipation constraints. However, the VI characteristic of the arrester is
not significantly affected by multiple impulses of low energy [152, 153].
5.8.5 Equivalent circuit of ZnO material
5.8.5.1 Review
The main objective in an equivalent circuit is to be able to represent and to model the
ZnO surge arrester under different conditions of voltage, temperature and frequency.
Furthermore, equivalent circuit representations should be simplified as far as possible,
consistent with the arrester element characteristics, to facilitate their application. The
proliferation of different equivalent circuits has arisen in response to data progres-
sively acquired fromnewmeasurements. Unexpected behaviour and properties of the
ZnOelement suchas temperature dependence, time lag of current growth, polarisation
currents, frequencydependence andfast transient response characteristics have helped
to refine the characterisation of the material.
The simplest model is an RC parallel circuit where R is defined to be highly
non-linear with increasing voltage; a similar non-linear behaviour, although of less
degree, is assigned to the capacitance. Such a circuit can be useful for engineering
modelling of arrester behaviour [82, 83]. However, as will be demonstrated in this
section, this simple equivalent circuit, even with non-linear elements, is ultimately
non-representative of the complete behaviour of the ZnO arrester. Experimental data
have shown that a more complex equivalent circuit is needed in order to take into
account the various parameters which affect the arrester characteristics.
Starting with the idealised microstructure of the block consisting of cubic conduct-
ing ZnO grains surrounded (coated) by a segregation layer (the intergranular layer)
which is responsible for the non-linear behaviour of the material [37, 39, 115], early
equivalent circuits consisted of a small resistance representing the ZnOgrains in series
with an inductance and a parallel RC circuit representing the voltage and frequency-
dependent resistance and capacitance of the intergranular layer (Figure 5.25a) [37].
The capacitance of ZnOgrains is sufficiently small to be neglected at normal working
voltages [39, 114, 120]. The series inductance represents the inductance of the arrester
body and accounts for the response to steep currents [154]. Aslightly different equiv-
alent circuit, which consists of a resistance R
g
representing the ZnO grains in series
with two parallel RC circuits, was suggested in Reference 37. One RC circuit was
used to represent the intergranular material component and the other to account for the
interface between ZnO grains and the intergranular layer. Other circuits represented
the ZnO blocks by three parallel branches [155], a resistance R which is essentially
frequency independent and is given by the VI characteristic, a capacitance C which
is weakly dependent on the applied voltage, temperature and frequency, and a resis-
tance Z which is nearly voltage independent but strongly dependent on frequency.
It is specified that at low voltages losses are determined by Z but at higher voltages
losses are governed by R. Other earlier proposed circuits [140, 141], accounting for
the measured polarisation currents proposed [55], consisted of:
(i) an inductance associated with the varistor body itself
(ii) a voltage and temperature-dependent resistance to represent losses due to
conduction
(iii) a voltage, temperature and frequency-dependent resistance to represent polar-
isation with losses
R
Ig
R
grain
C
Ig
intergranular layer
resistance
intergranular layer
capacitance
zinc-oxide grain
resistance
L
0
L
1
R
1
R
0
A
0
A
1
C
L
b
R
grain
L
c1
L
c2
R
Ig
C
Ig
R
c
a
b
c
Figure 5.25
a Simple equivalent circuit for ZnO material
b Equivalent circuit for fast impulses. (Courtesy of IEEE working
group 3.4.11: Modelling of metal oxide surge arresters, Trans-
action on Power Delivery, 1992, 7, (1), pp. 302309, 2003
IEEE)
c Equivalent circuit to account for the dynamic conduction at high-
magnitude fast impulses
(iv) a frequency and temperature-dependent capacitance also associated with
polarisation with losses
(v) a capacitance to account for polarisation without losses.
An equivalent circuit based on laboratory test data included the series grain resistance
and inductance and contains a three-branch parallel circuit [128]:
(i) a non-linear resistance which is a function of temperature
(ii) a voltage and temperature dependent capacitance which is also affected by the
rate of change of the voltage
(iii) a turn on element, which will account for the dynamic charge distribution at the
grain boundary; this is a function of voltage rate of rise of voltage and the time
constant t to reach the equilibrium of electrons and holes at the grain boundary.
Recent research on zinc oxide surge arresters has concentrated on the very fast
transient characterisation [156, 157], and two main equivalent circuits were pro-
posed. In addition to the above elements, these circuits include parallel branches
(Figures 5.25b and 5.25c) with inductances to account for the fast transient
measurement data.
5.8.5.2 Determination of the equivalent circuit main components
Low voltage tests and/or computation [158] can be used to estimate the parallel
capacitance C
ig
, and high current impulse data can be used to calculate the ZnOgrain
R
g
and intergranular layer R
ig
resistances. The voltage V
ig
across the intergranular
layer resistance is related to current by:
V
ig
V
B
=
_
I
I
B
_
(5.17)
where = 1/, V
B
and I
B
are the base values.
The intergranular layer resistance is:
R
ig
=
V
ig
I
=
V
B
I
B
_
I
I
B
_
1
= R
B
_
I
I
B
_
1
(5.18)
This yields the resistive voltage current behaviour for the arrester:
V = I(R
g
+R
ig
) = I
_
R
g
+R
B
_
I
I
B
_
1
_
(5.19)
which gives
V = I(R
g
+kI
1
) (5.20)
with
k =
R
B
I
1
B
Such an expression is readily applicable [159] in computer packages such as Spice.
5.9 Monitoring of ZnO surge arresters
Zinc oxide surge arresters are designedtolast a useful lifetime of at least 20to30years.
Over this period, the arrester is expected to absorb a large number of surges and limit
the voltage to a safe level, cope with the harsh environment and withstand temporary
overvoltages and system voltage fluctuations. Long term accelerated ageing tests,
which are conducted with elevated applied voltage and temperature, have indicated
that when the resistive component of current and the power consumption in a ZnO
surge arrester increases to more than twice the initial value, the arrester has degraded
to a stage equivalent to the end of its useful life. This criterion is used in many
condition monitoring techniques for surge arresters. Most of these techniques require
the measurement of applied voltage and leakage current through the surge arrester.
One technique to assess arrester health is to perform offline laboratory tests such
as DC or AC current injection to monitor [159] the changes in the arresters VI
characteristic over a wide range of current (0.5 Ato 1 mA). This method, however,
is not cost effective as it requires outages and careful transport arrangements.
Other more cost-effective methods are based upon online monitoring which
involves leakage current analysis [160162]. These techniques use harmonic analy-
sis of the total leakage current to obtain the third harmonic of the resistive current.
Subsequent scaling allows reconstruction of the resistive current level. Voltage mea-
surement in some of these techniques is based on a capacitive probe pick up. The
analysis compensates for voltage harmonic content as well as influence of adjacent
phases. Errors of more than 30 per cent were observed with these techniques, which
occur because at systemvoltages, the arrester leakage current is predominantly capac-
itive. The resistive component is generally less than five per cent of the total leakage
current. Considering this small ratio and the highly non-linear conduction in ZnO
material, accurate discrimination of the resistive component is not easy. Also, the
third harmonic is a small fraction of the resistive component which may impose
a demanding resolution limit for the transducers used in arrester leakage current
measurements. Furthermore, any harmonic contents in the voltage will give rise to
complex current waveshapes. In addition, most existing techniques for current dis-
crimination assume a constant linear capacitance behaviour of the ZnOsurge arrester,
which is inconsistent with the measured non-linear ZnO capacitance.
The ZnO element may be represented by a parallel RC circuit, both R and C
being non-linear, and the current can be resolved into conduction and displacement
components. A conventional laboratory method of obtaining the two components of
current uses a constant loss-free high voltage capacitor to compensate for the capac-
itive current [17, 32, 55, 87, 120]. This compensation technique and/or the standard
Schering bridge method may be used to estimate the ZnO parameters. The additional
need in these methods for a high voltage capacitor may be avoided with the atten-
uator compensation technique [163, 164] or by means of special electronic circuits
[91, 165]. However, these methods do not allowfor the ZnOcapacitance being voltage
dependent. Another method with similar assumptions [24] uses Fourier analysis to
resolve the measured current into in-phase and quadrature components. For a detailed
description of leakage current measurement and diagnostic indicators of ZnO surge
arresters, refer to BS EN 60099-5:1997 or its equivalent EN60099-5:1996 including
amendment A1:1999 [166].
Recently, a point-on-wave method [167], which requires voltage and current
traces, has identified variations during the voltage cycle not only in the equivalent
resistance of the sample, but also in its capacitance. The method is based on the
expression of the average power and assumes a single valued voltage conduction
current characteristic.
The total current for an RC parallel equivalent circuit is:
I
t
(t ) = I
c
(t ) +I
r
(t ) = C
dV(t )
dt
+I
r
(t ) (5.21)
where I
c
and I
r
are the capacitive and resistive components of the total leakage
current I
t
.
It was shown [167] that the instantaneous resistive current magnitude at two
instants t
1
and t
2
corresponding to the same level of voltage on the cycle can be
expressed as:
I
r
(t
1
) = I
r
(t
2
) =
_
1
2
(I
t
(t
1
) +I
t
(t
2
))
_
_
1
2
(I
t
(t
1
) I
t
(t
2
))
_
_
dV(t
1
)/dt +dV(t
2
)/dt
dV(t
1
)/dt dV(t
2
)/dt
_
(5.22)
This equation allows the calculation of the resistive current around the cycle for any
waveshape. If the voltage is sinusoidal with no harmonic content or with only odd
harmonics without phase shift, it can be shown that:
I
r
(t
1
) = I
r
(t
2
) =
1
2
(I
t
(t
1
) +I
t
(t
2
)) (5.23)
Figure 5.26 shows typical examples obtained with this discrimination technique on
a 15 kV surge arrester in the low and high conduction regimes. As can be seen, both
the resistive and capacitive components show non-linearities.
5.10 Standards and application guidelines
This section is a general summary of important aspects of the existing international
standards (IEC60099 and IEEE/ANSI C62.110).
5.10.1 Standard definitions of important parameters
The following standard definitions are used for both the testing and selection of zinc
oxide surge arresters.
5.10.1.1 Continuous operating voltage (U
c
)
U
c
is the maximum permissible value of a sinusoidal power frequency voltage which
may be continuously applied across the arrester terminals. The normal systemvoltage
1
0.75
0.5
0.25
0
0.25
0.5
0.75
1
0 4 8 12
time, ms
V
max
= 0.62 kV
16 20
0.2
0.15
0.1
0.05
0
0.05
0.1
0.15
0.2
I(t)
I
r
(t)
V(t)
v
o
l
t
a
g
e
,

k
V
c
u
r
r
e
n
t
,

m
A
I(t)
I
r
(t)
V(t)
2
1.5
1
0.5
0
0.5
1
1.5
2
0 4 8 12
time, ms
V
max
= 1.85 kV
16 20
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
v
o
l
t
a
g
e
,

k
V
c
u
r
r
e
n
t
,

m
A
time, ms
V
max
= 4.58 kV
v
o
l
t
a
g
e
,

k
V
c
u
r
r
e
n
t
,

m
A
6
4
2
0
2
4
6
1.5
1
0.5
0
0.5
1
1.5
0 4 8 12 16 20
I(t)
I
r
(t)
V(t)
v
o
l
t
a
g
e
,

k
V
c
u
r
r
e
n
t
,

m
A
I(t) I
r
(t)
V(t)
6
4
2
0
2
4
6
9
6
3
0
3
6
9
0 4 8 12 16 20
time, ms
V
max
= 5.62 kV
a b
c d
Figure 5.26 Discriminated leakage current components using the point-on-wave
technique
should not exceed this value. Some manufacturers use the terms maximumcontinuous
operatingvoltage (MCOV) andcontinuous operatingvoltage (COV), the former being
an upper limit for the arrester permissible continuous voltage.
5.10.1.2 Continuous current (I
c
)
I
c
is the continuous r.m.s. or peak magnitude current which flows through the arrester
when the continuous operating voltage, U
c
is applied across its terminals.
5.10.1.3 Rated voltage (U
r
)
U
r
is the maximum permissible short duration r.m.s. value of power frequency volt-
age between the arrester terminals. In IEC60099-4:1991 [77], the rated voltage is
defined as the 10 s power frequency voltage used in the operating duty test involv-
ing high current and long duration impulses. However, it should be noted that some
national standards use a number of impulses at nominal current with superimposed
power frequency voltage in order to determine the rated voltage which may pro-
duce different values from those recommended by IEC. The standards bodies are
now considering these discrepancies. Table 5.3 reproduces the recommended rated
voltages [77].
5.10.1.4 Discharge current
The discharge current is used to describe the impulse current which flows through
the arrester under test.
Table 5.3 Recommended values for ZnO
surge arrester rated voltages [77]
Range of rated voltage Steps of rated voltage
(kV r.m.s.) (kV r.m.s.)
<3 under consideration
330 1
3054 3
5496 6
96288 12
288396 18
396756 24
5.10.1.5 Switching current impulse
The switching current impulse is the peak magnitude of discharge current having a
recommended standard impulse shape, a virtual front time greater than 30 s but less
than 100 s and a virtual time to half value on the tail of roughly twice the virtual
front time.
5.10.1.6 Nominal discharge current (I
n
)
I
n
is the peak magnitude of a standard lightning current impulse (8/20) that an arrester
is rated for and used for its classification. It is considered as the main parameter for
determining the protective characteristic and the energy absorption capability of an
arrester.
5.10.1.7 High current impulse
The high current impulse is the peak magnitude of discharge current having a 4/10
impulse shape which is used to test the stability of an arrester if subjected to direct
lightning strokes.
5.10.1.8 Residual voltage (U
res
)
U
res
is the peak value of the impulse voltage that appears between the terminals of
an arrester during the flow of discharge current. Some countries still use the term
discharge voltage to describe the residual voltage.
5.10.2 Classification of ZnO surge arresters
5.10.2.1 Line discharge class
ZnO surge arresters with 10 kA and 20 kA nominal discharge currents are given
a standard class number (1 to 5) according to their energy absorption capability.
Table 5.4 gives the specified classes of IEC99-4:1991 [77].
Table 5.4 ZnO surge arrester classes as defined by
IEC60099-4:1991 [77] using the line discharge test
Arrester Line Surge impedance Virtual Charging
classification discharge of the line () duration voltage
class of peak (s) (kV DC)
10 000 A 1 4.9 U
r
2000 3.2 U
r
10 000 A 2 2.4 U
r
2000 3.2 U
r
10 000 A 3 1.3 U
r
2400 2.8 U
r
20 000 A 4 0.8 U
r
2800 2.6 U
r
20 000 A 5 0.5 U
r
3200 2.4 U
r
Table 5.5 Peak currents for switching impulse
residual voltage tests [77]
Arrester classification Peak currents (A)
20 000 A, line discharge 500 and 2000
classes 4 and 5
class 3
classes 1 and 2
5.10.2.2 Pressure relief class
The pressure relief class is determined by the capability of an arrester to withstand
internal fault currents without violent shattering of the housing; this is particularly
applicable to porcelain housed surge arresters.
5.10.3 Other important arrester characteristics
5.10.3.1 Protective characteristic and protective levels of an arrester
The protective characteristic of an arrester includes three main protective levels:
(i) Slow front overvoltages (switching impulse protective level): the maximum
residual voltage at the specified switching impulse current. Table 5.5 repro-
duces the standard recommended peak magnitudes for the switching currents
to determine the switching impulse protective level.
(ii) Fast front overvoltages (lightning impulse current, 8/20, protective level): the
maximum residual voltage at nominal discharge current.
(iii) Veryfast transients (steepcurrent protective level): the residual voltage for steep
currents whose front times are between 0.9 and 1 s and tail times no longer
than 20 s. Areliable measurement procedure for the precise determination of
the residual voltage under steep current is under consideration. The technique
should take into account inductive effect errors as described in section 5.7.2.
5.10.3.2 Power frequency withstand voltage versus time characteristic of an
arrester
Corresponds to the maximum duration for which the arrester can withstand a given
level of continuous power frequency voltage without sustaining any damage or
thermal instability. This characteristic is usually above the temporary overvoltage
characteristic.
5.10.3.3 Pollution withstand characteristics
The pollution performance of an arrester housing is determined according to relevant
IEC507:1991 [168] and IEC815:1986 [169] standards. In addition, the arrester has
to withstand temperature gradients and internal partial discharges which may appear
on the arrester column due to surface pollution. An extensive review of pollution
characteristics of zinc oxide surge arresters is given in Reference 170.
5.10.4 Standard tests
The current standards specify the following tests (for a detailed description of these
tests, refer to IEC 99-4:1991 [77]):
(i) type or design tests, which are made on a new arrester design to establish its
performance and compliance with the relevant standard; these tests will be
required on subsequent arresters with a similar design
(ii) routine tests are made on each arrester to ensure that the product meets the
design specifications
(iii) acceptance tests, which are agreed between the manufacturer and the
purchaser.
In the operating duty cycle, correct selection of the test sample depends on the correct
selection of the reference voltage and current. The reference current, I
ref
, of an
arrester is defined as the peak value of the resistive component of the total current
flowing through the arrester. The reference current ranges between 0.05 mA and
1.0 mA per square centimeter of disc area for single column arresters. The voltage
obtained at the reference current is defined as the reference voltage, U
ref
.
5.10.5 Recommended arrester identification
IEC60099-5:1999 recommends that metal oxide surge arresters shall be iden-
tified by the following minimum information which shall appear on a nameplate
permanently attached to the arrester:
continuous operating voltage
rated voltage
rated frequency, if other than one of the standard frequencies
nominal discharge current
pressure relief rated current in kA r.m.s. (for arresters fitted with pressure relief
devices)
the manufacturers name or trade mark, type and identification of the complete
arrester
identification of the assembling position of the unit (for multiunit arresters only)
the year of the manufacture
serial number (at least for arresters with rated voltage above 60 kV).
5.11 Selection of gapless metal oxide surge arresters
Figure 5.27 reproduces the procedure of arrester selection specified by the standards
[166]. During the selection process of a metal oxide surge arrester for a particular
application, the following arrester parameters need to be determined: the continuous
operating voltage, the rated voltage, the nominal discharge current and the protective
characteristic, the line discharge class, the pressure relief class, the pollutionwithstand
and live washing capabilities and the mechanical properties [166]. The following are
particularly important.
(i) Continuous operating voltage U
c
: must be higher than the maximumattainable
power frequency systemvoltage including any harmonic content. It is proposed
to use a safety factor of 1.05 to account for harmonic content. The continuous
operating voltage level to be used depends on the system earthing and the earth
fault clearing arrangements.
(ii) Rated voltage: is determined from the temporary overvoltage levels where
the arrester is to be installed. The power frequency voltage characteristic of
the arrester should be higher than the temporary overvoltage characteristic
of the system. The standards recommend considering temporary overvoltages
due to earth fault overvoltages, load rejections, resonance effects and Ferranti
effect. Table 5.6 reports current rating for the whole range of standard rated
voltages.
(iii) Nominal discharge current: is normally selected according to the lightning
discharge current through the arrester. Table 5.7 summarises the recommended
values of arrester nominal discharge current for the range of voltages in use in
power systems.
(iv) Energy absorption capability: zinc oxide surge arresters must be able to
absorb the energy due to a variety of transient overvoltages arising from clos-
ing or reclosing long lines; disconnection of capacitor banks or cables with
restriking circuit breakers; and lightning. Formulae for the calculation of the
energy absorbed by the arrester for the different overvoltages are given in
Reference 166.
(v) Coordination of ZnO arresters: ZnO surge arresters are now used as the pri-
mary overvoltage protection of power systems equipment such as transformers.
highest
operating voltage
temporary
overvoltage
lightning
discharge current
fault current
discharge
energy
continuous
operating voltage
surge arrester
lightning impulse
protection level
coordination
lightning impulse
withstand voltage
rated insulation level
coordination
switching impulse
withstand voltage
switching impulse
protection level
acceptable
+
nominal
discharge current
pressure relief
class
line discharge
class
rated voltage
yes
arrester selected
no no
+
+
+
conductor length
between arrester
and protected object
station
layout
station
layout
prospective slow
front overvoltage
representative impinging
lightning overvoltage
Figure 5.27 Recommended procedure for the selection of ZnO surge arresters
1
Table 5.6 Arrester classification and rated voltages [166]
Standard nominal discharge current
20 000 A 10 000 A 5000 A 2500 A 1500 A
rated 360 < U
r
756 3 U
r
360 U
r
132 U
r
36 this low
voltage U
r
voltage range
(kV r.m.s.) is now under
consideration
Table 5.7 Practical examples of arrester ratings (extracted
from [166])
System voltage level Recommended nominal
discharge current (kA)
up to 72 kV: short lines (less than 5 km) 5
up to 72 kV: low flash density and effectively 5
shielded lines
up to 72 kV: high flash density and earth 10
resistance
above 72 kV and up to 245 kV 10
above 245 kV and up to 420 kV 10
above 420 kV 20
The transformer bushings are normally protected by air gaps which need to
be coordinated with the surge arrester. At distribution voltages, some of the
arresters may be in vulnerable positions which requires careful examination
of the gap/arrester assembly. It has been shown that the gap distance may be
adjusted to act as a relief device for the arrester when an unexpectedly high
magnitude surge hits the system [171].
(vi) Multicolumn surge arresters: the energy dissipation capability of ZnO surge
arresters depends on their surface area; larger ZnO blocks are required for
larger amounts of energy. However, there is a block size limit that can be
manufactured economically with a guaranteed good homogeneity and uniform
current distribution. For this reason, multicolumn surge arresters are preferred.
Usually, the matching of the parallel columns is a difficult aspect of the arrester
construction since a good balance of the current sharing must be achieved. The
normal practice is to match the voltagecurrent characteristics of individual
columns at two points; usually at 1 mA and at the nominal discharge current.
Matching at other points is not certain and the characteristics can be up to
six per cent different for matched columns [172]. Such imbalance could lead
to premature failure following service degradation, which can be caused by
leakage current, pollution and other weather effects, solar radiation and ambient
temperature differences.
5.12 Location and protective distance of surge arresters
The selection process described in the previous section allows specifying the best
choice of a ZnO surge arrester to protect against prospective overvoltages at a given
site/location. However, the surge arrester will offer the optimised protection level
only if it is installed directly across the terminals of the equipment to be protected.
Unfortunately, in practice, it is not always possible to locate the arrester close to the
equipment, and an inevitable separation distance will be required. Protection against
switching overvoltages is normally guaranteed for most applications since travelling
wave effects can be neglected on short distances, and the switching protective level of
the arrester will appear across the equipment terminals. For fast transient overvoltages,
however, conductor inductance and travelling wave effects can cause significant
voltage differences between the arrester and the equipment. The following three main
approaches were adopted to predict the effectiveness of a surge arrester protection at
a given location.
5.12.1 Effect of distance on protective level
When a surge arrester of total length a (including its leads) is located a distance d away
from the protected equipment, estimation of the voltage level, V
equ
, at the equipment
for a given overvoltage steepness S and a propagation speed u can be achieved using
the following simplified expression:
V
equ
= V
a
+
2 (d +a) S
u
(5.24)
where V
a
is the residual voltage of the surge arrester, and d is the distance between
the arrester and the equipment including the lengths of connection leads.
Equation (5.24) indicates that the voltage level at the equipment will increase
with separation distance and steepness of the incoming surge. This method, however,
should be used with some caution since it does not take into account the character-
istics of the equipment to be protected, e.g. equivalent capacitance in the case of
transformers or shunt reactors. Computer simulations of such systems indicate that
this simplified approach exaggerates the increase in protective level (Figure 5.28).
5.12.2 Calculation of separation distance
IEEE Standard C62.22-19:1991 [173] recommends a calculation procedure of the
maximum allowable separation distance d between the equipment to be protected
and the surge arrester connection point for which an adequate overvoltage protection
would still be provided. The following empirical equation was derived fromextensive
0 2 4 6 8 10 12 14
simplified method
U=30 kV + [2 (40 kV/us) d]/(300 m/us)
computer calculations for
transformer C=2 nF
computer calculations for
transformer C=0 nF
distance from arrester connection point d, m
v
o
l
t
a
g
e

a
t

t
r
a
n
s
f
o
r
m
e
r
,
U
,

V
30500
30900
31300
31700
32100
Figure 5.28 Effect of distance on protective characteristic of ZnO surge arrester
(15 kV-rated surge arrester, S = 40 kV/s, V
res
= 30 kV, u =
3 10
8
m/s)
EMTP simulations of single transformer substations with system voltages ranging
from 69 kV to 765 kV:
d
_
0.1155 uV
res
S
__
0.957 V
BIL
V
res
2.92 V
res
0.957 V
BIL
_
(5.25)
where V
BIL
is the basic lightning impulse insulation level (BIL) of the trans-
former (kV), u is the surge velocity (m/s), V
res
= V
a
+ L
a
dI/dt is the residual
voltage across the arresters and its leads (kV) and S is the steepness of the incom-
ing surge at the surge arrester connection point (kV/s). L
a
is the inductance of the
arrester branch of length a.
5.12.3 Calculation of arrester protective zones
IEC 99-5:1996 [166] recommends the technique proposed in IEC 71-2:1996 [174]
for estimating the coordination lightning impulse withstand voltage, and proposes the
calculation of the protective zone, L
p
= d +a, of the arrester using:
d =
N
A
__
V
BIL
1.15
V
a
__
(L
sp
L
f
) a (5.26)
where N is the number of lines connected to the substation (usually N = 1 or 2), Ais
a factor describing the lightning performance of the overhead line connected to the
substation (see Table 5.8 reproduced fromReference 174), V
BIL
is the rated lightning
impulse withstand voltage, V
a
is the lightning impulse level protection level of the
surge arrester, L
sp
is the span length on the incoming overhead line and L
f
is the
Table 5.8 Factor A for various overhead lines [166, 174]
Type of line Factor A (kV)
distribution lines (phase-to-phase flashovers)
with earthed cross-arms (flashover to earth at low voltage) 900
woodpole lines (flashover to earth at high voltage) 2 700
transmission lines (single-phase flashover to earth)
single conductor 4 500
double conductor bundle 7 000
four conductor bundle 11 000
six and eight conductor bundle 17 000
length of the overhead line section with outage rate equal to acceptable failure rate.
For more details on these factors see Reference 166.
5.13 Note
1
The author thanks the International Electrotechnical Commission (IEC) for
permission to reproduce: Figure 5.166 from its International Standard IEC 60099-5
Edition 1.1 2000-03 Surge arresters Part 5: Selection and application recommenda-
tions. All such extracts are copyright of IEC, Geneva, Switzerland. All rights reserved.
Further information on the IEC is available from www.iec.ch. IEC has no responsi-
bility for the placement and context in which the extracts and contents are reproduced
by the author; nor is IEC in any way responsible for the other content or accuracy
therein.
5.14 References
1 KAUFMANN, R.H., and HALBERG, M.N.: System overvoltages causes
and protective measures, in BEEMAN, D. (Ed.): Industrial power system
handbook (McGraw-Hill Book Co., London, 1955, 1st edn.), pp. 278336
2 BURGER, U.: Surge arresters with spark gaps, in RAGALLER, K.
(Ed.): Surges in high voltage networks (Plenum Press, New York, 1980),
pp. 251282
3 SAKSHAUG, E.C., KRESGE, J.S., and MISKE, Jr. S.A.: A new concept in
station arrester design, IEEE Trans. PAS, 1977, PAS-96, (2), pp. 647656
4 KOBAYASHI, M., MUZINO, M., AIZAWA, T., HAYASHI, M., and
MITANI, K.: Development of zinc oxide non-linear resistors and their
applications to gapless surge arresters, IEEE Trans. PAS, 1978, PAS-97, (4),
pp. 11491158
5 BARGIGIA, A., MAZZA, G., CARRARA, G., and DE NEGRIS, M.:
Evolution of long duration current impulse withstand test on surge arresters for
HVAC systems, IEEE Trans. Power Deliv., 1986, PWRD-1, (4), pp. 173183
6 ABB HIGH VOLTAGE TECHNOLOGIES LTD: Dimensioning, testing and
application of metal oxide surge arresters in medium voltage networks.
Manufacturers booklet, 1984
7 See general literature and websites of surge arrester manufacturers: Bowthorpe
EMP, Ohio Brass, Raychem, Siemens, etc.
8 STANDLER, R.B.: Protection of electronic circuits from overvoltages (John
Wiley & Sons, New York, 1989)
9 MUENCH, F.J., and DUPONT, J.P.: Coordination of MOV type lightning
arresters and current limiting fuses, IEEE Trans. Power Deliv., 1990, 5, (2),
pp. 966971
10 LAT, M.V., and KORTSCHINSKI, J.: Distribution arrester research, IEEE
Trans. PAS, 1981, PAS-100, (7), pp. 34963505
11 SHIH, C.H., HAYES, R.M., NICHOLS, D.K., KOCH, R.E.,
TIMOSHENKO, J.A., andANDERSON, J.G.: Application of special arresters
on 138 kV lines of Appalachian power company, IEEE Trans. PAS, 1985,
PAS-104, (10), pp. 28572863
12 KERSHAW, S.S., GAIBROIS, G.L., andSTAMP, K.B.: Applyingmetal-oxide
surge arresters on distribution systems, IEEE Trans. Power Deliv., 1989, 4,
(1), pp. 301303
13 HILEMAN, A.R., ROGUIN, J., and WECK, K.H.: Metal oxide surge arresters
in AC systems, part V: protection performance of metal oxide surge arresters,
Electra, 1991, (133), pp. 132144
14 BURKE, J.J., SAKSHAUG, E.C., and SMITH, S.L.: The application of
gapless arresters on underground distribution systems, IEEETrans. PAS, 1981,
PAS-100, (3), pp. 12341243
15 NIEBUHR, W.D.: Protection of underground systems using metal-oxide surge
arresters, IEEE Trans. Ind. Appl., 1982, IA-18, (2), pp. 188191
16 REID, W.E., SECREST, S.R., HASSLER, S.P., and POTTER, M.E.: MOV
arrester protection of shield interrupts on 138 kV extruded dielectric cables,
IEEE Trans. PAS, 1984, PAS-103, (11), pp. 33343341
17 PARMIGIANI, B., QUAGGIA, D., ELLI, E., and FRANCHINA, S.: Zinc
oxide sheath voltage limiter for hv end ehv power cables: field experience and
laboratory tests, IEEE Trans. Power Deliv., 1986, PWRD-1, (1), pp. 164170
18 POTTER, M.E., and SOKOLY, T.O.: Development of metal oxide varis-
tors for gas-insulated surge arresters, IEEE Trans. PAS, 1982, PAS-101, (7),
pp. 22172220
19 MILLER, D.B., REUTHER, R.W., NOURAI, A., and TAHILIANI, V.:
Development of gapless metal oxide arresters for 362 kV gas insulated
substations, IEEE Trans. PAS, 1982, PAS-101, (7), pp. 21782186
20 IMATAKI, M., UJITA, K., FUJIWARA, Y., ISHIBE, S., and NITTA, T.:
Advanced metal oxide surge arrester for gas insulated switchgear (GIS),
IEEE Trans. PAS, 1984, 103, (10), pp. 29902998
21 STEINSTROM, L., LINDBERG, P., and SAMUELSSON, J.: Testing proce-
dure for metal oxide varistors protecting EHV series capacitors, IEEE Trans.
Power Deliv., 1988, 3, (2), pp. 568583
22 HAMANN, J.R., MISKE, S.A., JOHNSON, I.B., and COURTS, A.L.: Azinc
oxide varistor protective system for series capacitors, IEEE Trans. PAS, 1981,
PAS-100, (3), pp. 929937
23 COURTS, A.L., HINGORANI, N.G., and STEMLER, G.E.: A new series
capacitor protectionscheme usingnon-linear resistors, IEEETrans. PAS, 1978,
PAS-97, (4), pp. 10421052
24 McGRANAGHAN, M.F., REID, W.E., LAW, S.W., and GRESHAM, D.W.:
Overvoltage protection of shunt capacitor banks using MOV arresters, IEEE
Trans. PAS, 1984, PAS-103, (8), pp. 23262336
25 RIBEIRO, J.R., and McCALLUM, M.E.: An application of metal oxide surge
arresters in the elimination of need for reclosing resistors in EHV breakers,
IEEE Trans. Power Deliv., 1989, 4, (1), pp. 282291
26 HARDER, J.E.: Metal oxide arrester ratings for rotating machine protection,
27 NIEBUHR, W.D.: Application of metal-oxide-varistor surge arresters on
distribution systems, IEEE Trans. PAS, 1982, PAS-101, (6), pp. 17111715
28 WESTROM, A.C., and SCOTT, W.L.: Increasing the reliability of MOV
arresters, IEEE Trans. Power Deliv., 1990, 5, (4), pp. 18231829
29 KOCH, R.E., TIMOSHENKO, J.A., ANDERSON, J.G., and SHIH, C.H.:
Design of zinc oxide transmission line arresters for application on 138 kV
towers, IEEE Trans. PAS, 1985, PAS-104, (10), pp. 26752680
30 KOCH, R.E., and SONGSTER, H.J.: Development of a non-fragmenting
distribution surge arrester, IEEE Trans. PAS, 1984, PAS-103, (11),
pp. 33423352
31 HARUMOTO, Y., TSUDA, Y., KOBAYASHI, M., MIZUNO, M., and
AOKI, T.: Evaluation for application of built-in type zinc gapless surge
arresters for power system equipments, IEEE Trans. Power Deliv., 1987,
PWRD-2, (3), pp. 750757
32 VOGLER, L., SMITH, C.T., and TAYLOR, T.F.: Some aspects of design
and performance of metal oxide surge arresters. IEE conference on Lightning
protection, 1984, pp. 107113
33 HAHN, E.E.: Some electrical properties of zinc oxide semiconductor, J. Appl.
Phys., 1951, 22, (7), pp. 855863
34 MATSUOKA, M., MATSUYAMA, T., and IIDA, Y.: Voltage nonlinearity of
zinc oxide ceramics doped with alkali earth metal oxide, Jpn. J. Appl. Phys.,
1969, 8, pp. 12751276
35 MASUYAMA, T., and MATSUOKA, M.: Current dependence of voltage non-
linearity in SiC varistors, Jpn. J. Appl. Phys., 1968, 7, p. 1294
36 LEVINE, J.D.: Theory of varistor electronic properties, CRC Crit. Rev. Solid
State Sci., 1975, 5, pp. 597608
37 MATSUOKA, M.: Nonohmic properties of zinc oxide ceramics, Jpn. J. Appl.
Phys., 1971, 10, (6), pp. 737746
38 LEVINSON, L.M., and PHILIPP, H.R.: ZnO varistors for transient
protection, IEEE Trans. Parts Hybrids Packag., 1977, PHP-13, (4),
pp. 338343
39 LEVINSON, L.M., and PHILIPP, H.R.: The physics of metal oxide varistors,
J. Appl. Phys., 1975, 46, (3), pp. 13321341
40 PHILIPP, H.R., and LEVINSON, L.M.: Low-temperature electrical studies
on metal-oxide varistors, a clue to conduction mechanisms, J. Appl. Phys.,
1977, 48, (4), pp. 16211627
41 BERNASCONI, J., KLEIN, H.P., KNECHT, B., and STRASSLER, S.: Inves-
tigation of various models for metal oxide varistors, J. Electron. Mater., 1976,
5, (5), pp. 473495
42 ROSSINELLI, M., BLATTER, G., and GREUTER, F.: Grain boundary
properties of ZnO varistors, Proc. Br. Ceram. Soc., 1986, pp. 117
43 EDA, K.: Grain growth control in ZnO varistors using seed grains, J. Appl.
Phys., 1983, 54, (2), pp. 10951099
44 BOWEN, L.J., and AVELLA, F.J.: Microstructure, electrical properties and
failure prediction in low clamping voltage zinc oxide varistors, J. Appl. Phys.,
1983, 54, (5), pp. 27642772
45 TRONTELJ, M., KOLAR, D., and KRASEVEC, V.: Properties of ZnO
ceramics containing TiO
2
, Proc. Br. Ceram. Soc., 1986, pp. 143151
46 ASOKAN, T., NAGABHUSHANA, G.R., and IYENGAR, G.N.K.: Improve-
ment of non-linear characteristics of multicomponent ZnO-based ceramics
containing Nb
2
O
5
, IEEE Trans. Electr. Insul., 1988, 23, (2), pp. 279287
47 SWEETANA, A., GUPTA, T., CARLSON, W., GREKILA, R., KUNKLE, N.,
and OSTERHOUT, J.: Gapless surge arresters for power systems applications,
vol. 1: development of 500 and 1200 kV arresters. EPRI EL 3166, 1, Project
657-1, final report, September 1983
48 MUKAE, K., TSUDA, K., and SHIGA, S.: Zinc oxide-praseodymium
oxide elements for surge arresters, IEEE Trans. Power Deliv., 1988, 3, (2),
pp. 591597
49 MUKAE, K., TSUDA, K., and NAGASAWA, I.: Capacitance vs. voltage
characteristics of ZnO varistors, J. Appl. Phys., 1979, 50, (6), pp. 44754476
50 BI-CHIOU CHIOU, and JIH, F.W.: Studies on ZnO-doped varistors with
conventional pellet form and thick film form, Proc. Br. Ceram. Soc., 1986,
pp. 129141
51 GREUTER, F., PERKINS, R., ROSSINELLI, M., and SCHMUCKLE, F.: The
metal-oxide resistor at the heart of modern surge arresters. ABB Rev. Surge
Arresters, 1989, 1
52 MORRIS, W.G.: Electrical properties of ZnO-Bi
2
O
3
ceramics, J. Am. Ceram.
Soc., 1973, 56, (7), pp. 360364
53 MAHAN, G.D., LEVINSON, L.M., and PHILIPP, H.R.: Theory of conduction
in ZnO varistors, J. Appl. Phys., 1979, 50, (4), pp. 27992812
54 MENTH, A., STREIT, P., and KNECHT, B.: Present state and development
potential of solid state varistors, in RAGALLER, K. (Ed.): Surges in high
voltage network (Plenum Press, New York, 1980), pp. 283298
55 TOMINAGA, S., SHIBUYA, Y., FUJIWARA, Y., and NITTA, T.: Electri-
cal properties of zinc oxide valve element for a surge arrester. IEEE Power
Engineering Society Summer Meeting, paper A78 595-1, Los Angeles, CA,
USA, 1621 July 1978, pp. 18
56 BLATTER, G., and GREUTER, F.: Carrier transport through grain boundaries
in semiconductors, Phys. Rev. B, 1986, 33, (6), pp. 39523966
57 JONSCHER, A.K.: Hopping losses in polarisable dielectric media, Nature,
1974, 250, pp. 191193
58 MANUEL, P.: Modelisation des mecanismes de conduction non lineaires dans
les varistances ZnO, Rev. Phys. Appl., 1987, 22, (9), pp. 971983
59 SATO, K., TAKADA, Y., and MAEKAWA, H.: Electrical conduction of ZnO
varistors under continuous DC stress, Jpn. J. Appl. Phys., 1980, 19, (5),
pp. 909917
60 SUZUOKI, Y., OHKI, A., MIZUTANI, T., and IEDA, M.: Electrical prop-
erties of ZnO-Bi
2
O
3
thin-film varistors, J. Phys. D, Appl. Phys., 1987, 20,
pp. 511517
61 BLATTER, G., and GREUTER, F.: Carrier transport through grain boundaries
in semiconductors, Phys. Rev. B, 1986, 33, (6), pp. 39523966
62 PIKE, G.E., and SEAGER, C.H.: The DC voltage dependence of semi-
conductor grain boundary resistance, J. Appl. Phys., 1979, 50, (5),
pp. 34143422
63 PIKE, G.E.: Grain boundaries in semiconductors, in PIKE, SEAGER and
LEAMY, (Eds): Grain boundaries in semiconductors. Proceedings of the
Annual Meeting Mat. Res. Proc., 1981, 5, pp. 369379
64 EDA, K.: Zinc oxide varistors, IEEE Electr. Insul. Mag., 1989, 5, (6),
pp. 2841
65 TOMINAGA, S., SHIBUYA, Y., FUJIWARA, Y., IMATAKI, M., and
NITTA, T.: Stability and long termdegradation of metal oxide surge arresters,
66 MIZUNO, M., HAYASHI, M., and MITANI, K.: Thermal stability and
life of the gapless surge arrester, IEEE Trans. PAS, 1981, PAS-100, (5),
pp. 26642671
67 FUJIWARA, Y., SHIBUYA, Y., IMATAKI, M., and NITTA, T.: Evaluation
of surge degradation of metal oxide surge arresters, IEEE Trans. PAS, 1982,
PAS-101, (4), pp. 978985
68 KAN, M., NISHIWAKI, S., SATO, T., KOJIMA, S., and YANABU, S.: Surge
discharge capability and thermal stability of a metal oxide surge arrester, IEEE
Trans. PAS, 1983, PAS-102, (2), pp. 282289
69 VICAUD, A.: ACvoltage ageing of zinc-oxide ceramics, IEEE Trans. Power
Syst., 1986, PWRD-1, (2), pp. 4958
70 KIRBY, P., ERVEN, C.C., and NIGOL, O.: Long term stability and energy
discharge capacity of metal oxide valve elements, IEEE Trans. Power Deliv.,
1988, 3, (4), pp. 16561665
71 LAT, M.V.: Thermal properties of metal oxide surge arresters, IEEE Trans.
PAS, 1983, PAS-102, (7), pp. 21942202
72 YEANFANG, W., YUYI, Z., and YUGIN, L.: Study of thermal stability of
metal oxide surge arrester. Fifth international symposium on High voltage
engineering, paper 83.01, Braunschweig, FRG, August 1987
73 KIRKBY, P.: The long term stability of valve elements used in metal oxide
surge arresters. CEAreport 199 T 425, July 1986
74 ST-JEAN, G., and PETIT, A.: Metal oxide surge arrester operating limits
defined by a temperature-margin concept, IEEE Trans. Power Deliv., 1990, 5,
(2), pp. 627633
75 TOMINAGA, S., AZUMI, K., NITTA, T., NAGAI, N., IMATAKI, M., and
KUWAHARA, H.: Reliability and application of metal oxide surge arresters
for power systems, IEEE Trans. PAS, 1979, PAS-98, (3), pp. 805812
76 NISHIKAWA, H., SUKEHARA, M., and YOSHIDA, T., et al.: Thyristor
valve gapless arrester element lifetime performance and thermal stability,
77 IEC99-4:1991 incorporating amendment No.1:1998: Surge arresters, part
4: metal oxide surge arresters without gaps for AC systems. IEC standard,
1998
78 JEC-217: Metal oxide surge arresters. Standard of the Japanese Electrotech-
nical Committee, published by the IEEJ, 1984
79 RICHTER, B.: Investigation of MOsurge arrester under constant high voltage
stress by application of potential-free measuring devices. Fifth international
symposium on High voltage engineering, paper 82.10, Braunschweig, FRG,
1987
80 RICHTER, B., and HEINRICH, B.: Investigation of the stresses of MO surge
arresters for high voltage application. Sixth international symposium on High
voltage engineering, paper 26.02, New Orleans, USA, 1989
81 YUFANG, W., YUJING, L., YUNYI, Z., and LIBO, H.: Study of thermal
stability of 500 kV MOAprorated section. Sixth international symposium on
High voltage engineering, paper 26.10, New Orleans, USA, 1989
82 HINRICHSEN, V., and PEISER, R.: Simulation of the electrical perfor-
mance and thermal behaviour of metal oxide surge arresters under AC stress.
Sixth international symposium on High voltage engineering, paper 26.04,
Braunschwieg, FGR, 1987
83 HINRICHSEN, V., and PEISER, R.: Simulation of the AC performance
of gapless ZnO arresters. Fifth international symposium on High voltage
engineering, paper 82.09, Braunschwieg, FGR, 1987
84 PHILIPP, H.R., and LEVINSON, L.M.: Degradation phenomenon in zinc
oxide varistors: a review, Advances in Ceramics, 1984, 7, pp. 121
85 NAWATA, M., KAWAMURA, H., and IEDA, M.: Studies on degradation
mechanism of zinc oxide ceramics varistor by thermally simulated current.
Fifth international symposium on High voltage engineering, paper 82.06,
Braunschweig, FRG, 1987
86 STUCKI, F., BRUESCH, P., and GREUTER, F.: Electron spectroscopic
studies of electrically active grain boundaries in ZnO, Surf. Sci., 1987,
189/190, pp. 294299
87 DE MORAIS, H.M., DA COSTA, E.G., and NOWAKI, K.: Ageing of zinc
oxide blocks through partial discharges. Fifth international symposium on
High voltage engineering, paper 83.05, Braunschweig, FRG, 1987
88 SAKSHAUG, E.C., KRESGE, J.S., MARK, D.A., and KARADY, G.G.:
Contamination and hot wash performance of zinc oxide station arresters,
89 OYAMA, M., OHSHIMA, I., HONDA, M., YAMASHITA, M., and
KOJIMA, S.: Life performance of zinc-oxide elements under DC voltage,
90 ZANDER, W.: Influence of material heterogeneity on the ageing of ZnO
arrester material. Fifth international symposium on High voltage engineering,
paper 82.05, Braunschweig, FRG, 1987
91 MIZUKOSHI, A., OZAWA, J., SHIRAKAWA, S., and NAKANO, K.: Influ-
ence of uniformity on energy absorption capabilities of zinc oxide elements as
applied in arresters, IEEE Trans. PAS, 1983, PAS-102, (5), pp. 13841390
92 BRONIKOWSKI, R.J., and DUPONT, J.P.: Development and testing
of MOVE arrester elements, IEEE Trans. PAS, 1982, PAS-101, (6),
pp. 16381643
93 NISHIWAKI, S., KIMURA, H., SATOH, T., MIZOGUSHI, H., and
YANABU, S.: Study of thermal runaway/equivalent prorated model of a ZnO
surge arrester, IEEE Trans. PAS, 1984, PAS-103, (2), pp. 413421
94 VONG, N.M., MILLER, R., RYDER, D.M., DOONE, R.M., and
SPARROW, L.: Deterioration in the performance of zinc oxide surge arrester
elements due to partial discharges. Sixth international symposium on High
95 KNOBLOCH, H.: The influence of the surrounding medium on the service
behaviour of metal oxide varistors. Sixth international symposium on High
96 DEMYANENKO, K.B., and SERGEEV, A.S.: Stability of highly non-linear
zinc oxide resistors under the prolonged action of a commercial frequency
potential, Elektrotekhnika, 1984, 55, (9), pp. 4651
97 SAKSHAUG, E.C., BURKE, J.J., and KRESGE, J.S.: Metal oxide arresters on
distribution systems fundamental considerations, IEEE Trans. Power Deliv.,
1989, 4, (4), pp. 20762089
98 YUFANG, W., YUJIN, L., and YUYI, Z.: Working stability of MOV in the
oil. Sixth international symposium on High voltage engineering, paper 41.02,
New Orleans, USA, 1989
99 EDA, K.: Destruction mechanism of ZnO varistors due to high currents,
J. Appl. Phys., 1984, 56, (10), pp. 29482955
100 SHUFANG, L., YUJIN, L., and YUYI, Z.: Energy absorption character-
istic of metal oxide elements under transient overvoltage. Fifth interna-
tional symposium on High voltage engineering, paper 83.02, Braunschweig,
FRG, 1987
101 DAKIN, T.W.: Electrical insulation deterioration treated as a chemical rate
phenomenon, AIEE Trans., 1948, 67, pp. 113122
102 CARLSON, W.G., GUPTA T.K., and SWEETANA, A.: A procedure
for estimating the lifetime of gapless metal oxide surge arresters for AC
application, IEEE Trans. Power Syst., 1986, PWRD-1, (2), pp. 6774
103 HADDAD, A., NAYLOR, P., METWALLY, I., GERMAN, D.M., and
WATERS, R.T.: An improved non-inductive impulse voltage measurement
technique for ZnO surge arresters, IEEE Trans. Power Deliv., 1995, 10, (2),
pp. 778784
104 KIZEBETTER, V.E., SEGEEV, A.S., and FIRSOV, A.V.: Calculation of
the distribution of voltage and current along the elements of an overvoltage
limiter with contaminated and wet covering, Elektrotekhnika, 1987, 58, (3),
pp. 2127
105 PEIBAI, Z., GUOHANG, W., HUIMING, M., JIALU, S., CIZHANG, F.,
and NAIXIANG, M.: Analysis of the potential distribution of gapless
surge arrester. Sixth international symposium on High voltage engineering,
paper 26.09, New Orleans, LA, USA, 1989
106 BENZAOUA, F., HADDAD, A., ROWLANDS, A.R., and WATERS, R.T.:
Acircuit model for polluted ZnOpolymeric surge arresters. 10th international
symposium on High voltage engineering (ISH), Montreal, Canada, 1997, 3,
pp. 317320
107 ROUSSEAU, A., GOURMET, P., NIKLASCH, H., and BURET, F.: A new
use of pollution tests on ZnOarresters. Sixth international symposiumon High
voltage engineering, paper 27.17, New Orleans, LA, USA, 1989
108 LENK, D.W.: An examination of the pollution performance of gapped and
gapless metal oxide station class surge arresters, IEEE Trans. PAS, 1984,
PAS-103, (2), pp. 337345
109 LENK, D.W., STOCKUM, F.R., and GRIMES, D.E.: Anew approach to dis-
tribution arrester design, IEEE Trans. Power Deliv., 1988, 3, (2), pp. 584590
110 OYAMA, M., OHSHIMA, I., HONDA, M., YAMASHITA, M., and
KOJIMA, S.: Analytical and experimental approach to the voltage distribution
on gapless zinc-oxide surge arresters, IEEE Trans. PAS, 1981, PAS-100, (11),
pp. 46214627
111 KOJIMA, S., OYAMA, M., and YAMASHITA.: Potential distributions of
metal oxide surge arresters under various environmental conditions, IEEE
Trans. Power Deliv., 1988, 3, (3), pp. 984989
112 RICHTER, B.: Application of fibre optics to the measurement of leakage
current in metal oxide surge arresters. Fourth international symposium on
High voltage engineering, paper 64.05, Athens, Greece, 1983
113 HIDAKA, K., and KOUNO, T.: Simultaneous measurements of two orthog-
onal components of electric field using a Pockel device, Rev. Sci. Instrum.,
1989, 60, (7), pp. 12521257
114 LEVINSON, L.M., and PHILIPP, H.R.: High frequency and high-currents in
metal-oxide varistors, J. Appl. Phys., 1976, 47, pp. 31173121
115 MATSUURA, M., and YAMAOKI, H.: Dielectric dispersion and equiva-
lent circuit in nonohmic ZnO ceramics, Jpn. J. Appl. Phys., 1977, 16, (7),
pp. 12611262
116 LEVINSON, L.M., and PHILIPP, H.R.: Long time polarization currents in
metal oxide varistors, J. Appl. Phys., 1976, 47, (7), pp. 31773181
117 LAGRANGE, A.: Les varistances a base doxide de zinc: des elements de
protection a letat solide en plein development, Rev. Gen. Electr., 1986, (9),
pp. 918
118 DANIEL, V.: Dielectric relaxation (Academic Press, London and NewYork,
1967)
119 CORDARO, J.F., MAY, J.E., and SHIM, Y.: Bulk electron traps in zinc oxide
varistors, J. Appl. Phys., 1986, 60, (12), pp. 41864190
120 ELLI, E., WARNOTS, E., and NASCIMENTO, L.F.M.: Laboratory tests
to characterize metal-oxide surge arresters, Energ. Elettr., 1986, (2),
pp. 8188
121 HORIUCHI, S., ICHIKAWA, F., MIZUKOSHI, A., KURITA, K., and
SHIRAKAWA, S.: Power dissipation characteristics of zinc-oxide arresters
for HVDC systems, IEEE Trans. Power Deliv., 1988, 3, (4), pp. 16661671
122 DANG, C., PARNELL, T.M., and PRICE, P.J.: The response of metal oxide
surge arresters to steep fronted current impulses, IEEE Trans. Power Deliv.,
1986, PWRD-1, (1), pp. 157163
123 RICHTER, B., KRAUSE, C., and MEPPELINK, J.: Measurement of the UI
characteristic of MO-resistors at current impulses of different wave shapes
and peak values. Fifth international symposium on High voltage engineering,
paper 82.03, Braunschweig, Germany, 1987
124 BREILMANN, W.: Protective characteristic of complete zinc-oxide arrester
and of single elements for fast surges. Fifth international symposium on High
voltage engineering, paper 82.04, Braunschweig, Germany, 1987
125 EDA, K.: Transient conduction phenomena in non-ohmic zinc oxide
ceramics, J. Appl. Phys., 1979, 50, (6), pp. 44364442
126 MODINE, F.A., and WHEELER, R.B.: Fast pulse response of zinc-oxide
varistors, J. Appl. Phys., 1987, 61, (8), pp. 30933098
127 TUA, P.F., ROSSINELLI, M., and GREUTER, F.: Transient response of elec-
trically active grain boundaries in polycrystalline semiconductors, Phys. Scr.,
1988, 38, (3), pp. 491497
128 SCHMIDT, W., MEPPELINK, J., RICHTER, B., FESER, K., KEHL, L., and
QIU, D.: Behaviour of MO-surge arrester blocks to fast transients, IEEE
129 PHILIPP, H.R., and LEVINSON, L.M.: ZnO varistors for protection
against nuclear electromagnetic pulses, J. Appl. Phys., 1981, 52, (2),
pp. 10831090
130 HADDAD, A., NAYLOR, P., METWALLY, I., GERMAN, D.M., and
WATERS, R.T.: An improved non-inductive impulse voltage measurement
technique for ZnO surge arresters, IEEE Trans. Power Deliv., 1995, 10, (2),
pp. 778784
131 HADDAD, A., and NAYLOR, P.: Dynamic impulse conduction in ZnO
arresters. 11th international symposium on High voltage engineering (ISH),
London, UK, 1999, 2, pp. 254257
132 HADDAD, A., FUENTES-ROSADO, J., GERMAN, D.M., andWATERS, R.T.:
Characterisation of ZnO surge arrester elements with direct and power
frequency voltages, IEE Proc., 1990, 137, (5), pp. 269279
133 HADDAD, A., ELAYYAN, H.S.B., GERMAN, D.M., and WATERS, R.T.:
ZnO surge arrester elements with mixed direct and 50 Hz voltages, IEE Proc.
A, Sci. Meas. Instrum. Manage. Educ., 1991, 138, (5), pp. 265272
134 TREGUIER, J.P.: Equipment de telecommunications: varistances a base
doxyde de zinc avec dautres composants pour un meilleur niveau de
protection, Rev. Gen. Electr., 1986, (9), pp. 2431
135 EKSTROM, A.: Application guide for metal oxide arresters without gaps for
HVDC converter stations. CIGRE Working Group 33/14-05, June 1988
136 PIKE, G.E.: Electronic properties of ZnO varistors, a new model, in Grain
boundaries in semiconductors. Annual Meeting Material Research Society
Proceedings, Boston MA, 1981, 5, pp. 369379
137 PIKE, G.E., GOURLEY, P.L., and KURTZ, S.R.: Impact ionisation near GaAs
grain boundaries, Appl. Phys. Lett., 1989, 43, (10), pp. 939941
138 EMTAGE, P.R.: The physics of zinc oxide varistors, J. Appl. Phys., 1977, 48,
(10), pp. 43724384
139 KORN, S.R., MARTZLOFF, F.D., MAY, J., and WOLFF, B.L.: Transient
voltage suppression, in SMITH, M.C., and McCORMICK, M.D. (Eds):
Manual of semiconductors (Production Dept. (G.E.C.), 1982)
140 HONG-YANG ZHENG: Study on the power loss characteristics of ZnO
arrester elements used in H.V.D.C. converter station. Fifth international sym-
posium on High voltage engineering, paper 83.03, Braunschweig, FRG,
1987
141 YU-JIN LIANG, YU-YI ZHAO, and HONG-YAN ZENG: Survey of low
current characteristics of metal oxide varistor. Fifth international symposium
on High voltage engineering, paper 82.11, Braunschweig, FRG, 1987
142 STOKES, A.D.: High voltage line-dropping switching stresses, IEE Proc. C,
Gener. Transn. Distrib., 1992, 139, (1), pp. 2126
143 CHI, H.: The transient behaviour of ZnO varistors under DC service voltage.
Seventh international symposium on High voltage engineering, paper 85.01,
Dresden, Germany, 1991
144 HADDAD, A., METWALLY, I., NAYLOR, P., TONG, Y.K., and
WATERS, R.T.: Impulse response of ZnO arresters with pre-existing direct
voltage stress. Eighth international symposium on High voltage engineering,
paper 77.04, Yokohama, Japan, 1993, pp. 417420
145 HADDAD, A., GERMAN, D.M., METWALLY, I., NAYLOR, P.,
GRIFFITHS, H., and WATERS, R.T.: The transient response of ZnO surge
arresters to sudden voltage reversal. 28th universities Power engineering
conference, Staffordshire University, UK, 1993, pp. 482485
146 NAYLOR, P., HADDAD, A., METWALLY, I., TONG, Y.K., GERMAN, D.M.,
and WATERS, R.T.: Trapped charge effects on the transient response of surge
arresters. 29th universities Power engineering conference, Galway, Ireland,
1994, pp. 441444
147 HADDAD, A., NAYLOR, P., TONG, Y.K., MARLEY, W.A.,
GERMAN, D.M., and WATERS R.T.: Direct voltage and trapped charge
effects on the protective characteristic of ZnO surge arresters, IEE Proc., Sci.
Meas. Technol., 1995, 142, (6), pp. 442448
148 HADDAD, A., NAYLOR, P., TONG, Y.K., METWALLY, I., GERMAN, D.M.,
and WATERS R.T.: Trapped-charge and point-on-wave effects on the
protective characteristic of surge arresters. 9th international symposium
on High voltage engineering (ISH), Graz, Austria, 1995, 4, pp. 7558-1 to
7550-4
149 ANSI/IEEE C62.11:1987: IEEE Standard for metal-oxide surge arresters for
ac power circuits. American National Standard, 1987
150 SARGENT, R.A., DUNLOP, G.L., and DARVENIZA, M.: Effects of
multiple impulse currents on the microstructure and electrical properties
of metal oxide varistors, IEEE Trans. Electr. Insul., 1992, 27, (3),
pp. 586592
151 DARVENIZA, M., MERCER, D.R., and TUMMA, L.R.: Effects of
multiple stroke lightning impulse currents on ZnO arresters. CIRED, 1993,
pp. 2.25.12.25.5
152 HADDAD, A., ABDUL-MALEK, Z., GERMAN, D.M., and WATERS, R.T.:
Double impulse testing of ZnO surge arresters. 23rd international conference
on Lightning conference (ICLP), Florence, Italy, 1996, II, pp. 696702
153 HADDAD, A., andABDUL-MALEK, Z.: Anewrotating spark gap, Rev. Sci.
Instrum., 1997, 68, (12), pp. 15
154 OZAWA, J., OOSHI, K., SHIRAKAWA, S., NAKANO, K., MUZUKOSHI, A.,
and MARUYAMA, S.: Fast transient response and its improvement of metal
oxide surge arresters for GIS. Sixth international symposium on High voltage
engineering, paper 26.03, New Orleans, USA, August 1989
155 KNECHT, B.: Solid-state arresters, in RAGALLER, K. (Ed.): Surges in high
voltage networks (Plenum Press, New York, 1980) pp. 299321
156 IEEE working group 3.4.11: Modelling of metal oxide surge arresters, IEEE
157 KIM, I., FUNABASHI, T., SASAKI, H., HAGIWARA, T., and
KOBAYASHI, M.: Study of ZnO arrester model for steep front wave, IEEE
158 HADDAD, A., and NAYLOR, P.: Finite element computation of capacitance
networks in multiple electrode systems: application to ZnO surge arresters,
IEE Proc., Sci. Meas. Technol., 1998, 145, (4), pp. 129135
159 HADDAD, A., and NAYLOR, P.: Dynamic impulse conduction in ZnO
arresters, 11th international symposium on High voltage engineering (ISH),
London, UK, 1999, 2, pp. 254257
160 BREDER, H., and COLLINS, T.: Supervision of gapless zinc oxide surge
arresters, IEE Conf. Publ. (236), London, 1984
161 SHIRAKAWA, S., ENDO, F., and KITAJIMA, H., et al.: Maintenance of
surge arrester by portable arrester leakage current detector, IEEETrans. Power
Deliv., 1988, 3, (3), pp. 9981003
162 LINDQUIST, J., STRENSTROM, L., SCHEI, A., and HANSEN, B.: New
method for measurement of the resistive leakage currents of metal oxide
surge arresters in service, IEEE Trans. Power Deliv., 1990, 5, (4),
pp. 18111819
163 DEEP, G.S., and NOWACKI, K.: Measurement of resistive leakage current of
metal-oxide surge arresters. Fourth international symposium on High voltage
engineering, paper 64.07, Athens, Greece, 59 September 1983
164 NAIDU, S.R., and SRIVASTAVA, K.D.: The validity of circuits for measur-
ing the resistive leakage current of zinc oxide arresters. Fifth international
symposium on High voltage engineering, paper 82.12, Braunschweig, FRG,
2428 August 1987
165 VICAUD, A., and EGUIAZABAL, D.: Characterisation electrique sous faible
courant de ceramique ZnO pour parafoudres. Societe des electriciens et
des electroniciens (SEE) conference on Varistance a base doxide de zinc,
paper B-11, pp. 8587, Ecole Superieure dElectricite, Gif-sur-Yvette, France,
6 March 1986
166 IEC 60099-5:1996 incorporating amendments 1:1999: Surge arresters, part 5:
selection and application recommendation. IEC Standard, 1996
167 SPELLMAN, C., and HADDAD, A.: A technique for on-line condition
monitoring of ZnO surge arresters. 10th international symposium on High
voltage engineering (ISH), Montreal, Canada, 1997, 4, pp. 151154
168 IEC507:1991: Artificial pollution tests on high-voltage insulators to be used
on AC systems. IEC Standard, 1991
169 IEC815:1986: Guide for the selection of insulators inrespect of polluted
conditions. IEC Standard, 1986
170 CIGRE Task Force 33.04.01: Polluted insulators: a review of current
knowledge. Brochure 158, 2000, pp. 134185
171 HADDAD, A., ABDUL-MALEK, Z., GERMAN, D.M., and WATERS, R.T.:
Coordination of spark gap protection with zinc oxide surge arresters, IEE
Proc., Gener., Trans. Distrib., 2001, 148, (1), pp. 2128
172 ABDUL-MALEK, Z., HADDAD, A., GERMAN, D.M., and WATERS, R.T.:
Current sharing in multi-column surge arresters. 10th international
symposium on High voltage engineering (ISH), Montreal, Canada, 1997,
5, pp. 145148
173 ANSI/IEEEC62.22: 1991: IEEEguide for the application of metal-oxide surge
arresters for alternating-current systems. IEEE Standard, 1991
174 IEC 71-2:1996: Insulation co-ordination, part 2: application guide.
IEC Standard, 1996
Chapter 6
Insulators for outdoor applications
D.A. Swift
6.1 Introduction
The product developments that have taken place since John Looms published, in
1988, his excellent book [1] that covered much of this subject have mainly concerned
polymeric materials especially silicone rubber. Although the principal modifications
made to porcelain and glass insulators over this period are those required for high
voltage DC applications, much more has nonetheless been learned about the
performance of such insulators under AC energisation. The discussed proposals of
some years ago for ultra high voltage transmission systems resulted in the production
of very high mechanical strength insulators of the cap and pin design. However, as
such proposals generally failed to come to fruition, at least one insulator manufacturer
now has a product that is ready, should a market become available in the future.
The design, the specification for a certain usage and the dimensioning of the
insulator for that usage in an outdoor environment are usually dominated by the need
to take account of wetted pollution on its surface. Often, the consequential problem
is that of flashover across this polluted surface when the leakage current exceeds a
certain value. However, some potentially dangerous defects in the structure of such
insulators have been known to occur at values of leakage current well below that
of the critical flashover limit. The associated mechanisms are quite involved and,
although studied in some detail already, are still not fully understood.
Essentially, the troublesome pollution comes in two forms. The soluble compo-
nents e.g. salts from the sea and industrial gases that result in weak acids being
formed produce an electrolyte when dissolved in water. This water is provided by
condensation from humid air, fog, mist and drizzle. Rain is usually helpful, but not
always so especially when heavy. Generally, it washes the pollution off the surface
before it reaches the troublesome magnitude. The exceptions to this general situation
have needed certain insulator profiles to be used or palliatives to be adopted to
minimise the probability of flashover under such heavy wetting conditions.
Paradoxically, insulators in some desert regions are severely affected. This occurs
because the sand can contain a very high concentration of salt. For example, in parts of
Tunisia chemical analyses have widely confirmed the aggressiveness of desert sand
because it may contain up to 18 per cent of salt (Znaidi, private communication).
Presumably such situations arise because, at some time in the past, such land was
below sea level.
Afurther point worthy of general mention is pollution flashover caused by snow
and ice. Here, the main problem lies with the difficulty in standardising an artificial
pollution test. As the salt fog test [1] took many years to be accepted internationally,
it can be reasoned that only young people have sufficient time available in their
working lifetime to attempt such an enterprise if they wish to reap benefit in personal
recognition of their research efforts!
Before proceeding to explain the specific details of leakage current, it is useful
to mention the state of the polluted surface from the point of view of its propensity
to become wetted. The usually used terms are hydrophilic i.e. surface readily wets
so that a film of water is produced and hydrophobic i.e. water resides as discrete
beads. This difference is due to the magnitude of the free surface energy; the lower
value being for the hydrophobic case.
Many thousands of scientific papers, covering both practical and theoretical
issues, have been published on this subject. The rate of publication continues to
increase as the years go by. At the International Symposium on High Voltage Engi-
neering, which is a biennial event, at least ten percent of the papers often being at
least five hundred in total cover this subject. One wonders why is this so? The reason
is related to the strategic importance of power transmission to the wellbeing of all
industrial nations and the complex issues involved in trying to explain the flashover
and some longer lasting failure processes of the insulating components of this system.
A major review of polluted insulators, fairly recently published as a brochure [2] by
the International Conference on Large High Voltage Electric Systems (CIGRE, which
is the acronym of the French version of this title) has been prepared by a taskforce
of 15 experts from 12 countries. This CIGRE publication comprises 185 pages that
summarise research published in 382 references. The same international committee is
currently working on a companion brochure dealing with guidelines for the selection
and dimensioning of insulators for use in polluted environments [3] hopefully, to be
published in early 2005. Their research findings are currently being used by the Inter-
national Electrotechnical Commission (IEC) to revise an old document concerning
polluted insulators [4] probably not to be completed until at least 2006. Although
these new CIGRE and IEC documents deal essentially with the same subject, the aim
of the IEC publication is to achieve a more rule of thumb approach than that being
adopted by CIGRE.
6.2 Role of insulators
Insulators for an overhead power line (OHL), in many substation applications and
on the overhead electrification systems of railways, must, primarily, support the
Insulators for outdoor applications 259
conductors. Also important, as already mentioned, is the need to avoid frequent
flashover events from occurring.
Although the total mechanical failure of such an insulator is, fortunately, a rare
event, its occurrence may be very serious. For example, should a vertical insulator
of an OHL (often referred to as a suspension unit) break, then its conductor could
be supported by the insulators of the neighbouring support structures (often called
towers) at either side. Then, it is possible that this conductor could be reenergised but
with little ground clearance!
The consequences of a flashover vary from being annoying to being very costly.
For example, the damage resulting from the external flashover of the insulating
housing of a high power circuit breaker during a synchronising operation, when
the voltage across the polluted surface can increase to twice the normal value, could
be extremely large.
6.3 Material properties
Table 6.1 provides a summary of the directly relevant properties in a generic way of
the mechanical properties of materials that are currently in use for outdoor insulators.
These values will be referred to later as various aspects of designs and problems are
discussed.
Because silicone rubber nowseems tobe preferredfor insulators that are employed
in highly polluted environments, their properties with respect to hydrophobicity war-
rant a special mention. High temperature vulcanised (HTV) silicone rubber, which is
now the usual variant of this polymer for constructing insulators, is commonly com-
posed of polydimethysiloxane and alumina trihydrate (ATH) filler. This filler adsorbs
the heat from electrical discharges and thereby prevents surface tracking from taking
place. Many researchers think that the hydrophobicity of silicone rubber insulators is
due to the low molecular weight (LMW) silicone chains, comprising methyl groups
attached to silicone atoms, that have diffused to the surface. Upon breaking this
attachment (e.g. by electrical discharges), the hydrophobicity of this surface is lost
until it becomes recovered by a new lot of LMW. As ATH affects the mobility of this
oil-like substance, it plays an important part in the recovery rate of hydrophobicity.
Table 6.1 Mechanical properties of the dielectric components of high
voltage insulators [2]
Units Porcelain Glass Polymer* RBGF**
Tensile strength MPa 30100 100120 2035 13001600
Compressive strength MPa 240820 210300 80170 700750
* silicone and carbon based
** resin-bonded glass fibre used for the core of a polymeric insulator
That is, the different formulations used by the various manufacturers may result in
different recovery rates. When the silicone rubber surface becomes polluted, it is
variously thought that, in time, these LMWs cover the pollutants by a process due to
evaporation or diffusion, or are related to the capillary effect (i.e. surface tension).
Therefore, the size and shape of the solid particles (e.g. spherical, flakes) may well
affect the rate of this transfer. Another complicating factor is that LMWs come in two
forms: one linear, the other cyclic. The former seems to be more effective in covering
the pollutants. Obviously, the available quantity of LMW is important. In newly pro-
duced HTVsilicone rubber, it is about three per cent by mass. Therefore, a pessimistic
assessment is that the rate and the amount of LMW diffusion from the bulk to the
surface of the silicone rubber, and hence onto the pollutants, decreases as the material
ages under service stress and weathering. However, there is a well supported postulate
that the quantity of LMWs is maintained because the bulk material tries to achieve
thermodynamic equilibrium. That is, silicone rubber is a dynamic material and so the
effects of ageing have to be assessed from this viewpoint. Asilicone rubber coating is
gaining increasing use as a way of enhancing the pollution flashover performance of
porcelain insulators. Typically, such coating materials are made from room tempera-
ture vulcanised (RTV) silicone rubber. In its newly produced state, the LMWcontent
is about five per cent by mass, but it is of the cyclic form.
Some comments concerning the resin-bonded glass fibres (RBGF) used in the
core of a polymeric insulator are also necessary. Although this structure has enor-
mous tensile strength when compared with that of glass and porcelain (Table 6.1)
a somewhat disturbing type of failure usually called brittle fracture has occurred
in a few cases. This defect appears as a very sharp break in the cross-section of the
rod as though it had been sliced with a knife. An elegant explanation [5], which still
remains to be proved, involves the replacement of the aluminium and calcium ions in
the glass with hydrogen ions from an acid. Various acids have been suggested includ-
ing oxalic acid resulting from electrical discharges within the structure that produce
reactive oxygen and oxides of nitrogen that combine with polymer carbon. In this ion
exchange reaction, the proton is smaller than the replaced metal ion thereby inducing
a tensile stress in the surface of the glass that causes spontaneous cracking. Such
fractures have been shown to happen readily on non-coated fibres of standard E-type
glass. Although numerous attempts have been made in the laboratory to produce this
type of failure using commercially produced RBGF rod and realistic acid concentra-
tions, none is known to have succeeded. Therefore, it is tempting to conclude that
there is no problem until the glass in the fibre is stripped of its coating of epoxy resin.
Perhaps a weakness has been built into a few insulators during the manufacturing
stage. A French/Swiss research team has fairly recently proposed [6] that incorrect
curing of the resin may cause such a weakness.
6.4 Examples of design
In general terms, the various shapes of insulators can be divided into two groups:
discs and cylinders. Of the former, the cap and pin designs for OHLconstructions are
the most numerous. The latter group embraces such designs as longrods, of both the
porcelain and the polymeric types, posts and barrels.
6.4.1 Cap and pin insulators
For areas inwhichthere is little pollutionthe standardprofile, Figure 6.1a, is adequate.
When, however, pollution is a major problem then two measures are adopted. Either
the leakage path length is increased by having substantially deeper ribs as shown
in the anti-fog profile of Figure 6.1b, or the surface is made very smooth to minimise
the catch and retention of pollutants as illustrated in Figure 6.1c by the aerodynamic
pin
a
b
c
cap
shell
mortar
Figure 6.1 Typical examples of the profile of a cap and pin insulator
a standard design
b anti-fog design
c aerodynamic design
profile. Often, the anti-fog design is used in coastal and heavy industrial regions and
the aerodynamic design is employed in deserts.
The advantage, from the mechanical viewpoint, of the cap and pin design is that
it employs the insulating material glass or porcelain in its compressive mode,
because these materials are very strong (see Table 6.1) under this type of mechanical
stress. This desirable arrangement to support the axial load is obtained by the con-
ical shape of the part of the steel pin embedded within the mortar that fixes it to a
wraparound cast iron cap.
The seemingly obvious shortcoming of the cap and pin insulator is the large
amount of metal mainly the caps that occupies the axial length of a string of such
units. However, there are two compensating aspects. One is the substantial amount of
leakage path length of ceramic that can be obtained per unit axial length, and the other
is associated with the small radial dimension of this insulating section around the pin
cavity which can account for an appreciable amount of the surface resistance. Also,
having some metal in the path of a propagating arc root can be beneficial due to its
ability to arrest such movement [7].
6.4.1.1 Some problems
Fromtime to time, defective insulators have been found on operational lines that have
necessitated changes in design. In other cases, such defects suggest that quality control
in manufacturing may need to be improved. Sometimes insulators are in service for
many years before faults reveal themselves thereby making the reasoned cause very
difficult to prove. These problems have puzzling features, which have stimulated
much research, but aspects still warrant further study.
6.4.1.2 Glass shattering under DC energisation
When insulator designs that have low failure rates on AC lines were installed on DC
lines, the shattering of the glass shell occurred much more frequently than was antic-
ipated. An obvious difference between AC and DC energisation is the unidirectional
flowof ioncurrent inthe latter. For soda-lime glass, ionic conductivitydepends mainly
on temperature, but it is also a function of the amount and the type of the alkali that
is present in the silica network. Further, impurity inclusions can cause electric stress
enhancement in the glass as they often do in other dielectrics thereby leading to a
concentration of the ions. Although various mechanisms, based on these phenomena,
have been proposed for causing this high shatter rate under DC enegisation, none has
been universally accepted. However, the technological remedial approach that has
been shown to be highly successful is to improve the quality of both the raw material
and the manufacturing process. Such improvements have achieved a high resistivity
toughened glass [8]; such comparative resistivity values are shown as a function of
temperature in Figure 6.2.
An aspect that is worthy of a brief mention is that of thermal runaway. It is readily
apparent from Figure 6.2 that glass is more conducting as its temperature increases.
Therefore, if this temperature increase is caused by ionic conduction, a positive
feedback situation arises resulting in a continuously increasing ionic current for a
21
20
19
18
17
16
15
14
13
12
2.9 3.4 4.0
l
o
g

r
e
s
i
s
t
i
v
i
t
y
,

o
h
m
.
c
m
temperature 1000/T, K
(
1
-
2
1
1
C
)
D
C

g
l
a
s
s
A
C

g
l
a
s
s
Figure 6.2 Resistivitytemperature relationship for glass insulators developed for
AC and DC lines [8]
constant electric field. Laboratory experiments have shown that this can occur with
a somewhat simple unit constructed from impure glass. When, however, such an
insulator is in a string the situation may be different. If the other insulators are of
very high resistivity glass, these units control the current. That is, the electric field
within the more conducting glass is not constant but decreases as the glass becomes
hotter thereby resulting in a negative feedback condition. Obviously, if all the
units in the string were identical and of impure glass, then positive feedback would
result. Although it is not very likely that all such units would be identical they could,
nevertheless, be near enough for this latter situation to arise in some cases. This may
be the explanation of the glass shattering on the AC-type cap and pin insulators when
they were energised on the DC lines.
6.4.1.3 Corrosion under DC energisation
Electrolytic corrosion of insulators on DC lines is a well researched problem [9].
The positive electrode dissolves gradually into the electrolyte, produced by wetted
pollution, due to the flow of leakage current. For ferrous material, such corrosion
occurs through the following ionising reaction:
Fe Fe
2+
+2e
(6.1)
The loss of weight W(z) of the positive electrode due to a flow of current I(z) for
duration T was found to follow approximately Faradays Law:
W(z) = I(z)TZ (6.2)
where Z is the electrochemical equivalent (i.e. the mass of ion taken into the solution
by a current of one ampere flowing for one second).
zinc collar
zinc collar
a b
Figure 6.3 Use of sacrificial zinc to protect the metal parts of cap and pin insulators
for DC applications
a on the pin
b on the cap
Fromdata reported for DClines in Japan, the annual quantity of charge (i.e. I(z)T )
involved in such a process has been estimated as 200300 C for inland regions and
500 C for coastal ones. For a very severe region in New Zealand, the corresponding
value was 15003000 C.
From the practical viewpoint, the pin is the component that suffers the most.
When its metal is substantially reduced, the insulator string can no longer support the
conductor.
The method successfully developed to provide protection of the steel pin in
addition to normal galvanising was to use a zinc sleeve as a sacrificial electrode [9].
This sleeve is fused around the pin in the position shown in Figure 6.3a. A corres-
ponding method of protecting the cap [8] by bonding a pure zinc collar around its
rim is shown in Figure 6.3b.
6.4.1.4 Radial cracks in porcelain under AC energisation
During the refurbishment of some UK lines, a considerable number of 400 kV
suspension insulators were found to be defective [10]. In one of the two major
problems, radial cracks had developed in the part of the shell external to the head
region. Most of these faulty units were at the high voltage end of the string. Subsequent
investigations showed that the corrosion of the part of the pin embedded within the
mortar for about 10 mm in from the air surface, see Figure 6.4 was the cause
[11]. The corrosion products resulted in an increase in the radial dimension of the
pin, thereby causing a tensile hoop stress to develop in the porcelain. Interestingly,
the corrosion of the pin external to the mortar was much less than that found for
DC insulators.
An assessment of an infra-red method for detecting such radial cracks in situ [12]
showed that under humid conditions the caps of sound units at the high voltage
end of the string were a few degrees warmer than the caps of the defective insulators.
Follow-up research revealed that adjacent dry band discharges produced this heating
and that similar discharges occurred around the pin cavity. By simulating this latter
Corroded surface
Figure 6.4 Corroded pin (with some mortar removed)
condition in the laboratory, using a construction in which the pin surface was covered
with a collection of electrodes, the current distribution was found to be that shown in
Figure 6.5 [unpublished findings of the author].
The comparison of Figures 6.4 and 6.5 strongly suggests that the leakage current
flowthrough the mortar causes, first, corrosion of the galvanised layer and then of the
underlying steel. Findings from complementary metallurgical experiments support
this conclusion. However, many aspects of this problem remain to be explained
properly including the part played by the stray capacitance current [13] in causing
the concentration of the cracked units at the high voltage end of the string.
As zinc corrodes less quickly than does steel, the design of the replacement
insulator includes a collar which is similar to that shown in Figure 6.3a for the
DC case.
6.4.1.5 Head cracks in porcelain under AC energisation
In contrast to the radial cracks, the second major problem noted on some of the UK
lines involved cracks that were hidden beneath the cap. For complete insulators, such
10
9
8
7
6
5
4
3
2
1
proportion
of current,
moist
mortar
%
10
air/mortar
interface
3
4
5
6
5
5
11
2
9
40
C
L
Figure 6.5 Diagrammatic representation of a pin fitted with electrodes, and the
corresponding current distribution for a dry-band discharge terminating
on the mortar surface adjacent to the porcelain
cracks can only be detected electrically. In the worst case, one third of the units at the
high voltage end of the suspension string were defective.
Upon removing the caps, typical examples of the cracks and the puncture path
through the porcelain were found to be those shown in Figure 6.6. Frombreaking open
a substantial number of supposedly sound units, a few partially completed puncture
paths were found. These varied from being very short to being ones that had nearly
developed through the complete thickness of the porcelain.
The findings of subsequent research [14] indicate that incipient defects, that were
built into the porcelain at the manufacturing stage, can extend in length under an
appreciable voltage while in service. The occurrence of such an excess voltage at the
high voltage end of the string is most likely to happen as polluted insulator surfaces
start to become moistened when, for example, RH > 70 per cent. Prior to this stage,
the electrostatic voltage distribution along the string applies and so because of stray
capacitance the voltage drop, V, across the high voltage end units is appreciably
greater e.g. by factor 2 than that across the units in the middle of the string.
For a resistance R(u) of the surface of a unit, the power dissipated in that surface
a
b
puncture path
Figure 6.6 Cracks and puncture path in the head of the porcelain shell
a view of top of the head
b view of the face of the cracked porcelain
is V
2
/R(u). Therefore, if all the units in the string have the same value of R(u) at
the start of this transition from the electrostatic to the resistive voltage distribution,
the power available to effect counteracting drying is greatest for the high voltage end
units. As a consequence, the value of R(u) for the high voltage end insulators becomes
larger than that of the others in the string and so the resistive voltage distribution
readjusts to enhance the value of V across these units. Sometimes the increase in
V
2
is substantial and so positive feedback applies for a while. By this process, the
value of V across a high voltage end unit continues to increase until it becomes
a large portion of the total voltage across the string. However, this situation may only
be short lived because moistening often becomes sufficiently large to swamp this
differential drying process.
Generally, head cracks are more disruptive than radial ones. Should an insulator
string become so severely polluted that a flashover occurs, the flow of such a large
current through a crack in the head of the porcelain shell will cause a very large rise
in pressure. Then the cap usually splits. Occasionally it does not; instead, the whole
contents of the insulator are ejected like shot from a canon. In this situation, it seems
very likely that the current path was essentially along the axis of the insulator. In both
cases, the insulator string breaks. In contrast, an arc within a radial crack in the shell
outside the cap can very rapidly escape into free air.
It would thus seem wise that, from time to time, sample units especially those
from the high voltage end of the string should be removed and subjected to a high
voltage test, particularly for those lines in highly polluted areas.
6.4.2 Longrods
The obvious way to minimise the quantity of metal along the axial length of the
insulator is seen in Figure 6.7 for the longrod design. For the porcelain version, there
is the additional advantage that any corrosion products will subject this material to a
compressive stress which it is well able to withstand (see Table 6.1). Of course, the
rod section of this design has to carry a tensile load and these diagrams clearly show
the advantage of the glass fibre version (again, see Table 6.1) in that the polymeric
insulator is much slimmer and hence much lighter than the porcelain one.
Both versions of this design can have alternate long and short sheds. Although
the leakage path length is obviously less than that of corresponding designs having
all sheds as large as possible, it is an optimum arrangement to achieve the benefit of
increasing the spacing between the longer sheds. This larger axial spacing is useful
when a substantial amount of wetting occurs e.g. heavy rain, live washing because
the run-off stream of water has a chance to break up before contact is made with the
surface below.
6.4.3 Posts
The multiple cone version of this insulator, Figure 6.8a, is worthy of a brief mention
because again it shows an important difference between a glass and a porcelain
construction. In some cases, such porcelain posts have developed radial cracks
(Figure 6.8b), which drastically reduce the path length of the leakage current and
thereby the pollution flashover strength (Figure 6.8c). The reduced amount is approx-
imately equal to the proportion of the post so affected (Figure 6.8d) as seen by the
close correlation between the experimental and theoretical curves (unpublished work
of the author). These cracks are probably caused by the volume growth of the mortar
that bonds these cones together. Such mortar growth is known to cause radial cracks
to develop in the porcelain shells of some cap and pin insulators [15]. The correspond-
ing shattering of glass cones does not seem to occur. This may be due to the much
metal fitting
a b
RBGF rod
polymeric
weathershed
porcelain
Figure 6.7 Examples of longrods
a ceramic version (based on [2])
b polymeric version (courtesy of Sedivers Mechanical application
guide for composite line post insulators, 1991)
greater tensile strength of toughened glass to that of porcelain. Another difference
is that Portland cement is used for the porcelain version whereas Fondu cement is
employed for glass ones.
It is worthy of note, that much of this failure phenomenon with the porcelain-type
of multiple post is similar to the radial cracking in porcelain cap and pin insulators.
As a consequence of the cracking of these porcelain cones, the solid core porce-
lain post was developed. As per the longrod line insulator, this design avoids putting
porcelain into a tensile mode should there be any metal corrosion, or cement,
growth.
Adesign of a polymeric post is shown in Figure 6.9 [16]. The general arrangement
is depicted in Figure 6.9a, and the ability to include a fibre optic link [1] within such
a construction is illustrated in Figure 6.9b.
Aparticular advantage of having a RBGC core lies both with its high mechanical
strength and its flexibility. Both of these features enable such posts to withstand
more readily than their porcelain equivalent the short circuit forces that occur under
mortar
glass or porcelain cone
metal end fitting
a
b
c
normal leakage path
shorted-out
leakage path
crack face
sectioned view
of a healthy cone
Figure 6.8 Multiple cone post
a general construction
b typical example of a radial crack in a porcelain cone
c probable conducting path across the face of the crack
d
0
10
20
30
40
50
60
70
80
90
100
10
experimental
theoretical
20 30
number of cracked cones expressed as
percentage of the thirty-three in the post, %
w
i
t
h
s
t
a
n
d

s
a
l
i
n
i
t
y

r
a
t
i
o
,

%
40 50
Figure 6.8 Continued
d withstand salinity in artificial salt fog of a 400 kV post comprising
various numbers of cracked cones; expressed as percentage ratio
of that of a completely sound post
high current fault conditions and the violent vibrations that happen under seismic
activity.
6.4.4 Barrels
Such hollowinsulators have a number of important applications; e.g. weather housing
of bushings, surge arrester enclosures, circuit breaker interrupter heads, cable sealing
ends. Sometimes they are pressurised to greater than 50 kPa gauge. The various
standards relating to porcelain, annealed glass and toughened glass for use with
voltages greater than 1000 V have, in recent years, been incorporated into a single
document: that is, IEC 62155.
The design that is now attracting much interest is the polymeric unit, especially
ones having silicone rubber sheds bonded onto a glass fibre tube for example,
see Figure 6.10. Special features are that such hollow insulators can be made into
HT end fitting
silicone rubber
or EPDM sheds
RBGF core
metal fitting
core
glass fibre
cable
cable slot
in core
a b
L
C
Figure 6.9 Example of a polymeric post
a general arrangement
b inclusion of a light link
long lengths with large diameters. Also, such insulators have an excellent pollution
flashover performance.
The resin bonded glass fibre tube is constructed on a mandrel, as per standard
practice. In one design of bushing manufacture [16] the silicone rubber sheds are
applied to this tube. The end fittings are attached while the silicone rubber is still
pliable and this complete assembly is then vulcanised in an autoclave. Final machining
is carried out when this assembly has cooled down. Such hollow core insulators can
currently be constructed with a maximuminternal diameter of 600 mmand an overall
maximum length of 6 m. The significance of this additional size over that of the
corresponding porcelain insulator will be discussed briefly later with regard to circuit
breaker interrupter heads (section 6.6.3).
6.5 Flashover mechanisms
Relative to the breakdown of the corresponding length of an unbounded air gap, the
flashover of a polluted insulator is usually a slow process that takes place at a fairly
low value of average electric field.
6.5.1 Surface wettability
A good measure of the wetting of an insulating material is the contact angle, , of a
sessile water drop residing on its surface (Figure 6.11). This angle is controlled by
silicone rubber sheds on
RBGF tube
Figure 6.10 Example of a polymeric hollow core insulator
air
water
0
insulator
Figure 6.11 Sessile water drop on an insulator surface
three coefficients of interfacial tension, , as related by the YoungDupre equation:
wa
cos =
ia
iw
(6.3)
where subscripts a, i and w refer to air, insulator and water, respectively.
For a clean and newsurface, is about 30
for ceramic and about 100
for silicone
rubber. However, should this surface become coated with a layer of pollution albeit
only dust then this angle becomes essentially zero. The difference between these
current flow heating of unit area
=(current density)
2
(resistivity)
electric field=(current density) (resistivity)
reference
point
1
D(I )
Figure 6.12 Developed surface of an arbitrary hydrophilic insulator that is polluted
and wetted
two types of material is that, for such a polluted surface, always stays zero for
ceramic (i.e. the hydrophilic case) but it will, in time, return to a substantial value for
silicone rubber (i.e. the hydrophobic one).
6.5.2 Hydrophilic case
When a film of polluted water covers the complete surface (see Figure 6.12) of an
energised insulator, this electrolyte conducts a leakage current. Joule heating causes
some of the water to evaporate. The corresponding power density is j
2
, where j
is the surface current density and is the surface resistivity. In a region where this
power density is a maximum or the thickness of the water is less than a critical value,
increases with time and the electric field E in the electrolyte at this point also
increases; that is, because E = j. The electric field in the air immediately above
this point has approximately the same value. Once the ionisation level in this air is
reached, a discharge occurs. Such a region is usually referred to as a dry band, but
strictly speaking it need not be completely dry. This discharge can be a spark, a glow
or an arc depending upon its duration and the magnitude of the current it conducts.
However, if a pollution flashover is to take place it has to be an arc whose roots
propagate over the surface of the electrolyte and along the length of the insulator.
On a vertically mounted string of cap and pin insulators, such discharges originate
around the pin cavity where j is a maximum and around the cap where is
often high due to the washing effect of rain.
From Kirchoffs law, the voltage V and current I relationship of an arc plus
electrolyte on such an insulator surface (Figure 6.13) can be expressed as:
V = IR +V(e) +If ()f (X) (6.4)
R is the resistance of the arc; V(e) is the sum of the voltage drops at the junctions
between the arc and electrolyte (i.e. anode phenomenon about 200 volts and cathode
dry
band
electrolyte
arc
I
V
Figure 6.13 Equivalent circuit for an arc spanning a dry band plus a portion of the
neighbouring electrolyte
one about 700 volts); f () is a function of , the value of which varies along the
length of the electrolyte principally because of enhanced heating around the arcs
root. Note: decreases as the electrolytes temperature increases until saturation of
the solution occurs; thereafter, it increases again as evaporation of the water continues
further. f (X) is a function of the arc length, X, the sum of the electrolytes length
plus that of the dry band together, designated L and the effective size of the arcs
root on the electrolytes surface; obviously, this size can have two values one for
the anode, and the other for the cathode.
From Suits arc equation [17]:
R = I
(1+a)
_
X
0
A(x) dx (6.5)
where a is a constant and A(x) has a value at position x from the beginning of the
arc that depends upon the proximity of that part of the arc to the electrolytes surface,
the amount of water on that surface and the extent that thermalisation has taken place
within the arc. To a first approximation, A can be considered to have an effective
value i.e. a weighted average that applies for the whole length of the arc.
Because the resistance of the arc decreases as its current increases, its length
extends when in so doing its current increases. That is, the criterion for the propagation
of the arcs root to effect this increase in length can be stated as:
dI
dX
> 0 (6.6)
For theACcase, the flashover process usually occurs over a fewhalf cycles and so the
reignition of the arc after current zero needs to be taken into account. This constraint
can be specified [18] as a reignition voltage V(r) of the form:
V(r) = NXI
n
(6.7)
Again to a first approximation, effective values for the constants n and N relating to
the surface and shape of a practical insulator rather than a gap in free air should
be used.
The various mathematical models that have been adopted to solve these equations
for simple insulator shapes i.e. the function f (X) is then simple have been well
reviewed by Rizk [19].
leakage
current in electrolyte
air
breakdown
surface arc
Figure 6.14 Illustration of how the breakdown of various air gaps on a practical
insulator can lead to the shortening of an arc that is starting to flash
over that surface
For the more complex insulator shapes, e.g. like the anti-fog design shown in
Figure 6.1, the first additional aspect that needs to be taken into account is the form
factor, F, of its surface. F can be readily obtained by considering the developed view,
Figure 6.12, of that surface to become:
F =
1
_
L
0
dl
D(l)
(6.8)
In this equation, D(l) is the diameter of the insulator at leakage position l from the
reference point and L is now the total leakage path of the insulator.
The second additional aspect that needs to be taken into account for practical
situations is that the arc does not necessarily hug the surface of the insulator over
its entire length an example is illustrated in Figure 6.14. The sparkover of the air
gaps between sheds and the ribs of a shed can effectively shorten the arcs length,
and buoyancy forces can result in either its shortening or its lengthening, depending
upon the part of the insulators surface under consideration. Mathematically, this can
be represented by stating that the arcs length X(a) is related to the corresponding
spanned length X(i) of the insulator by:
X(a) = kX(i) (6.9)
where k < 1 for the shortening case and k > 1 for the lengthening one.
Experience indicates that once the critical conditions have been reached, flashover
is likely to occur even though an appreciable length of the insulator remains to be
spanned by the arc. Therefore, up to this stage the resistance of the electrolyte around
the arc root is much smaller than that of the rest of the electrolytes length that still
remains to be spanned and so for all practical purposes can be taken to have
a constant value and f (X) can be considered to be of the form:
f (X) = F
_
1
X
L
_
(6.10)
The solution to obtain the flashover criterion is straightforward for the DC case. The
value of the flashover voltage V(FO, DC) is determined as being:
V(FO, DC) (F)
a/1+a
(kAL)
1/1+a
(6.11)
In contrast, the rigorous solution for the AC case is quite involved [19]. However,
a method of obtaining a simple relationship for an approximate criterion is to make
the following assumptions [20]. The first one is that n = a; the physical argument
being that both constants are strongly related to the temperature of the arc. The second
assumption is to make V equal to V(r) in Equation (6.4), thereby obtaining I in terms
of X; that is, the least value of the supply voltage is that required to cause reignition.
The third assumption, and the one not so readily justified, is that the critical value
of X is about 2L/3. This value is that of the DC case when the electrolyte is a
long rectangular shape. Furthermore, for the AC case, it is consistent with the value
obtained by making a more rigorous analysis. With these assumptions, the flashover
voltage for the AC case becomes:
V(FO, AC) 0.7N(F/2(N A))
n/1+n
(kL)
1/1+n
(6.12)
In this equation, V(FO) represents a value somewhat less maybe only slightly
so than that of the peak of the waveform.
By comparing the calculated values with the measured ones, the effective quan-
tities for the various constants can be determined. Those for a and n can be obtained
from simple bench-top experiments. Data obtained in artificial pollution tests made
on the simpler designs of post and longrod insulators can be used to find the quantities
of A and N. The corresponding values for k are best derived from the testing of insu-
lators that have the anti-fog profile. Ballpark values are a = n = 0.5, A = 10 volts
per mm square root amps and N = 100 volts per mm square root amps. For anti-fog
cap and pin insulators, the value for k is found [21] to vary between 0.4 and 1; it
depends upon the pollution severity smaller values relate to higher resistivities
and whether AC or DC energised smaller values apply to the DC case.
An examination of Equations (6.11) and (6.12), using the parameter values quoted
above, readilyshows that V(FO, DC) is muchlower thanV(FO, AC) for the same level
of pollution severity. That is, the lack of the reignition process makes DC insulators
substantially longer than those for the corresponding AC case if their flashover is to
be avoided.
For the AC case involving flashover in less than one half cycle i.e. on a short
insulator then, obviously, reignition does not apply and so the process is more akin
to that of the DC situation. For this reason, linearity between AC flashover voltage
and insulator length does not apply over the complete range of length. This means that
reliable information about the pollution flashover performance under ACenergisation
is best obtained by testing fullscale insulators at, or close to, the relevant operating
voltage. From the theoretical viewpoint, this restriction does not apply to the DC
case which is an aspect that needs to be considered further by experimentation.
Because such pollution flashover tests on high voltage insulators are both costly
and time consuming, the ability to calculate with a reasonable degree of accu-
racy the flashover voltage pollution severity relationship would be very beneficial.
Therefore, further theoretical studies directed towards this end are highly warranted.
6.5.3 Hydrophobic case
When the water drops remain discrete, which is the distinctive feature of this case, the
flashover path involves the intervening air gaps [22]. Here the water drop plays two
major roles. In one, it acts as a stress enhancer, due to its much larger permittivity than
that of both the surrounding air and the insulating material on which it resides. In the
other, it reduces the length of the air path across the dry part of the insulator. In this
second role, the flashover path is through the drop when the water is highly conducting
or across its surface when it is not so. Acontributing feature is the deformability of the
drop under the influence of the electric field, such that stress enhancement becomes
greater and the air path between the drops in the direction of the electric field becomes
smaller. Sometimes the drops can coalesce to form rivulets. Also, the migration of
soluble pollutants from the seemingly dry part of the surface to the drops is another
important factor [23].
Consequently, the flashover mechanism of a hydrophobic surface is appreciably
different from that of a hydrophilic one. For one thing, it develops faster. More
important from the practical viewpoint, the average electric stress is substantially
greater than that of the corresponding hydrophilic surface but, unfortunately, no
authoritative method is yet available to enable this difference to be calculated.
6.5.4 Ice and snow conditions
The processes responsible for the flashover of ice-covered insulators and those on
which snow has settled are not yet fully understood. Nonetheless, the following
observations apply.
A flashover caused by ice accretions usually occurs when a water film forms on
that ice during its thawing stage. This water film has a substantially lower resistance
than that part of the insulator on which ice was either not present or has fallen away.
Most of the voltage is then impressed across the latter i.e. it is somewhat similar
to the dry band situation described earlier. Depending upon the relative lengths of
these two parts, the arcs that are produced across the more highly stressed sections
can extend to produce a total electrical failure. The probability for such a flashover
to happen may be significantly increased when fog, drizzle or rain is also present at
the critical moment. Pollution on the insulators surface prior to the formation of this
ice also reduces the insulators strength.
The quantity of snow that covers an insulator is usually the greatest when it is
horizontally mounted. The density of this snowand the resistance of the melted water
are the main parameters that affect the value of the leakage current. The parts of
the snow that have a high current density are likely to melt more quickly due to
Joule heating. This snow may then drop away from the insulator thereby leading to a
non-uniform voltage distribution along its length. Depending on the resistance of the
remaining snow and the length of the insulator it does not cover, arcs may bridge the
latter and thereby provide the precursors of a possible flashover event.
6.6 Electrical characteristics
Although the largest population of insulators that are subjected to severe pollution is
by far on operational lines, the published information concerning their flashover
performance is very limited. Areason is, possibly, that the Electricity Supply Industry
does not wish to be embarrassed when their failure statistics are higher than they had
designed for in the quest to provide a very reliable supply at minimumcost. However,
some information does become known via networking sources especially within
technical meetings, suchas CIGREWorkingGroup33.04.13therebyaidingresearch
thinking.
The research findings on high voltage insulators have been obtained using natural
pollution testing stations and artificial pollution test chambers. Testing the same
insulator using both of these facilities provides the most thorough investigation. The
severity of the pollution can be expressed in a number of ways. The main ones are:
(i) the density of the pollutants on the insulator surface
(ii) the surface conductivity of those pollutants when adequately wetted
(iii) the quantity of the polluting medium used in the relevant artificial test.
An example of the latter is the amount of sodium chloride per unit volume of water
used to produce a salt fog.
For many years, only the quantity of the soluble component of the pollution that
settled on an insulators surface in service was considered important [4], it often being
expressed as the equivalent salt deposit density (ESDD) where the salt is sodium
chloride. However, it has recently been recommended [3] that the corresponding
non-soluble deposit density (NSDD) should also be measured, because this non-
soluble component helps to retain surface water. With silicone rubber insulator, the
non-soluble material plays another important part in that it temporarily masks the
water-repellent properties of the rubber. Both the type and the density of this deposit
need to be known to assess the consequences of having such an insulator so covered.
The advantage of the natural pollution investigations is that they fully replicate
the service conditions. They show both the extent to which an insulator collects and
retains the pollution froma certain environment plus the ability of this energised insu-
lator to withstand this pollution deposit when wetted under fog, mist and, drizzle. The
disadvantages are that they take some years to conduct, they are costly when compre-
hensively done and the results apply directly only to that type and severity of pollution.
The attractions of the artificial pollution test are:
(i) the severity of the pollution can be varied over a wide range
(ii) it is performed fairly quickly
(iii) it is cheaper than the making of the corresponding test conducted under natural
conditions.
However, with regard to this last point, the test source is costly because of the need
to provide a current of some amperes often more than ten with negligible voltage
regulation. The corresponding principal disadvantage from the practical viewpoint is
that the test shows only the ability of the insulator to cope electrically with a controlled
amount of wetted pollution. Consequently, care must be taken in interpreting the
findings. For example, if the flashover voltage pollution severity curves for an anti-
fog type of insulator and a corresponding aerodynamic one are shown in the same
diagram, then a cursory inspection suggests that the pollution flashover of the latter
is much inferior to that of the former. The more discerning observer realises, of
course, that from the practical viewpoint these two curves should be substantially
separated along the pollution severity axis to take account of the much smaller amount
of pollutants that the aerodynamic insulator catches in service.
Major research programmes most employing artificial pollution but a substantial
number using natural pollution have involved the testing of more than 120 types of
ceramic insulator and nearly 30 types of polymeric ones. Acomprehensive record of
the findings is provided in the CIGRE review [2] and a summary of the main features
is provided in the rest of this section.
6.6.1 Performance under natural pollution
Under natural pollution, the flashover statistics of a test insulator are obtained over a
number of years relative to those of a control one. Also, the flashover frequency can
be determined by slightly changing the length of the insulator for the same test volt-
age. Some good examples for marine pollution have been obtained from the results
obtained at the Brighton Insulator Testing Station (BITS) on the south coast of Eng-
land. Upon the closure of this long operating station, much of the test equipment was
relocated about 55 miles to the east at Dungeness (to become DITS). The practice
adopted by the operators of BITS that is, Central Electricity Research Laboratories
(CERL) of the Central Electricity Generating Board was to publish many of its find-
ings [1]. In contrast, the tests conducted by the current owners of DITS National
Grid Company were made under contract conditions, thereby making their gen-
eral dissemination much more difficult. Regrettably this site is currently mothballed
awaiting the emergence of a new owner.
The results from these test stations can be expressed in the form of a figure of
merit (FOM). This FOM is defined as the length, L
c
of the control insulator divided
by the length L
t
of the test insulator for flashovers to occur across both of them under
the same voltage and environmental conditions. That is:
FOM =
L
c
L
t
(6.13)
Note: the length in this method is the axial value between fixing points. The insulator
with the largest FOM value has the best pollution flashover performance. That is, the
axial length of this insulator is equivalent to that of the control one divided by FOM.
Table 6.2 Figure of merit (FOM) values of some
porcelain cap and pin insulators when
AC energised at BITS [2]
Insulator type Orientation* FOM
Reference vertical 1.00
Standard horizontal 0.79
Anti-fog horizontal 1.35
Anti-fog angled 1.11
* horizontal is about 75
to the vertical; angled is about 45
to
the vertical
Table 6.3 Figure of merit (FOM) of some poly-
meric insulators; AC energisation at
BITS [2]
Insulator type Orientation* FOM
Silicone rubber horizontal >1.53
Silicone rubber vertical >1.53
EPDM horizontal 1.28
EPDM vertical 1.12
EPR horizontal 1.19
EPR vertical 1.17
* horizontal is about 75
to the vertical
Some values of FOM for strings of cap and pin insulators under various orienta-
tions that had been tested at 275 kV and 400 kV are shown in Table 6.2. These are
compared relative to that for a vertical string of reference insulators known as CERL
Reference A [2] which is the large anti-fog design. Table 6.3 gives corresponding
FOM values for polymeric insulators tested at 550 kV.
The larger FOM value for a horizontally mounted string of cap and pin insulators,
relative to that of a corresponding vertically mounted one, is reasoned to be due to the
beneficial effect of rain. That is, with the former both the pin side as well as the cap side
are well washed. This finding at BITS is supported from the general performance of
the insulators on UKoverhead lines. For example, the total leakage path of the tension
string on a 400 kVline is about 30 per cent less than that of the suspension string in the
same location. Further proof became obvious many years ago when the coastal region
of Kent was subjected to about a week of strong dry wind off the North Sea. After
this sustained build-up of salt on the insulators, the wind then changed direction to
become a prevailing south westerly involving a little mist. Some flashovers involving
these tension strings, but not the suspension ones, then occurred. Fortunately, this
south westerly wind also soon brought in rain so this problem was fairly short lived.
Of special note: the FOM for the silicone rubber insulators can be expressed only
as being greater than 1.53 because no flashover of this type of insulator occurred
during the four years of testing in this severe marine environment. In this last respect,
the pollution severity of BITS has been assessed as being equivalent to a salinity
of 60 kg/m
3
in an artificial salt fog test. This excellent performance of the silicone
rubber type is now being confirmed in service. For example, South Africa now has
no pollution flashovers [24, 25] on previously troublesome lines (they were then
insulated with either ceramic or epoxy resin insulators). Further, China nowhas about
600 000 silicone rubber insulators in service and, likewise, their pollution flashover
performance is reported as being excellent [26].
Afurther use of the FOM concept is that it enables the flashover statistics at a site
to be determined even though the data come fromdifferent insulator types. That is, the
axial flashover stress of a test insulator is normalised to that of the control insulator
by dividing its stress by FOM. By this technique, the statistics of the flashover events
at BITS are shown in Figure 6.15 as a cumulative frequency versus normalised axial
flashover stress E
n
.
The relationship shown in Figure 6.15 can be expressed in terms of the number
of flashovers per year N(FO) of this equivalent control insulator. That is:
N(FO) = K
_
E
n
E
o
E
o
_
m
(6.14)
For an insulator type of similar profile to that shown in Figure 6.1b and of overall
diameter of 394 mm, the constants K, mand E
o
have values of 16.1, 2.1 and 99 kV/m
(based on system voltage), respectively.
This equation indicates that the distribution of flashover events is truncated with
respect to axial stress. That is, when E
n
is equal to or less than E
o
there should be
no flashover at this site. This is not strictly true, however, because E
o
is the mean
170
160
150
140
130
120
110
100 n
o
r
m
a
l
i
s
e
d

f
l
a
s
h
o
v
e
r

s
t
r
e
s
s
k
V
,

s
y
s
t
e
m

m
1
,

o
v
e
r
a
l
l

l
e
n
g
t
h
90
0.01 0.1 1.0
cumulative frequency, flashover/insulator/year
EHV
UHV
10
Figure 6.15 Cumulative frequency distribution of 275 kV and 400 kV ceramic insu-
lators, at normalised flashover axial stress, for a test period of about
eight years at BITS. (Courtesy of CIGREs paper 33-01, published in
1982)
value to which must be related a standard deviation. Unfortunately, the data obtained
at BITS were not sufficient to determine this standard deviation with an acceptable
degree of accuracy. Consequently, this is an area that warrants further research.
6.6.2 Performance under artificial pollution
The two types of test standardised by IEC [4] are the salt fog method and the solid
layer method; the latter now often referred to as the clean fog test because of the
technique used to wet the pollution layer. The former simulates coastal pollution
where a thin conductive layer formed by the sea salt covers the insulators surface.
In practice, this layer usually contains little if any insoluble matter. The same test
also applies to industrial pollution principally gases that result in the formation of
acids when there is essentially no inert material present on the surface. In contrast,
the solid layer (clean fog) method applies when there is an appreciable amount of
non-soluble pollution in addition to that forming ions when wetted.
The current versions of these two methods IEC 507 and IEC 1245 for AC and
DCenergisations, respectively were developed for the testing of ceramic insulators.
The modification of these tests so as to apply also to polymeric insulators has been
considered by CIGREWG33. 04 .13. Unfortunately, there is as yet no internationally
accepted method to simulate snow and ice although some tests have been developed
for research purposes. The development of a test that IEC will ultimately recommend
is no mean task. For example, the first one to be produced the salt fog method
involved collaborative research by the European utilities of CEGB/EdeF/ENEL
lasting many years (perhaps as many as ten!). There are some other noteworthy
methods that are currently of non-standardised status; e.g. the dry salt layer (DSL) [2].
The major difficulty with the testing of polymeric insulators especially those of
the silicone rubber variety is the hydrophobic aspect. The actual insulators tested at
BITS were also subjected to the salt fog test. It was then realised that the ranking in
performance of the four types of insulator referred to in Table 6.3, was different in this
artificial pollution [27] from that found under natural pollution conditions. Anumber
of reasons for this difference have been suggested but, as yet, no explanation seems
certain. The temporary loss and then the subsequent recovery of hydrophobicity for
the silicone rubber material are of particular concern. Because the dynamic nature of
this material has a very large influence on the pollution flashover performance, there
is currently some uncertainty that the full potential of silicone rubber for the solving
of a number of important problems can be realised. However, some useful results
obtained using a solid layer method have been reported [3].
The extent of the change in the flashover strength in a salt fog test of a hori-
zontally mounted porcelain insulator that had been coated with silicone rubber is
exemplified in Table 6.4. These results show that over a period involving 11 attempts
to obtain a sensible average value of the flashover voltage, FOV, the individual val-
ues progressively decreased by factor two. After resting this insulator unenergised
for twenty-four hours the initial strength was fully recovered. Obviously, by this pro-
cedure the lowest value of the withstand voltage, WSV, could be obtained. However,
this is a very pessimistic value to adopt for dimensioning purposes as it came about by
having had repeated flashovers which, hopefully, would never happen in practice.
Table 6.4 Salt fog test on a horizontal porcelain insulator
coated with silicone rubber
Voltage* Test number
kV 1 2 3 4 5 6 7 8 9 10 11
68 WS WS
71 WS FO FO FO
74 WS
77 FO FO
80 WS WS WS
83 WS FO FO
86 WS WS FO
89 WS WS
92 WS FO FO
95 WS
104 FO
* each voltage level was maintained for five minutes
FO is flashover
WS is withstand
On the other hand, it is probably not prudent to assume that the upper safe limit is
the first flashover value because constant energisation with severe pollution and long
lasting surface discharges may result in the temporary loss of some hydrophobicity.
Consequently, much research is required to establish what the meaningful upper value
should be.
The situation with the clean fog test may be less serious but, as yet, this is not
fully proven as, however, some aspects of the procedure still remain to be investigated.
The main difficulty from the test viewpoint and one that may also be of concern
in some practical situations is that immediately after a hydrophobic silicone rubber
surface is covered with pollution its outer surface is then hydrophilic. Although it is
known that water repellent properties are transferred from the silicone rubber to the
adhering pollutants, the rate and extent at which this transfer occurs and the effect
on the pollution flashover strength during this transitional period have not been fully
investigated. An example of how the leakage current across a new silicone rubber
insulator decreases as a function of elapsed time after an application of slurry
comprising kaolin, sodium chloride, a wetting agent and tap water is shown in
Figure 6.16. In these tests [28], the dry insulator was placed in a clean fog chamber
after this elapsed time. Then it was energised and a few minutes later the mist was
applied; the contact angle of a water dropplacedonthe surface of the pollutedinsulator
was measured prior to the start of these tests. These results clearly indicate that the
transfer of hydrophobicity from the silicone rubber to the outer layer of the pollution,
as manifested by the increase in this contact angle and the decrease in the leakage
current, can take an appreciable time to occur. Also, it is very dependent upon the
200
20g/l
100 1000 10000
time, mins
100000
0
10
20
30
c
o
n
t
a
c
t

a
n
g
l
e
,

d
e
g
r
e
e
s
40
50
60
20g/l
40g/l
80g/l
80g/l
160g/l
40g/l
160g/l
150
100
50
25
15
h
i
g
h
e
s
t

p
e
a
k

c
u
r
r
e
n
t
,

m
A
10
5
3
1
0.5
0.3
160g/l pk current 80g/l pk current 40g/l pk current 20g/l pk current
160g/l angle 80g/l angle 40g/l angle 20g/l angle
Figure 6.16 Peak current and contact angle for a silicone rubber insulator
as a function of elapsed time from pollution application, at four
concentrations of kaolin in the polluting slurry
concentration of the kaolin in the slurry; the corresponding densities of the soluble
and non-soluble components of the pollution as indicated by ESDD and NSDD,
respectively on the insulators surface are shown in Table 6.5. As a consequence
of this phenomenon, a standardised test will have to specify clearly the time that
must pass before such a polluted insulator should be subjected to a withstand test.
Acomplicating feature is that the transfer rate of this water repellent property is also
a function of the type of the non-soluble pollutant [29].
6.6.2.1 Effect of soluble component
Generally, the critical stress E(s) at flashover of a ceramic insulator, and pollution
severity S, can be expressed in the form:
E(s) S
p
(6.15)
Table 6.5 ESDD and NSDD measurements
for silicone rubber insulator used
for investigations into hydrophobic-
ity transfer [29]; results presented in
Figure 6.16 apply
Kaolin concentration ESDD NSDD
g/l mg/sq.cm mg/sq.cm
20 0.2 0.1
40 0.3 0.3
80 0.3 0.5
160 0.4 1.3
10
3.5 5 7 10 14 20 28 40 56 80 112 160
15
20
25
30
withstand salinity, kg/m
3
3
4
11
2
12
19
6
9
15
7
5
13
10
14
18
16
20
1
17
8
21
number next to point
refers to the insulator
type shown in [2]
cap and pin insulators
longrod
post insulators (core diameter > 75 mm)
s
p
e
c
i
f
i
c

c
r
e
e
p
a
g
e

d
i
s
t
a
n
c
e
,

m
m
/
k
V
Figure 6.17 Electric strength of different AC hydrophilic insulators in the salt fog
test as a function of pollution severity [2]
The value of p lies within the range 0.1 to 0.6. Insulators with plain open shedding
tend to have the higher values of p, and for many other cases p is often about 0.2.
From the mathematical viewpoint, the value of p can be considered as a weighted
average of the one for the electrolyte surface (e.g. p = 0.3 for brine) and that for the
air breakdown (p = 0) between parts of the insulator surface.
Some typical relationships of specific creepage length i.e. the inverse of surface
stress versus pollution severity for various insulator types are shown in Figure 6.17.
6.6.2.2 Effect of non-soluble component
The withstand voltages, WVS, of a cap and pin insulator, a porcelain longrod insulator
and of a porcelain post type insulator are shown in Figure 6.18 as a function of NSDD
when the non-soluble component is Tonoko material.
0.8
0.6
0.4
0.2
0
1
0.1 1 0.3 3 30 10 100
NSDD, mg/cm
2
inert material: Tonoko
longrod: SDD=0.5 mg/cm
2
LP : SDD=0.12 mg/cm
2
regression curve
c
o
m
p
a
r
a
t
i
v
e

c
o
n
t
a
m
i
n
a
t
i
o
n

w
i
t
h
s
t
a
n
d

v
o
l
t
a
g
e
0.8
0.6
0.4
0.2
0
1
0.1 1 0.3 3 30 10
NSDD, mg/cm
2
inert material: Tonoko
disc : SDD=0.4 mg/cm
2
2
2
2
regression curve
c
o
m
p
a
r
a
t
i
v
e

c
o
n
t
a
m
i
n
a
t
i
o
n

w
i
t
h
s
t
a
n
d

v
o
l
t
a
g
e
Figure 6.18 Influence of NSDDon withstand voltage of some hydrophilic insulators
[2]
This relationship can be expressed in the form:
WSV (NSDD)
0.15
(6.16)
which applies for the NSDD range of 0.110 mg/cm
2
.
6.6.2.3 Influence of average diameter for hydrophilic insulators
Some of the data indicate that at flashover, or withstand, the relationship between
specific length SL, i.e. the inverse of stress, and average diameter D(a) is of the form:
SL D(a)
q
(6.17)
The best support for Equation (6.17) occurs when the insulators are of the same profile
and only the diameter is varied. An excellent example is shown in Figure 6.19 for
a DC porcelain housing that had been subjected to a clean fog test with an ESDD of
0.12 mg/cm
2
; the value of q is 0.16.
Although this reduction of pollution flashover performance with average diameter
is appreciable, field experience indicates that for cylindrical-type insulators the
pollution collected on the surface of the insulator also decreases as the average
diameter increases. A conservative relationship for the relative value of its severity,
ESDD(r), where ESDD(r) = 1 for an average diameter of 115 mm can be
expressed [2] as:
ESDD(r) = 2.6D(a)
2.1
(6.18)
Therefore, for dimensioning an insulator for practical use, both of the aspects covered
by Equations (6.17) and (6.18) need to be taken into account.
100.0
10.0
2
30.0
2 3 4 5 6 7 8 9
1000.0
s
p
e
c
i
f
i
c

a
x
i
a
l

l
e
n
g
t
h
,

m
m
/
k
V
average insulator diameter, mm
Figure 6.19 Specific axial lengthversus insulator diameter, for aporcelaininsulator
under DC energisation in a clean fog test [2]
6.6.3 Interrupter head porcelains
Improvements in the design of the AC circuit breaker, particularly the use of SF
6
in
lieu of air, have resulted in the interrupter head being more highly stressed in some
cases than has hitherto been the situation. Of particular concern is the head of the
horizontally mounted version.
The axial electric stress necessary to cause external flashover of the interrupter
porcelain of various designs is shown in Figure 6.20 as a function of the severity
of an artificial salt fog. The much inferior performance of the horizontal head (i.e.
H values) when compared with that of a corresponding vertically mounted porcelain
(i.e. V values) is reasoned to be associated with the way that wetted pollution drains
from the insulators surface [30].
The design of SF
6
circuit breakers has now advanced to the stage that a two-
break version is adequate for 400 kV (cf. ten breaks or more for the corresponding
air blast breaker). That is, the axial stress on the H1 porcelain for the SF
6
breaker
would be about 180 kV r.m.s. per metre during arduous synchronising conditions.
The results shown in Figure 6.20 plus supporting tests conducted at BITS show
that even small amounts of pollution result in flashover for this condition which is
also so even for normal service stressing. The scope for substantially increasing the
length of these large diameter porcelains that are suitable for pressurised use is very
limited. Two possibilities that warrant study for future development are: (i) porcelain
when coated with a hydrophobic material and (ii) other types of barrel construction
e.g. silicone rubber sheds on a resin bonded glass fibre reinforced tube. This latter
option is covered in section 6.4.4 and Figure 6.10.
90
80
70
60
50
40
5
l
i
g
h
t
p
o
l
l
u
t
i
o
n
c
l
a
s
s
i
f
i
c
a
t
i
o
n
10
a
v
e
r
a
g
e
20
h
e
a
v
y
40
v
e
r
y
h
e
a
v
y
80
fog salinity, kg/m
3
a
x
i
a
l

s
t
r
e
s
s
,

k
V
/
m

r
.
m
.
s
.
H2 H3
H1
H4
V1
Figure 6.20 Axial stress to cause flashover against fog salinity for various
interrupter porcelains under AC energisation [2]
6.6.4 AC versus DC
Although not many publications contain information on the same or very similar
insulators that have been subjected toACand DCenergisations for an extensive range
of pollution severity, those provided by Electric Power Research Institute (EPRI) for
AC [31] and DC [32] are very useful. Data for identical insulators are shown in
Figure 6.21, where it is seen that the AC strength (based on peak value) is about
70 per cent greater than that for DC at the same pollution severity condition.
6.6.5 Transient overvoltages
A useful comparison is the ratio of the impulse flashover voltage to that for normal
stressing. Such information for cap and pin insulators is illustrated in Figure 6.22 as
a function of the duration of the impulse waveform (stated as the time for which the
voltage is greater than 50 per cent of the peak value). Also included in this diagram
are some results for the corresponding temporary AC overvoltage condition (in this
case, the duration is the time for which the 50 Hz voltage is applied before flashover
occurs).
These and related findings [2] indicate that the pollution flashover problemis asso-
ciated with the normal stressing condition. Although the magnitude of the transient
overvoltage can be considerably greater especially for the lightning case than
225
200
175
150
125
100
75
50
50 75 100 200 350 500
DC case
AC case
(peak stress value)
c
r
i
t
i
c
a
l

s
u
r
f
a
c
e

s
t
r
e
s
s
,

k
V
/
m
surface resistivity, k
Figure 6.21 Flashover data for identical insulators under ACand DCenergisations
lightning wave
switching surge
notes:
1 length of horizontal lines represents the duration
that the voltage is greater than 50 per cent of the
peak value.
2 length of vertical lines represents the spread of
the results
10 100 1000
1
2
3
r
a
t
i
o

o
f

p
e
a
k
:

s
h
o
r
t

d
u
r
a
t
i
o
n

F
/
O

v
o
l
t
a
g
e
t
o

p
e
a
k

n
o
r
m
a
l

A
C

F
/
O

v
o
l
t
a
g
e
4
5
6
10000 100000
AC overvoltage
1
duration of energisation, s
1000000
Figure 6.22 Comparison of the short duration stressing for positive polarity
impulses with the normal AC stressing strength for a nine-unit string
of cap and pin insulators; the surface was very heavily polluted [2]
that of the normal stressing, the slowness of the pollution flashover process is the
important feature. That is, the magnitude of this overvoltage has reduced appre-
ciably before the arc can extend by the critical amount over the length of the
insulator.
6.6.6 Iced insulator
The flashover voltage of insulators that are covered with ice depends greatly upon the
type, the densityandthe thickness of the ice plus the conductivityof the freezingwater,
the overall covering of the surface and the number of icicles. Wet grown ice (glaze)
with a density of about 0.9 g/cm
3
has been found to be more dangerous than other
types of atmospheric ice accretions [2]. The maximum withstand stress, expressed
per unit leakage length, is shown in Table 6.6 for various insulator types. There are
porcelain cap and pin (IEEE and antifog types), EPDM longrod and porcelain post.
The influence of freezing water conductivity as measured at 20
C on the withstand
stress is shown in Figure 6.23. These results are for a six-unit string of standard cap
and pin insulators when tested with an air temperature of 12
C. In such tests, the

ice thickness, , as recorded on a rotating monitor cylinder was two centimetres.
Table 6.6 AC withstand stress for some insulator types covered with ice
grown in wet and in dry regions [2]
Type of ice Ice density
(g/cm
3
)
Withstand stress (kV/m)*
IEEE** anti-fog EPDM post
Glaze with icicles 0.87 70 84 96 90
Rime <0.3 >148 >146 >168 >197
* maximum withstand voltage for an insulator of 1 m length
** small standard-type insulator
140
130
120
110
100
90
80
70
60
0 50 100
at each conductivity position,
the upper stress values are
generally for flashover;
the others are for withstand
uniform wet-grown ice
c =2cm
water conductivity, S/cm
m
a
x
i
m
u
m

w
i
t
h
s
t
a
n
d

s
t
r
e
s
s
,

k
V
/
m
150
Figure 6.23 AC withstand stress of a short vertical string of cap and pin insulators
as a function of freezing water conductivity [2]
50
0.1 0.2 0.3 0.4
volume density, g/cm
3
w
i
t
h
s
t
a
n
d

v
o
l
t
a
g
e
,

k
V
/
m

(
i
n
s
u
l
a
t
o
r

s
t
r
i
n
g

l
e
n
g
t
h
)
0.5 0.6
100
150
naturally snow covered
artificially snow covered
S/cm
20
30
50
100
Figure 6.24 AC withstand voltage of a vertical string of porcelain cap and pin
insulators as a function of snow density [2]
Some researchers have reported that the withstand voltage of insulators that had
their shed spacing completely bridged by artificial icicles was only 60 per cent of that
for the case without ice. When the ice thickness is much lower than one millimetre
and the insulators are precontaminated, ions from the pollution tend to dominate the
total electrical resistivity of the ice. For this latter situation, a flashover stress along
the insulators surface has been found to be as low as 20 kV/m.
When the insulator is energised during the formation stage of the ice, the ices
distribution along the insulator is affected. Often, some sections of the insulator may
be free of ice. This condition is particularly the case, in practice, for long insulator
strings, and has also been observed in laboratory tests. The cause is reasoned to be due
to the heating effect of the surface arcs and/or an increase in the temperature of the air.
6.6.7 Snow on insulators
The AC withstand or flashover voltage decreases with increasing water conductivity
from the melted snow, and with increasing snow density up to 0.5 g/cm
3
. Thereafter,
it remains constant. An example is shown in Figure 6.24 for a string of cap and pin
insulators; a few of the results are for a natural snow covering of the insulator but
most are for the case where the snow has been artificially produced.
6.7 Selection and dimensioning
Selection and dimensioning is partly dealt with in IEC Publication 815 [4], which
has been in existence for more than fifteen years and is based on limited practical
experience and some test results. It applies only to conventional porcelain and glass
insulators when used for AC applications. Although the approach specified in this
document is attractive because of its simplicity, it can lead to poor design.
This IEC 815 concept is based on a specific creepage distance. That is, for a
qualitatively defined level of pollution, a corresponding minimum creepage distance
is specified for the surface of the glass (or porcelain) along the length of the insulator.
The unit of this parameter is millimetres per kilovolt (phase-to-phase) of the maximum
value of the system voltage. Just four pollution levels are specified to cover the
whole range of typical environments. These pollution levels and the corresponding
specific creepage distances are called: light, 16 mm/kV; medium, 20 mm/kV; heavy,
25 mm/kV; very heavy, 31 mm/kV. In addition, some limiting geometrical features
of the insulators profile are stated essentially to avoid too much creepage distance
being crammed into a given axial length. Otherwise, this would lead to excessive
sparking between sheds and hence short circuiting of the leakage path.
A comprehensive analysis conducted by CIGRE Taskforce 33-13-01 has found
the specific creepage distance concept to be wanting in quite a number of aspects.
For example, salt fog tests conducted on twenty-seven insulator types of porcelain
and glass insulators comprising different designs and sizes for the cap and pin,
longrod insulator, posts and barrels have found a range, at flashover, from 20.6
to 36.1 mm/kV for the same salinity representing very heavy pollution. This large
variation found under very controlled laboratory conditions for vertically mounted
insulators, which comply with IEC815 profile restrictions, is supported by many other
results that have been obtained by conducting natural pollution tests. Consequently,
it can now be concluded that factors other than creepage path length greatly influence
the pollution flashover performance of porcelain and glass insulators under the same
environmental and electrical conditions.
In general, the CIGRE review [2] perceives IEC 815 to be weak and so in need
of revision in the following areas:
(a) performance of polymeric insulators
(b) insulator orientation
(c) extension of applicability to voltages above 525 kVAC system voltage
(d) design for DC applications
(e) insulators with semiconducting glaze
(f ) surge arrester housing performance, particularly with reference to polymeric
materials
(g) longitudinal breaks in interrupter equipment
(h) radio interference, television interference and audible noise of polluted
insulators
(i) effect of altitude
(j) effect of heavy wetting.
From this above list, it can be appreciated that a comprehensive revision of IEC 815
(now called 60815) is a mammoth task. Although it is well under way, the changes
proposedtodate have yet tobe approvedbythe international vettingauthority. Further,
many substantial aspects have not yet been dealt with. Therefore, no details are likely
be released for quite sometime.
However, the CIGRE document on this subject is nearing completion. It is based
on a flow chart similar to that shown in Figure 6.25; the aspects relevant to its various
basic data
application system parameters environment
project
objective
task
definition
detailed data
considerations
service experience
performance monitoring
inspection
research
insulator characteristics
constraints
site severity
measurements
laboratory tests
test certificates
candidate insulator types
evaluate
cost and time of the
dimensioning process
service
experience?
want to specify
insulator dimensions
yes
no
want to specify
a pollution test
insulator user
insulator supplier
both user and supplier
material profile maintenance lifetime
acceptable insulators
unacceptable
acceptable
asset management
approved insulator
evaluate
cost/risk
select field
trial site
select field
trial duration
insulation
specification
(dimensions)
evaluate
submitted
dimensions
select
laboratory
test method
select
laboratory
test severity
insulation
specification
(function)
qualification
Figure 6.25 Possible flowchart for the selection and dimensioning of outdoor
insulators
boxes are comprehensively dealt with in the forthcoming Applications Guidelines
document [3].
Essentially, this CIGREapproach entails first selectingcandidate insulators for the
intended installation by using the information contained within their two companion
documents [2, 3]. Firstly, the pollution severity to which these insulators will be
subjected at this installation has to be determined. Thereafter, the lengths of these
candidate insulators are estimated to have an acceptable pollution performance using
either relevant service experience or from conducting tests. Obviously, the most
thorough testing is carried out in a natural pollution environment that is very similar
to that of the intended installation. Unfortunately, this is not always possible due
to the required time and the cost of such tests, and so the use of relevant artificial
pollution is usually much more likely.
For an AC installation, the aerodynamic catch from air is the principal process by
which the insulators are contaminated [1]. In contrast, the electric field around the
insulator helps to greatly enhance this catch for a corresponding DC installation.
The determination of the severity of the pollution collected on the surface of
a candidate insulator at the site of the installation is the most difficult part of this
new procedure. Frequently, a standard insulator will to be used in lieu of candidate
insulators for this pollution assessment because, for various reasons for example,
size it is more convenient to do so. Then, the site pollution severity can be defined
as follows.
The site pollution severity (SPS) is the maximum value of either ESDD and
NSDD, or leakage current, recorded over an appreciable period of time i.e. one
or more years on a vertically mounted standard insulator comprising a string of
cap and pin units or a porcelain longrod. If rain occurs during this measuring period,
the measurements should be repeated at appropriate intervals; SPS is then the largest
value recorded during this series of measurements. Note:
(i) even if the largest values of ESDD and NSDD do not occur at the same time,
then SPS is nonetheless taken as the combination of these largest values
(ii) when there is no natural washing during the measuring period, the maximum
value of ESDD and NSDD can be estimated from the plot of deposit density as
a function of the logarithm of time.
The pollution deposit density on a certain type of insulator relative to that on a refer-
ence insulator (profile similar to that shown in Figure 6.1a, having an overall diameter
of 254 mm) is provided in Table 6.7 for vertically installed disc insulators. The corre-
sponding relative value for DC andAC voltages is given in Table 6.8. As horizontally
installed insulators generally collect less pollution than do vertically mounted
ones in the same location, a conservative approach is to use the same information as
for the vertical case.
Other aspects of this correlation process are dealt with in the CIGREdocument [3].
Prudent managers of high voltage systems will conduct routine inspections,
possibly supported with sample tests, of its outdoor insulators over their anticipated
long lifetime e.g. fifty years. It is particularly pertinent that such inspections should
be carried out on polymeric insulators to see if any detrimental ageing effects are
Table 6.7 Ratio of pollution deposit density on an insulator type
to that on the standard IEEE type [3]; for the vertical
mounting condition
Insulator type Desert Coastal* Industrial/agricultural Inland
Aerodynamic 0.4 0.3 0.6 0.3 0.7 0.3 0.6 0.2
Anti-fog 0.8 0.3 0.8 0.3 0.85 0.3 1.08 0.2
Longrod 0.4 0.3 1.6 0.4 0.4 0.3 0.4 0.3
* during rapid pollution build-up (e.g. typhoon or strong wind off the sea)
Table 6.8 Ratio of pollution deposit density for DC energisation to that for the
corresponding AC voltage
Ratio Site conditions
11.2 natural pollution only, such as sea and desert
1.31.9 natural pollution plus industrial pollution whose source is a few kilometres away
2.03.0 industrial pollution only, a few kilometres away and considered clean w.r.t.
AC voltage
taking place. If so, then this adverse behaviour obviously has to be taken into account
when a follow-up selection and dimensioning exercise is carried out.
6.8 Supplements, palliatives and other mitigating measures
Occasionally the pollution flashover performance is not good enough, which may be
due to a number of reasons. For example: an unforeseen shortcoming in the original
design, an increase in the pollution level e.g. introduction of new industry into
the location or the strategic importance of the supply facility is raised after its
construction. Often, the insulators cannot be changed for longer ones be it due to
cost or installation constraints, such as the size of the support structures especially
when it would reduce the ground clearance of the conductors below a safe level.
Fortunately, a whole raft of measures can now be adopted to improve the reliability;
some are essentially permanent while others need to be renewed from time to time.
6.8.1 Booster sheds
The booster shed [33] was developed to reduce the incidence of flashover under heavy
wetting conditions.
It is constructed from a copolymer of silicone rubber and polyethylene that is
radiation cross-linked. Such a polymer has now been shown to have a long and
excellent service history in very severe pollution, often saline.
fuse wire
booster shed
porcelain
Figure 6.26 An arrangement to initiate a sandwiched arc between a booster shed
and a porcelain surface
The advantage of this supplement is twofold. When it is fitted to every other or
every third porcelain shed of a vertically mounted insulator, cascading water will
break up before such booster sheds are bridged. As a stream of water drops has a
much greater resistance than that of a continuous column, the flashover withstand
strength is much improved. Second, when an arc extends across the part of the
insulators surface that is covered with this shed, it is more highly cooled i.e.
greater value of A(x) applies in Suits equation, see section 6.5 than it would be
if the polymer were not present. Therefore, as seen from Equation (6.5), the arcs
resistance is correspondingly larger. The value of A(x) increases as the gap between
the polymer and the porcelain decreases [33]. However, this gap cannot be made
too small; otherwise any propagating discharge would severely damage the adjacent
materials. The optimum spacing is about 6 mm. As a consequence of fitting booster
sheds, a higher value of pollution severity can be withstood before flashover occurs.
Should, nonetheless, a flashover take place there is some concern about the resul-
tant damage especially having seen the consequence of such an arc confined within
a cracked head of a cap and pin insulator (see section 6.4.1.1). To investigate this
possibility, such a sandwiched arc was initiated using fuse wire, as illustrated in
Figure 6.26 (unpublished research of the author). High speed photographs revealed
that an arc carrying a current of some kilo amperes rapidly transferred from being so
sandwiched to become free to move in the air above the booster shed. The damage
was minimal and mainly to the polymeric material. Although extensive blackening of
its surface had occurred, it was easily rubbed off to leave essentially sound material.
The reason for this fortunate situation lies with the low inertia of the arc, the high
buoyancy forces of hot gases and the extensive wriggles of such a sandwiched arc
as witnessed in a complementary experiment involving an arc conducting a current
of just one ampere [33].
6.8.2 Shed extender
In contrast to the previous case where the polymeric shed was supported just above
the surface of the insulator, the polymeric shed of the extender is bonded to the surface
of the shed see Figure 6.27 in order to lengthen the leakage path. The adhesive
porcelain
shed
polymeric
extender
Figure 6.27 A silicone rubber shed extender fitted to a porcelain insulator
used to achieve this bond is a crucial component of this supplement, because surface
discharges in any air gaps will damage both the porcelain and the polymer.
Although the maximum increase in the leakage path is obtained by having all
these extenders of equal size, there is scope for a long and short shed arrangement
to be provided. That is, it helps to cope with polluted cascading water even though
the water run off is far from being in the category of heavy wetting.
Of course, with this construction, there will be a mixture of hydrophobic and
hydrophilic surfaces. That is, from the leakage current viewpoint, films and drops
of water are now in series. This is another aspect that has so far received little study
with regard to the flashover process and so it becomes a research topic that warrants
some investigation.
A fairly recent development [34] that is now coming into service avoids this
problem by covering the porcelain with a silicone rubber coating (see sections 6.3
and 6.8.4) after the extender has been added. By this process, a somewhat hybrid
hydrophilichydrophobic surface has been converted into a fully hydrophobic one.
6.8.3 Shed protector
Should an appreciable part of a porcelain shed be broken, the pollution withstand
strength of that insulator is proportionally reduced. Two measures that have been
adopted to protect the most vulnerable part of the shed against vandal attack using
hard projectiles, such as stones, are shown in Figure 6.28.
In one, Figure 6.28a, the complete shed is covered with a polymeric shield which
is essentially a mini booster shed. In the other, Figure 6.28b, the circumference of the
shed is protected with an impact absorbing polymeric tyre.
6.8.4 Coatings
A long established method to enhance the pollution flashover performance of the
porcelain insulator is to coat its surface with a layer of grease, which has two benefits.
First, it provides a hydrophobic surface and, second, it encapsulates and/or imparts
water repellent properties to the pollutants. The two types of grease are hydrocarbon
and silicone. The former is restricted to the temperate zone because of its fairly low
sliding temperature, but the latter is used throughout the world. The hydrocarbon
version is relatively inexpensive, applied as a substantial layer and can last up to
porcelain
shed
porcelain
shed
polymeric
shield
polymeric
tyre
a
b
Figure 6.28 Two examples of shed protectors
a polymeric shield
b polymeric tyre
about ten years. Silicone grease is fairly expensive, applied as a thin layer and may
have to be replaced after a year. Such greases are normally applied by hand or by
using a brush or spray. The disadvantage lies with the removal of dirty grease before
a new coating is applied. If this change is not done in good time, trapped electrical
discharges that can occur beneath the greases surface may damage the porcelain. The
indication of the need for such a change is the presence of dry band discharges on the
surface of the grease.
The good service record of silicone rubber insulators for outdoor applications
has stimulated the development of RTV rubber coatings. All known commercially
produced variants consist of polydimethylsiloxane polymer, alumina trihydrate
or alternative filler to increase the tracking and erosion resistance, a catalyst and a
cross-linking agent. Such materials are dispersed in either naphtha or trichlorethylene
solvent, which acts as the carrier to transfer the RTVrubber to the insulators surface.
End of life is not easy to determine accurately [35]. Service experience with both AC
and DC systems has found that the useful life of a RTV coating can vary from less
than a year to many years. Basically, when such a coating loses some of its water
repellence, it may be washed and/or cleaned using a dry abrasive such as crushed
corncobs or walnut shells and then recoated provided the existing coating is well
bonded to the ceramic surface.
6.8.5 Washing
Because dry pollutants rarely cause leakage current, the obvious way to overcome
the flashover problem is to remove them before they reach the critical level albeit
if that means getting them wet to do so!
If the insulator can be deenergised, then washing with tap water is a simple
process. However, insulators generally must be kept in service and so live washing
sometimes referred to as hot line washing has to be undertaken. For this procedure
to be conducted without promoting flashover, two conditions are usually essential.
The water should be demineralised (resistivity > 50 000 ohm cm). Also, the washing
sequence should be such as to avoid the concentration of the dissolved pollutants.
That is, for a vertically mounted insulator, the bottom section should be washed clean
before the middle section is wetted and finally, the top section is wetted only when
the rest of the insulator is clean.
The four methods of live insulator washing that are most often used which differ
mainly in the type of nozzle arrangement that is employed to project the water are:
(i) portable handheld jet nozzles
(ii) fixed spray nozzles
(iii) remote controlled jet nozzles, often automated using robots
(iv) helicopter mounted nozzles.
The many safety procedures that must be followed have been extensively dealt with
in the CIGRE review document [2].
6.8.6 Provisions for ice and snow
For increasing insulator reliability under these special albeit relatively infrequent
conditions, it is generally necessary to increase the length of the insulators. When
ground clearance poses a problem, the usual practice is to use an inverted Vinsulator
mounting arrangement i.e. two insulators are hung at 45
to the vertical in lieu of

one normally suspended. Other measures are dealt with briefly in the CIGRE review
document [2].
6.9 Miscellaneous
A number of relevant aspects that do not fit readily into the previous sections
are discussed briefly below. These range from physical processes to proposals,
speculations, future needs and financial matters.
6.9.1 Cold switch-on and thermal lag
As can be readily reasoned from section 6.5, the pollution flashover strength is at a
minimum when the insulator is well wetted but not sufficiently so for the washing
of the surface to have occurred. Such a vulnerable condition can happen when an
insulator is reenergised after it has been out of service for some time e.g. following
maintenance/repair or after switch-out overnight for voltage control purposes. It was
just such a situation that resulted in the flashover of strings containing some cracked
porcelain insulators that were referred to in section 6.4.1.4.
Even with energised insulators, enhanced moistening happens when the insula-
tors temperature is only a few degrees lower than that of the surrounding air [36].
Generally, this is an early morning situation when the ambient temperature rises
quickly and is especially pertinent in deserts.
Although there is insufficient evidence to quantify this problem accurately,
guidance nonetheless is provided by considering some results obtained for
semiconducting glazed insulators [2]. For example, the flashover voltage of a cold
switch-on for a soaking wet condition (0.15 Mper suspension unit) was 40 per cent
lower than that for the same insulators when only damp (1000 M per unit) and
fourfold less than the value for the dry case (15 000 M per unit).
The above quoted conditions apply to hydrophilic surfaces. As yet, there are no
corresponding data for the hydrophobic case. Hence, there is a particular need for
conducting such a study with the silicone rubber insulator, because its thermal mass
and thermal conductivity are appreciably different from those of porcelain.
6.9.2 Semiconducting glaze
A metallic oxide, such as that produced from tin, is doped to provide semiconduct-
ing properties. When this material is applied to the surface of a porcelain insulator
as a glaze to achieve a continuous current of about 1 mA under normal operating
conditions, there is sufficient heat on the insulators surface to keep it essentially dry
in dew or fog.
Such an insulator has a withstand voltage of three to four times that of the
corresponding ordinary glazed unit when moistened. Although this insulator type
may, therefore, seem a solution to the severe pollution problem especially that
arising from thermal lag cold switch-on remains a difficulty. However, the latter
may be overcome by energising such circuits when thermal equilibrium has become
established between the air and the insulator.
Post insulators coated with such a glaze operated very successfully for over ten
years at the Brighton Insulator Testing Station but, unfortunately, a type of cap and
pin version did not. However, a manufacturer now claims that an improved design
has been produced for AC applications by substantially increasing the thickness of
this glaze around the pin cavity. These insulators may not be suitable for use on
DC systems because of electrolytic corrosion of the glaze.
6.9.3 Live line working
To reduce costs, working on energised OHLs is becoming an increasing practice.
If the insulators have to support the weight of the linesmen, their structural integrity
is obviously of paramount importance. Questions have, therefore, been raised
about the possibility, albeit probably rare, of there being the beginning of a brittle
fracture in the RBGF rod of the polymeric longrod see section 6.3. Consequently,
there is a need to develop a test for detecting such an eventuality.
Apossibility of coping with this uncertainty is to measure the vibrational signature
of a polymeric insulator in situ; perhaps by using the method developed for detecting
radial cracks in porcelain cap and pin insulators [12]. Basically, a small electro-
magnetic vibrator is fixed to the base of the tower and the minute movement of the
insulators surface is measured at ground level by using a laser Doppler vibrometer.
By slowly changing the frequency of the vibrator, the resonances of the insulator are
determined. From comparing this signature with that of a sound unit, abnormalities
can hopefully be identified.
6.9.4 Visual annoyance, audible noise and electromagnetic compatibility
Even without causing flashovers, the flow of leakage current across polluted and
wetted insulators can be a source of nuisance to people and can interfere with
communication systems.
Apollution severity of only one tenth of that necessary for flashovers to happen is
sufficient for appreciable gas discharges and audible noise (AN) to occur [2]. Visible
discharges can cause annoyance, especially where an OHLpasses through a populated
area. Although AN does not pose a serious problem under normal surface discharge
conditions for either ACor DCsystems, single unit sparkovers under DCenergisation
produce periodic loud bangs that can continue for hours.
The signal to noise ratio for acceptable reception is 20 dB and 35 dB for radio and
television, respectively; this difference arises because the eye is more sensitive than
the ear tointerference. The acceptable upper limits onradiointerference (RI) produced
by overhead lines and substations are covered generally in CISPR publication 18-2
[37] and for polluted insulators in CISIR publication 18.2, Amendment 1 [38]. For
normal surface discharges on AC and DC insulators, RI and television interference
(TVI) are not usually severe and do not increase much as the pollution severity
gets greater. However, single unit sparkovers on DC insulators can be troublesome.
Some small-scale investigations [40] have found that the electromagnetic interfer-
ence under AC energisation for polluted and wetted silicone rubber insulators
was substantially lower than that for two corresponding ceramic insulators one a
short string of glass cap and pin units, the other a porcelain longrod and that for
polymeric insulators made of EPDM and epoxy resin.
Any of these problems can be overcome completely, or be very substantially
reduced, by adopting some of the measures covered in the previous section namely:
using greases, silicone rubber coating, semiconducting glaze and by washing.
6.9.5 Electric field distributions
Capacitor-type internal insulation in bushing and current transformers is designed to
provide uniform radial and axial electrical stress distributions on the inside of their
weather housings. Also, the varistor arrangement of a surge arrester is such as to
achieve a similar situation. As can be readily reasoned from section 6.5, an uneven
distribution of the pollutants and/or the water will result in a non-uniformvoltage dis-
tribution along the outside surface of this insulator. Of special note, the heating from
the components within this housing can increase the degree of this uneven wetting.
Such differential axial voltage distributions produce a radial electric field. This
resultant radial field may cause electrical discharges to occur within this insula-
tor. In extreme situations, the insulator may puncture if the internal stress-relieving
measures are not properly designed [40].
6.9.6 Interphase spacers
These insulators are mainly used to prevent mid-span flashovers of overhead lines
during conditions of galloping and conductor jumping following ice release. In addi-
tion, phase-to-phase spacers may be required to achieve compact line constructions,
to reduce phase spacing so as to decrease the magnetic field levels or to improve the
aesthetics of the line.
Such spacers are made from either porcelain or polymers. The procedure adopted
for obtaining their dimensioning is the same as that outlined in section 6.7 for the
line to ground insulation, except that the withstand voltage is

3 greater. Further,
considering the dire consequences of a phase-to-phase flashover on the operation of
a transmission line, an additional safety margin may need to be included.
6.9.7 Compact and low profile lines
Wayleave sometimes called right of way requirements are becoming more difficult
to obtain in some countries and so there is a growing pressure to make OHLs as
narrow as possible. Also, in order to gain permission to erect lines across areas of
outstanding natural beauty there is often a need to keep these lines as low as possible.
Hence, there is a desire to achieve compact line arrangements, thereby providing
scope for the designing of more imaginative insulators/support structures possibly
making some components act as both functions.
One wonders, therefore, if the good pollution flashover performance of silicone
rubber could be exploited more than has happened so far. Currently, there is a reluc-
tance to design an insulator around the incorporation of the RTV coating because the
latter will probably need to be renewed from time to time. However, if the benefit
of its application to the surface of a rigid insulator e.g. porcelain, epoxy resin, a
substantial section of glass fibre rod outweighs this disadvantage then its use may
be justified.
6.9.8 Financial and related matters
Although insulators form a relatively small part of the total cost of a new electricity
supply project, the potential consequences of their bad performance can be costly.
It is, therefore, important to consider this aspect when selecting insulators for a
particular function. One method of approach is to define an operational risk, which
is the likelihood of failure either flashover or mechanical breakage multiplied by
the severity of the consequences of that failure.
In addition to the aspect covered in section 6.7, the general specifications of
the minimum technical requirements of insulators and the tests to verify them
are normally obtained from international and/or national standards, e.g. IEC, IEEE,
BritishStandards Institute. Further, the users practice mayplayits part inthis defining
process. Insulator costs, even when expressed in relative terms, can vary markedly
over a short period of time. Such variations can depend on matters other than technical
ones e.g. if a new manufacturer wishes to break into the market or cheap labour
becomes available in an emerging economy.
Even though the electricity supply industry usually expects its insulators to last
for thirty plus years, there are special situations in which a much shorter lifetime
is still very useful. A good example drawn from the authors experience is in
the KwaZulu-Natal Province of South Africa. The long and strategic 400 kV OHLs
have been reinsulated a number of times in an attempt to reduce the number of
serious flashovers caused by pollution. It was estimated that if such events could
be eliminated, or substantially reduced, for just seven years by the use of silicone
rubber insulators, their cost would be recouped by avoiding the need to live wash
(using helicopters) the existing porcelain insulators. It is encouraging to note that the
pollution flashover performance of these silicone rubber insulators has been excellent
for their nearly ten years now (2004) in service.
Asset management, in its formalised form, has of late become very fashionable in
many industries especially those that have been deregulated. Therefore, it is timely
that a paper [41] was presented at the 2002 CIGREconference that deals with a survey
conducted by CIGRE Taskforce 23.18 and to note that this group of experts propose
to publish a CIGRE brochure to present their guidelines in 2005. Also, at the thir-
teenth International Symposium on High Voltage Engineering (held in 2003 at Delft,
Holland), the organisers included for the first time a subject entitled Asset manage-
ment of HV equipment: strategies and tools. Further, such a topic may be discussed
at such symposiums in the future. Hopefully, these two respected bodies will pro-
vide clear guidelines on how outdoor insulators should be monitored, maintained and
replaced when deemed necessary. In the meantime, it is suggested that prudent man-
agers should note the remarks concerning inspections that are included in section 6.7.
6.10 References
1 LOOMS, J.S.T.: Insulators for high voltages (IEE Power Engineering Series,
No. 7, Peter Peregrinus Ltd., 1988)
2 CIGRE Taskforce 33.04.01: Polluted insulators: review of current knowledge.
CIGRE technical brochure 158, June 2000
3 CIGRE Taskforce 33.13.01 (formally 33.04.01): Guidelines for the selection
and dimensioning of insulators for outdoor applications: part 1, General prin-
ciples and the AC case. CIGRE technical brochure, publication is likely in
2005
4 IEC Publication 815: Guide for the selection of insulators in respect of polluted
environments. 1986
5 CHANDLER, H.D., JONES, R.L., and REYNDERS, J.P.: Stress corrosion
failure on composite longrod insulators. 4th international symposium on High
voltage engineering (ISH), Athens, Greece, paper 23.09, 1983
6 DE TOURILLE, C., PARGAMIN, L., THENVENET, G., PRAT, S., and
SIAMPIRINGUE, N.: Brittle fracture of composite insulators: the new expla-
nation and a field case study. 12th international symposium on High voltage
engineering (ISH), Bangalore, India, 2001, 3, pp. 683686
7 SWIFT, D.A.: Arresting arc propagation with narrow metal strips, IEE Proc.
A, Phys. Sci. Meas. Instrum. Manage. Educ. Rev. 1980, 127, pp. 553564
8 SEDIVER BROCHURE.: HRTG insulators for HVDC applications. 500ARP
99/MRG, 79 avenue Francois Arago, 92017 Nanterre Cedex, France, December
1999
9 CRABTREE, L.M., MACKEY, K.J., KITO, K., NAITO, K., WATANABE, A.,
and IRIE, T.: Studies on electrolytic corrosion of hardware of DC line
insulators, IEEE Trans., 1985, PAS-104, (3), pp. 645654
10 MADDOCK, B.J., ALLNUTT, J.G., and FERGUSION, J.M., et al.: Some
investigations of the ageing of overhead lines. CIGRE, paper 22.09, 1986
11 CROUCH, A.G., SWIFT, D.A., PARRAUD, R., and DE DECKER, D.: Ageing
mechanisms of AC energized insulators. CIGRE, paper 22-203, 1990
12 SWIFT, D.A.: Cap and pin insulators: an assessment of techniques for detecting
defective porcelain. International conference on Revitalising transmission and
distribution systems, IEE Publ., 273, 1987, pp. 17
13 SWIFT, D.A., SPELLMAN, C.A., JAMES, G., and HADDAD, A.: Stray capac-
itance of cap and pin insulators in suspension strings: analytical and computer
techniques. 36th universities Power engineering conference (UPEC), Swansea,
UK, paper 1C-1, 2001
14 BARCLAY, A.L., and SWIFT, D.A.: Cap and pin insulators: electrical puncture
of porcelain under AC energization. 5th international conference on Dielectric
materials, measurements and applications, University of Kent at Canterbury,
UK, 2730 June 1988
15 CHERNEY, E.A., and HOOTON, R.D.: Cement growth failure mechanism in
porcelain suspension insulators, IEEE Trans., 1987, PWRD-2, (1), p. 249
16 MARTIN, R.: Silicone rubber station posts and bushings. Maclean Power
Systems brochures, 11411 Addison Street, Franklin Park, 1L 60131-11300,
USA, 2002
17 SUITS, C.G.: High pressure arcs in common gases in free convection, Phys.
Rev., 1939, 55, pp. 561567
18 CLAVERIE, P.: Predetermination of the behaviour of polluted insulators, IEEE
Trans., 1971, PAS 90, (4), pp. 19021908
19 RIZK, F.A.M.: Mathematical models for pollution flashover, Electra, 1981,
(78), pp. 71103
20 HOLTZHAUSEN, J.P.: Acritical evaluation of AC pollution flashover models
for HV insulators having hydrophilic surfaces. Ph.D. dissertation, University
of Stellenbosch, 1997
21 HOLTZHAUSEN, J.P., and SWIFT, D.A.: The pollution flashover of AC and
DC energized cap and pin insulators: role of shortening of the arc. 11th inter-
national symposium on High voltage engineering (ISH), London, UK, 1999, 4,
pp. 333336
22 SWIFT, D.A.: AC flashover mechanism for water droplets on an hydrophobic
insulator. 8th international symposium on High voltage engineering (ISH),
Yokohama, Japan, paper 44.09, 1993, pp. 113116
23 KARADY, G.G., SHAH, M., and BROWN, R.L.: Flashover mechanisms of
silicone rubber insulators used for outdoor insulation part I and II, IEEE Trans.
Power Deliv., 1995, 10, (4), pp. 19651978
24 RAVERA, C.N., BRITTEN, A.C., and SWIFT, D.A.: Service experience
with polymeric insulators in ESKOM, South Africa. Conference on Appli-
cations on non-ceramic outdoor insulation, Paris, France, 78 June,
1994
25 RAVERA, C.N., OLIVIER, P.J., BRITTEN, A.C., and SWIFT, D.A.: Silicone
rubber insulators on ESKOMS AC transmission lines. IEE Conf. Publ. 423,
AC and DC power transmission, 29 April3 May, 1996
26 LIANG XIDONG, WANG SHAOWU, HUANG LENGCENG, SHEN
QUINGE, and CHENGXUEQUI: Artificial pollution test and pollution perfor-
mance of composite insulators. 11th international symposium on High voltage
engineering (ISH), London, UK, 1999, 4, pp. 337340
27 HOULGATE, R.G., and SWIFT, D.A.: Polymeric insulators: AC flashover
voltage under artificial salt-pollution of new and naturally-aged units compared
to porcelain. 6th international symposium on High voltage engineering (ISH),
New Orleans, USA, paper 47.03, 1989
28 SWIFT, D.A., SPELLMAN, C.A., and HADDAD, A.: Leakage current
on silicone rubber insulators and a RTV coating in clean fog: hydropho-
bicity transfer to pollution. 12th international symposium on High voltage
engineering (ISH), Bangalore, India, 2001, 3, pp. 607610
29 WANGSHAOWU, LIANGXIDONG, GUANZHICHENG, and WANGXUM:
Hydrophobicity transfer properties of silicone rubber contaminated by different
kinds of pollution. CEIDP, Victoria, Canada, 2000, pp. 373376
30 HOULGATE, R.G., and SWIFT, D.A.: ACcircuit breakers: pollution flashover
performance of various types of interrupter head, IEE Proc., Gener. Transm.
Distrib., 1997, 144, (1), pp. 5056
31 Electric Power Research Institute: Transmission line reference book, 345kV
and above (Second Edition, 1982)
32 Electric Power Research Institute and Bonneville Power Administration:
Transmission line reference book: HVDC to 600 kV, undated
33 ELY, C.H.A., LAMBETH, P.J., LOOMS, J.S.T., and SWIFT, D.A.: Discharges
over wet, polluted polymers: the booster shed. CIGRE paper 15-02, 1978
34 Mace Technologies Brochure: Insilcure. PO Box 739, Lonehill 2062, RSA,
undated
35 CHERNEY, E.A., and GORUR, R.S.: RTVsilicone rubber coatings for outdoor
insulators, IEEE Trans. Dielectr. Electr. Insul., Oct. 1999, 6, (5), pp. 605611
36 ORBIN, D.R.H., and SWIFT, D.A.: Surface resistivity of a cool polluted insu-
lator when exposed to warm ambient air. Proceedings of 4th international
conference on Properties and applications of dielectric materials, Brisbane,
Australia, paper 7244, July 38, 1994
37 IEC Publication 437: Radio interference tests on high voltage insulators, 1973
38 CISPR Publication 16-1: Specification for radio disturbances and immunity
apparatus and methods, 1993
39 SWIFT, D.A., and BRITTEN, A.C.: Electromagnetic interference from high-
voltage insulators: a comparison of hydrophobic and hydrophilic cases.
International conference on Electromagnetic compatibility, Rome, Italy,
September 1720, 1996, pp. 846851
40 MELIK, G.: private communication
41 BARLETT, S.: Asset management in a de-regulated environment. CIGRE
paper 23-303, 2002
Chapter 7
Overvoltages and insulation coordination
on transmission networks
D.M. German and A. Haddad
7.1 Introduction
In this chapter we review some aspects of insulation coordination and overvoltages
on transmission networks. The purpose of insulation coordination is to ensure that
the probability of insulation breakdown is limited to an acceptable value and that
any breakdown is restricted to self-restoring insulation. It is based on computing
the most severe overvoltages occurring on the network and relating these to the
breakdown characteristics of the insulation through appropriate margins to obtain
withstand voltages for the network components together with the statistical risk of
insulation failure. The procedure is summarised in Figure 7.1.
Fromthe network configuration and component data, an appropriate simulation of
the network is formulated and used to compute the overvoltages, taking into account
the effect of control and protective devices. The breakdown voltage characteristics
are based on the type and configuration of the insulation. The coordination procedures
relate these two components to give withstand voltages and to quantify the risk of
insulation failure.
Coordination is an integral part of the design process, giving criteria that are
satisfied through insulation design and through overvoltage control and protection.
Although simple in concept, involving comparisons between maximumnetwork over-
voltages and minimum insulation breakdown voltages, in practice the process is
complex.
Network overvoltages are usually the result of some sudden change in conditions,
typically switching operations, lightning strokes or faults. They have a very wide
network configuration
component data
insulation characteristics network simulation
coordination
network computation
withstand voltages
risk of failure
Figure 7.1 Insulation coordination
range of waveshapes and durations, the most important being:
temporary overvoltages lightly damped oscillations at supply frequency
switching overvoltages damped oscillations at frequencies of <10 kHz
lightning overvoltages damped oscillations at frequencies <100 kHz.
The three types of insulation gaseous, liquid and solid have different breakdown
characteristics, which are dependent on the configuration and environment of the
insulation and on the waveshape and duration of the applied voltage. The coordination
process relates these various characteristics to the temporary, switching and lightning
overvoltages on the network.
Air still provides the major insulation for transmission networks. An important
property is that, across the long air gaps necessary for transmission voltages, the
breakdown voltage is dependent on the voltage rise time, and is a minimum for the
relatively long rise times associated with switching overvoltages. Coordination is
therefore based primarily on switching overvoltages, although it is still important
to confirm that the performance under both lightning and temporary overvoltages is
acceptable.
A second important property of air insulation is that for a given rise time of the
applied voltage the actual breakdown value will vary randomly about a mean value.
This characteristic is represented by the cumulative probability function P(U) where
P(U = U
b
) is the probability of breakdown occurring with a crest voltage U = U
b
and shown in Figure 7.2.
In computing switching overvoltages, it is often not possible to specify the combi-
nation of parameter values that will result in the maximum overvoltage. In particular
the circuit breaker may close at any point on the AC cycle and, due to circuit breaker
tolerances, the instants at which each pole closes will vary randomly about this point.
By computing the switching overvoltages for a large number of switching operations
Overvoltages and insulation coordination on transmission networks 311
1.5
1
0.5
0
0 1 2 3 4 5
U p.u.
F(U)
P(U)
F(U)
P(U)
Figure 7.2 Cumulative overvoltage distribution F(U) and breakdown voltage
distribution P(U)
with different instants of pole closing a switching overvoltage distribution F(U)
is obtained (Figure 7.2), with F(U = U
e
) giving the probability of the switching
overvoltage exceeding U
e
.
From the probability distributions of F(U) and P(U) shown in Figure 7.2,
statistical methods are used to specify insulation withstand voltage and risk of failure.
The International Standard IEC 60071: Insulation coordination [1, 2] provides
a comprehensive guide to coordinating procedures and specifies coordination factors
and standard insulation levels.
Although coordination is based primarily on determining withstand voltages and
risks of failure from overvoltage and insulation characteristics, other factors are
involved. The economic factor is always important and includes cost comparisons
between overvoltage protection, insulation levels, risks of failure and network out-
ages. Another factor of increasing importance is the environmental one and this
has resulted in the design of compact lines having both reduced visual impact and
transmission corridor width.
Thus, insulation coordination takes into account a range of factors and has param-
eters that are interdependent. An optimum design therefore involves an iterative and
integrated analysis of the network.
7.2 System overvoltages
Increase in demand for electrical energy has led to increase in both the size and
capacity of electrical power systems. High voltage AC systems of up to 750 kV are
now in operation in several countries; future extension of the working voltage to the
megavolt (MV) range is beinginvestigatedbymanyresearchers andelectrical utilities.
Moreover, the advances in solid-state electronics have resulted in the development of
high voltage high power thyristors and the expansion of HVDCtransmission systems.
However, such advances are inevitably accompanied by technical problems.
The cost of equipment and the system as a whole is closely related to the insula-
tion level which has been adopted for the network. The choice of insulation level
should take into account both the probability of occurrence and the severity of stress
that the system may be subjected to. Therefore, the performance of any protective
device inserted into the system should be defined as quantitatively as possible.
Overvoltages represent a major threat to security and continuity of supply. Ashort
summary of the sources of these overvoltages is given in this section. Fortunately,
a controlled limitation of overvoltage surges is possible by protective devices. Surge
arresters are widely used in electrical power systems, and the superior modern metal
oxide surge arresters have renewed interest and widespread use of surge arresters in
protection practice (see Chapter 5).
For economic design of equipment and safe operation of power systems, a detailed
knowledge of types and sources of overvoltages on power systems is required.
Figure 7.3 shows a classification of the different types of overvoltage that are likely to
occur on todays modern power systems. As can be seen in the figure, and according
to their origin, two main categories of overvoltages can be distinguished: (i) external
and (ii) internal overvoltages.
7.2.1 External overvoltages
External overvoltages are generated by sources that are external to the power system
network. Their magnitude is essentially independent of the system. Two types of
external overvoltage have been identified.
7.2.1.1 Lightning overvoltages
Lightning overvoltages are characterised by very high peak currents and relatively
low energy content. They are responsible for nearly half of all short circuits on lines
in systems of 300 kV and above [3, 4]. Up to approximately 300 kV, the system
insulation has to be designed primarily to withstand lightning surges.
The lightning mechanism is now well understood [57] to be due to charged
bipolar cumulo-nimbus clouds which induce an opposite charge on the ground. When
the local electric field reaches a critical value of about 30 kV/cm, an electric discharge
is initiated. Anegative polarity discharge advances in steps in a channel known as the
stepped leader. When the strongly ionised channel bridges the path between cloud
and earth, a very high current discharge takes place; this is known as the return
stroke. Subsequent discharges of dart leaders and return strokes may follow along
the ionised path at irregular intervals of time for periods of time less than a second.
Up to 40 strikes of decreasing magnitude have been observed. It is postulated that
these subsequent strikes correspond to discharging deeply located charges inside the
cloud (for a more detailed reading on lightning phenomena, refer to Chapter 3).
Lightning discharges may be of positive or negative polarity (according to the
polarity of stepped leader tip) and ascending or descending type, i.e. upwards or down-
wards going discharges. Positive strikes, in general, consist of one long discharge and
are observed to discharge currents of very high amplitudes (up to 250 kA), in contrast,
contact with circuits of higher voltages
neutral inversion
arcing ground phenomenon
resonance phenomena
closing/reclosing of line
interruption of capacitive/inductive currents
switching on transformers
fault initiation or clearing
direct
induced
backflashover
nuclear accident
magnetic solar storms
load rejection
Ferranti effect
self-excitation
saturation effects
ground faults
asymmetric faults
transient temporary steady state
lightning NEMP NNEMP
overvoltages
external internal
Figure 7.3 Classification of overvoltages
negative strikes are of lower intensity (less than 80 kA). Field observations and mea-
surements have shown that up to 90 per cent of lightning discharges are of negative
polarity and descending type. From the point of view of power system overvoltages,
the important parameters of the lightning stroke are the current amplitude and shape.
Both are statistically distributed. Field measurements [8] have shown that the modal
value of current is about 32 kA. Typical rise times are about 2 s.
Lightning affects overhead lines in three ways, that is (i) a direct strike on a phase
conductor or (ii) induced overvoltage if the impact point of lightning is in the vicinity
of the line, which is particularly dangerous for low and medium voltage systems.
In addition, (iii) a backflashover may occur due to a rise in potential of the tower top
when lightning hits the tower or the earth wire of shielded lines.
The overvoltage, which is determined by the line surge impedance and the
discharge current of lightning, is often over one million volts. These surges travel
along the line and may cause internal breakdown of insulation of terminal equipment
systems. They also generate corona which can be considered as a protective effect.
It is the practice that protection of power systems against lightning overvoltages is
usually decided on the most probable lightning current shape which might hit the
system and the keraunic level or density of lightning flashes in the location area of
the system (flashes per kilometer square per year). Up to 0.46 strikes/km/year were
recorded and lightning discharges of crest current of approximately 40 kAwere found
to be the most frequent in the United Kingdom [9]. These figures, although substan-
tially smaller than those of other parts the world [10] (mainly Europe, South America
and Africa), show that lightning overvoltages are a serious threat for power systems
in the UK.
7.2.1.2 Nuclear and non-nuclear electromagnetic pulses (NEMP, NNEMP)
NEMP are also an external source of power system overvoltages [11]. They are char-
acterised by very short front times (<10 ns) and tens of kiloamperes in amplitude.
Their occurrence is fortunately rare and protecting against themis usually not consid-
ered in power systems design. An attractive feature of ZnOmaterial is its fast response
to limit overvoltages. It has been shown that ZnO material could be used for effi-
cient protection against NEMP[12]. NNEMPhave been reported [13] to affect power
systems during magnetic storms associated with exceptional solar disturbances.
7.2.2 Internal overvoltages
Unlike external overvoltages, internal overvoltages are principally determined by the
system configuration and its parameters. They are mainly due to switching opera-
tions and following fault conditions. According to their duration, which ranges from
few hundreds of microseconds to several seconds, three main categories of internal
overvoltage can be defined.
7.2.2.1 Transient switching surges
Compared with lightning surges, switching surges contain a much higher amount of
energy but are of lower amplitude. Several causes have been identified for switch-
ing transient overvoltages, a brief account of which will be given here; detailed
descriptions can be found elsewhere [1416]. The following major causes have been
reported:
(i) Closing and reclosing of a line: the resistance, capacitance and induc-
tance of transmission lines form a distributed parameter (oscillatory) network
[17, 18]. Therefore, any switching operation involves transient phenomena.
For example, switching an unloaded line which is open at the other end pro-
duces up to 2 p.u. overvoltage, but this value could be even higher if a circuit
breaker restrikes on a line with residual voltage. Such voltages are known to
exist because of the presence of charges trapped in transmission lines after
interruption of supply voltage.
(ii) Switching of capacitive currents: this includes capacitor banks, unloaded over-
head lines and cables. It involves interruption of small capacitive currents at
peak voltage. The recovery voltage may reach 1.5 p.u. across the contacts of
the circuit breaker [15]. Half a cycle later, when the voltage of the feeding side
attains its maximum, a voltage of 2.5 p.u. is then established across the circuit
breaker contacts which may lead to restrike of the circuit breaker.
(iii) Switching of inductive current: when interrupting the small inductive currents
of unloaded transformers and shunt reactors, current chopping may occur with
the current forced to zero before the natural current zero. The high di/dt asso-
ciated with current chopping results in high induced voltage in the inductive
circuit. This type of operation produces overvoltages of 2 to 3 p.u. in mod-
ern transformers [15]; however, with transformers loaded with shunt reactors,
values up to 5 p.u. may be reached which necessitates the use of surge arresters
for protection.
(iv) Initiation and clearing of system faults: the most frequent fault on power sys-
tems is the short circuit phase-to-earth which is often accompanied by an
increase of neutral potential. If the fault occurs at peak voltage, then up to
2.7 p.u. overvoltage can be generated. Reducing the ratio (X
0
/X
1
), however,
will limit significantly the overvoltage; e.g. for a configuration having a ratio
equal to 1, no overvoltage has been observed [15]. The single phase fault which
causes an asymmetry is therefore the most dangerous fault. The clearing of a
fault by circuit breakers, in particular a three-phase fault, will also generate
crest overvoltages of up to 3 p.u.
The magnitude of switching surges is affected by the following factors:
(i) line parameters including dimensions, earth resistivity, trapped charges,
terminating network and coupled energised circuits
(ii) the circuit breaker performance
(iii) the source network.
Many techniques have been developed [15, 19] and are widely used to reduce the
peak value of switching transients to values lower than 2 p.u., which helps economical
design of higher voltage systems (400 kVand above). Among these techniques are:
(i) switching resistors
(ii) controlled synchronised closing of circuit breakers
(iii) shunt reactors
(iv) drainage of trapped charges before reclosing by provision of leak resistors.
In addition, protective measures, such as surge protection capacitors and surge
arresters, are adopted. Modern metal oxide surge arresters are particularly efficient
for such duties because of their fast switching response, energy absorption capability
and excellent voltagecurrent non-linearity. Corona losses and earth return attenua-
tion have, however, little effect on reducing switching surge overvoltages except on
long transmission lines.
7.2.2.2 Temporary overvoltages
Adefinition adopted by many authors [20] describes a temporary overvoltage (TOV)
as an oscillatory phase-to-ground or phase-to-phase overvoltage of relatively long
duration at a given location which is undamped or weakly damped in contrast to
switching and lightning overvoltages which are usually highly or very highly damped
and of short times. Temporary overvoltages have also been classified into three
groups according to whether the frequency of oscillation is lower, equal to or higher
than the working voltage frequency. In relation to operating power system networks,
CIGRE WG 33.10 and IEEE Task Force on TOV [21] propose to define temporary
overvoltages as an overvoltage higher than the highest system voltage and lasting
for more than 2 cycles.
The most frequent causes of temporary overvoltages are Ferranti effect, load rejec-
tion, ground faults, saturation effects and ferroresonance. Most of these overvoltages
range between 1.2 and 1.5 p.u., but in severe conditions they may reach 2 p.u. [19].
The highest TOV levels are known to occur when long lines are connected to a weak
feeding system. Table 7.1 [21] summarises the most important causes of TOV and
their properties.
Table 7.1 Summary of important TOV causes and characteristics [21]
Temporary Important Overvoltage Typical Methods of control
overvoltage parameters magnitudes durations
phenomena
Common causes
fault fault location 1.01.4 p.u. 210 cycles usually not
application system X
0
/X
1
ratio necessary
fault current
magnitude
load rejection power flow 1.01.6 p.u. seconds switched reactors
system short SVS
circuit MVA
system capacitance generator controls
machine automatic
voltage regulators
line energising line capacitance 1.01.2 p.u. seconds switched reactors
system short SVS
circuit MVA generator controls
Table 7.1 Continued
Temporary Important Overvoltage Typical Methods of control
overvoltage parameters magnitudes durations
phenomena
line dropping/ fault conditions 1.01.5 p.u. <1 second shunt reactors
fault clearing line capacitance relaying
shunt reactors SVS
breaker opening
sequence
reclosing line capacitance 1.01.5 p.u. seconds shunt reactors
shunt reactors relaying
trapped charge SVS
levels
fault conditions
transformer system short 1.01.5 p.u. 02 seconds switched reactors
energising circuit MVA SVS
transformer saturation harmonic filters
characteristics breaker
frequency response closing res.
characteristics
system voltage level
Special cases
parallel line coupling capacitance 1.02.0 p.u. steady state neutral reactors
resonance between circuits switched reactors
shunt reactor values
and saturation
line corona losses
uneven breaker circuit capacitance 1.02.0 p.u. steady state neutral reactors
poles shunt reactor switched reactors
values and saturation
line corona
losses
ferroresonance circuit capacitance 1.01.5 p.u. steady state operating
transformer procedures
saturation
transformer
characteristics
backfeeding cable or line 1.02.0 p.u. seconds operating
capacitance procedures
system short shunt motors
circuit MVA
frequency response
characteristics
Temporary overvoltages due to the Ferranti effect and load rejection are limited
by shunt reactors, but because these shunt reactors remain connected to the system
under normal working conditions, a problemof reactive power consumption is raised.
Thyristor-controlled reactive compensation and shunt reactors with flat magnetising
characteristics have also been utilised [20].
7.2.2.3 Steady state overvoltages
Steady state overvoltages are of supply frequency and are sustained for long time
periods. They are often related to earthing and neutral arrangements within the
system: examples are the neutral displacement in badly designed star-connected
voltage transformers, arcing ground phenomena in systems with insulated neutrals
and the resonance phenomena appearing as a result of the open-circuiting of one or
two phases in a three-phase system which can be initiated by faulty circuit breakers
or broken conductors. Most systems are designed so that such overvoltages do not
occur.
7.3 Network simulation and analysis
Since switching overvoltages are a major factor in determining insulation levels for
transmission networks and as the most severe magnitudes result from line reener-
gisation, the simulation of the network under these conditions is a prerequisite to
overvoltage computations.
The selection of models for the network components is based on:
the frequency range to be simulated; this depends on the type of overvoltage
being computed, with lightning overvoltages requiring a frequency <100 kHz,
switching overvoltages <10 kHz and temporary overvoltages <1 kHz
the extent of the network about the source of the overvoltage requiring a detailed
simulation; this again depends on the type of overvoltage and varies from
a minimum for lightning to a maximum for temporary overvoltages
a compromise between model accuracy, complexity and the availability of
numerical data for the model parameters.
Transmission networks are usually simulated by circuit models for overvoltage
studies and the computations carried out in the time domain.
7.3.1 Transmission lines
Models are based on the transient analysis of the line, a subject extensively covered
in the literature [16, 2224], and three models the phase model, the 0, , model
and the modal model will be summarised.
7.3.1.1 Line parameters
A multiphase line with one conductor/phase and parallel to an earth plane of zero
resistivity is described by the following per unit length parameters [16, 17]:
[L] full matrix of conductorearth self and mutual inductances
[P] full matrix of conductorearth self and mutual potential coefficients
[C] = [P]
1
full matrix of conductorearth self and mutual capacitances
[R] diagonal matrix of conductor resistances
length of line.
The order of the matrices is equal to the number of phases and [L] and [C] are
related by:
[L] = (2)
1
[G] (7.1)
[C] = 2 [G]
1
(7.2)
where [G] is obtained from the phase geometry.
7.3.1.2 Phase model
Under transient conditions the line is usually considered as a lossless line with losses
represented by resistance networks at each end as shown in Figure 7.4. For a more
accurate representation, or to obtain voltages at points along the line, the line may be
subdivided and each section represented as in Figure 7.4.
The phase voltages [V(x, t )] and phase currents [I(x, t )] at a point distance x
along the line s
are related by the lossless line equations [23]:

d
2
dx
2
[V(x, t )] = [L][C]
d
2
dt
2
[V(x, t )] (7.3)
d
2
dx
2
[I(x, t )] = [C][L]
d
2
dt
2
[I(x, t )] (7.4)
and from Equations (7.1) and (7.2):
[L][C] = [C][L] = [U] (7.5)
where [U] is the unit matrix. The phase variables [V(x, t )] in Equation (7.3) and
[I(x, t )] in Equation (7.4) are therefore decoupled and a solution in terms of travelling
s
0.5 l [R] [L], [C], l 0.5l [R]
[I(t)]
s
[I(t)]
s
[I(t)]
r
[I(t)]
r
[V(t)]
r
[V(t)]
r
[V(t)]
s
[V(t)]
s
resistance network lossless line resistance network
Figure 7.4 Diagram of circuit model for multiphase line
waves of voltage and current readily obtained as [23]:
[I(x, t )] = [I(x vt )]
f
+[I(x +vt )]
b
(7.6)
[V(x, t )] = [V(x vt )]
f
+[V(x +vt )]
b
= [Z]{[I(x vt )]
f
[I(x +vt )]
b
}
(7.7)
where suffixes f and b refer to waves travelling in directions s
to r
and r
to s
,
respectively.
The voltage and current waves are related through the phase surge impedance
matrix:
[Z] = [L]
1/2
[C]
1/2
=
1/2
1/2
[G] (7.8)
and all waves have the same velocity:
[v] = [L]
1/2
[C]
1/2
=
1/2
1/2
[U] (7.9)
From the travelling wave solutions for the lossless line the voltages and currents
at the terminals s
and r
are obtained in a form suitable for computation as [25]:

[V(t )]
s
= [V(t T )]
r
+[Z]{[I(t )]
s
+[I(t T )]
r
} (7.10)
[V(t )]
r
= [V(t T )]
s
+[Z]{[I(t )]
r
+[I(t T )]
s
} (7.11)
where, T = ( )
1/2
l is the travel time for the line.
These equations define a model for the lossless line and when combined with the
conductor resistance can be represented from the line terminals by the circuit shown
in Figure 7.5. The line is characterised by the surge impedance [Z], which includes
coupling terms between phases, the travel time T and the conductor resistance [R].
7.3.1.3 Earth resistivity
When the earth resistivity = 0, earth currents penetrate into the earth and their
distribution is not readily determined. Numerous studies into earth current have been
reported [2630], and most involve field analysis in the frequency domain assuming
a homogeneous earth. Some studies have considered a multilayer earth with different
[I(t)]
s
[I(t)]
s [I(t)]
r
[V(t)]
r
[V(t)]
r
[I(t)]
r
[Z] [Z]
r s
0.5l R 0.5l R
[V(t)]
s
[V(t)]
s
[V(t T)]
r
[V(tT)]
s
[I(t T)]
r
[I(t T)]
s
Figure 7.5 Diagram of circuit model for line at s and r (sending and receiving end)
layer resistivities. The overall effect of the current distribution in the earth is to modify
both [L] and [R] to equivalent values [L]
e
and [R]
e
where:
[L]
e
= [L] +[L()] (7.12)
[R]
e
= [R] +[R()] (7.13)
Power series are widely used to compute [R()] and [()] [33], with the
correction terms evaluated at some representative frequency. This frequency can
be approximated by f = l/4T where T is the travel time for the lossless line.
Equations (7.3) and (7.4) now become:
d
2
dx
2
[V(x, t )] = [L]
e
[C]
d
2
dt
2
[V(x, t )] (7.14)
d
2
dx
2
[I(x, t )] = [C][L]
e
d
2
dt
2
[I(x, t )] (7.15)
[L]
e
[C] and [C][L]
e
are non-diagonal matrices and the phase variables in both
Equations (7.12) and (7.13) are not decoupled. A simple travelling wave solution
is not available and a transformation of variables is necessary to obtain decoupled
equations.
7.3.1.4 0, , transformation
The 0, 1, 2 or 0, , transformations used in fault studies can be used to decouple the
line equations [3133]. The latter is preferred because all the elements in the transfor-
mation have real rather than complex values. In addition to analytical advantages, the
transformed variables provide a simple insight into the network response. However,
to be effective, the arrangement of phase conductors and earth must be assumed to
be symmetrical. This requires an approximation, replacing all mutual terms and all
diagonal terms by their average values to give, for a three-phase line:
[L]
e
=
_
_
L
S
L
M
L
M
L
M
L
S
L
M
L
M
L
M
L
S
_
_
(7.16)
[C]
e
=
_
_
C
S
C
M
C
M
C
M
C
S
C
M
C
M
C
M
C
S
_
_
(7.17)
The values of the elements of the transformation matrix relating the 0, ,
components to the phase variables depend on the convention adopted to define
the components and any normalisation procedure used. With the transformation
defined by:
_
_
V
0
V
_
_
=
_
_
_
_
_
_
_
_
1
3
1
3
1
3
1
2
1
2
0
1
6
1
6
2
6
_
_
_
_
V
a
V
b
V
c
_
_
(7.18)
Equation (7.14) then transforms to:
d
2
dx
2
_
_
V
0
(x, t )
V
(x, t )
V
(x, t )
_
_
=
_
_
L
0
0 0
0 L
0
0 0 L
_
_
_
_
C
0
0 0
0 C
0
0 0 C
_
_
d
2
dt
2
_
_
V
0
(x, t )
V
(x, t )
V
(x, t )
_
_
(7.19)
where the 0, , parameters are given by:
L
0
= L
s
+2L
m
L
= L = L
s
L
m
C
0
= C
s
2C
m
C
= C = C
s
+C
m
Using the same transformation for currents transforms Equation (7.15) to an identical
form to Equation (7.19).
The 0, , variables are decoupled and the travelling wave solution gives surge
impedances:
Z
0
=
_
L
0
C
0
_
1/2
Z
= Z
=
_
L
_
1/2
And travel times:
T
0
= (L
0
C
0
)
1/2
= T
= (L
)
1/2
The ideal line can now be represented by uncoupled 0, , networks, each network
has terminal equations of the same form as Equations (7.8) and (7.9).
In a similar way the resistance matrix [R]
e
now includes off diagonal elements
due to the equivalent resistance of the earth return paths, and will also transform to
a diagonal matrix with:
R
0
= R
s
+2R
m
and
R
= R
= R
s
R
m
Although there is no coupling between the 0, , networks, the terminal condi-
tions can result in an interconnection of the networks. As in symmetrical component
analysis of unbalanced faults where the interconnected sequence networks represent
the constraints between voltages and currents at the fault points, in a similar way,
under sequential switching conditions, the 0, , networks are interconnected to
satisfy the constraints at the line ends [34]. Thus, if the line is energised from one
end by closing phase a with phases b and c open, the constraints between the phase
variables at this end are:
V
a
= E
a
, the source voltage
I
b
= I
c
= 0
And between the 0, , variables
E
a
=
V
0
3
+
V
2
+
V
3I
0
=
2I
6I
These constraints are satisfied by an interconnection of the 0, , networks at the

energised end through coupling transformers as shown in Figure 7.6. In a similar
way, the interconnected network for other sequential switching combinations can be
obtained.
I
s
0
(t)
V
s
0
(t)
V
r
0
(t)
E
a
Z
0
T
0
1: 3
0.5l R
0
0.5l R
0
I
s
:
(t)
V
s
:
(t)
V
r
:
(t)
Z
:
T
:
Z
[
T
[
1: 2
0.5l R
:
0.5l R
:
I
s
[
(t)
V
s
[
(t)
V
r
[
(t)
1: 6
0.5l R
[
0.5l R
[
Figure 7.6 Interconnection of 0, , network to simulate single pole closing
7.3.1.5 Modal transformation
When the assumption of the symmetry used in Equations (7.14) and (7.15) is not
acceptable and a more accurate model is required then modal transformations can
be used to transform the variables in Equations (7.3) and (7.4) from the phase to
the modal domain and diagonalise the equations [33]. Different transformations are
required for voltage and current and are given by:
[V
(x, t )]
p
= [T ]
v
[V
(x, t )]
m
(7.20)
[I
(x, t )]
p
= [T ]
i
[I
(x, t )]
m
(7.21)
where [T ]
v
is the eigenvector matrix of [L]
e
[C] and [T ]
i
the eigenvector matrix of
[C][L]
e
with suffixes p and m denoting phase and modal variables, respectively.
The lossless line equations then become:
d
2
dx
2
[V
(x, t )]
m
= []
d
2
dt
2
[V
(x, t )]
m
(7.22)
d
2
dx
2
[I
(x, t )]
m
= []
d
2
dt
2
[I
(x, t )]
m
(7.23)
where [] is the diagonal eigenvalue matrix of [L]
e
[C] and of [C][L]
e
.
These equations are of the same formas Equation (7.19) and have similar solutions
giving:
modal travel time [T ]
m
= []
1/2
modal surge impedance [Z]
m
= []
1/2
Again, there is no coupling between modes and Figure 7.7 shows a representation
of the model. The eigenvalues and hence the modal velocities are uniquely defined.
However, the eigenvectors can be multiplied by any selected non-zero constant with-
out altering their properties and the modal surge impedances are dependent on the
particular eigenvector matrix selected.
The lossy networks cascaded at each end of the lossless line may be transformed to
modal variables as in Figure 7.7, or left outside the transformation as phase variables.
7.3.1.6 Corona losses
Under transient conditions, corona discharges will occur from the conductor surface
when the voltage exceeds the critical corona inception voltage E
c
and will establish a
space charge around the conductor. As the voltage increases the corona will propagate
modal
transformation
resistance
network
lossless
line
resistance
network
modal
transformation
[T]
v
[R]
[Z]
m
[T]
m
[R]
m
[T]
i
[T]
v
[T]
i
s
r
[V
s
(t)]
p
[V
r
(t)]
p
[V
s
(t)]
m
[V
s
(t)]
m
[V
r
(t)]
m
[V
r
(t)]
m
Figure 7.7 Modal simulation of line
conductor
C
c
E
c
Figure 7.8 Circuit model for corona on line
E
c
= corona inception voltage
C
c
= corona capacitance
further into the space around the conductor until the conductor voltage reaches a max-
imum and at this point the corona will be suppressed by the space charge. The corona
loss represents an energy loss from the line resulting in distortion of the surge front
and, for fast-front, lowenergy surges, significant surge attenuation. This phenomenon
may be simulated by the circuit model shown in Figure 7.8 [36]. The corona capaci-
tance C
c
can be obtained from impulse voltage tests on the conductors or estimated
from the corona inception field and conductor geometry. In using this model, the line
must be divided into short cascaded sections and the corona model included at the
terminals s
and r
of each section.
7.3.1.7 Bundled conductors
When each phase consists of several conductors in parallel, [L], [C] and [R] may
be formulated by considering each conductor rather than each phase and then matrix
reduction methods used to obtain equivalent phase matrices [33]. In a similar way,
the earth conductor and earth return can be combined. These reductions are frequency
dependent and usually evaluated at a representative frequency.
7.3.1.8 Lightning and temporary overvoltages
Although the circuit models are described for the computation of switching overvolt-
ages, they can also be applied to lightning overvoltages except that earth correction
parameters are evaluated at a higher frequency. For strikes to earth wires, line spans
close to the strike are each represented by a circuit model, and the towers simulated
by a surge impedance [35].
For temporary overvoltages with a much lower maximum frequency, pi circuit
representations are adequate [23], but a more extensive section of network should be
simulated.
7.3.1.9 Trapped charge
When an energised line is isolated there will be a trapped charge left on the line
and this will discharge rapidly through any transformer windings still connected
to the line, or more slowly through leakage over insulation, and it is necessary to
simulate this charge especially for high speed reclosure. Any trapped charge present
on reclosure can have a significant effect on switching voltages and is simulated as
an initial voltage on the line. This voltage can be computed from the previous line
opening sequence or, assuming each pole of the circuit breaker clears at a current
zero and interphase capacitance is neglected, then the most severe conditions will be
V
a
(0) = V
b
(0) = V
ph
and V
c
(0) = V
ph
, where V
ph
is the crest value of the phase
voltage.
7.3.1.10 Summary of line models
The three models for transient analysis of lines are, in order of increasing complexity:
Phase model: = 0, characterised by full surge impedance matrix [Z] and
a single travel time T .
0, , model: = 0, symmetrical line characterised by surge impedances Z
0
,
Z
= Z
and travel times T

0
and T
= T
. Transformation 0, , a, b, c
variables defined.
Modal model: = 0, asymmetrical line characterised by diagonal surge
impedance matrix [Z]
m
and diagonal modal travel time matrix [T ]
m
. Transfor-
mation modal abc computed.
Accuracy of transmission line modelling is limited by uncertainties in parameter
values and for general overvoltage studies the 0, , model is adequate.
7.3.2 Cables
Cables may be simulated by similar circuit models to lines with all conductors, core,
sheath and earth represented and with cross bonding and earthing taken into account.
7.3.3 Circuit breakers
Circuit breakers are usually simulated by ideal switches with each pole closing at some
specified nominal time. In circuit breakers without point-on-voltage wave control the
closure can occur with equal probability at any point on the voltage wave, although
with pre-arcing it would be biased towards points of maximum voltage. In addition,
circuit breaker tolerances will result in a statistical spread of individual pole closing
times about the nominal time. To simulate these effects a large number of switching
operations are computed with the individual pole closing times varied in a random or
systematic manner [33]. Although more accurate models including pre-arcing can be
formulated they are not normally necessary for network overvoltage computations.
primary
terminals
secondary
terminals
[C]
p
[C]
ps
[C]
s
Figure 7.9 Circuit model for transformer capacitance
7.3.4 Transformers
For overvoltage computations, the simplest simulation is an equivalent capacitance
network representing the interwinding and self-winding capacitance as shown in
Figure 7.9.
When it is relevant to simulate the magnetic couplings between windings the
model must account for both the electrical constraints due to the interconnection of
the windings and the magnetic constraints imposed by the core construction. Amodel
based on the duality between electric and magnetic circuits meets these requirements.
The method is illustrated in Figure 7.10, which shows a three-limb core with one
winding on each limb, the windings being connected in star [37]. The isolating trans-
formers are necessary to ensure that both the electrical and the magnetic constraints
are satisfied simultaneously. The limb and the yoke inductance are defined by flux-
current curves based on the magnetic characteristics of the core material and the core
dimensions. The characteristic can include recoil paths within hysteresis loop and are
readily simulated in piecewise linear form. The model simulates the varying levels
of saturation occurring in different sections of the core, especially during transient
conditions. It is readily extended to represent several windings on each limb and to
include core losses.
7.3.5 Network reduction
Sections of the network can be replaced by a Thevenin equivalent circuit. Ideally, the
frequency response of this circuit should correspond with that of the actual circuit up
to an appropriate frequency. Generally, there are insufficient network data available
for this and a simple alternative is to use a circuit such as shown in Figure 7.11, where
L is the inductance obtained from the fault level at the section terminals and Z the
surge impedance looking into the section from these terminals.
7.4 Computed switching overvoltages
The switching overvoltages of greatest relevance are those associated with energising
a line terminated in an open circuit, and variables having a significant effect on these
a b c n
limb a limb b limb c
a b c n
l : l l : l l : l
l
a
l
b
l
c
L
a
L
ab
L
bc
L
b
L
c
a
b
Figure 7.10 Circuit model of three-limb, transformer core with star connected
winding
a physical arrangement of core and winding
b circuit model
L
a
, L
b
, L
c
= limb inductances
L
ab
, L
bc
= yoke inductances
l
a
, l
b
, l
c
= leakage inductance
overvoltages are:
trapped charge on each phase at the instant of energisation
point on voltage wave at which each phase is energised
earth resistivity
line length.
These effects will now be reviewed.
For the simple case of a lossless line with a travel time of T = 1 ms, energised
from a source voltage V
s
= cos(t +) p.u. with the circuit breaker closed at t = 0,
Figure 7.12 shows the variation of the maximum overvoltage V
r
at the open circuit
termination with the switching angle . Curve a is for zero trapped charge with V
r
[Z]
[L]
n
n
[E]
section of
network
Figure 7.11 Equivalent circuit representing section of network
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
switch closing time, ms
o
v
e
r
v
o
l
t
a
g
e
,

p
.
u
.
b
a
Figure 7.12 Variation of switching overvoltage with instant of switch closing t = 0,
when source voltage on phase a is maximum
Curve a zero trapped charge
Curve b trapped charge = 1 p.u. voltage
having a maximum value of 2.0 per unit for = 0
and a minimum value of 1.2 p.u.

at = 90
. Curve b shows the effect of a trapped charge represented by an initial

voltage of 1.0 p.u. giving an increase in the maximum voltage to 3 p.u. for = 0
with the minimum value being 1.0 p.u. at = 180
.
The relative effects of trapped charge, pole closing times and earth resistivity are
shown in Figure 7.13. They are based on a 150 km single circuit line and computed
using the 0, , model and EMTP [33]. The line has four conductors/phase and
a single earth wire, and the 0, , parameters are given in Table 7.2 with earth
resistivity having greatest effect on the 0 parameters. The operating conditions are
shown in Table 7.3.
Figure 7.13(i) shows the voltage waveshapes at the line termination for the
ideal case of = 0, simultaneous pole closing at the instant when E
a0
= 1 p.u.,
4
3
2
1
0
1
2
3
4
20 25 30 35 40
t, ms
V
r
a
p
.
u
.
4
3
2
1
0
1
2
3
4
20 25 30 35 40
t, ms
V
r
b
p
.
u
.
1
0
1
2
3
4
20
(i)
25 30 35 40
t, ms
V
r
c
p
.
u
.
4
3
2
Figure 7.13 Phase voltages at line termination
i simultaneous pole closing, zero trapped charge, = 0
E
b0
= E
c0
= 0.5 p.u. and with zero trapped charge. The phase voltages V
a
(t ),
V
b
(t ) and V
c
(t ) are linear combinations of the 0, , travelling waves appearing at
the termination and with simultaneous pole closing on to a symmetrical source, no
zero components are present. The voltages have well defined waveshapes and show
clearly the travelling wave components arriving simultaneously. The maximumvalue
of the overvoltage V
ra
= 2 p.u. on phase a.
Figure 7.13(ii) shows the effect of sequential switching. The closing times t
a
, t
b
and t
c
are selected within a pole spread of 5 ms and switch at voltages of E
a0
= 1 p.u.,
E
b0
= 0.87 p.u. and E
c0
= 1 p.u. The first pole to close, a results in V
ra
(t ) of
the same waveshapes and magnitudes as case (i), and produces coupled voltages on
phases, b and c. When c and then b close, these coupled voltages present differ-
ent initial conditions from those seen by phase a resulting in modified waveshapes
following closure and with V
rc
(t ) now reaching a maximum value of 2.3 p.u.
4
3
2
1
0
1
2
3
4
20 25 30 35 40
t, ms
4
3
2
1
0
1
2
3
4
20 25 30 35 40
t, ms
4
3
2
1
0
1
2
3
4
20
(ii)
25 30 35 40
t, ms
V
r
a
p
.
u
.
V
r
b
p
.
u
.
V
r
c
p
.
u
.
ii sequential pole closing, zero trapped charge, = 0
Figure 7.13(iii) shows the effects of = 0 resulting in longer travel times for the
zero waves. This results in further distortion of the waveshapes and greater attenuation
because of the increased value of R
0
. The maximum overvoltage is now 2.35 p.u. on
phase b.
Figure 7.13(iv) shows the effect of trapped charge represented by V
a0
= V
b0
=
1 p.u. and V
c0
= 1 p.u. The wave magnitudes are increased with a maximum of
3.78 p.u. appearing on phase b.
Finally, Figure 7.13(v) is as for case (iv) but includes the effect of a surge arrester
connected between each phase and earth at the line termination. The arresters have
a protective level of 2.38 p.u., cause further distortion of the wave shapes and limit
the maximum voltage to 2.07 p.u.
4
3
2
1
0
1
2
3
4
20 25 30 35 40
t, ms
4
3
2
1
0
1
2
3
4
20
(iii)
25 30 35 40
t, ms
4
3
2
1
0
1
2
3
4
20 25 30 35 40
t, ms
V
r
a
p
.
u
.
V
r
b
p
.
u
.
V
r
c
p
.
u
.
iii sequential pole closing, zero trapped charge, = 100 m
Sequential pole closing, earth resistivity and trapped charge can all result in
high values of the switching overvoltage, however it is not possible to specify the
combination that will result in the most severe overvoltages, hence it is necessary to
consider a statistical analysis of the overvoltage distribution.
As an example the effect of random pole closing on the overvoltages was com-
puted using a statistical switching model with random times computed for each pole
and based on a normal distribution about a nominal closing time of 20 ms and a stan-
dard deviation of 0.83 ms. Three operating conditions to show the effect of trapped
charge and of surge arresters were considered using the data in Table 7.4.
Figure 7.14 shows the cumulative distribution of the overvoltages F(U) together
with the density distribution f (U). These distributions are based on 1000 switching
4
3
2
1
0
1
2
3
4
20 25 30 35 40
t, ms
4
3
2
1
0
1
2
3
4
20
(iv)
25 30 35 40
t, ms
4
3
2
1
0
1
2
3
4
20 25 30 35 40
t, ms
V
r
a
p
.
u
.
V
r
b
p
.
u
.
V
r
c
p
.
u
.
iv sequential pole closing, trapped charge 1.0,1.0, 1.0 p.u.,
= 100 m
operations taking the highest overvoltage from each switching operation. The mean
values of the overvoltages together with the standard deviations are given in Table 7.5.
The distribution with zero trapped charge is shown in Figure 7.14 case (vi) with
overvoltages ranging from 1.75 p.u. to 2.95 p.u. The effect of trapped charge is to
move the distribution to a higher value ranging from 2.75 p.u. to 4.90 p.u. but with
a similar distribution as shown in Figure 7.14 case (vii). When surge arresters are
included at the line termination the overvoltages are limited to the narrowdistribution
of 2.05 p.u. to 2.25 p.u. seen in Figure 7.14 case (viii).
7.5 Insulation coordination
Although coordination between switching overvoltages and the strength of self-
restoring insulation can be based on the maximum value of overvoltage and the
4.5
1.5
0
1.5
3.0
4.5
20 25 30
t, ms
35 40
3.0
4.5
1.5
0
1.5
3.0
4.5
20 25 30
t, ms
35 40
3.0
4.5
1.5
0
1.5
3.0
4.5
20 25 30
t, ms (v)
35 40
3.0
V
r
a
p
.
u
.
V
r
b
p
.
u
.
V
r
c
p
.
u
.
v sequential pole closing, trapped charge 1.0, 1.0, 1.0 p.u., surge
arresters included, = 100 m
minimum value of breakdown voltage, this takes no account that the probability of
both occurring simultaneously is very low and it is more realistic to use statistical
data for coordination.
IEC 60071: Insulation coordination [1, 2] specifies procedures for the selection
of withstand voltages for phase-to-earth, phase-to-phase and longitudinal insula-
tion. A full statistical analysis would consider all variables that may influence the
overvoltage distribution, including various configurations of the network. Such an
analysis is bothtime consuminganddifficult tooptimise andmost studies consider one
or two configurations. The procedures include correction factors based on operational
experience on transmission networks.
Statistical data are obtained from computations of overvoltages and labora-
tory/field tests on insulation characteristics. It is generally obtained as: F(U)
Table 7.2 0, , line parameters
(m) Z
0
() T
0
(ms) R
0
(/km) Z
() T
(ms) R
(/km)
0 442 0.50 0.065 256 0.50 0.043
100 553 0.63 1.6000 259 0.51 0.048
Table 7.3 Operating conditions
Case (m) t
a
(ms) t
b
(ms) t
c
(ms) V
a0
(p.u.) V
b0
(p.u.) V
c0
(p.u.)
(i) 0 20.0 20.0 20.0 0 0 0
(ii) 0 20.0 25.0 23.3 0 0 0
(iii) 100 20.0 25.0 23.3 0 0 0
(iv) 100 20.0 25.0 23.3 1 1 1
(v)
100 20.0 25.0 23.3 1 1 1
includes surge arrester at line termination

Table 7.4 Operating conditions
Case (m) V
a0
(p.u.) V
b0
(p.u.) V
c0
(p.u.)
(vi) 100 0 0 0
(vii) 100 1 1 1
(viii)
100 1 1 1
includes surge arresters at line termination

the cumulative distribution of switching overvoltages, where F(U = U
e
) is the
probability that the maximum switching overvoltage U
e
U, f (U) is the density
distribution where f (U) = dF(U)/dU and f (U = U
e
)dU is the probability that
U
e
U U
e
+ dU; P(U) the cumulative distribution of breakdown voltage
where P(U = U
b
) is the probability of breakdown with a crest voltage U = U
b
.
The probability of breakdown occurring at voltage U is f (U)P(U)dU and the
risk of failure R for all values of U is given by [38, 39]:
R =
_
0
f (U)P(U)dU (7.24)
Because there is an upper limit to the switching voltage U
et
and a lower limit to the
breakdown voltage U
b0
the integration limits 0 can be replaced by U
b0
and
U
et
, where F(U = U
et
) = 0 and P(U = U
b0
) = 0.
0
0.02
0.04
0.06
0.08
0.1
1
.
7
3
1
.
8
8
2
.
0
3
2
.
1
8
2
.
3
3
2
.
4
8
2
.
6
3
2
.
7
8
2
.
9
3
3
.
0
8
3
.
2
3
3
.
3
8
3
.
5
3
3
.
6
8
3
.
8
3
3
.
9
8
4
.
1
3
4
.
2
8
4
.
4
3
4
.
5
8
4
.
7
3
4
.
8
8
U, p.u.
0
0.2
0.4
0.6
0.8
1
f (U) F(U)
f (U) F(U)
f (U) F(U)
case (vi)
U, p.u.
case (vii)
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
1
.
7
2
5
1
.
8
7
5
2
.
0
2
5
2
.
1
7
5
2
.
3
2
5
2
.
4
7
5
2
.
6
2
5
2
.
7
7
5
2
.
9
2
5
3
.
0
7
5
3
.
2
2
5
3
.
3
7
5
3
.
5
2
5
3
.
6
7
5
3
.
8
2
5
3
.
9
7
5
4
.
1
2
5
4
.
2
7
5
4
.
4
2
5
4
.
5
7
5
4
.
7
2
5
4
.
8
7
5
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
U, p.u.
case (viii)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1
.
7
2
5
1
.
9
2
5
2
.
1
2
5
2
.
3
2
5
2
.
5
2
5
2
.
7
2
5
2
.
9
2
5
3
.
1
2
5
3
.
3
2
5
3
.
5
2
5
3
.
7
2
5
3
.
9
2
5
4
.
1
2
5
4
.
3
2
5
4
.
5
2
5
4
.
7
2
5
4
.
9
2
5
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Figure 7.14 Density and cumulative density distributions of overvoltages
case (vi): zero trapped charge, no arresters
case (vii): 1.0 p.u. trapped charge, no arresters
case (viii): 1.0 p.u. trapped charge, arresters at line end
Insulation coordination is then achieved by optimising the overvoltage data
through protection and control, and the breakdown voltage data using different
insulation levels in order to obtain an acceptable risk of failure. An acceptable risk
value is based on satisfactory operational experience on similar networks and is
generally within a range of 10
1
to 10
4
per switching operation [2].
7.5.1 Analytical expressions for F(U) and P(U)
Although the risk of failure is best estimated fromF(U) and P(U), often insufficient
data is available and analytical expressions based on limited data are widely used. The
expressions should give a good fit over the region U
b0
to U
et
and IEC71 recommends
Table 7.5 Overvoltage data
Case U
e50
(p.u.)
e
(p.u.)
(vi) 2.21 0.231
(vii) 3.64 0.449
(viii) 2.16 0.035
the use of Weibull probability functions of the form:
F(U) = 1 e
((U
et
U)/)
(7.25)
and
P(U) = 1 e
((UU
b0
)/)
(7.26)
where are scale parameters and shape parameters. These expressions are readily
evaluated and differentiated.
F(U) can be readily characterised using the U
e50
switching overvoltage, where
F(U = U
e50
) = 0.5 and
e
the standard deviation of F(U). Then, with the
assumptions [2] that:
U
et
= U
e50
+3
e
(7.27)
and
U
e2
= U
e50
+2.05
e
(7.28)
where F(U = U
e2
) = 0.02 a 2 per cent probability that U
e2
U. Equation (7.25)
becomes:
F(U) = 1 e
((3
e
+U
e50
U)/)
(7.29)
giving:
F(U = U
e50
) = 0.5 = 1 e
(3
e
/)
(7.30)
F(U = U
e2
) = 0.02 = 1 e
(0.95
e
/)
(7.31)
and from Equations (7.30) and (7.31) we obtain: = 3.07 and = 3.4
e
.
Equation (7.29) now becomes:
F(U) = 1 e
((3
e
+U
e50
U)/3.41
e
)
3.07
F(U) = 1 0.509
(1+(U
e50
U)/3
e
)
3.07
(7.32)
and with acceptable accuracy can be represented by:
F(U) = 1 0.5
(1+(U
e50
U)/3
e
)
3
(7.33)
and
f (U) =
ln(0.5)
e
_
1 +
U
e50
U
3
e
_
2
0.5
(1+(U
e50
U)/3
e
)
3
(7.34)
Similarly P(U) is characterised using U
b50
and
b
, and with the assumptions that:
U
b0
= U
b50
4
b
U
b16
= U
b50
b
where P(U = U
b16
) = 0.16; Equation (7.26) becomes:
P(U) = 1 0.5
(1+(UU
b50
/4
b
))
5
(7.35)
These expressions can then be used to estimate the risk of failure of the
insulation.
7.5.2 Risk of failure
Using the data in Table 7.5 and Equations (7.33) and (7.34) the distributions F(U)
and f (U) for cases (vi), (vii) and (viii) are obtained and are shown in Figure 7.15, as
a, b and c. For each case a distribution P(U) is obtained from Equation (7.35), using
U
b50
= 2.8 p.u. and
b
= 0.168 and is also shown in Figure 7.15. The risk of failure
is then evaluated as:
case (vi), no trapped charge, R = 0.02
case (vii), trapped charge, R = 0.94
case (viii), trapped charge, surge arrester, R = 5 10
06
The risk of failure for case (vi) is rather high, whereas for case (vii) the risk is
unacceptably high and for case (viii) the insulation level too high. For each case the
distribution P(U) can be optimised to give an acceptable risk value.
7.5.3 Simplified method
A simplified method is available [2] which assumes that F(U) and P(U) are each
defined by one value:
U
e2
the statistical overvoltage where F(U = U
e2
) = 0.02, i.e. a 2 per cent
probability of being exceeded
U
10
the statistical withstand voltage where P(U = U
b10
) = 0.10, i.e. a 10 per cent
probability of breakdown and 90 per cent probability of withstand.
Coordination is achieved through the statistical coordination factor K
es
, where:
U
b10
= K
es
U
e2
K
es
is selected to meet a required risk of failure from data for typical transmission
networks. Figure 7.16 shows typical variations of the risk of failure with the statistical
coordination factor K
es
[2]. Although the curve will be modified by the shape of the
1.5
1
0.5
0
1.5
1.0
0.5
0
0 1 2 3 4 5
P(U)
F(U)
P(U)
F(U)
P(U)
F(U)
f (U)
1.5
1.0
0.5
0
f (U)
f (U)
U, p.u.
1.5
1
0.5
0
4
3
2
1
0
4
3
2
1
0
0 1 2 3 4 5
0 1 2 3 4 5
U, p.u.
U, p.u.
a
b
c
Figure 7.15 Weibull representation of f(U), F(U) and P(U)
a case (vi), no trapped charge, no arrester, R = 0.02
b case (vii), trapped charge, no arrester, R = 0.94
c case (viii), trapped charge, arrester, R = 5.4E-06
overvoltage K
cs
has a low sensitivity to changing R. Table 7.6 shows the statistical
withstand voltage obtained for R = 0.005.
7.5.4 Withstand voltage
Having optimised F(U) and P(U) to obtain an acceptable risk of failure, then U
b10
,
the statistical breakdown voltage, is available. P(U) is based on laboratory/field tests
where the conditions may differ from those on the actual network. Factors such as
manufacturing tolerances, environmental conditions and ageing can modify P(U).
These factors are accounted for through a safety factor K
s
to give the withstand
voltage U
rw
= K
s
U
b10
, where typically K
s
= 1.05 [2].
Afurther correction for altitude may be necessary (see Chapter 2).
1
5
3
2
10
1
5
3
2
10
2
5
3
2
10
3
5
3
2
10
4
0.4 0.6 0.8 1.0 1.2 1.4
U
e2
= 3.5
= 2.5
= 1.5
K
cs
R
Figure 7.16 Variation of risk of failure R with statistical coordination factor
K
cs
[2]
1
Table 7.6 Statistical withstand
voltages
Case U
e2
(p.u.) K
es
U
b10
(p.u.)
(vi) 2.67 1.04 2.78
(vii) 4.54 1.04 4.72
(viii) 2.23 1.04 2.32
7.6 Compact transmission lines
One area where both insulation coordination and environmental factors have had a
combined impact is that of compact lines. In the normal development of transmission
networks load growth and relocation of load centres will at some time necessi-
tate a reinforcement of the network involving additional lines/substations at present
voltage levels or at a higher voltage level. Such developments are often opposed
because of the high level of public concern about their effect on the environment,
including visual impact, land requirement and feared health risks in electric and mag-
netic fields. This has led to the design of compact lines of reduced dimensions thus
lessening both the visual impact and land requirement, but still providing the reli-
ability of conventional networks, and also allowing their upgrading [40, 41]. The
reductions have been achieved through:
improvements in overvoltage protection and control through using modern surge
arresters
developments in insulation including polymeric insulation
changes in design such as insulating crossarms on towers.
The principal gains obtained using compact lines are:
maximisation of power transfer through available transmission corridor
minimisation of environmental impact.
7.6.1 Insulation
Composite insulators made of polymeric materials such as EPDMand silicone rubber
are used extensively on compact networks. They have a high mechanical strength
under tensile, compressive and bending loads and are lighter than porcelain insulators.
Their performance under polluted conditions is superior due to their hydrophobic-
ity and this is highly important on compact systems where reduced clearances can
increase their sensitivity to pollution (for further reading, refer to Chapter 6).
Insulating crossarms are widely used to reduce tower dimensions and typically
consist of a glass fibre core with silicon rubbers sheaths and sheds. They are usually
mounted in a horizontal Vor line post arrangements as shown in Figure 7.17 [45] and
prevent lateral movement of the conductor at the tower with some designs allowing
limited longitudinal movement to relieve torsional stress. A 400 kV line with insu-
lating crossarms and requiring a transmission corridor width of 19 m compared with
30 m for a conventional line is described in Reference 42.
Insulating interphase spacers are sometimes necessary along compact line spans
to limit conductor movement, and consist of silicone rubber sheaths moulded to an
epoxy reinforced glass fibre rod [43]. The spacers allow further compaction of the
line but increase the risk of phase-to-phase breakdown under polluted conditions.
7.6.2 Surge arresters
Surge arresters provide the protection necessary to limit switching overvoltages at
line terminations to less than 2.0 p.u. Arresters are usually connected phase-to-earth
and are less effective against phase-to-phase overvoltages. Phase-to-phase arresters
may be required on compact lines and can be integral with interphase spacers.
Phase-to-phase arresters have been considered as a feasible alternative to resistance
switching [44].
7.6.3 Comparison between compact and conventional network
Table 7.7 is based on the data taken from Reference 1 and shows standard values of
impulse withstand voltages and minimum air clearances.
rigid connection to pole pivoted connection to pole
line post
horizontal V
rigid connection to pole
(braced line post)
pivoted connection to pole
(articulated braced line post)
Figure 7.17 Typical insulating crossarm [45] (courtesy of paper 400-04 from the
Leningrad CIGRE symposium, 35 June, 1991)
Table 7.7 Standard insulation levels
Nominal voltage Switching impulse Lightning impulse Minimum clearance
(kV r.m.s.) withstand voltage withstand voltage conductor-structure
(kV pk) (kVpk) (m)
275 750 950 1.6
275 850 1050 1.8
400 850 1175 1.8
400 950 1300 2.2
400 1050 1425 2.6
Switching impulse withstand voltage on 400 kV networks are typically at the
upper limits of 1050 kV. With appropriate overvoltage protection this can be reduced
to 850 kV, reducing the minimum clearances from 2.6 to 1.8 m. Compact networks
take advantage of these reductions to provide a network with a similar risk of failure
factor to that of a conventional design.
A comparison between 275 kV line with a switching impulse level of 850 kV
and a compact 400 kV line with a switching impulse level of 850 kV indicates that
a conventional 275 kV transmission corridor could accommodate a compact 400 kV
line with a doubling in transmission capacity.
a
c c
b
Figure 7.18 Busbar arrangement
Table 7.8 Busbar dimensions
Conventional Compact
substation substation
SIWL (kVpk) 1050 850
Mini ph earth clearance (m) 2.6 1.8
Mini ph ph clearance (m) 3.6 2.6
Insulator height a (m) 4.0 3.2
Support height b (m) 3.0 3.0
Bubsar centres c (m) 4.6 3.6
Comparable gains are feasible with compact air insulated substations as is demon-
strated by the simple busbar arrangements shown in Figure 7.18. Table 7.8 shows the
principal dimensions for both networks allowing the same margins above the mini-
mumclearances. The data indicate a reduction of 11 per cent in height and 21 per cent
in width.
The effect of compaction on the profile of the electric and magnetic fields for the
above geometry is shown in Figure 7.19 and is based on a symmetrical busbar voltage
of 400 kVand symmetrical currents of 2 kA/phase and assumes zero earth resistivity.
Compaction has resulted in maximumvalues for both the electric and magnetic fields
being some ten per cent higher than those for the conventional layout. The magnetic
field is dependent upon the earth resistivity
e
and profiles are shown for
e
= 0 and
e
= .
distance from busbar centre, m
B: conventional line j=0 B: compact line j=0
B: conventional line j= B: compact line j=
E: conventional line E: compact line
distance from busbar centre, m
0
a
b
5 10 15 20
0
2
4
6
8
10
0
10
20
30
40
50
60
70
80
0 5 10 15 20
e
l
e
c
t
r
i
c

f
i
e
l
d
,

k
V
/
m
m
a
g
n
e
t
i
c

f
i
e
l
d
,

u
T
Figure 7.19 Electric and magnetic field profiles under conventional and compact
busbars
a electric field profiles
b magnetic field profiles
These simple comparisons showadvantages for compact networks, however these
will be modified because compact lines are usually designed for a particular situation
and there is limited standardisation, this often introduces cost penalties.
7.7 Acknowledgement
The authors wish to thank Niamat Ullah for his help in obtaining and preparing the
computation results.
7.8 Note
1
Figures from British Standards reproduced with the permission of BSI under
licence number 2003SK/0157. British Standards can be obtained from BSI Customer
Services, 389 Chiswick High Road, London W4 4AL (Tel +44 (0) 20 8996 9001).
7.9 References
1 IEC 60071-1: 1993: Insulation co-ordination part 1: definitions, principles
and rules
2 IEC 60071-1: 1993: Insulation co-ordination part 2: applications guide
3 RAGALLER, K., BURGER, U., and RUOSS, E.: Introduction and survey
in high voltage networks, in RAGALLER, K. (Ed.): Surges in high-voltage
networks (Plenum Press, London, 1980), pp. 124
4 BERGER, K.: Lightning surges, in RAGALLER, K. (Ed.): Surges in high
voltage networks (Plenum Press, London, 1980), pp. 2562
5 GOLDE, R.H.: Lightning, vol. 1: physics of lightning (Academic Press,
London, 1977)
6 GOLDE, R.H.: Lightning, vol. 2: lightning protection (Academic Press,
London, 1977)
7 GARY, C.: La foudre, in EYROLLES, E.D.F. (Ed.): Les properties dielec-
triques de lair et les tres hautes tensions (Paris, 1984), pp. 92229
8 ANDERSON, R.B., ERIKSSON, A.J., KRONINGER, H., MEAL, D.V., and
SMITH, M.A.: Lightning and thunderstorm parameters. Proceedings of the
IEE conference on Lightning and power systems, Conf. Publ. 236, 1984,
pp. 5766
9 ERSKINE, A.: Lightning flash current magnitudes and ground flash density in
the UK. Proceedings of the 24th UPEC, Belfast, 1989, pp. 565568
10 GALLAGHER, T.J., and PEARMAIN, A.J.: High voltage measurement,
testing and design (John Wiley & Sons, 1983)
11 AGUET, M., and IANOVICI, M.: Phenomenes transitoires en haute tension,
in Traite dElectricite, Vol. XXII Haute Tension, E.P.F., GIORGI Lauzanne,
1982, pp. 67109
12 LEVINSON, L.M., and PHILIPP, H.R.: ZnOvaristors for transient protection,
IEEE Trans. on Parts Hybrids Packag., 1977, PHP-13, (4), pp. 338343
13 SMITH, J.: Geomagnetic. Proceedings of the 25th UPEC, Aberdeen, 1990,
pp. 358360
14 LE ROY, G.: Les contraintes internes, in EYROLLES, E.D.F. (Ed.): Les
proprieties dielectriques de lair et les tres hautes tensions (Paris, 1984),
pp. 190
15 ERCHE, M.: Switching surges, in RAGALLER, K. (Ed.): Surges in high-
voltage networks (Plenum Press, London, 1980), pp. 6397
16 BICKFORD, J.P.: Electric power transmission systems (Peter Peregrinus Ltd,
London, 1976)
17 EATON, J.R.: Electric power transmission systems (Prentice-Hall, Inc,
Englewood Cliffs, 1972)
18 CSUROS, L.: Overvoltage protection, inThe Electricity Council (Ed.): Power
system protection 2 systems and models (Peter Peregrinus Ltd, Stevenage,
1981), pp. 276316
19 DIESENDORF, W.: Temporary overvoltages, in High-voltage electric power
systems (Butterworths, London, 1974)
20 GLAVITSCH, H.: Temporary overvoltages, in RAGALLER, K. (Ed.): Surges
in high voltage networks (Plenum Press, London, 1980), pp. 131163
21 CIGRE working Group 33.10 and IEEE Task force on TOV: Temporary
overvoltages causes, effects and evaluation, CIGRE conference 1990 session,
paper 33-210, 1990
22 DOMMEL, H.W., and MEYER, W.S.: Computation of electromagnetic
transients, Proc. IEEE, 1974, 62, pp. 983993
23 GREENWOOD, A.: Electrical transients in power systems (John Wiley &Son,
Chichester, 1991)
24 CHOWDHURI, P.: Electromagnetic transients in power systems (John Wiley
& Son, Chichester, 1996)
25 DOMMEL, H.W.: Digital computer solutions of electromagnetic transients in
single and multiphase networks, IEEE Trans., 1969, PAS-88, pp. 388389
26 CARSON, J.R.: Wave propagation in overhead wires with ground return,
Bell Syst. Tech. J., 1926, 5, pp. 539554
27 NAKAGAWA, M., and IWAMOTO, K.: Earth return impedance for the
multilayer case, IEEE Trans., 1976, PAS-95, pp. 671676
28 DERI, A., TEVAN, C., SEMLYEN, A., and CASTANHEIRA, A.: The complex
ground return plane, a simplified model for homogeneous and multi-layer earth
return, IEEE Trans., 1981, PAS-10, pp. 36863693
29 WEDEPOHL, L.M., and EFTHYMIADIS, A.E.: Wave propagation in trans-
mission lines over lossy ground: a new complete field solution, Proc. IEEE,
1978, 125, pp. 505510
30 MARTI, J.R.: Accurate modelling of frequency dependent transmission lines
in electromagnetic transient simulations, IEEE Trans., 1982, PAS-101,
pp. 147157
31 CLARKE, E.: Circuit analysis of a.c. power systems (John Wiley & Son,
Chichester, vol. 1, 1950)
32 TAVARES, M.C., PISSOLATO, J., and PORTELA C.M.: Mode domain mul-
tiphase transmission line model-ore in transient studies, IEEE Trans., 1999,
P-D 14, pp. 15331540
33 DOMMEL, H.W.: Electromagnetic transients program reference manuel
(EMTP theory book). Bonneville Power Administration, Oregon, 1986
34 BRANDEO FARIA, J.A., and HILDEMARO BRICERO, J.: On the modal
analysis of asymetrical three-phase transmission lines using standard transfor-
mations, IEEE Trans., 1997, PWRD-12, pp. 17601765
35 CHISHOLM, W.A., and CHOU, Y.L.: Travel time of transmission towers,
IEEE Trans., 1985, PAS-104, pp. 29222928
36 AL-TAI, M.A., ELAYAN, H.S.B., GERMAN, D.M., HADDAD, A.,
HARID, N., and WATERS, R.T.: The simulation of corona on transmission
lines, IEEE Trans., 1989, PD-4, pp. 13601369
37 GERMAN, D.M., and DAVIES, A.E.: The simulation of transformer feeders
following switching operations, IEEE Trans., 1982, PAS-100, pp. 45104514
38 JONES, B., and WATERS, R.T.: Air insulations at large spacings, IEE Proc.,
1978, 125, pp. 11521176
39 ELOVAORA, J.: Risk of failure determination of overhead line phase-to-earth
insulation under switching surges, Electra, 1978, (56), pp. 6987
40 ELAHI, H., PANELI, J., STEWART, J.R., and PUENTE, H.R.: Substa-
tion voltage uprating design and experience, IEEE Trans., 1991, PD-6,
pp. 10491057
41 TONG, Y.K., BENHACKE, R.H., CLARK, A.M., SPARROW, L.T.,
WHITE, H.B., and OPASCHASTAT, P.: 500kV compact line designs for the
greater Bangkok area. CIGRE international session, group 22, 1998
42 AMMAN, M., and PAPAILOU K.O.: Long term experience with silicon
composite insulators in the h.v. lines of Energie Overt Suisse. Mid power
conference, Cyprus, 1998
43 KARADY, C.G., SCHNEIDER, H.M., and HALL, J.F.: Utilisation of compact
insulators for compacting transmission lines, CIGRE symposium, Compacting
overhead transmission lines, June 1991, Leningrad, paper 400-05
44 RIBEIRO, J.R., LAMBERT, S.R., and WILSON, D.D.: Protection of compact
transmission lines with metal oxide arresters. CIGRE symposium, Compacting
overhead transmission lines, June 1991, Leningrad, paper 400-06
45 PARIS, L., and PARGAMIN, L.: Application of composite insulators for
overhead transmission lines. CIGRE symposium, Compacting overhead
transmission lines, June 1991, Leningrad, paper 400-04
Chapter 8
Earthing
H. Griffiths and N. Pilling
8.1 Introduction
Three-phase power systems are earthed by connecting one or more selected neutral
points to buried earth electrode systems. Such earths are referred to as system earths.
At electrical installations, all non-live conductive metallic parts are interconnected
and also earthed to protect people against electric shock, and in this role, the earth
is referred to as a protective earth. Under normal conditions, there is only a residual
current or no current at all in the earth path. However, very high magnitudes of
current return to source via the earth path under fault conditions. The earth also
conducts lightning currents and the current path may involve part of a power system
either directly or by induction. The earthing system, or part of it, may therefore also
be specifically designed to act as a lightning protective earth [1].
The earth is a poor conductor and, therefore, when it carries high magnitude
current, a large potential gradient will result and the earthing system will exhibit an
earth potential rise (EPR). Earth potential rise is defined in the recently published
CENELECdocument HD637 S1:1999 [1] as the voltage between an earthing system
and reference earth. In the UK, this quantity has previously been referred to as the
rise of earth potential (ROEP) or the earth electrode potential. In the US, it is referred
to as the ground potential rise (GPR). Here, this quantity will be referred to as the
earth potential rise (EPR).
Soil and rock resistivity may vary considerably from region to region, and it
is rarely constant either vertically or horizontally in the area of interest around an
electrical installation. This variability makes the construction of earth models for
the prediction of earth potential rise a very difficult task. The magnitude of power
frequency earth fault currents can range from a few kA up to 2030 kA, and earth
impedances of high voltage substations may lie in the range from 0.05 to over
1 . Although higher fault current magnitudes are generally associated with lower
magnitude earth impedances, earth potential rises can be as high as several tens
of kV. Consequently, there is a potential risk of electrocution to people in the vicinity
of power systems during earth faults, and damage to equipment may also occur
unless measures are taken to limit the earth potential rise and/or to control potential
differences in critical places. Lightning transients can also generate currents of several
tens of kAin the earthing system, and this requires the power system to be protected
against overvoltages. The discharge of transient current into earth may also present
an electrocution hazard but the tolerable limits are less well defined compared with
power frequency currents.
In the past, earthing systems were designed to achieve earth resistances below
a specified value or on a particular density of buried conductor. Current practice,
however, dictates that such systems are designed to control potential differences
within and around the electrical installation. These potential differences are specifi-
cally referred to as step and touch voltages. For example, in 1992, the UK Electricity
Association introduced a new earthing standard (EA-T.S.41-24) [2], with specific
step and touch safe voltage levels, replacing Engineering Recommendation S.5/1 [3].
In North America, the change to quantify safe voltages occurred earlier [4]. In some
standards, consideration is also given to the maximum earth potential rise of the
earthing system. The extent of the rise of potential on the ground surface around a
substation can be described in terms of a hot zone, and this is used to identify whether
third parties in the vicinity of the installation are affected.
Nowadays, there is great awareness of the safety issues concerning the earthing of
electrical installations. Some recent developments have also highlighted the impor-
tance of this subject and the new challenges that face earthing system designers:
a Restricted land area: replacing or upgrading substations in urban areas restricts
the options available to control earth potential differences.
b Urban encroachment: urban encroachment of transmission lines has raised
concerns about earth potential rise at transmission line tower bases.
c Mobile communication base stations: the rapid growth in mobile telecommunica-
tions has resulted in power transmission line towers hosting GSM (global system
for mobile communication) base stations. Such installations bring the transmis-
sion system into close proximity to the distribution systems supplying the base
stations, and introduce challenges for insulating and protecting against faults on
the high voltage side.
d Modernisation of railway electrification systems: new electrification schemes
involve siting transmission substations in close proximity to the railway net-
work (typically 400 kV/25 kV) where, traditionally, the earthing systems of the
respective systems were segregated.
e Windfarms: the design of earthing systems for windfarms requires special consid-
eration because of the distributed nature of the electrical network and the absence
of a low impedance earth at the power system interface.
This chapter aims to outline the current state of the subject of earthing covering both
power frequency and transient aspects. To begin with, the main earthing system com-
ponents at transmission and distribution levels are described and different methods
Earthing 351
of earthing the neutral are considered. The next section is devoted to earth resistivity,
and conduction mechanisms are described together with site investigation techniques
for determining earth structure and resistivity. In reviewing power frequency per-
formance, the procedure for designing safe earthing installations is explained based
on information contained in the latest standards and published literature. This pro-
cedure has a number of stages which include calculations to determine (i) the earth
impedance of different earthing system components, (ii) earth fault current magni-
tude and the proportion of this current that will flow through the earthing system
and (iii) step and touch voltages and their maximum acceptable values. Each of these
stages can be carried out with different degrees of complexity and detail depending on
the level of accuracy required. Accordingly, earthing system design may be carried
out entirely using analytical formulae using a homogeneous earth model or may
require the use of specialised earthing software and complex earth models. Asection
is devoted to the description of the measurement of earthing system impedance and
potentials which are important for the validation of the performance of installed earth-
ing systems. To conclude the treatment of power frequency aspects, a new approach
to earthing system design is introduced which uses risk management techniques to
identify priorities for investment to improve safety.
The final section of this chapter deals with the transient behaviour of earthing
systems. A summary of current earthing design recommendations from standards is
provided and the subject of soil ionisation under high current magnitudes is consid-
ered in some detail. Circuit models and simulation software for evaluating the high
frequency and transient performance of earthing systems are also described.
8.2 Earthing system components and system earthing methods
8.2.1 Transmission system
The 400 kV/275 kV transmission system operated by the National Grid in the
UK comprises over 250 substations interconnected by 15 000 km of double circuit
overhead lines, supported by about 26 000 transmission towers [5]. Approximately
200 faults per annum occur on the transmission system, of which almost 90 per cent
can be classified as earth faults [6]. Although few in number, such faults produce
current magnitudes of between 4 kA and 35 kA flowing through substation earthing
systems, depending on the level of generation, the system configuration and the type
and location of the fault [7]. Normally, faults are cleared by main protection systems
in less than 160 ms, the precise speed depending on the technology of the circuit
breaker. Backup protection would normally provide clearance within 500 ms [8].
At transmission level, earthing systems consist of two main components: main
earth grids [2] and extended earth electrode systems formed by transmission tower
lines. The main earth grid of an outdoor transmission substation will cover a substan-
tial area, typically 30 000 m
2
, although indoor substations can be much more compact
with less scope for a substantial earth grid. The earth wire of transmission tower lines
also forms an important part of the transmission earthing system in addition to its
lightning shielding role. The earth wire interconnects the latticed steel towers and
is terminated at the substation earth grids at the ends of each line. In this way, an
extended earthing system is formed by virtue of the connections made to earth by the
tower footings. The presence of the earth wire, therefore, reduces the earth impedance
of the system at the main substation or indeed at any point along the transmission
line. As a result, any rise in earth potential is transferred over a much larger area. The
impedance of the extended earthing systemformed by the transmission line earth wire
and towers is commonly referred to as the chain impedance. The earthing systems of
transmission substations directly supplying distribution systems (typically at 132 kV
in the UK) are also connected to the lower voltage shielded tower line circuits. In
addition to lowering the earth impedance of the system, a shielded circuit system also
results in a significant reduction of the proportion of earth fault current flowing back
to the source through earth. This is due to the direct conductive path formed by the
shield and also the inductive mutual coupling between the faulted phase conductor
and the shield.
In urban areas, the transmission substation may be situated close to subtransmis-
sion and distribution substations. Usually, the earthing systems of these substations
will be bonded together. At such sites, the overall earth impedance can be very low
due to the effect of the extended earthing systems of the lower voltage systems.
However, although this interconnection reduces the earth impedance of the transmis-
sion system, potentials developed due to high magnitude fault currents on the higher
voltage system will be transferred over wide areas to the lower voltage earthing
systems.
8.2.2 Distribution system
In the UK, the distribution system uses voltages from 132 kV down to 6 kV. It com-
prises approximately 180 000 km of overhead lines and 133 000 km of underground
cables. It is estimated that there are approximately 5000 substations operating from
132 kV to 20 kV and about 170 000 substations distributing power from 11 kV to the
low voltage system [9].
On the distribution system, the fault rate is significantly higher compared with the
transmission system. With the considerably greater route length of the distribution
system, it is not surprising that there are significantly more faults occurring at this
level of the system. It is estimated that there may be 25 000 faults per annum on the
HV distribution system, 70 per cent of which involve a connection to earth [10, 11].
Typically, earth fault current magnitudes are 25 kA at 132 kV, 12 kA at 33 kV and
11 kV substations. Below 66 kV, the system is designed to restrict earth fault current
magnitudes. Although fault current magnitudes on the distribution system are gen-
erally much lower compared with the transmission system, fault clearance times are
considerably longer. As a general rule, the fault clearance time will increase the lower
the distribution voltage level. On 11 kV systems employing IDMT protection, fault
clearance times may exceed 1 s. It is important to establish accurate fault clearance
times because the tolerable level of current passing through the human body depends
on the shock exposure time, which is normally assumed equal to the fault duration.
The longer the fault, the lower the tolerable body limit.
Earthing 353
The 132 kV shielded double circuit system in the UK originally functioned as
a transmission grid. Over a period of time, it evolved into a distribution system as
generation displaced to the Super Grid at 275 kV and 400 kV [12]. These days, new
circuits at 132 kVare often constructed using unearthed wood pole single circuits and
the vast majority of overhead line construction at 66 kV and below has always been
unearthed. Hence, distribution substations connected to overhead line circuits may
not always have the benefit of an extended earthing system.
On the other hand, most distribution circuits in urban areas will consist of under-
ground cables. The metallic sheaths of these cables forman extended earthing system
by virtue of their connection to earthing points along the circuit. This system can be
highly interconnected and cover a large area providing low earth impedance values.
Also, similar to the action of the shielding wire on overhead lines, the sheath of a
cable provides an alternative path for the earth fault current to return to source. The
mutual coupling between the core and sheath of a cable is considerably greater than
that between a phase conductor and earthwire of an overhead line, and therefore this
mutual effect is much greater for cables.
Older underground cables were constructed with lead sheaths and covered by an
insulating layer of bitumen-impregnated hessian (PILCSWA paper-insulated lead
sheathed steel wire armoured). It has been found that, over a period of time, this type
of insulation degrades leaving the lead sheath of the cable effectively in direct contact
with soil. Such cables, therefore, provide an additional fortuitous earth connection
[13]. These PILCSWAcables are steadily being replaced by plastic insulated cables,
which will result in a general increase in earth impedances for such urban systems.
8.2.3 Methods of system earthingtreatment of neutral
8.2.3.1 Background
Inthe earlyperiodof development of three-phase power systems, it was commonprac-
tice to isolate neutral points and operate the system in an unearthed state. In the UK,
the majority of the high voltage systems were operated this way until 1912 [14, 15],
and in Germany this was the case until 1917 [16]. However, as power systems grewin
size, problems with this method of operation emerged because the magnitude of earth
fault current in an unearthed system increases with the phase-to-earth capacitance
of the network. Above a certain current threshold, persistent intermittent arcing will
occur during the fault resulting in damage to equipment close to the arcing fault. Also,
damage can occur to other parts of the network as a result of high magnitude over-
voltages that are developed [1618]. Therefore, the permanent single-phase-to-earth
fault on such systems becomes unmanageable and fast fault detection and isolation,
or alternatively a method of suppressing the arc, are required [16].
Nowadays, some parts of high voltage networks still operate unearthed because it
can be advantageous, under certain circumstances, to continue operating in the pres-
ence of an earth fault [1921]. However, the most commonly recommended practice
is to earth at least one neutral point of the network, and there are important advantages
in operating the system in this way. The main advantage is that the power system is
safer because earth faults are also much easier to detect, to clear and to locate [21].
Also, there is a reduction in overvoltage magnitude during earth faults. The Electri-
city Safety, Quality and Continuity Regulations 2002 state Agenerator or distributor
shall, in respect of any high voltage network which he owns or operates, ensure that
the network is connected with earth at, or as near as is reasonably practicable to the
source of voltage. . .. [22]. The most convenient way of making this connection is to
use the neutral formed by the star point of the supply transformer. For delta-connected
equipment, an earthing transformer may be employed for this purpose [21].
Accordingtohowthe neutral is connectedtoearth, earthedsystems are categorised
as either solidly earthed or impedance earthed. Impedance earthed systems can be
classified as resistance, reactance or resonant type.
8.2.3.2 Solid earthing
Solid earthing describes the connection of the system neutral point directly to
the installation earth. This results in high earth fault levels but minimises system
overvoltages.
8.2.3.3 Resistance earthing
With resistance earthing, a transformer neutral point and earth are connected together
through a resistor. On systems at and above 11 kV a liquid earthing resistor (LER) is
normally employed. The resistance value is chosen such that earth fault current levels
are reducedconsiderably, typicallyto1000Aat 11kV. However, althoughfault current
flow is restricted, sustained phase voltages during the earth fault are not increased
substantially above those for solidly earthed systems. Transient overvoltages are also
contained to within acceptable levels. Occasionally, high resistance earthing is used
to limit overvoltages on previously ungrounded systems [21].
8.2.3.4 Reactance earthing
With this method of systemearthing, a reactor is placed between the neutral and earth.
When reactance earthing is used, it is recommended that the earth fault level should
not be reduced to below 60 per cent of the three-phase fault level, in order to avoid
high transient overvoltages [21]. Therefore, reactance earthing is not considered as
an alternative to resistance earthing. Reactance earthing can be employed in systems
that have an earth fault level exceeding the three-phase fault level per phase [21].
8.2.3.5 Resonant earthing
An alternative method of neutral earthing is achieved by inserting a reactor between
the neutral point and earth such that it compensates the phase to earth capacitance of
the systemoperating under the earth fault condition. This method is known as resonant
earthing [23] or Petersen coil earthing, named after its founder. It is also sometimes
referred to as arc suppression coil earthing or ground fault neutraliser earthing. The
device, the arc suppression coil, is therefore also known as a Petersen coil or a ground
fault neutraliser. Resonant earthing offers two main advantages. First, during a single-
phase-to-earth fault, arcing is suppressed automatically without the need for current
interruption. Second, if required, the network can be operated with the fault present
Earthing 355
because the capacitive earth fault current is compensated. A disadvantage of this
method is that the transient and steady state voltages on healthy phases during the
earth fault are greater than those on low impedance or solidly earthed systems. Also,
the detection and location of an earth fault is more difficult because conventional
overcurrent protection cannot be applied.
8.2.4 Application of different system earthing methods
In the UK power system, solid earthing has been adopted on networks operating
at 66 kV up to 400 kV. At the transmission level (275 kV and 400 kV), the power
system is also multiple earthed because it operates as a meshed system [24]. This
practice reduces transformer capital costs, where graded insulation can be used and
surge arresters of lower temporary overvoltage rating can be employed [11, 2527].
At voltage levels from 6.6 kVto 66 kV, the system is normally radial configured, and
therefore it is earthed at a single point corresponding to the neutral or derived neutral
of the supply transformers. On networks at 33 kV and below, low resistance earthing
is used.
A system is classified as effectively earthed when the single-phase-to-earth fault
current is at least 60 per cent of the three-phase short circuit. This is achieved when the
zero sequence resistance is not greater than the positive sequence reactance (R
0
X
1
)
and the zero sequence reactance is not more than three times the positive sequence
reactance (X
0
3X
1
) [21]. The main advantages of effectively earthed systems are
that arcing faults are very rare, overcurrent earth fault relays can be used to detect and
isolate the faulty circuit and the transient and steady state neutral and phase voltages
with respect to earth are suppressed [16].
8.3 Earth resistivity and measurement techniques
8.3.1 Conduction mechanisms and resistivity
Electrical conductioninearthis predominantlyelectrolytic conductioninthe solutions
of water-bearing rocks and soils. Metallic conduction, electronic semiconduction
and solid electrolytic conduction can occur but only when specific native metals
and minerals are present [28]. Accordingly, the resistivity of soil or rock normally
depends on:
a the degree of porosity or fracturing of the material
b the type of electrolyte
c temperature.
Figure 8.1 indicates the effect of variations in salt, moisture and temperature on
resistivity [31].
8.3.2 Resistivity data of soils and rocks
At power frequency, the series impedance of the conductors forming a concentrated
earth electrode is small, and when the soil has fully consolidated around the installed
a b
20 0 40
c
10
10
2
10
2
10
3
5 10
added salt, %
0 15
moisture, %
20 60
temperature, C
r
e
s
i
s
t
i
v
i
t
y
,
m
r
e
s
i
s
t
i
v
i
t
y
,
m
10
r
e
s
i
s
t
i
v
i
t
y
,
m
1
10 20 0 30
10
10
2
10
3
1
Figure 8.1 Variation of resistivity with salt (a), moisture (b) and temperature (c)
(reproduced from IEEE Std.81 1983, IEEE Copyright 1983 and IEEE
Std.80 2000. All rights reserved [31])
system the contact resistance between electrode and the soil can also be neglected.
Therefore, it is the resistivity of the surrounding earth that determines the earth resis-
tance of the electrode. The resistivity and earth structure will also determine the
potential distribution on the surface of the ground around the electrode, and it may be
necessary to construct very accurate earth models in order to predict step and touch
voltages around an installation.
Figure 8.2 [28] shows the wide range of resistivity for different soil and rock
types. As can be seen in the figure, resistivity can vary from about 10 m (clays) to
10
6
m (granites), and, in general, resistivity increases with the age of the geological
formation [29]. Also noticeable is that the wide resistivity ranges for nearly all types
of soil and rock overlap. This illustrates the difficulty in estimating earth resistivity
from a geological classification. Accordingly, it is recommended that practical inves-
tigations be carried out on site in order to measure directly the earth resistivity for
each specific electrical installation [30]. However, the variable qualities of natural
Earthing 357
j in m c
r
(dry rock)
rock type
water content
water content
increasing decreasing
increasing
10
1
10
3
10
5
10
7
20 15 10 5 0
magmatic
granite
syenite
diorite
gabbro, diabase
basalt
peridotite
metamorphic
quartzite
gneiss
phyllite
sedimentary
dolomite
limestone
sandstone
sand
clay
Figure 8.2 Mean value ranges of resistivity and permittivity for different rock types
(reproduced from [28])
soil and rock make this task difficult. As expected, variations in resistivity with depth
are common. Lateral variations can also be significant even across one site, although
this effect is sometimes neglected [31]. In addition, seasonal variations in water con-
tent will also result in changes in earth resistivity and temperature variations at the
soil surface may also have an effect.
8.3.3 Site investigation and measurement techniques of earth resistivity
and structure
It is important to determine an accurate earth resistivity model as a basis for designing
an earthing system, and there are a number of different but complimentary methods
and sources of information available for this purpose. Many of these, such as geolog-
ical maps, borehole data, seismic testing and ground penetrating radar, are useful for
identifying physical boundaries in the earth. However, they do not quantify accurately
earth resistivity. On the other hand, only earth resistivity measurements carried out
on the ground surface may not be sufficiently adequate to define the boundaries of
a earth model uniquely. Earth resistivity surveying used in conjunction with one or
more of the methods for identifying earth region boundaries may offer a much greater
degree of confidence in the derived earth models.
8.3.3.1 Geological maps
Geological maps are available in solid and drift formats. Solid maps describe the
underlying bedrock, and drift maps detail superficial deposits. The British Geological
Survey produced a series of maps for the UK, including 1:625000 scale maps, showing
the geology of the UK landmass. Acomprehensive series of 1:50000 scale solid and
drift maps is also available with supporting literature [32]. Although these maps are
two-dimensional, data for certain vertical cross-sections are provided which gives the
depths of different layers.
8.3.3.2 Borehole data
At electrical installations requiring detailed civil engineering works, borehole sam-
ples can provide information about changes in the earth structure and type with depth.
Althoughthis type of informationis useful for defininglayer boundaries inearthanaly-
sis, measurement of the resistivity of the removed earth sections is not recommended
because the compaction and moisture content of the samples will be significantly
affected by the extraction process and may no longer be representative [31].
8.3.3.3 Seismic surveying
To perform a seismic survey, an acoustic wave is generated on the ground surface
(e.g. using a sledgehammer) and the reflected waves from earth layer boundaries
are measured using geophones placed in a line away from the source. By analysing
the travel times and amplitudes of the returning waveforms, estimates of the thickness
and density of the subsurface layers can be obtained. An estimation of the depth to
bedrock or the level of the water table is particularly useful for earthing investigations
[33, 34].
8.3.3.4 Ground penetrating radar
A radar set is used to generate short bursts of VHF electromagnetic waves in the
frequency range 35 MHz to 900 MHz. Reflections of these waves are produced from
boundaries between materials with contrasting resistivity and/or dielectric properties.
By estimating the velocity of propagation, it is possible to determine the depth to these
boundaries. Using this technology, depths of investigation may be limited to tens of
metres in dry sands, and as little as 1 m in wet clay [34].
8.3.3.5 Earth resistivity measurements
The most widely recommended and commonly used earth resistivity technique in
earthing investigations is the Wenner method [30, 3538]. This method uses four
equally spaced electrodes, as shown in Figure 8.3 and its characteristic feature is a
constant spacing a between adjacent electrodes. Current is circulated between the
outer two electrodes and the potential difference is measured between the inner two
electrodes. A DC instrument is used to energise the circuit but in order to overcome
the effects of telluric currents, electrochemically-produced currents and interference
fromDCpower sources, a periodic reversal in the direction of the current is necessary.
Earthing 359
C
2
P
2
P
2
C
1
a a a
P
2
P
1
C
a a
A M N B
l l
P
2
P
1 C
a
l
C
2
P
2
P
1
P
0
C
1
a
a
2 2
a
a
P
2
P
0
P
1
C
a
2
a
2
a
a. Wenner configuration
surface
b. Lee configuration
surface
c. asymmetrical Wenner configuration
surface
d. asymmetrical Lee configuration
surface
e. Schlumberger configuration
surface
f. Lgn configuration
surface
Figure 8.3 Various soil resistivity test electrode configurations (reproduced
from [38])
Connection of the voltage terminals can also be delayed until any switching transients
have decayed. Telluric currents flow in the earth over large areas with changing
magnitude and direction and varying periods. They correspond to regular and irregular
changes in the geomagnetic field and are particularly high during magnetic storms.
Electrochemically-produced currents can be much higher in magnitude than telluric
currents and are produced, for example, by differential oxidation processes. Lee
records the use of instruments such as the Megger in earth resistivity measurement as
far back as 1928 [38], and a modern version is used extensively in the UK nowadays
[39]. Anumber of alternative earth resistivity instruments are available fromdifferent
manufacturers including those specifically designed for geophysical exploration. The
specifications of these instruments vary considerably.
Earth resistivity measurements taken in the vicinity of existing electrical instal-
lations may be significantly affected if test transects are close to the buried earth
grid or the routes of the substations underground cable circuits. The presence of
buried metallic objects would, as expected, result in a general underestimation of the
resistivity of the native ground. However, such distortion in the apparent resistivity
readings may not be easily recognisable and could be interpreted as a variation due
to the natural heterogeneity of the earth. Therefore, it is advisable to identify the
location and avoid measurements around cable circuit routes. If the orientation of an
earth resistivity transect is varied around a centre point, the presence of underground
metallic structures may be indicated by changing trends in the resulting apparent
resistivity curves [31, 40].
(i) The Wenner method
For any given spacing a, the ratio of voltage to current will yield a specific value of
resistance. If the current electrodes are assumed as point sources and the resistivity,
, to be homogeneous, it can be shown by applying the principle of superposition that
the resistivity is related to the measured resistance, R, by the following formula [38]:
A
= 2a
V
I
= 2aR (8.1)
Even if the earth is heterogeneous, the quotient of V/I can be obtained and will
be related to a quantity that is dimensionally equivalent to resistivity. This quantity
can be considered as an index indicative, in a complex way, of the different resis-
tivities present beneath the surface. This index is known as apparent resistivity. If a
number of readings are taken for different constant interelectrode spacings, a, an
apparent resistivity curve can be plotted as a function of a. The curve can be com-
pared against theoretically obtained curves to derive approximate earth models, and
the apparent resistivity will usually lie within the range of actual material resistivities.
As electrode spacing is increased, the current path will involve a larger volume of
earth and encompass deeper earth strata. Therefore, the apparent resistivity as a func-
tion of probe spacing provides an indication of the change in resistivity of the earth
as a function of depth. However, the depth rule, which states that the electrode spac-
ing is equal to the depth of penetration, should be applied with caution because the
correspondence between spacing and depth is only approximate. There are a number
of variants of the Wenner configuration and these are also shown in Figure 8.3 [38].
Adistinguishing feature of these alternative methods is that the potential measurement
is made across a much smaller distance and this has two implications:
a The voltage measurement at small distances can be approximated to a measure
of the electric field at that location and it is, therefore, easier to detect geological
changes. In contrast, if the potential is measured across a large distance, effects
of geological anomalies tend to be attenuated.
b The measurement of such small voltage magnitudes with sufficient accuracy
may be difficult in practice either because of limitations in the resolution of the
instrumentation or due to high levels of background noise in the measured signal.
Taking into account these considerations, the Wenner method is well suited to appli-
cations requiring reasonable accuracy but where the detection of local heterogeneity
is not of primary concern.
Earthing 361
(ii) Other methods
The SchlumbergerPalmer method is another configuration that uses unequally
spaced probes. For this arrangement, the spacing between the potential electrodes
is increased such that they are much closer to the current electrodes [31]. In this way,
a higher potential difference is obtained compared with other arrangements and this
can be beneficial when the resolution of the resistivity meter is a limitation. With
both the Schlumberger and Wenner configurations, the requirement to investigate the
resistivity of deeper earth for large area substations introduces practical difficulties.
For such tests, a very large interelectrode separation is required which necessitates
long test leads and a relatively unobstructed path along the test route. In order to
overcome these difficulties, dipole techniques were used to survey to greater depths,
up to several kilometres, by separating the current and voltage circuits, which are
referred to as dipoles. Measurements carried out at a very large spacing (330 km)
indicated an upper limit to the resistivity of deep strata in the order of 10
4
m [41].
(iii) Resistivity surveying techniques
In order to obtain information on earth resistivity changes with depth, the centre point
of the test electrode setup is kept fixed and the interelectrode spacing is varied. This
is known as vertical profiling. If the whole configuration is moved along a traverse
at a fixed spacing, it is referred to as horizontal profiling [38].
8.3.3.6 Transient electromagnetic technique
In earthing applications, the depth of interest for determining resistivity may vary
from a few metres to several hundred metres depending on the size of the electrode
system. The transient electromagnetic technique (TEM) offers a practical method for
obtaining earth resistivity data at significant depths. In this technique, a large closed
transmitter loop is laid out on the earths surface. A DC current is established in the
loop which is abruptly interrupted. This sudden change induces currents in the earth
which decay according to the resistivity and structure of the earth. The decaying
currents produce a magnetic field which is measured by a receiver on the earths
surface. By analysing the nature of the transient decay, it is possible to determine
equivalent resistivity models [42].
8.3.3.7 Earth resistivity maps
Numerous earth resistivity measurements were carried out in carefully selected parts
of the UK with a view to ascertaining the range of values which are representative
of each of the principle geological formations [43]. These measurements were taken
around the time when the first 132 kV national grid network was established, for
the purpose of calculating telephone interference. The apparent resistivity values at
a spacing of 500 feet were used to categorise the resistivity of the solid geology. As a
result, two electrical resistivity maps were produced based on solid geological maps;
one covering England and Wales [44] (c.f. Geological map of the UK (South) [45])
and the other of Southern Scotland [46] and these are reproduced in Figures 8.4a
and 8.4b. It is stated that the maps are not suitable for quantifying the resistivity of
a
Figure 8.4a Electrical resistivity map of England and Wales (reproduced from both
a 1934 Ordnance Survey Map and [44])
subsurface layers. These maps need updating as good progress has been made since
their establishment in terms of measurement techniques and analysis of data.
8.3.3.8 Formation of earth models
(i) Apparent resistivity curves
The recommended practice for earth resistivity investigations in earthing system
design is to carry out vertical profiling at a number of positions around the
Earthing 363
b
Figure 8.4b Electrical resistivity map of Southern Scotland (reproduced from
John Bartholomew, ERA, IPR [46])
installation [30]. Test results of vertical profiles taken at different sites are shown
in Figure 8.5 [47]. Curve A of Figure 8.5 illustrates a case of when the apparent
resistivity varies little with spacing which indicates reasonably homogeneous earth
conditions. However, all other curves indicate that changes occur in apparent resis-
tivity as a result of the non-uniformities in the earth. Curve E represents a two-layer
earth with the top layer consisting of a higher resistivity than the lower layer. By
contrast, curve D shows a low resistivity near the surface and suggests very high
resistivity at depth. The reflection coefficient k is used to describe the contrast in
resistivity between the upper layer (
1
) and the lower layer (
2
):
k =

2
1
+
2
(8.2)
Curves B and C are characteristic of three-layer earth structures: low resistivity near
the surface, an intermediate layer of higher resistivity and lower resistivity at depth.
Habberjam [48] refers to such three-layer curves as K-type curves, which is one of
four different classifications of three-layer earth models. Apparent resistivity curves
may be obtained which indicate earth structures consisting of more than three layers.
However, as the complexity of the earth structure increases, it becomes more difficult
to construct reliable models.
value from
computed model
measured results
interelectrode spacing, m
10
1
10
1
10
2
10
3
10
0
10
1
10
2
10
3
a
p
p
a
r
e
n
t

r
e
s
i
s
t
i
v
i
t
y
,

o
h
m
-
m
B
C
C
B
A
E
A
E
D
A
B
C
D
E
Figure 8.5 Apparent resistivity versus interelectrode spacing (reproduced
from [47])
(ii) Simple models
Some earth structures may be very complex and practical testing can yield apparent
resistivity curves with many maxima and minima. It may therefore be difficult to
establish a suitable earth model for power system earthing applications and the pub-
lished literature contains many different approaches to this problem. The simplest
method that can be applied is to obtain an effective or equivalent value of earth resis-
tivity from the apparent resistivity data [36, 4951]. The selection of a single value
resistivity is convenient because many analytical formulae for computing earth resis-
tance and ground surface potentials are based on homogeneous earth assumptions and
some earthing standards still employ such formulae [36, 52]. Recent studies [53] have
shown that although it may be possible to obtain acceptable homogeneous equivalent
earth models for the calculation of earth resistance of concentrated earth structures,
their use for the prediction of earth surface potentials is questionable. Therefore, a
homogeneous model would only be acceptable if site measurements indicate a small
variation in resistivity over a wide range of interelectrode spacing. Recent measure-
ments [54] of apparent resistivity (Wenner method vertical profiling) at over 80 sites
around the UK showed that significant variations occur over interelectrode spacing
in the range from a few metres up to several tens of metres.
(iii) Standard models
To deal with more complex earth structures, the IEEE Std.80 standard recommends
the formation of a two-layer earth model or if necessary a multilayer earth model
Earthing 365
according to the number of maxima and minima on the apparent resistivity curve.
A number of techniques for obtaining such models are described [30]. Dawalibi and
Barbeito [55] analysed practical earth grids and concluded that non-uniform earth
requires the use of multilayer earth models to predict earth grid performance. They
also noted the significant effect of vertical discontinuities and stressed the importance
of obtaining resistivity readings of sufficiently wide spacing in order to establish the
asymptotic value of earth resistivity at depths of the order of the earthing system
dimensions.
(iv) Two-dimensional/three-dimensional models
In two or multilayer models, it is assumed that the resistivity changes only with
depth and that there are no lateral variations. These are sometimes referred to as one-
dimensional models as shown in Figure 8.6. If vertical profiling is combined with
horizontal profiling along one line, it is possible to obtain a two-dimensional model. If
the measurements are extended such that the horizontal profiling is carried out in two
orthogonal directions, a three-dimensional model may be obtained. Typically, one-
dimensional measurements may require 20 readings and three-dimensional surveys
would require thousands. In order to make such measurements feasible, large num-
bers of electrodes, multicore cables and automatic switching systems interfaced with
computerised data acquisition systems are required. The data froma two-dimensional
survey can be plotted using the pseudosection contouring method such that the
horizontal location of the plot point corresponds to the mid-point of the particular
electrode configuration. This gives an approximate picture of the subsurface earth
in a similar way to the apparent resistivity curves from one-dimensional surveys. To
analyse data from such detailed surveys, an inversion procedure is required which
constructs a model to fit the measured data. Commercial measurement systems
with inversion software are available to analyse both two- and three-dimensional
measurements [56].
8.4 Power frequency performance of earthing systems
8.4.1 Standards recommendations
The determination of the performance of an earthing system under power frequency
earth fault conditions, for the purpose of developing a safe earthing design, requires
a number of aspects to be investigated. The main procedure outlined in various stan-
dards is the same and is shown below, although the detailed methodology, estimation
of parameters and safety criteria differ to some extent:
(i) measurement of earth resistivity and formulation of earth model
(ii) calculation of overall earth impedance including main earth grid resistance and
impedance of extended earth electrodes using the model developed in (i)
(iii) earth fault current calculation for various fault locations
(iv) fault current distributioncalculationtodetermine the proportionof fault current
passing through the earth
1D model 2D model 3D model
C
1
P
1
P
2
p
1
p
1
p
1
p
2
p
2
p
3
p
2
p
3
C
2
a
b
c
electrode spacing
1.0 10.0 100.0 1000.0
0.5 5.1 51.0 510.0
depth of layers
depth iteration 3 r.m.s. error = 5.4 %
0.0 48.0 96.0 144 192 m
inverse model resistivity section
resisitivity in ohm.m unit electrode spacing 3.0 m
42.2 59.6 84.3 119 168 238 336 475
1000.0
app.
100.0
resis.
10.0
1.0
+
iteration 6
cal. app. res.
obs. app. res
% r.m.s. error 0.17
resistivity
apparent
inversion method
Gauss Newton
least squares
computed model
0.8
9.4
17.6
30.3
39.2
Figure 8.6
a three different models used in the interpretation of resistivity
measurements
b a typical 1Dmodel used in the interpretation of resistivity sounding
data (reproduced from [56])
c an example of a 2D model
(v) estimation of earth potential rise and the safety hazard voltages
(vi) specification of safety limits and determination of the safety level of the
earthing design
(vii) measurements at installation to verify calculated impedances and potentials
and, subsequently, to ensure continued integrity.
Apart fromsatisfying electrical safety, the earthing systemshould also have sufficient
mechanical strength and be corrosion resistant, and the system should have adequate
thermal capability for carrying the maximum fault current. These aspects are dealt
with in detail in most earthing standards.
Earthing 367
8.4.2 Earth impedance
8.4.2.1 Calculation of earth resistance of concentrated earth electrodes
Analytical expressions of earth resistance for simple forms of earth electrodes are
well established [57, 58]. Formulae for various geometries including the sphere, rod,
plate, strip, disk, ring, grid, grid and rods are specified in many standards [30, 36, 52].
All these expressions for earth resistance assume homogeneous earth.
The resistance of a hemispherical electrode of radius r in earth of resistivity is
given by:
R =

2r
(8.3)
This expression encapsulates the two main features of the power frequency behaviour
common to all concentrated earths; i.e. that earth resistance is proportional to earth
resistivity and inversely proportional to the extent of the electrode. Based on a realistic
range of earth resistivity (section 8.3), the earth resistance of an earth grid may
therefore vary from several tens of ohms for small grids to as little as 0.01 for
large systems in low resistivity earth. Unless the earthing system is very extensive or
situated in very low resistivity earth, the series impedance of the conductors forming
the earthing system is very small in comparison with the earth resistance at power
frequency. Accordingly, in most cases, it can be assumed that all metallic points on
the electrode are approximately at the same potential.
Recently, there have been considerable developments in techniques for evaluating
the earth resistance of practical systems. Work has focused on improvements and
extensions of the analytical formulae for simple electrode systems, in particular, to
account for two-layer horizontal earth models [5961]. Also, computer programs have
been developed which model earth electrodes as systems of interconnected cylindrical
conductor segments [4, 62]. Features of more advanced programs include the ability
to model more accurately the asymmetry of practical earthing system designs. Such
computer programs are now commonly employed to design grids and are used in
conjunction with earthing standards.
New approaches to the modelling of earthing systems include the application of
the boundary element method [6365]. Using this method, it is possible to analyse
earthing systems that are located in earth which exhibits both horizontal and vertical
variations in resistivity and which may also include local inhomogeneities.
8.4.2.2 Chain impedance of extended earthing systems of tower lines
The impedance of the extended earthing system formed by a transmission line earth
wire, commonly referred to as the chain impedance, depends on a number of factors.
These are the size and type of the earth wire, the size and design of the tower footings
and the nature of the earth along the route of the line [66]. Figure 8.7 shows the
50 Hz earth impedance of a tower line for four different earth resistivity conditions
as a function of line length. As can be observed, the earth impedance seen from the
sending end decreases as the line length increases. For a given earth resistivity, there is
a particular length of line, known as the effective length, which results in a minimum
(i) j=10 km
(iv) j=10 m
(iii) j=100 m
(ii) j=1km
c
h
a
i
n

r
e
s
i
s
t
a
n
c
e
,

circuit model computer model

0.1
1
10
100
1000
number of spans
1 11 21 61 31 41 51
Figure 8.7 Chain impedance of a shielded transmission line as a function of line
length [67]
tower footing resistance R
t
, ohms
0 5 10 15 20
c
h
a
i
n

i
m
p
o
r
t
a
n
c
e

Z
c
h
,

o
h
m
s
132 kV line
0=34 approx
275 kV line
0 =41 approx
400 kV line
0= 46 approx
1.0
2.0
3.0
Figure 8.8 Chain impedance and phase angle for long lines as a function of tower
footing resistance (reproduced from [36], courtesy of Energy Networks
Association (ENA))
value equal to the characteristic impedance [67]. The chain impedance is commonly
expressed for an infinite half line as a function of tower footing resistance, which is
assumed constant along the route of the line. Figure 8.8 shows typical values of chain
impedance (Z
ch
) for different line construction types [36]. The chain impedance of
Earthing 369
DC
AC 50 Hz
100 1000
2
4
6
8
10
r
e
s
i
s
t
a
n
c
e
/
i
m
p
e
d
a
n
c
e
,
0
12
10 10000
soil resistivity, m
Figure 8.9 Comparison between DC chain resistance and AC (50 Hz) chain
impedance for a 400 kV tower line (reproduced from [67])
an infinite half line can also be estimated using the following formula [52]:
Z
ch
=
_
Z
s
R
t
(8.4)
where Z
s
, the longitudinal self-impedance per span, can be calculated using the
CarsonClem formula or by using the complex image method [68, 69].
The tower footing resistance R
t
can be estimated from standard earth resistance
formulae (see section 8.4.2.1).
Figure 8.9 shows both the DC chain resistance and AC chain impedance of
a line as functions of earth resistivity corresponding to the asymptotic values of
the impedance/resistance length curves. The curves indicate a significant difference
between the DCand the ACcase that increases with earth resistivity. It is important to
appreciate this relationship because sometimes the AC chain impedance is estimated
from DC tests.
More detailed circuit models of the earthing system formed by a transmission
line can be established to account for the resistances of individual tower footings.
However, in practice, such data are often not available.
8.4.2.3 Earth impedance of extended systems with sheathed cables
Urban distribution systems, in particular those at the lower end of the high voltage
range (typically 11 kV), comprise extensive interconnected networks of underground
cables. The sheaths of these cables interconnect the main substation earth grid to
other substation earths and provide an extended earthing system. In addition, for
older construction types of cable employing lead sheaths and hessian servings, direct
contact is established between the soil and the lead sheath over a period of time.
As a result, such cables can be considered as horizontal extensions to the grid [36].
100 1000
length, m
10 10000
10m
100m lkm
1
10
100
i
m
p
e
d
a
n
c
e

m
a
g
n
i
t
u
d
e
,

0.1
1000
Figure 8.10 Earth impedance of an 11 kV 185 mm
2
uninsulated cable sheath as a
function of length for different earth resistivities [47]
Figure 8.10 shows the earth impedance magnitude of an uninsulated sheath as a
function of cable length. The graphs showthat sheath earth impedance is proportional
to earth resistivity and, for a given resistivity, the impedance decreases in magnitude
up to a certain length. Beyond this length, referred to as the effective length, the
impedance is practically constant. The effective length can extend to about 1 km in
100 m earth [47]. Plastic insulated cables do not provide this fortuitous earth effect
and, consequently, the gradual replacement of older cables would, over a period of
time, increase the earth impedance of urban systems. Detailed analyses of the earthing
contribution of underground cable circuits can be found elsewhere [7073].
8.4.3 Interactions between fault currents and earthing systems
8.4.3.1 Fault current magnitude
Fault current calculations are primarily required to specify the interrupting capacity
of circuit breakers at various points on the system, to determine the short time thermal
rating of equipment and to enable the setting of protective relays. These calculations
are also required to determine the current magnitude that will flow through earthing
systems under fault, which in turn determines the earth potential rise.
Techniques for calculating both symmetrical and total asymmetrical fault currents
are described in IEEE/ANSI [7477] standards and in IEC909 [78]. These standards
employ quasi steady state equations and a number of simplifying empirical factors
[7981]. Recent studies have shownthat suchapproaches produce highlyconservative
Earthing 371
results when compared with detailed dynamic fault current calculation techniques
[8284]. The need to employ more detailed fault calculation algorithms has been
identified [85, 86].
For an accurate assessment of earth potential rise, it is necessary to account for the
dynamic components in the fault current waveform. These dynamics include the DC
offset and the decayingACcomponent that is present if the fault is near to a generator.
In contrast to standard fault current calculations, which require the determination of
current magnitude at a particular point in time, in earthing system assessments the
cumulative effect of the current is required to determine the electrical shock severity.
Accordingly, the r.m.s. equivalent value of the total asymmetrical earth fault current
over the fault period should be calculated. In one standard [30], in order to represent
the most severe condition, it is recommended that the maximum DC offset is used
and that the AC component is assumed to remain constant at its initial subtransient
value. For any given earthing system under study, different earth fault locations
should be examined because these can have a considerable effect on the earth return
current magnitude. For example, at substations, studies should examine faults on both
the high and low voltage sides and faults external to the substation should also be
considered [2, 30].
8.4.3.2 Fault current distribution
When an earth fault occurs at a substation, the current returns through the earth along
the path of least impedance and enters the earth grid connected to the neutral at the
source end to complete the circuit. If the earthing system of the source and faulted
substation are interconnected, either by the earth wire of a shielded transmission
line or the sheath of an underground cable, an additional return path is established
for the fault current. Also, as a result of the mutual coupling between the faulted
phase conductor and the parallel earth conductor, this additional metallic return path
can carry a considerable proportion of the total fault current. In this way, the earth
return current is reduced and, hence, the rise of earth potential at the substation.
Generally, to determine the earth return current at the fault point as a proportion
of the total earth fault current, it is not essential to model the entire system. For
practical purposes, the relevant part of the network is that electrically near to the fault
and the use of a two-substation model is sufficiently accurate. This arrangement is
shown in Figure 8.11 where the faulted substation and equivalent source substation
are interconnected by a single shielded circuit. For this circuit, the earth return current,
I
gr
, can be solved in terms of the total fault current, the impedance parameters of the
interconnecting circuit and the substation earth impedances:
I
gr
= I
f
z
c
z
mp,c
z
c
+((Z
eA
+Z
eB
)/l
AB
)
(8.5)
where I
f
is the phase conductor fault current, z
c
is the self-impedance/unit length of
the shield conductor, z
mp,c
is the mutual impedance/unit length between the phase and
shield conductors, Z
eA
and Z
eB
are the earthing system impedances of the sending
end and receiving end substations, respectively, and l
AB
is the circuit length.
Z
eB
S
A
Z
eA
I
f
I
gr
l
AB
I
AB
sheath
phase
S
B
z
mp,c
z
c
Figure 8.11 Fault current distribution between sheath and ground return for a
two-substation arrangement employing a three-core cable (reproduced
from [87])
This expression shows that when the circuit is relatively short, as is generally the
case for cable circuits, the earth return current magnitude is strongly influenced by the
values of the substationgridimpedance. Accordingly, whenconsideringcable circuits,
it is necessary to evaluate fault current distribution case-by-case [87]. Earthing stan-
dards provide both nomograms and analytical relations for a number of two-substation
arrangements employing three-core and single-core paper-insulated cables [36, 52].
On the other hand, shielded overhead lines are generally quite long and the magni-
tudes of the substation earth impedances are sufficiently low in comparison to the
system longitudinal impedances not to affect current distribution. This is why, in the
case of overhead lines, a shielding factor can be applied which is determined only by
the line parameters for a particular tower line type [36]. For a more detailed analysis of
fault current distribution, a number of different analytical methods have been devel-
oped considering both overhead lines [88, 89]and cable circuits [90, 91]. Specialised
circuit analysis software for analysing complex systems is also available [4].
8.4.4 Measurement of earth impedance and potentials
For new installations, the measurement of the earth resistance or impedance of an
earthing installation confirms whether the system performs according to its designed
value. For existing installations the measurement is, to some extent, a method of
establishing the continued integrity of the earthing system.
Measurements are made difficult due to a number of factors including stray
currents, the settling of the earth grid after installation and distortion of ground sur-
face potentials by the presence of buried metallic objects. In the case of large area
grounding systems or earth grid connected to extended earthing systems, the reactive
component may be significant, and, therefore, it is imperative to carry out AC rather
Earthing 373
than DC testing. For earthing systems with low impedance, the mutual coupling
effect between test leads becomes a serious issue. Low magnitude current testing,
the preferred option for safe and non-invasive measurements, is made difficult by the
presence of relatively high levels of power frequency background noise.
8.4.4.1 The fall-of-potential (FOP) earth resistance/impedance measurement
technique
The most representative measurement of the earth impedance of an installation is
the staged fault test. This test produces realistic fault current magnitudes and, by
using a remote voltage reference, the rise of earth potential and hence the earth
impedance can be calculated. As an alternative to the staged fault, one circuit feeding
a substation can be taken out of service and used as a test circuit. The staged fault test
or the out-of-service circuit test is seldom performed due to economic penalties and
system operational constraints. For this reason, it is necessary to measure the earth
impedance of an energised substation without any interruption to the supply.
(i) Fall-of-potential set up
The measurement is achieved by injecting current into the earthing system using a
temporary test circuit. This type of test is generically referred to as the fall-of-potential
(FOP) method [92], although many variants of this technique have been developed.
The FOP method is established as the most suitable method for measuring the earth
resistance of high voltage substation earthing systems and is preferred to techniques
such as the two-point or three-point methods [93]. The principle of the FOPmethod is
shown in Figure 8.12. Atest current is injected into the earth electrode under test and
returns via an auxiliary current electrode at some distance C away. The difference in
potential between the test electrode and an auxiliary potential electrode is measured
at successive points, identified a distance P along a straight line close to the current
carrying lead. Such an arrangement is referred to as an in-line FOP test. The ratio of
voltage to current, known as the apparent resistance, is plotted against the distance
P to form an earth resistance curve shown in Figure 8.13 [94]. If the distance C is
r
P
C
1
2 3
I
P=0.618C
V
Figure 8.12 Fall-of-potential test electrode arrangement
curve 2
curve 1
earth resistance of electrode determined by
apparent resistance in the plateau region
auxiliary current electrode spacing, m
a
p
p
a
r
e
n
t

r
e
s
i
s
t
a
n
c
e

V
/
I
,
Figure 8.13 Earth resistance curves with and without plateau regions
sufficiently large in comparison with the dimensions of the test electrode (approxi-
mately ten times the maximum dimension of the test earthing system), the apparent
resistance curve (curve 1) has a definite plateau. A plateau is formed only when the
resistance zones of the electrode under test and the auxiliary electrode do not overlap.
The value in this region of the plateau yields the earth resistance of the electrode
under test. If C is comparatively small, there is no plateau region in the apparent
resistance curve (curve 2) and estimating the earth resistance from such a curve is
more difficult. To address this problem, Curdts [95] developed an analytical model
of the FOP method. The model was based on a hemispherical shaped test electrode
of radius r in uniform earth, and it was shown that the apparent resistance at any one
distance P is equal to:
R =

2
_
1
r
1
P
1
C r
+
1
C P
_
(8.6)
With the approximation that r is small compared with C, the solution of Equation (8.6)
which yields the earth resistance of a hemispherical electrode R = /2r is given
by the condition P = 0.618C. This method of analysing the fall-of-potential curve is
therefore known as the 61.8 per cent rule (it is interesting to note that the 61.8 per cent
rule has a numerical counterpart in pure geometry, known as the golden rule). The
61.8 per cent rule offers a technique for estimating the earth resistance of large area
electrodes when it is impractical to set the auxiliary electrode far enough away such
as to obtain a plateau region in the apparent resistance curve. Nevertheless, it is
recommended that C should still be about ten times r. If the test electrode and the
auxiliary electrode were of the same size, as expected, the correct position to place
the probe would be at 50 per cent.
One of the practical difficulties in applying the 61.8 per cent rule is the requirement
to measure the distances C and P from the electrical centre of the earthing system.
Tagg developed a number of techniques which addressed this issue culminating in
the slope method to enable the earth resistance to be determined without requiring the
Earthing 375
position of the electrical centre of the grid to be known [9698]. The slope method
assumes a distance x between the chosen point of reference, e.g. the substation fence,
and the true electrical centre of the earthing system under test. The test requires three
apparent resistance readings to be taken corresponding to 20, 40 and 60 per cent of C.
Using these values of resistance, x can be evaluated by solving three simultaneous
equations and the required auxiliary potential electrode location and corresponding
resistance is determined using the 61.8 per cent rule. Tagg also considered that the
slope method would provide an indication whether there was sufficient distance to
the auxiliary current electrode.
(ii) Effect of non-uniform earth on FOP measurements
The effect of earth non-uniformity on the fall-of-potential technique has been inves-
tigated for two-layer earths [35, 96, 99]. It was demonstrated that a two-layer earth
causes a shift in the auxiliary potential electrode location yielding the earth resistance
of the electrode under test i.e. the 61.8 per cent method does not apply. For positive
reflection coefficients (high resistivity lower layer), the auxiliary potential electrode
would need to be situated beyond the 61.8 per cent position, towards the auxiliary
current electrode. Recent work by Ma et al. [100] has identified the need to carry
out accurate earth resistivity measurements in conjunction with FOP tests in order to
construct accurate earth models to assist in the interpretation of the FOP curve.
(iii) Effect of test lead mutual coupling on FOP measurements
The potential difference measured in AC fall-of-potential tests consists of two com-
ponents: the actual voltage difference between the earthing system under test and the
auxiliary potential electrode, and the induced potential due to alternating current flow
in the current test loop.
The effect of mutual coupling between current and voltage test leads can be con-
siderable and, for an in-line FOP arrangement in particular, there can be a significant
overestimation of the impedance value of the system under test.
Measurements on large area earthing systems are more susceptible to the effects
of mutual coupling for two reasons: (i) the value of earth impedance to be measured
is quite low, typically less than 1 , (ii) large area earthing systems, by virtue of their
large resistance zones, require long test lead distances.
Typically, mutual impedance can be about 0.5 /km but the precise value is
dependent on the test lead spacing [36], and it is recommended that this effect should
always be accounted for when measuring earthing systems with a resistance of less
than 1 [31]. Recent work by Ma and Dawalibi [101] examined the mutual coupling
effect under different conditions and confirmed that when the operating frequency is
high or the earthing systemis very large, i.e. having lowearth resistance, measurement
results are severely affected. The influence of mutual coupling on earth impedance
measurements of shielded transmission lines, for different test lead configurations, is
quantified by Harid et al. [102].
If the earthing system under test has extended earths such as tower lines, there
may also be mutual coupling between the extended earth and the potential measuring
circuit. The mutual coupling effect can be calculated using Carsons formulae for
infinitely long conductors [103] or simplified versions such as the CarsonClem
formulae [68]. Velazquez et al. [104] and White and Rogers [69, 105], using the
complex image method [106], developed equations for calculating the coupling effect
between finite length parallel and angled conductors.
As would be expected, the mutual coupling reaches a maximumwhen the auxiliary
potential test leadis laidinparallel withthe auxiliarycurrent test lead, andthe coupling
reduces as the spacing between the circuits is increased, and falls to a minimum
when the auxiliary potential test lead is at 90
to the auxiliary current test lead. If

the angle between test leads is greater than 90
, the measured impedance will be

lower due to the mutual coupling. This means that the so-called 180
test, where
the potential lead is laid out in the opposite direction to the current lead, results in
underestimation of the earthing impedance of the test system. For large area earthing
systems, it is recommended that the earth impedance be measured using angled test
lead arrangements, preferably at 90
.
(iv) Practical testing considerations
Measurements are much more difficult on lowimpedance earthing systems due to the
presence of background noise and mutual coupling effects. Background noise arises
fromunbalanced loading on the three-phase system, the presence of harmonics or any
other mechanism that causes a current to circulate through the grid. This noise can
be quantified by the standing voltage present on the earthing system with respect to a
remote earth and magnitudes are typically in the range from100 mVto several tens of
volts. The background noise can be measured using a frequency-selective voltmeter
or a spectrum analyser. Figure 8.14 shows an example of voltage noise measured
on the earthing system of a 400 kV substation over a range of frequencies. The
background noise may also fluctuate according to changing system conditions, for
example at substations supplying electrified railway systems. Figure 8.15 shows the
background voltage measured on the earthing system of a 400/25 kV railway supply
substation as a train was passing. An assessment of the background noise conditions
on an earthing systemwill enable selection of the minimumrequired magnitude of the
test current in order to achieve an adequate signal to noise ratio. As an alternative, the
earth impedance of high voltage energised substations can be measured at a frequency
away from the power frequency in order to reduce or eliminate the effects of noise
[107, 108]. Figure 8.16 shows earth impedances at frequencies either side of 50 Hz
measured at a 275 kV substation using an impedance measurement system (IMS).
The required 50 Hz impedance is obtained by interpolation [94].
(v) Test instrumentation
So-called composite instruments designed to measure earth resistance of concen-
trated earths are not recommended for measuring systems of impedance of less than
1 . Larger earthing systems with extended earths will exhibit reactance as well as
resistance and may require the separate measurement of current and voltage using
specialised instrumentation [2, 94].
Earthing 377
100
(i) P=1000m V
max
(ii) P=1000 m V
median
(iii) P=1000m V
min
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
2.1
10 1000
test frequency, Hz
b
a
c
k
g
r
o
u
n
d

n
o
i
s
e
,

V
0.0
2.2
Figure 8.14 Background noise as a function of frequency measured at a 400 kV
transmission substation
8.4.5 Maintenance and integrity testing of earthing systems
The importance of regular testing (e.g. every five or six years) to prove the contin-
ued integrity and performance of an earthing system is recognised in many earthing
standards [2, 30]. As a part of this exercise, a detailed up-to-date site plan of the
earthing system is required [1]. Earthing systems should be inspected and tested to
a
b
Figure 8.15 Background noise present on railway earthing system
a remote-track voltage: 2 V/div, 5 ms/div, train passing
b remote-track voltage: 4 V/div, 5 ms/div, train passing
ensure that all joints and connections are sound and secure. Joint resistances should
be measured and proven to be of negligible value. In particular, earthing and bonding
connections to plant and earth mats should be checked and neutral to earth connections
verified. If great reliance is placed on extended earthing systems or interconnection
with other substations, such connections should also be verified. The integrity test is
typically carried out using a variable voltage source capable of delivering currents
up to 300 A[1]. Various methods [1, 109] for evaluating the integrity of the earthing
system include:
radio frequency injection to establish conductor routes or breaks in conductors
comparing voltage drops across different parts of a grid
comparing current sharing through grid
comparing earth impedances measured from different points on the grid
point-to-point continuity measurements
joint resistance measurements.
It should be appreciated that where an earth grid is defective, measurements may be
hazardous.
Earthing 379
100 10 1000
test frequency, Hz
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
a
p
p
a
r
e
n
t

i
m
p
e
d
a
n
c
e

m
a
g
n
i
t
u
d
e

,

0.00
1.00
Figure 8.16 Earth impedance magnitude against frequency measured at a 275 kV
substation [94]
8.4.6 Special installations
A number of additional considerations for earthing design apply to the following
special installations.
8.4.6.1 GIS substation
Due to the much smaller area occupied by GIS installations and the fast transients
that are generated under switching operations, special measures are necessary for the
design of earthing systems for such installations [30].
8.4.6.2 Earthing systems for electric railways
In the UK, Engineering Recommendation P24 [110] specifies the earthing arrange-
ments for AC traction supplies from 25 kVto 132 kV. However, recent developments
in upgrading the electrified rail system in the UK have resulted in the siting of
400/25 kV substations in close proximity to the rail system. This means that detailed
evaluations are required in order to investigate the effect of faults; and in particular,
the effect of a fault on the 400 kV side on the rail earthing system. A recent system-
wide analysis of safety hazards arising fromearth faults on a railway supply systemis
reported by Natarajan et al. [111]. The safety threshold limits applicable to coupling
into telecommunication systems fromACelectric power andACtraction installations
is provided in ITU standard K.33 [112].
8.4.6.3 Cellular phone systems on transmission line towers
The increased use in recent years of high voltage towers for hosting GSM transmitter
stations has required special earthing arrangements at the tower bases. Such instal-
lations bring into close proximity the high voltage system supported by the tower
(e.g. 400 kV) and the lower voltage supply system of the mobile station (e.g. 11 kV).
Anewengineeringrecommendation(G78) [113] addresses the designof suchsystems.
8.4.6.4 Wind farms
Land-based wind farms are often located in remote elevated areas with high earth
resistivity. The substation that interconnects the wind farm to the main power system
is usually quite compact, and, therefore, the substation earth grid may not provide a
sufficiently low earth resistance. In such cases, the earthing system can be enhanced
by providing a series of buried earth interconnections between the wind turbines
extending over the whole area of the farm. This not only lowers the earth impedance
at the main substation but also at each individual turbine. Earthing systems may be
required at the base of each turbine to control step and touch voltages and for lightning
protection [114].
8.5 Electrocution hazards and safety issues
8.5.1 Step and touch potentials
8.5.1.1 Background
Once the earth impedance seen from the point of fault and the proportion of fault
current flowing through the earth path have been calculated, the rise of earth poten-
tial of the earthing system with respect to remote earth can be estimated. Potential
differences across points at locations in and around the installation can also be deter-
mined to establish whether a human body would suffer electrocution if it bridged these
points. Such potential differences are classified as touch, step and transfer potentials,
and examples of them are illustrated in Figure 8.17 [1]. With reference to this figure,
the earthing system and any metal structure connected to it will experience the full
rise of earth potential U
E
. The figure also illustrates the fall of potential on the ground
Earthing 381
reference earth
(in sufficient
distance)
U
E
U
S
T
U
ST
U
SS
U
TST
U
TSTE
E
1m
1m
1m
S1
S2
S3
E
E
without potential
grading with potential
grading
cable having a continuous metallic
sheath insulated throughout but
with both ends exposed; sheath
is connected to earth at the substation
E
S1, S2, S3
earth electrode
potential grading earth elec-
trodes (e.g. ring earth electrodes),
connected to the earth electrode E
U
E
earth potential rise
U
SS
source step voltage
U
ST
source voltage for touching
U
TST
transferred source voltage for
touching, if the sheath is not earthed
at the remote end
U
TSTE
transferred source voltage for
touching, if the sheath is earthed at
the remote end as well
earth surface potential
Figure 8.17 Illustration of earth surface potential profile at a substation under earth
fault conditions and resulting hazard voltages (reproduced from [1])
surface with distance from the earthing system, which tends to zero at some distant
point. As a result, a touch potential, U
ST
, is experienced if a person standing on
the ground also touches metalwork connected to the earthing system. Similarly, if a
person stands on the ground in a direction such as to bridge a potential difference,
they are said to experience a step voltage, U
SS
. Transferred potentials describe the
potential difference which could be bridged by human contact as a result of either:
a an insulated conductor connected to the earthing systemunder fault extending out
into a region of zero or low potential (U
T ST
or U
T ST E
)
b an insulated conductor connected at a remote point of low potential reaching into
the substation area of high potential but not bonded to the earthing system.
The magnitudes of these potentials at a given installation and for a fault current
magnitude will depend on the location of the person and what they are touching.
When making an assessment of safety at an installation, it is not practical to consider
every prospective step and touch situation, and, therefore, normally an assessment is
made of the maximum touch and step voltages that occur.
Step voltages are normally less hazardous than touch voltages for two reasons:
(i) the human body can tolerate higher voltages across the foot-to-foot current path
(step) compared to the hand-to-feet path (touch). In particular, less current flows
through the path of the heart for the foot-to-foot electrocution scenario, and this
is expressed in terms of a heart current factor; also, if additional resistances are
taken into account (e.g. footwear, surface chippings), the current is limited to a
greater extent by the series connection of these additional resistances [115]
(ii) for any given position, the step voltage is lower than the prospective touch
voltage.
This generality can lead to the step voltage being neglected as a hazard in its own
right. For example, EA Technical Specification 41-24 [2] states, if the earthing
system is safe against touch potential it will be inherently safe against step potential.
However, IEEE Std.80 [30] recognises that even if safety is achieved within the
substation through the control of potential gradients, it may still be possible for step
voltage hazards to be present outside the substation. Therefore, it is recommended
that step voltages are also computed and compared. IEEE Std.80 (section 19.2) states
that the best assurance that a substation is safe would come from actual field tests of
step and touch voltages. However, it is noted because of the expense, few utilities
are likely to make these tests as a routine practice.
8.5.1.2 Establishing the locations of maximum values of touch and step
voltage
The maximum surface potential gradients around an earthing grid normally occur
along a diagonal line from the centre extending through a corner point of the earth
grid. Inside the grid, potential minima will occur at the surface of the ground near
the central point of each mesh as shown in Figure 8.18 for a four-mesh grid. The
maximum prospective touch voltages within the area of the grid will correspond to a
person standing centrally inside a mesh while touching earthed metalwork, and the
location of the highest touch voltage for a symmetrical grid corresponds to the centre
of the corner mesh. The touch voltage at this position is referred to as the mesh voltage.
IEEE Std.80 notes that the actual worst-case mesh voltage occurs slightly off-centre
towards the corner of the grid. Using the IEEE Std.80 standard, the required mesh
spacing for the grid can be determined to ensure that the mesh voltage is below the
tolerable touch voltage limit. Earthing design standards in the UK follow a different
approach where particular attention is given to the touch voltage established when a
person stands one metre diagonally out from the corner of the grid [2, 52] while no
specific calculation of the mesh voltage is made. The mesh voltage may be greater
or less than the corner touch voltage depending on the mesh density and whether
earth rods are used at the periphery of the grid [30]. However, provided no earthed
metalwork can be touched while standing outside the grid, i.e. if the grid fully covers
the area occupied by plant, then the corner touch voltage hazard should not arise.
Accordingly, it should only be necessary to calculate the mesh voltage. The location
of the worst-case step voltage is accepted by both UK and US standards to be the
potential difference across the ground surface one metre diagonally out from the
corner of the grid.
8.5.1.3 Magnitude of touch, mesh and step voltages
Until quite recently, touch, mesh and step voltages were calculated using approxi-
mate analytical expressions, which incorporated geometrical and empirical correction
Earthing 383
60.96 m
6
0
.
9
6

m
legend
grid conductor
ground rod
diagonal voltage
profile
p
1
=300m
p
2
=100m
H=4.572m
D=15.24m
d=0.01m
h=0.5m
rod length = 9.144m
corner
mesh
inner
mesh
40
100.0 85.7 71.4 57.1 42.9 28.6 14.3 .7E 08
60
80
p
o
t
e
n
t
i
a
l

%
100
120
X
Y
potential
GPR
100.0 85.7 71.4 57.1 42.9 28.6 14.3 .7E 08
Figure 8.18 Diagonal voltage profile for a 16-mesh square grid with nine rods in
a two layer soil (from IEEE Std.81 1983, IEEE Copyright 1983 and
IEEE Std.80 2000. All rights reserved [30])
factors. Examples of such expressions are available in a number of earthing design
standards [2, 30, 52]. More recently, computer programs have become available for
computing earth grid resistance and surface potentials around the grid [62, 116, 117].
The most recent edition of IEEE Std.80 [30] outlines the reasons to justifying the use
of such programs in carrying out earth design including:
a a complex earth structure requiring multilayer representation
b non-uniform grid geometry and the presence of other buried conductors or
metalwork
c a more specific examination of step and touch danger points.
UK standards currently do not take account of computer-based earthing design
programs. As an alternative to carrying out detailed estimations of prospective step
10000
5000
2000
1000
500
d
u
r
a
t
i
o
n

o
f

c
u
u
r
e
n
t

f
l
o
w

t
,
m
s
200
100
50
20
0.1 0.2 0.5 1 2 5 10 20 50 100 200 500 1000 2000 5000 10 000
10
a b
c
1
c
2
c
3
AC-4.1
AC-4.2
AC-4.3
AC-1 AC-2 AC-3 AC-4
70kg body
50kg body
Body current, I
b
Figure 8.19 Tolerable body current as a function of shock duration comparison
of IEC limits (curved lines) with IEEE Std.80 limits (straight lines)
[115, 30]
and touch voltage, the CENELEC document also allows for the estimation of touch
voltages by direct measurement [1].
8.5.2 Computation of tolerable voltages
Once the prospective hazardous voltages arising under earth fault conditions have
been computed for a given design of earthing system, it is necessary to determine
whether or not these exceed recognised safety thresholds known as tolerable voltages.
Ultimately, it is the level of current through the human body that determines whether
there is a risk of electrocution, and well defined limits apply. However, the tolerable
voltage for a particular shock scenario depends not only on the level of tolerable
current but also on the values of resistance that are present in addition to the human
body resistance. Accordingly, a suitable accidental circuit is also required in order to
obtain values of tolerable voltages.
8.5.2.1 Tolerable body currents
Electricity flowing through the human body can be perceived and can cause muscle
contraction for currents of a few tens of milliamperes. However, when the current
reaches a magnitude of the order of 50 mA, there is a risk of ventricular fibrillation. It is
Earthing 385
the fibrillation limit that forms the basis of most recommended earthing safety calcu-
lations [88]. The fibrillation threshold is not only dependent on current magnitude but
also on the current time duration, current path and frequency. Figure 8.19, reproduced
from IEC479 and based on work carried out by Biegelmeier and Lee [118], shows
the tolerable r.m.s. human body current at power frequency for a range of times for
the hand-to-feet body path. The variation in fibrillation threshold for different body
current paths is described by the heart current factor [115] and it is interesting to note
that the hand-to-hand shock scenario is less onerous than hand-to-foot. The different
curves shown in Figure 8.19 correspond to the probability of ventricular fibrillation
occurring. For currenttime values to the left of curve c1, the risk of ventricular fib-
rillation is negligible, and curves c2 and c3 delineate the 5 per cent and 50 per cent
probability boundaries. Some earthing standards [1, 52, 119] adopt curve c2, although
UK industry earthing standard EA TS 41-24 [2] uses curve c1 as the basis of safety
calculations. The American earthing standard IEEE Std.80 [30], on the other hand,
adopts a different specification for tolerable body current based on the findings of
Dalziel [120, 121]. In this case, an empirical formula is used to describe the tolerable
current I
b
, which applies to 99.5 per cent of a population; this is plotted in Figure 8.19
for comparison:
I
b
=
k
t
s
(8.7)
where t
s
is the shock duration and k is a constant related to body weight.
For a body weight of 50 kg and 70 kg, k = 0.116 and 0.157, respectively.
In IEEE Std.80, it is suggested that the 70 kg figure may be applied if this is the
average population weight within the area of the electrical installation. However, it
is considered that the 50 kg figure is more suitable when applied to areas outside
controlled environments (e.g. outside a substation fence).
Whichever current curve is used, the time duration is related to the speed of the
circuit protection. The estimation of shock duration may require careful judgement on
systems that employ fast acting main but slower backup protection and also circuits
using fast autoreclosure.
8.5.2.2 Accidental earth circuit
Once a suitable permissible body current has been selected, the permissible touch
voltage can be determined from the accidental shock circuit (Figure 8.20) applicable
to either the step or touch voltage scenarios. This circuit comprises: (i) the body
impedance Z
b
(a resistance is sufficient at power frequency), (ii) additional resistances
such as footwear resistance, and (iii) the resistance to earth of the standing point or
the feet earth resistance. With reference to the circuit of Figure 8.20, it is important
to distinguish between the touch voltage (U
T
) and the source voltage for touching
or the prospective touch voltage (U
ST
). U
T
is the part of the earth potential rise
due to the earth fault which can be picked up by a person, assuming the current is
flowing via the human body from hand to feet, and U
ST
is the voltage which appears
during an earth fault between the contact point and earth when these parts are not
Z
B
U
Tp
I
b
U
STp
R
a1
R
a2
R
a
Figure 8.20 Simple equivalent circuit for the touching shock scenario [1]
U
STp
voltage difference acting as a source in the touching circuit
with a limited value that guarantees the safety of
a person when using additional known resistances
(e.g. footwear, standing surface insulating material)
Z
b
total body impedance
I
b
current flowing through the human body
U
Tp
permissible touch voltage, the voltage across the human body
R
a
additional resistance (R
a
= R
a1
+R
a2
)
R
a1
for example resistance of the footwear
R
a2
resistance to earth of the standing point
s
resistivity of the ground near the surface in an installation
(in m)
t
F
fault duration
being touched [1]. It is noticeable that there are considerable differences between
standards in the values adopted for each of the resistance components, which are
described in more detail in the following sections.
(i) Human body resistance
The resistance of the human body is assumed to be equal to a constant 1 k in both
UK and American standards [2, 52, 115]. However, the new CENELEC standard [1]
adopts a more precise voltage-dependent resistance model based on IEC data. The
characteristics of this model are shown in Figure 8.21, and it can be seen that the IEC
model yields higher values of resistance than the constant resistance model at lower
voltages [1, 115].
(ii) Footwear resistance
In IEEE Std.80 [30], footwear resistance is neglected. UK standards [2, 52], on the
other hand, assume 4 k per shoe and the ITU-T K33 standard allows for a footwear
resistance of between 250 and 3 M per shoe depending on the shoe type and
condition [122].
Earthing 387
5000
4000
3000
2000
1000
200 500 700 1000 2000 3000 4000
parameter,
percentage of population
95 %
50 %
5 %
b
o
d
y

i
m
p
e
d
a
n
c
e
,

6000
0
0 5000
touch voltage U
t
, V
Figure 8.21 Statistical values of human body impedance [1] (copyright 1994,
IEC, Geneva, Switzerland. www.iec.ch)
(iii) Resistance to earth of standing point
In a more accurate model, the resistance to earth of the standing point should be
considered as a component of a Thvenin equivalent resistance across the two points
that the human body bridges [30, 123, 124]. However, in most standards, a simplified
approach is adopted that assumes an additional resistance due only to the earth resis-
tance of the feet. The resistance of the human foot R
f
is calculated by representing
it as a metallic disc of equivalent area and is given by:
R
f
= 3 (8.8)
If the earth surface resistivity is artificially increased, the tolerable touch or step
voltage will be higher. Accordingly, it is standard practice in most outdoor substations
to add a thin surface layer of high resistivity material, such as gravel. A number of
different analytical techniques and computer models are available to account for this
effect, and it should be noted that the addition of a surface layer of different resistivity
will also alter the surface potential distribution [30].
8.5.2.3 Typical tolerable touch voltages
From the preceding sections, it can be seen that there are considerable differences
between standards in both the definition of tolerable body current and also the con-
figuration and parameter values of the accidental circuit. Tolerable touch voltages are
therefore very dependent on the model chosen and examples of ranges of tolerable
voltages are shown in Figure 8.22.
8.5.2.4 Transferred potential limits
A metallic circuit, connected to the main grid, and extending outside the area of the
substation, can transmit transferred potentials. The danger normally arises from a
(5)
(4)
(3)
(2)
(1)
2000
1000
500
200
100
0.1 0.2 0.5 1 2 5 0.05 10
t
f
, s
5000
50
U
S
T
P
,
V
Figure 8.22 Examples of tolerable touch voltage curves for different assumed
additional resistances in the accidental circuit [1]
(1) without additional resistances
(2) R
a
= 750 (R
a1
= 0 ,
s
= 500 m)
(3) R
a
= 1750 (R
a1
= 1000 ,
s
= 500 m)
(4) R
a
= 2500 (R
a1
= 1000 ,
s
= 1000 m)
(5) R
a
= 4000 (R
a1
= 1000 ,
s
= 2000 m)
R
a1
= 1000 represents an average value for old and wet shoes
remote touch voltage scenario. The transferred potential will usually be much higher
in magnitude than local step and touch voltages, and can even exceed the rise of
potential of the main substation [30].
In the UK, limits are applied to such transferred potentials. For these situations,
which assume direct contact between the full rise of earth potential and remote earth,
safe limits of 430 V and 650 V are designated depending on the reliability and pro-
tection speed of the circuits connected to the substation [2]. These limits derive from
a set of CCITT directives (Comitte Consultatif International Telephonique et Tele-
graphique) [68]. In these directives, the 430 V and 650 V limits apply to voltages
induced on aerial telephone lines and not directly to transfer voltages at substations.
In a different publication [125], it is mentioned that these limits were adopted as a
result of a compromise involving economic considerations, and it can be assumed
that they were not derived from human tolerable body current assumptions. If the
limit was required to avoid danger to a person, it is considered that it should be set
to as low as 60 V for the longer fault clearance times. The continued application of
the 430 V and 650 V limits over many years may be explained by the extremely low
probability of the transfer potential hazard. This may be an example of how, during a
periodwhendeterministic approaches were used, a probabilistic considerationentered
the specification by artificially raising the safety threshold. The recently published
Earthing 389
standard IEC 61936-1 states that It must be recognised that fault occurrence, fault
current magnitude and presence of human beings are probabilistic in nature [126].
Section 8.5.4 of this chapter considers in more detail howprobabilistic considerations
affect the risk management of earthing systems.
By contrast, the approach adopted in the American standard IEEE Std.80 is to
acknowledge that it is impractical or impossible to restrict the rise of earth potential
to prevent dangerous voltages of this kind, and therefore, either bonding or isolating
best avoids such hazards [30].
8.5.3 Methods for limiting hazardous potential differences and
dimensioning of earthing systems
The area occupied by the substation or installation initially sets the overall dimensions
of an earth grid, and normally all the available area is occupied in order to achieve the
lowest resistance. The grid mesh density should be sufficient to facilitate connection
to the substation plant by short earthing conductors or downleads.
If the basic design of an earth grid is insufficient to satisfy the adopted safety
criteria, a number of different measures can be taken to either reduce step and touch
potentials or restrict current flow through the accidental human body circuit. These
measures can be grouped into four categories.
8.5.3.1 Reducing earth grid current
Impedance-earthed systems, which are described in detail in section 8.2.3 of this
chapter, enable the single line to earth fault current to be restricted. On existing
solidly-earthed systems, however, it may not be possible to introduce impedance
earthing because of limitations in the insulation withstand capabilities of the system.
As an alternative, instead of reducing the total earth fault current, the earth return
component of the fault current can be reduced by providing an additional above
ground metallic return path. For example, on unearthed overhead lines, a continuous
earth wire could be installed. On circuits already equipped with an earth wire, a
second earth wire could be provided.
8.5.3.2 Reducing earth impedance
If all the available land area has been utilised, making direct connections to nearby
existing grids can reduce the overall earth impedance. If extended earthing systems
formed by tower lines or cables are available, normally these should also be con-
nected to the main earth grid. Often, such extended earthing systems have much
lower earth impedances than the grid to which they are connected. As already noted
in section 8.4.2.1, grid earth resistance is directly proportional to the earth resistivity.
Therefore, if an earth resistivity survey indicates that the grid is in relatively high
resistivity earth and there is a nearby low resistivity earth region, it can be advanta-
geous to connect to a satellite grid installed in the low resistivity earth area. It should
be noted that although grid extensions or connections to extended earthing systems
are beneficial by lowering grid resistance, the earth potential rise is exported over a
much wider and often uncontrolled area.
Instead of extending the earthing system outwards, sometimes it is beneficial to
install deep-driven earth rods or piles or drilled earth wells. Such deep-driven earths
are particularly useful for penetrating deeper earth layers of lowand stable resistivity.
Finally, additives or treatments are sometimes employed to lower the earth resistivity
in the immediate vicinity of the earth electrodes. However, these may be of limited
benefit for large area earth grids because they have little effect on the resistivity of
the deeper earth which influences the grid resistance.
8.5.3.3 Limiting potential differences
A solid plate covering the entire surface area would provide the ultimate protection
against step and touch voltage hazards within a substation, but such an installa-
tion would be impractical and uneconomic [127]. However, the following practical
methods for the limitation of potential gradients may be adopted. In the first place,
the overall grid mesh density can be increased to provide a better general equipotential
area within the installation. It was noted in section 8.5.1.2 that the mesh voltage in an
equally spaced grid increases towards the perimeter. Accordingly, the grid mesh den-
sity may be made denser at the grid periphery in order to equalise the mesh voltages.
A more complete equipotential zone could be established by separating the operat-
ing or fault current carrying earth grid from the protective earth grid, which can be
achieved by installing a fine mesh counterpoise mat near the ground surface.
At often-frequented places inside the substation, such as switch and isolator posi-
tions, small fine mesh wire mats can be buried just beneath the surface chippings.
Afurther area requiring special attention in terms of earthing is the substation fence.
In many cases, the fence is metallic and can be touched fromthe outside by the public.
Therefore, suitable measures to control potential gradients must be employed.
At installations which have a limited earthing area and which may carry high fault
current including lightning currents, such as communication masts and transmission
line towers, there may be significant step voltage hazards as well as touch voltage risk
at the base of such installations. The step voltage risk, specifically, may be controlled
by the use of potential ramps [127], which are illustrated in Figure 8.23. Ramp rings
1
2
p
o
t
e
n
t
i
a
l
,

%
potential
curves
100
50
0
area of
fine mesh
elevation
ramp area of
wide mesh
top surface
of ground
Figure 8.23 Potential distribution above an earth grid with a potential ramp
(curve 1) and without ramp (curve 2) (from IEEE Std.81 1983, IEEE
Copyright 1983 and IEEE Std.80 2000. All rights reserved [30])
Earthing 391
are also recommended in BS6651 for the protection of masts, towers and columns
using single and multiple earth electrode systems [128].
8.5.3.4 Increasing resistance/insulation in the accidental path
One of the simplest and most widely used methods for increasing the resistance in the
accidental path is to cover the entire surface of the substation with gravel [30]. This
type of covering provides a high resistance barrier even under wet conditions but in
practice the integrity of the layer is often lost due to lack of maintenance and erosion
due to excessive vehicle travel. As an alternative or in addition to the gravel layer,
a plastic sheet can be placed under the ground surface [129]. If there are particular
routes in frequent use within an installation, wooden passageways can also provide
very good insulation.
The previous methods are concerned with providing additional resistance between
the ground surface and the body in the accidental path. A further option for
touch voltage conditions would be to provide insulation to earthing downleads
and to insulate or provide isolation from all aboveground metal work that can be
touched from ground level. In this way, the touch voltage hazard is effectively
eliminated.
There are various methods for protecting against the danger of transferred poten-
tials, depending on the nature of the circuit involved in transmitting the potential.
Such circuits are commonly formed by communications circuits, traction rails, low
voltage neutral circuits, potable equipment, piping and metallic fences [30].
8.5.4 Risk management approach to earthing safety
To recall, EATS41-24 [2] states that if the potential rise of the earth electrode exceeds
the appropriate CCITT level (430 V or 650 V),. . .measures shall be taken to counter
the risk of danger. This standard also specifies limits for touch and step voltages
corresponding to probabilities of heart fibrillation based on IEC 60479-1 and certain
assumed values of resistance in the accidental circuit. However, experience has shown
that, in many situations, it is not reasonably practicable to satisfy these requirements,
particularly in respect of third parties. As already mentioned in section 8.5.3, the
minimum impedance of a substation earthing system is often limited by the available
land area and ground resistivity, and the minimum earth fault current is limited by
the electrical system configuration. In many cases, the minimum achievable rise of
earth potential may exceed several thousands of volts, and high potential contours
may extend considerably beyond the substation. Consequently, transferred potentials
affecting third parties in the vicinity of a substation may exceed the levels specified
in the standards, and the costs involved in mitigating these potentials could be very
substantial.
8.5.4.1 Allowable risk levels
In the UK, legislation that covers the duties of employers to their employees and third
parties is set out in The Health and Safety at Work etc. Act 1974 [130]. This act
requires that it shall be the duty of every employer to conduct his (her) undertaking
in such a way as to ensure, so far as is reasonably practicable, that persons. . .are not
thereby exposed to risks to their health and safety. More recently, the Management
of Health and Safety at Work Regulations 1999 [131], which were enacted as a result
of European Union Framework Directive 89/391/EEC, states that Every employer
shall make a suitable and sufficient assessment of. . .the risks to the health and safety
of persons not in his employment arising out of or in connection with the conduct by
him (her) of his (her) undertaking.
What is reasonably practicable has been established by a number of legal test cases
and more comprehensively by the Health & Safety Executive (HSE) [132], and this
concept is referred to as the as low as reasonably practicable (ALARP) principle.
An individual fatality risk of 1 in 10 million per person per year is considered broadly
acceptable for members of the public. Between 1 in 1 million and 1 in 10 000 per
person per year, the ALARPprinciple must be applied; which means that the required
expenditure to reduce risk is dependent on the individual risk level, i.e. the higher the
risk, the higher the justifiable spend. The HSEhas attempted to reflect the price people
are prepared to pay to secure a certain averaged risk reduction and advise a benchmark
value for preventing a fatality of 1 m (2001 prices). This value corresponds to a
reduction in risk of 1 in 10
5
being worth about 10 to the average individual [133].
The Electricity Industry in the UK, through the offices of the Energy Networks
Association, is currently investigating applying risk management techniques to haz-
ards resulting from earth potential rise. It is expected that this work will result in the
establishment of an industry-wide policy recommending the necessary action to be
taken in respect of third parties in the vicinity of electrical installations. This will
include a revision to Engineering Recommendation S.36 [134], which covers the
provision of telecommunications services near to substations. The risk management
approach starts by recognising the probabilistic nature of exposure to earth potential
rise. In simple terms, this means that account is taken of the likelihood of the occur-
rence of an earth fault at the same time as an individual being in a position which
bridges a dangerous potential.
8.5.4.2 Case study
The following example will serve to illustrate the main steps involved in such an
approach. A member of the public in his or her garden and using the garden tap
is exposed to a transferred touch potential of 1500 V for 200 ms, as a result of
an earth fault on the transmission system. Assuming standard values of resistance
in the accidental circuit (see section 8.5.2.2), a fatality would be likely from this
exposure as this voltage is clearly higher than the safe levels specified in the standards.
Accordingly, from a deterministic perspective, this hazard would require mitigation.
However, let us now consider the probability that the individual will experience this
fatal electric shock. This overall probability P can be determined as the product of
three separate probabilities:
P = P
F
P
FB
P
C
Earthing 393
where P
F
is the probabilityof anearthfault onthe power system, P
FB
is the probability
of heart fibrillation and P
C
is the probability of contact i.e., the probability that the
individual is in an accidental circuit.
P
F
canbe estimatedfromhistorical fault records. Assumingthe typical probability
of an earth fault occurring which results in a significant earth potential at a transmis-
sion substation is 0.2 per annum; i.e. one significant earth fault every five years on
average.
P
FB
is the likelihood that, if a person were exposed to the earth potential rise, heart
fibrillation would occur. Probabilities of heart fibrillation corresponding to current
magnitude and duration (in this case 200 ms) are given in IEC 60479-1 [115]. The
current magnitude can be estimated from the accidental touch circuit and, in this
example, a body resistance R
b
of 487 (hand-to-feet current path) will be assumed
to cover at least 95 per cent of the population. The additional circuit resistances will
depend upon the specific exposure scenario, which depends on what the individual is
wearing and doing at the time of exposure. Assumptions may be made to determine
the exposure scenario: for example, it may be reasonable to assume that a person
would be wearing footwear while outdoors. The ITU standard T K33 [135] specifies
a resistance of 3000 for damp elastomer soled shoes on loose soil and so this
is adopted in our example. All other circuit resistances and the source impedance
are considered to be negligible in this case. In many exposure scenarios there will
be an additional insulation in the accidental circuit such as where an individual is
using an insulated power tool. In this example, such insulation is neglected. The total
resistance of the accidental circuit is calculated to be 3650 , which would result
in a current of 0.41 A flowing through the body for a touch voltage of 1500 V. At
this current magnitude, IEC 60479-1 predicts a probability of fibrillation of less than
five per cent or 0.05.
Finally, it is necessary to estimate P
C
, which is the likely time the person is
present in the accidental circuit. This will again depend upon the activity undertaken
by the exposed individual; in particular the proportion of time in contact with the
circuit elements. Let us assume that in this example the person is in contact with
the garden tap, on average, for one minute per day. Each day, therefore, the proba-
bility of contact is 7 10
4
. The estimated annual probability that this person will
experience a fatal electric shock as a result of using a garden tap can be estimated as
the product of the probabilities P
F
, P
FB
and P
C
, viz., 0.2 0.05 7 10
4
=
7 10
6
or 1 in 142 000. This level of individual risk falls within the lower
ALARP region, where the cost of mitigation must be balanced against the risk.
If mitigation in this case were not prohibitively expensive, it may be considered
worthwhile.
It should be noted that the above example covers one exposure scenario only.
To cover the total individual risk all reasonably foreseeable scenarios should be
considered and all risks added together.
In many cases, it is expected that although the exposure voltages may exceed
the limits prescribed in the standards, the level of individual risk does not warrant
significant expenditure on mitigation. However, it is worth noting that, in general,
it is not prohibitively expensive to design an earthing system to control touch and
step voltages inside a substation to within the levels specified in the standards and,
consequently, this approach is normally adopted.
Were industry to apply expensive mitigation, the costs would ultimately be borne
by society as a whole through increased electricity prices. This is because this issue
affects the industry as a whole and, therefore, the effect of competition or regulation
would be minimal. But according to the HSE, this price may not be considered
by society as worth paying. So, by using risk assessment to determine necessary
expenditure, the industry spends responsibly and can also prioritise investment with
the end result being improved safety. It should, however, also be remembered that
the risks resulting from earth potential rise are borne predominately by people living
and working at or close to substations, i.e. they are unevenly distributed within the
general population and, consequently, the perceived levels of tolerability of such risks
tend to be lower.
8.6 Impulse performance of earthing systems
In the preceding sections, earthing system design and performance under power
frequency conditions has been examined in detail and a number of standards which
provide very detailed design guidelines were referred to. In this section, to begin
with, the standards have been examined in respect of their consideration of earth-
ing requirements under fast transients and a summary of their recommendations is
presented.
8.6.1 Standard guidelines for transient earthing
8.6.1.1 IEEE Std.80 and IEEEStd.142
IEEEStd.80(Guide for safety in substation grounding) does not provide detailedguid-
ance for designing earthing systems subjected to lightning surges but considers that
grounding systems designed according to power frequency principles will provide
a high degree of protection against steep wave front surges. . .. This is based on the
assumption that the human body can withstand higher currents for very short duration.
This standard also recommends that surge arresters should always be provided with
a reliable low resistance ground connection and have as short and direct a path to
the grounding system as practical.
Some utilities provide separate downlead connections for surge arresters and
others use the metallic mounting structure because it offers a lower impedance path
to the earthing system [30].
8.6.1.2 EA-TS 41-24
The UK earthing design standard EATS 41-24 [2] (Guidelines for the design, testing
and maintenance of main earthing systems in substations), recognises that equipment
such as surge arresters and CVTs are more likely to pass high frequency current
due to the low impedance they present to steep fronted surges and unless a low
Earthing 395
impedance earth connection is provided. . .the effectiveness of the surge arrester could
be impaired.
This standard also recommends that the connection from the equipment to
earth should be as short, and as free from changes in direction, as is practicable
and that the effectiveness of the arrester can be improved. . .by connecting. . .(it)
to a high frequency earth electrode in the immediate vicinity, for example an
earth rod.
8.6.1.3 CENELEC HD 637 S1
The CENELEC harmonisation document, HD 637 S1 [1], which is likely to form
the basis of a new IEC standard, suggests measures to reduce the amount of inter-
ference created when surges are dissipated to earth. These include minimising the
inductance of the current paths by the significant meshing of earth electrodes and
earthing conductors. Where high transient currents are more likely to occur, the
density of the earthing mat should be increased.
8.6.1.4 BS 6651
A quantitative value applicable to transient earthing appears in BS 6651 (Protection
of structures against lightning) [136]. This standard recommends that earth-
ing systems designed for lightning protection should have an earth resistance
of less than 10 . Many electricity companies for power system earthing sys-
tems have adopted this value. In Japan, the 10 resistance figure is the
target value for transmission line tower footing resistance [137]. Further work
is required to establish values based on more detailed models which include
inductance.
8.6.2 Soil ionisation
Many factors influence the behaviour of soil under high current magnitudes. First,
the characteristics of the medium: the soil type and resistivity, the moisture content,
temperature and pressure. Second, the nature of the energisation: the waveform or
impulse shape, its magnitude and polarity.
The reduction in the earth impedance of earth electrode systems under high magni-
tude impulse currents has been long recognised and this phenomenon is nowgenerally
accepted to be as a result of soil ionisation. Since the performance of earthing systems
under fast transients is vital to overvoltage protection, this area has attracted much
interest. Early work concentrated on experimental studies of earth electrode systems
experiencing soil ionisation effects. Recent work not only involved earth electrode
studies, but also the development of soil ionisation models and experimentation to
understand the soil ionisation process.
A review of work investigating soil ionisation effects in earthing systems is pre-
sented. Four main areas of investigation are considered: soil ionisation effects on
earth electrode systems, estimation of critical electric field intensity, soil ionisation
models, and analysis of soil ionisation mechanisms.
30 30
25
20
15
10
5
0
25
20
15
10
5
0
c
u
r
r
e
n
t
,

A
v
o
l
t
a
g
e
,

k
V
r
e
s
i
s
t
a
n
c
e
,
0 25 50 75 100 125 150 175 200

time, s
R
t
R(t) =v (t)/i (t)
i
peak2
i
peak1
v (t)
i (t)
0
3000
2500
2000
1500
1000
500
Figure 8.24 Typical voltage and current records in the ionisation region, and the
ratio R(t ) derived from v(t )/i(t ) (water content, wc = 3%, V
ch
=
27 kV [154])
8.6.2.1 Response of earth electrode systems to high impulse currents
(i) Early work
The earlyexperimental workonthe characterisationof earthelectrodes under transient
conditions [138145] demonstrated that:
a transient impedance, defined as the ratio between instantaneous voltage and cur-
rent, was shown to decrease from an initial impedance, during the fast front of
the surge, to a minimum corresponding to maximum ionisation which is as little
as one third of the power frequency value and, subsequently, recovering to the
leakage resistance value at the end of the impulse tail (Figure 8.24)
b soil ionisation occurs above a certain current threshold, which can be related to
the electric field intensity in the soil
c the instantaneous voltagecurrent curves were shown to form the shape of a
hysteresis loop
d spiked electrodes could enhance the soil ionisation effect
e time delays in ionisation and deionisation required a dynamic modelling
approach.
(ii) Recent work
Work continued with investigations by Kosztaluk et al. [146] on tower footings and
further evidence of the complicated dynamics of the soil ionisation process was pro-
vided by Oettl and Geldenhuys [147]. They described the corona ionisation time
constant, and found it to be inversely proportional to the current magnitude. They
considered that this was due to the formation of individual streamers in the soil.
Earthing 397
Tests were carried out [148] on a meshed earthing system of a 77 kV power
substation and also a steel-reinforced concrete pole. Lightning impulses with current
magnitudes of up to 40 kAwere applied, and it was found that the transient impedance
during soil ionisation, of both the mesh and the concrete pole, decreased with increas-
ing peak current magnitude. This effect was more pronounced for the smaller concrete
pole. Subsequently, tests on single rod electrodes, grids, horizontal electrodes and
tower footings [149] showed that no further reduction in resistance occurred at mag-
nitudes of current above 20 kA. It was also established that the reduction in resistance
was more marked in high resistivity soils and that the dependency is less, the more
extensive the earthing systems. Tests on a 500 kVtransmission tower base in medium
resistivity soil did not show the marked reduction in resistance found with rods and
crowsfeet [137].
8.6.2.2 Estimation of critical electric field intensity
A key parameter in representing the ionisation threshold level in soils is the critical
electric field intensity, E
C
. This critical electric field, E
C
, can be calculated from the
product of the measured soil resistivity and the current density at the electrode surface
assuming uniform current distribution. The value of current used corresponds to the
instant when the impulse resistance decreased belowthe lowvoltage resistance value.
Awide range of E
C
values (between 1.3 and 17.3 kV/cm), depending on soil type, test
electrode geometry and homogeneity of soil have been reported [140, 142145, 150].
In reviewing these experimental-based investigations, some consider that an aver-
age E
C
value of 3 kV/cm is realistic [151] and others [152] showed that plots of E
C
values against the corresponding soil resistivity were largely scattered which indi-
cated that there was no definite relationship between E
C
and soil resistivity. Mohamed
Nor et al. [153, 154] obtained an E
C
value of 5.5 kV/cm for sand which was found
to be independent of moisture content. Loboda and Scuka [155], based on their
own experimental studies, considered that streamers played an influential role in the
soil ionisation process and demonstrated that the calculated E
C
values could be up
to 70 per cent higher than those determined by assuming a uniform and perfectly
conducting ionised zone.
8.6.2.3 The mechanisms of soil ionisation
Asignificant amount of research work to understand the mechanisms of soil ionisation
has been carried out in the last few decades. Two contending theories have emerged
to explain the initiation of ionisation.
(i) Electric field enhancement
Between 1981 and 1993, many studies [156159] of electrical breakdown character-
istics of small-scale soil samples have been carried out. Emerging from these is the
proposition that ionisation occurs in soil air voids due to electric field enhancement.
Key points from these investigations are that;
a the ionisation process takes a finite time to initiate fromapplication of the impulse,
and delay times range from 10 to 100 s
b delay times exhibited a statistical variation, but average delay times and their
deviation decreased as the applied voltage was increased
c delay times decreased with water content and applied voltage
d breakdown thresholds with soil samples immersed in SF
6
were 2.5 times higher
than with air, which provided strong evidence in support of the theory.
(ii) Thermal process
In 1982, van Lint and Erler [160] investigated breakdown in earth samples under
high electric field stress using coaxial test rigs. Their results confirmed that initia-
tion delay times of ionisation decreased with increasing voltage. However, unlike the
results obtained by Flanagan et al. [157], initiation delay times of up to 1500 s were
reported. It was suggested that the field enhancement model could not explain such
long delay times. Instead, it was proposed that the negative temperature coefficient
of ionic solutions would produce a concentration of current flow in narrow parallel
water filaments. Further heating of these channels would result in vaporisation and
eventually breakdown along the filament path. Further investigations into the pro-
posed thermal mechanisms of soil breakdown [161164] suggested that neither the
thermal model nor the field enhancement model alone could account for the wide
ranging behaviour observed in soil conduction under high voltage. Srisakot et al.
[159], based on fast impulse experimental studies on sand samples, proposed that
two distinct conduction phases could be identified; the first associated with thermal
effects and the second due to soil ionisation.
8.6.3 Models of concentrated earth electrodes exhibiting soil ionisation
An example of a finite-element (FE) modelling approach was used by Nekhoul et al.
[165] to describe the behaviour of an earth rod exhibiting ionisation effects. Srisakot
et al. [159] carried out numerical electric field modelling of a simple hemispherical
earth electrode system using FE and boundary element BE software. Sekioka et al.
[166] developed linear (RL) and non-linear (RL) models to describe the earth
impedance of an earth grid, and these are compared with the simple single resis-
tance model. It is demonstrated that the inductance effects of a grounding grid should
be included to achieve an accurate lightning surge analysis and that the non-linear
model should be used for high magnitude, steep fronted impulses. It was also demon-
strated experimentally that reinforced concrete poles provide effective conduction
paths under impulse conditions [167]. A non-linear resistance can model such con-
centrated electrodes, and it is suggested that an energy-dependent resistance model
is more accurate than a current-dependent resistance model [168].
8.6.4 Models of earthing systems under high frequency and transient
conditions
Three different types of model have been applied in the analysis of high frequency and
transient performance of earthingsystems. These correspondtoincreasingcomplexity
according to the type of earthing system under study [169].
Earthing 399
R c
r L
Figure 8.25 Equivalent circuit model for a vertical earth electrode
Table 8.1 Formulae for the earth resistance [29],
inductance [170] and capacitance [35] of
a vertical earth electrode
R =

2
_
ln
2
a
1
_
(8.9)
L =

2
_
ln
2
a
1
_
(8.10)
C =
2
0
(ln(2l/a) 1)
(8.11)
8.6.4.1 Transmission line per unit length series impedance and shunt
admittance
This model, either in lumped parameter or distributed parameter form, can be used
to describe single horizontal and vertical earth conductors (Figure 8.25). The per unit
length impedance and admittance quantities can be computed using expressions
for earth resistance, shunt capacitance and conductor self-inductance. Approximate
formulae applicable to a vertical electrode of length l and radius a are shown in
Table 8.1.
Even at high frequency, the series resistance r of the earth conductor can be
neglected. The corresponding per unit length impedance and admittance can be used
to obtain the input impedance Z
oc
of the earth electrode:
Z
oc
=
_
Z
C
Y
C
coth(l
_
Z
C
Y
C
) (8.12)
where Z
C
is the per unit length series impedance, Y
C
the per unit length shunt
admittance and l the length of the electrode.
8.6.4.2 Network analysis of transmission line segments
For extended earthing systems that do not exhibit complex coupling of the inground
electrode, e.g. transmission tower line earthing systems, it is possible to build-up
circuit models from the transmission line components in section 8.6.4.1 using circuit
analysis techniques. In such models, the effects of soil ionisation have also been
included [171, 172].
8.6.4.3 Electromagnetic model
The electromagnetic model, first developed to analyse ground losses from antennas
due to the finite conductivity of earth, is based on the method of moments. It was fur-
ther developedandappliedtoanalyse groundingsystems [173, 174]. Suchanapproach
offers a more accurate theoretical treatment than the models in sections 8.6.4.1 and
8.6.4.2, but the model still requires a laterally homogeneous equivalent earth model.
Examples of software based on the electromagnetic field theory approach are found
in the literature [175177]. Olsen et al. [178] distinguish between the low frequency
or quasi-static model and an exact or full wave model valid at high frequency. They
point out that even low frequency models can be considered to be field theory models
in contrast to equivalent circuit models.
Acomparison of methods in sections 8.6.4.1 and 8.6.4.2 indicated that both meth-
ods were suitable for the simulation of simple structures such as horizontal electrodes
but that significant differences were apparent for large earthing grids [179]. The limits
of low frequency quasi-static models have been investigated and it has been shown
that if the buried electrode is greater than one-tenth of a wavelength in earth, then an
exact full wave model would be required [178]. Heimbach and Grcev [180] proposed
an alternative method of combining the frequency-dependent characteristics of an
earthing system in EMTP in order to evaluate overvoltages in power systems.
8.6.5 Simulations of earthing system performance under transient and
high frequency conditions
The performance of earthing systems under transient conditions is different from
power frequency behaviour because of the more significant influence of inductive and
capacitive effects. In contrast to the 50 Hz response, at high frequency the inductance
of a small earthing system, such as a rod, has a significant effect and the effective
length of such systems can be very small [175, 181185]. Inductive and capacitive
effects of rod earths are also evident under fast transients [186]. Figure 8.26 shows the
simulated voltage front time resulting from a 1/5 current impulse injected at the top
of a 5 m rod. For low values of resistivity, the inductance of the system has a greater
effect and a sharpening of the voltage impulse shape is seen. At high resistivity,
the magnitude of the displacement current becomes significant compared with the
conducted leakage current, and the effect of earth permittivity is influential on the
voltage rise time.
Figure 8.27 shows the frequency response of a 100 100 m
2
earth grid for a
range of soil resistivities. As can be observed, each curve has a lower frequency
range over which the impedance is nearly constant. For each earth resistivity value,
the impedance increases rapidly above a threshold frequency, and this behaviour can
be attributed to the inductance of the conductor. Transient simulations, of the same
grid in 100 m earth were carried out for a 1/5, 1 A current impulse. The results
Earthing 401
100 1000
c
r
=30
c
r
=10
c
r
=1
0.5
1
1.5
2
2.5
3
3.5
4
f
r
o
n
t

t
i
m
e
,

s
0
4.5
10 10000
resistivity, m
Figure 8.26 Voltage front time for a 1/5 current impulse on a 5 m rod, = 100 m
(reproduced from ERA report no. 2002-0113, published 2002)
1.E+02 1.E+03 1.E+04 1.E+05
j=10m
j=100km
j=10km
j=10km
1.E+01 1.E+06
frequency, Hz
0.1
1
10
i
m
p
e
d
a
n
c
e
,
0.01
100
Figure 8.27 Earth impedance of a four-mesh grid as a function of frequency for
different soil resistivities
(Figure 8.28) indicate significant attenuation of the surge as it propagates across the
grid and a slowing of the voltage rise time. The voltage drop across the down lead is
very high for the conditions considered [186].
Other investigations based on simulations using the electromagnetic
model support the principle of an effective area and confirm that increased grid
meshing in the vicinity of the injection point plays an important role in reducing the
peak voltage [187].
At low frequencies, the impedance of a grid is independent of the injection point
for the same earth conditions. However, above a critical frequency, the impedance
0
10
8
6
4
2
0
1 2 3 4 5
t, s
v
,

(
t
)
injection
origin
midpoint
corner
Figure 8.28 Voltage at different points on a grid for a 1/5 current impulse
(reproduced from ERA report no. 2002-0113, published 2002)
seen from the grid extremities compared with a central injection point increases
considerably [180].
8.7 References
1 HD637 S1: Power installations exceeding 1kVAC. European Committee for
Electrotechnical Standardisation (CENELEC), 1999
2 EA-TS 41-24: Guidelines for the design, installation, testing and maintenance
of main earthing systems in substations. Electricity Association, Technical
Specification, 1992
3 THEELECTRICITYCOUNCIL: Engineering recommendation S.5/1 earth-
ing installations in substations. Thirty-fourth chief engineers conference,
July 1966
4 SIMPSON, I.B.K., et al.: Computer analysis of impedance effects in large
grounding systems, IEEE Trans. Ind. Appl., 1987, 23, (3), pp. 490497
5 NATIONAL GRID: The seven year statement. Market Analysis and Design,
Commercial Systems and Strategy, Coventry, UK, 2000
6 NATIONAL GRID: FADS faults and defects database on the high volt-
age transmission system. Engineering and Technology, Leatherhead, UK,
2000
Earthing 403
7 JONES, P.: A guide for evaluating earth return currents at national grid
substations. Internal report, quality of supply, National Grid Company,
2000
8 NATIONAL GRID: Technical and operational characteristics of the NGC
transmission system. Issue 1.1, Coventry, UK, 2000
9 HMSO: Main prospectus for the sale of the regional electricity companies
share offers. November 1990
10 ACE REPORT NO. 51: Report on the application of engineering recom-
mendation P2/5 security of supply. British Electricity Boards (Electricity
Association), 1979
11 WESTINGHOUSE ELECTRIC CORPORATION: Electrical transmission
and distribution reference book (PA, 1950, 4th edn.)
12 B.E.I: Modern power station practice, vol. K(Pergamon Press, London, 1990,
3rd edn.)
13 SHREIR, L.L.: Corrosion, vol. 12 (G. Newnes, London, 1963)
14 BELL, H.: Protection of a.c. systems, J. IEE, 1960, 6, (70), pp. 571575
15 SAY, M.G.: Electrical earthing and accident prevention (G. Newnes, London,
1954)
16 WILLHEIM, R., and WATERS, M.: Neutral grounding in high voltage
transmission (Elesevier, New York, 1956)
17 BEEMAN, D.: Industrial power systems handbook(McGraw-Hill, NewYork,
1955)
18 VUKELJA, P., et al.: Experimental investigations of overvoltages in neu-
tral isolated networks, IEE Proc. C, Gener. Transm. Distrib., 1993, 140, (5),
pp. 343350
19 HANNINEN, S., and LEHTONEN, M.: Characteristics of earth faults in elec-
trical distribution networks with high impedance earthing, Electr. Power Syst.
Res., 1998, 44, (3), pp. 155161
20 SUGIMOTO, S., et al.: Thyristor controlled ground fault current limiting
systemfor ungrounded power distribution systems, IEEETrans. Power Deliv.,
1996, 11, (2), pp. 940945
21 IEEE Std.142-1991: IEEE recommended practice for grounding of industrial
and commercial power systems. IEEE Press, New York, 1991
22 THE ELECTRICITY SAFETY, QUALITY AND CONTINUITY REGULA-
TIONS, 2002 No. 2665
23 PETERSON, H.: Transients in power systems (Wiley, New York, 1951)
24 OPENSHAW TAYLOR, E., and BOAL, G.: Electric power distribution
415 V33 kV (Arnold, London, 1966)
25 KERSHAW, S.S., et al.: Applying metal-oxide surge arresters on distribution
26 SAY, M.: Electrical earthing and accident prevention (G. Newnes, London,
1954)
27 FRANKLIN, A.C., and FRANKLIN, D.P.: The J&P transformer book:
a practical technology of the power transformer (A.C. FRANKLIN &
D.P. FRANKLIN, 1983, 11th edn.)
28 SCHN, J.H.: Physical properties of rocks: fundamentals and principles of
petrophysics, in HELBIG, K., and TREITEL, S. (Eds): Handbook of geo-
physical exploration. Section 1, seismic exploration: vol. 18 (Elsevier Science
Ltd, Oxford, 1998)
29 SUNDE, E.D.: Earth conduction effects in transmission systems (Dover
Publications Inc., 1968)
30 IEEE Std.80-2000: IEEE guide for safety in AC substation grounding. The
Institute of Electrical and Electronic Engineers, New York, 2000
31 IEEE Std.81-1983: IEEE guide for measuring earth resistivity, ground
impedance, and earth surface potentials of a ground system. The Institute
of Electrical and Electronic Engineers, New York, 1983
32 BRITISH GEOLOGICAL SURVEY: URL:http://www.bgs.ac.uk [accessed
1 June 2002]
33 GRIFFITHS, D.H., and KING, R.F.: Applied geophysics for geologists and
engineers the elements of geophysical prospecting(Pergamon Press, Oxford,
1981, 2nd edn.)
34 BELL, F.G.: Ground engineers reference book (Butterworth, London, 1987)
35 TAGG, G.F.: Earth resistances (G. Newnes Ltd., England, 1964, 1st edn.)
36 THE ELECTRICITYASSOCIATION: Engineering recommendation S.34 a
guide for assessing the rise of earth potential at substation sites. The Electricity
Association, 1986, p. 23
37 WENNER, F.: Amethod for measuring earth resistivity. Bureau of Standards
scientific paper, 1915, no. 258 (Frank Wenner was a physicist at the American
Bureau of Standards)
38 VAN NOSTRAND, R.G., and COOK, K.L.: Interpretation of resistivity
data. Geological Survey professional paper 499, US Dept. of the Interior,
Washington, 1966
39 AVO International Ground/Earth resistance products [www]
URL:http://www.avo.co.uk/products/groundtster/ground.html [accessed
24 April 2002]
40 MA, J., and DAWALIBI, F.P.: Study of influence of buried metallic structures
on soil resistivity measurements, IEEE Trans. Power Deliv., 1998, 13, (2),
pp. 356365
41 BHATTACHARYA, P.K., and PATRA, H.P.: Direct current geoelectric
sounding part 9, principles and interpretation (Elsevier, London, 1968)
42 McNeill, J.D.: Applications of transient electromagnetic techniques. Techni-
cal note TN-7, Geonics Limited, 1980
43 THE BRITISH ELECTRICAL AND ALLIED INDUSTRIES RESEARCH
ASSOCIATION: Section M: communication interference notes on electri-
cal resistivity maps (Refs. M/T32 and M/T33). Reference M/T35, November
1934
ASSOCIATION: Electrical resistivity of England and Wales apparent resis-
tivity at a depth of 500 feet. Ref. D.263, February 1963 (Reproduced from
ERAReport M/T32, 1934)
Earthing 405
45 INSTITUTE OF GEOLOGICAL SCIENCES: Geological map of the UK
(South). 3rd edition solid, 1979
ASSOCIATION: Electrical resistivity map of Southern Scotland. (Repro-
duced from ERAreport M/T33, 1934)
47 CLAXTON, J.: Investigation into the performance of electrical earthing
systems. PhD thesis, Cardiff University, UK, 1997
48 HABBERJAM, G.M.: Apparent resistivity observations and the use of square
array techniques (Gebrder Borntraeger, Berlin, 1979)
49 BRITISH ELECTRICITY INTERNATIONAL: Modern power station
practice (Pergamon, London, 1992, 3rd edn.)
50 THAPAR, B., and GROSS, E.T.B.: Grounding grids for high voltage systems,
IEEE Trans. PAS, 1963, 82, pp. 782788
51 ENDRENYI, J.: Evaluation of resistivity tests for design of station grounds in
non-uniform soil, AIEE Trans., 1963, 82, pp. 782788
52 BS7354: 1990: Code of practice for the design of high-voltage open terminal
stations. British Standards Institution, 1990
53 MA, J., et al.: On the equivalence of uniform and two-layer soils to multilayer
soils in the analysis of grounding systems, IEE Proc., Gener. Transm. Distrib.,
1996, 143, (1)
54 TAYLOR, M.J.: Soil resistivity assessment for electrical earthing applica-
tions. PhD thesis, Cardiff University, UK, 2001
55 DAWALIBI, F., and BARBEITO, N.: Measurements and computations of the
performance of grounding systems buried in multilayer soils, IEEE Trans.
Power Deliv., 1991, 6, (4), pp. 14831487
56 LOKE, M.H.: Electrical imaging surveys of environmental and engineer-
ing studies. 1999, [www] URL:http://www.abem.com [accessed 27 April
2002]
57 DWIGHT, H.D.: Calculation of resistances to ground, Electr. Eng., 1936, 55,
pp. 13191328
58 SCHWARZ, S.J.: Analytical expressions for resistance of grounding systems,
IEEE Trans. Power Appar. Syst., 1954, 73, (13), pp. 10111016
59 NAHMAN, J., and SALAMON, D.: Analytical expressions for the resistance
of grounding grids in nonuniformsoils, IEEETrans. Power Appar. Syst., 1984,
103, (4), pp. 880885
60 SULLIVAN, J.A.: Alternative earthing calculations for grids and rods, IEE
Proc., Gener. Transm. Distrib., 1998, 145, (3), pp. 270280
61 CHOW, Y.L., ELSHERBINY, M.M., and SALAMA, M.M.A.: Resistance
formulas of grounding systems in two-layer earth, IEEE Trans. Power Deliv.,
1996, 11, (3), pp. 13301336
62 EPRI TR-100622: Substation grounding programs. Vol. 15, Electric Power
Research Institute, Palo Alto, US, 1992
63 XISHAN, W., et al.: Computation and experimental results of the grounding
model of Three Gorges power plant. 11th international symposium on High
voltage engineering, London, UK, 1999, p. 4
64 MA, J., and DAWALIBI, F.P.: Analysis of grounding systems in soils with
finite volumes of different resistivities, IEEE Trans. Power Deliv., 2002, 17,
(2), pp. 596602
65 COLOMINAS, I., et al.: A numerical formulation for grounding analysis in
stratified soils, IEEE Trans. Power Deliv., 2002, 17, (2), pp. 587595
66 ENDRENY, J.: Analysis of transmission tower potential during ground fault,
IEEE Trans. PAS, 1967, 86, (10), pp. 12741283
67 MGHAIRBI, A., GRIFFITHS, H., and HADDAD, A.: The chain impedance of
shielded transmission lines. 36th universities Power engineering conference,
Swansea University, 2001, session 6C1, p. 5
68 CCITT(ITU), THE INTERNATIONAL TELEGRAPH AND TELEPHONE
CONSULTATIVE COMMITTEE: Directives concerning the protection of
telecommunications lines against harmful effects fromelectric power and elec-
trified railway lines. Volume IIX, The International Telecommunications
Union, Geneva, 1989
69 ROGERS, E.J., and WHITE, J.F.: Mutual coupling between finite lengths of
parallel or horizontal earth return conductors, IEEETrans. Power Deliv., 1989,
4, (1), pp. 103113
70 NAHMAN, J.: Earthing effects of coated underground cables with metallic
shields bonded to earth electrodes, IEE Proc., Gener. Transm. Distrib., 1997,
144, (1), pp. 2230
71 POPOVIC, L.M.: Practical methods for the analysis of earthing systems with
long external electrodes, IEE Proc. C, Gener. Transm. Distrib., 1993, 140,
(3), pp. 213220
72 NAHMAN J., and SALAMON, D.: Effects of the metal sheathed cables upon
the performances of the distribution substations grounding systems, IEEE
73 MELIOPOULOS, A., and MASSON, J.: Modelling and analysis of URDcable
74 IEEE/ANSI C37: IEEE standard rating structure for AC high-voltage circuit
breakers rated on a symmetrical current basis. The IEEEInc., November 1985,
1986 edn.
75 IEEE Std.242-1975: IEEE recommended practice for protection and
co-ordination of industrial and commercial power systems, buff book. IEEE
Press, New York, 1980
76 IEEE/ANSI, C37.12-1981: Guide to specifications for AChigh-voltage circuit
breakers rated on a symmetrical current basis and a total current basis.
Approved on August 1981, The IEEE Inc., November 1985
77 IEEE Std.399-1980: IEEE recommended practice for power system analysis,
brown book. IEEE Press, New York, 1980
78 IEC909: Short circuit calculation in three phase ac systems. IECInternational
Electrotechnical Commission publication, 1988
79 BRIDGER, B.: All amperes are not created equal: a comparison of current
of high voltage circuit breakers rated according to ANSI and IEC standards,
IEEE Trans. Ind. Appl., 1993, 29, (1), pp. 195201
Earthing 407
80 KNIGHT, G., and SIELING, H.: Comparison of ANSI and IEC 909 short-
circuit current calculation procedures, IEEE Trans. Ind. Appl., 1993, 29, (3),
pp. 625630
81 BERIZZI, A., MASSUCCO, S., SILVESTRI, A., and ZANINELLI, D.:
ANSI/IEEE and IEC standards for short-circuit current evaluation: method-
ologies, computed values and results. 6th international symposium on
Short-circuit currents in power systems, September 1994, Liege, Belgium,
pp. 1.1.11.1.8
82 DUNKI-JACOBS, J.R., LAM, B.P., and STRATFORD, R.P.: A compari-
son of ANSI-based and dynamically rigorous short-circuit current calculation
procedures, IEEE Trans. Ind. Appl., 1988, 24, (6), pp. 11801194
83 ROENNSPIESS, O.E., and EFTHYMIADIS, A.E.: A comparison of static
and dynamic short-circuit analysis procedures, IEEE Trans. Ind. Appl., 1990,
26, (3), pp. 463475
84 BERIZZI, A., MASSUCCO, S., SILVESTRI, A., and ZANINELLI, D.: Short-
circuit current calculation: a comparison between methods of IEC and ANSI
standards using dynamic simulation as reference, IEEE Trans. Ind. Appl.,
1994, 30, (4), pp. 10991106
85 PROFESSIONAL GROUP P9: IEE colloquium on fault level assessment
guessing with greater precision?. The IEE, London, 30 January 1996, digest
no.: 1996/016
86 ARRIILAGA, J., and ARNOLD, C.P.: Computer analysis of power systems
(J. Wiley and Sons, London, 1990)
87 VIEL, E., and GRIFFITHS, H.: Fault current distribution in HV
cable systems, IEE Proc., Gener. Transm. Distrib., 2000, 147, (4),
pp. 231238
88 VERMA, R., and MUKHEDKAR, D.: Ground fault current distribution
in substations, towers and ground wires, IEEE Trans. PAS, 1979, 98,
pp. 724730
89 NAHMAN, J. M.: Proximity effects on the ground fault current distribu-
tion within the earthing system formed by a substation and the associated
transmission lines, IEE Proc. C, Gener. Transm. Distrib., 1988, 135, (6),
pp. 497502
90 GUVEN, A.N., and SEBO, S.A.: Analysis of ground fault current distribution
along underground cables, IEEE Trans. PD, 1986, PWRD-1, (4), pp. 915
91 POPOVIC, L.M.: Practical method for evaluating ground fault current distri-
bution in station supplied by an inhomogeneous line, IEEE Trans. PD, 1997,
12, (2), pp. 722727
92 PETERS, O.S.: Bureau of Standards technologic paper 108, 1918
93 MORGAN, A., and TAYLOR, H.G.: Measurement of the resistance of earth
electrodes, World Power, 1934, XXI, (CXXI), pp. 2227, 7681, 130135
94 JONES, P.: Electrical measurement of large area substation earth grids.
PhD thesis, Cardiff University, UK, 2002
95 CURDTS, E.B.: Some of the fundamental aspects of ground resistance
measurements, AIEE Trans., 1958, 77, (1), pp. 760767
96 TAGG, G.F.: Measurement of earth-electrode resistance with particular refer-
ence to earth-electrode systems covering a large area, IEE Proc., 1964, 111,
(12), pp. 21182130
97 TAGG, G.F.: Measurement of the resistance of an earth-electrode system
covering a large area, IEE Proc., 1969, 116, (3), pp. 475479
98 TAGG, G.F.: Measurement of the resistance of physically large earth-electrode
systems, IEE Proc., 1970, 117, (11), pp. 21852189
99 DAWALIBI, F., and MUKHEDKAR, D.: Ground electrode resistance mea-
surements in non uniform soils, IEEE Trans. Power Appar. Syst., 1974, 93,
(1), pp. 109116
100 MA, J., et al.: Ground impedance measurement and interpretation in various
soil structures. IEEE Power Engineering Society winter meeting, 2000, 3,
pp. 20292034
101 MA, J., and DAWALIBI, F.P.: Influence of inductive coupling between leads
on ground impedance measurements using the fall-of-potential method, IEEE
102 HARID, N., MGHAIRBI, A., and GRIFFITHS, H., et al.: The effect of
AC mutual coupling on earth impedance measurement of shielded transmis-
sion lines. 37th universities Power engineering conference, Staffordshire
University, 2002, pp. 822826
103 CARSON, J.R.: Wave propagation in overhead wires with ground return, Bell
Syst. Tech. J., 1926, 5, (10), pp. 539554
104 VELAZQUEZ, R., REYNOLDS, P.H., and MUKHEDKAR, D.: Earth return
mutual coupling effects in ground resistance measurements, IEEE Trans.
Power Appar. Syst., 1983, 102, (6), pp. 18501857
105 IEEE Std.81.2-1991: IEEE guide for measurement of impedance and safety
characteristics of large, extended or interconnected grounding systems. The
Institute of Electrical and Electronic Engineers, NewYork, 1991, appendices A
and B, pp. 8194
106 FOSTER, R.M.: Mutual impedance of grounded wires lying on the surface of
the earth, Bell Syst. Tech. J., 1931, X, (3), pp. 408419
107 BOAVENTURS, W.C., et al.: Alternative approaches for testing and eval-
uating grounding systems of high voltage energised substations. 10th
international symposium on High voltage engineering, Montreal, Canada,
1997, p. 4
108 WANG, G., and ZHOU, W.: A frequency conversion measurement sys-
tem of grounding resistance. 11th international symposium on High voltage
engineering, London, UK, 1999, p. 4
109 GILL, A.S.: High-current method of testing ground grid integrity, NETA
WORLD, International Testing Association, 10, (2), 198889
110 THE ELECTRICITY ASSOCIATION: Engineering recommendation P.34
AC traction supplies to British Rail. The Electricity Association, 1984,
Addendum 1990, pp. 2932
111 NATARAJAN, R., et al.: Analysis of grounding systems for electric traction,
Earthing 409
112 International Telecommunication Union (ITU): Limits for people safety
relatedtocouplingintotelecommunications systemfroma.c. electric power and
a.c. electrified railway installations in fault conditions. Series K: Protection
against interference, 1996
113 THE ELECTRICITY ASSOCIATION: Engineering recommendation G78
recommendations for lowvoltage connections tomobile telephone base stations
with antennae on high voltage structures. London, 2003, Issue 1
114 IEC/TR 61400-24: Wind turbine generator systems lightning protection.
International Electrotechnical Commission IEC publication, 2002
115 IEC 479-1: Guide to effects of current on human beings and livestock, part 1
general aspects. International Electrotechnical Commission IEC publication,
1994
116 DAWALIBI, F., and MUKHEDKAR, D.: Optimum design of substa-
tion grounding in two-layer earth structure: part I analytical study,
part II comparison between theoretical and experimental results and
part III study of grounding grids performance and new electrodes con-
figurations, IEEE Trans. PAS, 1975, PAS-9, (2), pp. 252261, 262266,
267272
117 HEPPE, R.J.: Computation of potential at surface above an energised grid or
other electrode allowing for non-uniform current distribution, IEEE Trans.
PAS, 1979, 98, (6), pp. 19781989
118 BIEGELMEIER, G., and LEE, W.R.: New considerations on the threshold of
ventricular fibrillation for a.c. shocks at 5060 Hz, IEE Proc. A, Phys. Sci.
Meas. Instrum. Manage. Educ. Rev., 1980, 127, (2), pp. 103110
119 DINVDE0141:1989: Earthing systems for power installations with rated volt-
ages above 1 kV. Technical Help to Exporters Translations, British Standard
Institution, Milton Keynes, UK
120 DALZIEL, C.F.: Dangerous electric currents, AIEE Trans., 1946, 65,
pp. 579585, also discussion on pp. 11231124
121 DALZIEL, C.F., and LEE, W.R.: Re-evaluation of lethal electric currents,
IEEE Trans. Ind. Gen. Appl., 1968, 4, (5), pp. 467476
122 INTERNATIONAL TELECOMMUNICATIONS UNION RECOMMENDA-
TION K.33: Limits for people safety related to coupling into telecommunica-
tions system from a.c. electric power and a.c. electrified railway installations
in fault conditions, ITU 1997
123 MELIOPOULOS, A.P.S.: Power system grounding and transients an
introduction (Marcel and Dekker, 1988)
124 DAWALIBI, F.P., SOUTHEY, R.D., and BAISHIKI, R.S.: Validity of conven-
tional approaches for calculating body currents resulting from electric shocks,
125 KLEWE, H.R.J.: Interference between power systems and telecommunication
lines. Report M/T 126, Electrical Research Association (E.R.A.), London,
1958
126 IEC61936-1: Power installations exceeding 1 kVa.c. part 1: common rules.
International Electrotechnical Commission IEC publication, 2002
127 KOCH, W.: Grounding methods for high-voltage stations with grounded
neutrals, Elektrotechnische Z., 1950, 71, (4), pp. 8991 (appendix J, IEEE
Std.80 1986)
128 BS6651:1999: Code of practice for protection of structures against lightning
129 THAPAR, B., and ZIAD, A.A.: Increasing of ground resistance of human
foot in substation yards, IEEE Trans. Power Deliv., 1989, 4, (3),
pp. 16951699
130 HMSO: The Health and Safety at Work etc. Act 1974. ISBN0105437743, UK
131 HSC, Management of health and safety at work regulations 1999, ACOPS,
HMSO 2000, ISBN 0 7176 2488 9
132 HSE: The tolerability of risk from nuclear power stations. HMSO 1992
133 HSE: Reducing risks, protecting people, HSEs decision-making process.
HMSO 2001
134 ENGINEERING RECOMMENDATION S.36: Procedure to identify and
record hot substations. The Electricity Association, 1988
135 INTERNATIONAL TELECOMMUNICATIONS UNION RECOMMENDA-
TION K.33: Limits for people safety related to coupling into telecommunica-
tions system from a.c. electric power and a.c. electrified railway installations
in fault conditions. ITU 1997
136 BS6651: 1999: Code of practice for protection of structures against lightning.
British Standard 1999
137 TAKEUCHI, M., et al.: Impulse characteristics of a 500 kV transmission
tower footing base with various grounding electrodes. 24th international
conference on Lightning protection, ICLP-98, Birmingham, UK, 1998,
pp. 513517
138 TOWNE, H.M.: Impulse characteristics of driven grounds, General Electric
Review, 1928, 31, (11), pp. 605609
139 BEWLEY, L.V.: Theory and tests of the counterpoise, Electr. Eng., 1934, 53,
pp. 11631172
140 DAVIS, R., and JOHNSTON, J.E.M.: The surge characteristics of
tower and tower-footing impedances, Journal of the IEEE, 1941, 88,
pp. 453465
141 BELLASHI, P.L.: Impulse characteristics of driven grounds, AIEE Trans.,
1941, 60, pp. 123128
142 BELLASHI, P.L., et al.: Impulse and 60-cycle characteristics of driven
grounds, part II, AIEE Trans., 1942, 61, pp. 349363
143 BERGER, K.: The behaviour of earthconnections under highintensityimpulse
currents. CIGRE paper 215, 1946
144 PETROPOULOS, G.M.: The high-voltage characteristics of earth resis-
tances, Journal of the IEE, Part II, 1948, 95, pp. 5970
145 LIEW, A.C., and DARVENIZA, M.: Dynamic model of impulse characteris-
tics of concentrated earths, Proc. IEE, 1974, 121, (2), pp. 123135
146 KOSZTALUK, R., LOBODA, M., and MUKHEDKAR, D.: Experimental
study of transient ground impedances, IEEE Trans. Power Appar. Syst., 1981,
PAS-100, (11), pp. 46534656
Earthing 411
147 OETTL, E.E., and GELDENHUYS, H.J.: Results of impulse tests on
practical electrodes at the high-voltage laboratory of the national electrical
engineering research institute, Trans. South Afr. Inst. Electr. Eng., 1988,
pp. 7178
148 MORIMOTO, A., et al.: Development of weatherproof mobile impulse volt-
age generator and its application to experiments on non-linearity of grounding
resistance, Electr. Eng. Jpn., 1996, 117, (5), pp. 2233
149 SEKIOKA, S., et al.: Measurements of grounding resistances for high impulse
currents, IEE Proc., Gener. Transm. Distrib., 1998, 145, (6), pp. 693699
150 ARMSTRONG, H.R.: Grounding electrode characteristics from model tests,
IEEE Trans. Power Appar. Syst., 1953, 73, (3), pp. 13011306
151 MOUSA, A.M.: The soil ionisation gradient associated with discharge of high
currents into concentrated electrodes, IEEE Trans. Power Deliv., 1994, 9, (3),
pp. 16691677
152 OETTL, E.E.: A new general estimation curve for predicting the impulse
impedance of concentrated earth electrodes, IEEE Trans. Power Deliv., 1988,
3, (4), pp. 20202029
153 MOHAMED NOR, N., HADDAD, A., and GRIFFITHS, H.: Impulse
characteristics of low resistivity soils. 36th universities Power engineering
conference, Swansea University, 2001, Session 3C, p. 5
154 MOHAMED NOR, N., SRISAKOT, S., GRIFFITHS, H., and HADDAD, A.:
Characterisation of soil ionisation under fast impulses. Proceedings of the
25th international conference on Lightning protection, Rhodes, Greece, 1822
September 2000, pp. 417422
155 LOBODA, M., and SCUKA, V.: On the transient characteristics of electrical
discharges and ionisation processes in soil, Proceedings of the 23rd interna-
tional conference on Lightning protection, Firenze, Italy, 1996, pp. 539544
156 FLANAGAN, T.M., et al.: Electrical breakdown properties of soil, IEEE
Trans. Nucl. Sci., 1981, NS-28, (6), pp. 44324439
157 FLANAGAN, T.M., et al.: Electrical breakdown characteristics of soil, IEEE
Trans. Nucl. Sci., 1982, NS-29, (6), pp. 18871890
158 LEADON, R.E., et al.: Effect of ambient gas on arc initiation characteristics
in soil, IEEE Trans. Nucl. Sci., 1983, NS-30, (6), pp. 45724576
159 SRISAKOT, S., GRIFFITHS, H., and HADDAD, A.: Soil ionisa-
tion modelling under fast impulse. Proceedings of the 36th universities
Power engineering conference (UPEC), Swansea (UK), 1214 September
2001
160 VAN LINT, V.A.J., and ERLER, J.W.: Electric breakdown of earth in coaxial
geometry, IEEE Trans. Nucl. Sci., 1982, NS-29, (6), pp. 18911896
161 ERLER, J.W., and SNOWDEN, D.P.: High resolution studies of the elec-
trical breakdown of soil, IEEE Trans. Nucl. Sci., 1983, NS-30, (6),
pp. 45644567
162 SNOWDEN, D.P., and ERLER, J.W.,: Initiation of electrical breakdown
of soil by water vaporisation, IEEE Trans. Nucl. Sci., 1983, NS-30, (6),
pp. 45684571
163 SNOWDEN, D.P., andVANLINT, V.A.J.: Vaporisation and breakdown of thin
columns of water, IEEE Trans. Nucl. Sci., 1986, NS-33, (6), pp. 16751679
164 SNOWDEN, D.P., BEALE, E.S., andVANLINT, V.A.J.: The effect of gaseous
ambient on the initiation of breakdown in soil, IEEE Trans. Nucl. Sci., 1986,
NS-33, (6), pp. 16691674
165 NEKHOUL, B., et al.: Calculating the impedance of a grounding system.
Collection de notes internes de la Direction des Etudes et Recherches Matrial
lectrique, transport et distribution dnergie, Electricit de France, Report
96NR00048, 1996.
166 SEKIOKA, S., et al.: A time and current dependent grounding impedance
model of a grounding net. International conference on Electrical engineering
(ICEE98), 1, pp. 832835
167 SEKIOKA, S., et al.: Surge characteristics of grounding system(s) of power
distribution lines in Japan with reference to a reinforced concrete pole. Inter-
national conference on Grounding and earthing, Belo Horizonte Brazil, 1821
June 2000, pp. 7377
168 SEKIOKA, S., et al.: Development of a non-linear model of a concrete pole
grounding resistance. International conference on Power system transients,
Lisbon, 37 September 1995, pp. 463468
169 GRCEV, L., and ARNAUTOVSKI, V.: Comparison between simulation and
measurement of frequency dependent and transient characteristics of power
transmission line grounding. 24th international conference on Lightning
protection ICLP-98, Birmingham, UK, 1998, pp. 524529
170 ROSA, E.B.: The self and mutual inductances of linear conductors, Bulletin
of the Bureau of Standards, 1907, 4, (2), p. 301
171 MENTER, F., and GRCEV, L.: EMTP-based model for grounding system
analysis, IEEE Trans. Power Deliv., 1994, 9, pp. 18381849
172 CIDRAS, J.: Nodal frequency analysis of grounding systems considering the
soil ionisation effect, IEEE Trans. Power Deliv., 2000, 15, (1), pp. 103107
173 GRCEV, L.: Computer analysis of transient voltages in large grounding
systems, IEEE Trans. Deliv., 1996, 11, (2), pp. 815823
174 GRCEV, L., and DAWALIBI, F.: An electromagnetic model for transients in
grounding systems, IEEE Trans. Power Deliv., 1990, 5, (4), pp. 17731781
175 GRCEV, L., and GRCEVSKI, N.: Software techniques for interactive opti-
misation of complex grounding arrangements for protection against effects
of lightning. 24th international conference on Lightning protection ICLP-98,
Birmingham, UK, 1998, pp. 518523
176 DAWALIBI, F., and SELBY, A.: An electromagnetic model of energised
conductors, IEEE Trans. Power Deliv., 8, (3), pp. 12751284
177 ANDOLFATO, R., et al.: Ariel and grounding system analysis by the shift-
ing complex images method, IEEE Trans. power Deliv., 2000, 15, (3),
pp. 10011009
178 OLSEN, R.G., et al.: A comparison of exact and quasi-static methods for
evaluating grounding systems at high frequencies, IEEE Trans. Power Deliv.,
1996, 11, (2), pp. 10711081
Earthing 413
179 HEIMBACK, M., and GRCEV, L.D.: Simulation of grounding structures
within EMTP. 10th international symposium on High voltage engineering,
Montreal, Canada, 1997, p. 5
180 HEIMBACH, M., and GRCEV, L.D.: Grounding systemanalysis in transients
programs applying electromagnetic field approach, IEEETrans. Power Deliv.,
1997, 12, (1), pp. 186193
181 MAZZETTI, C., and VECA, G.M.: Impulse behaviour of ground electrodes,
IEEE Trans. Power Appar. Syst., 1983, PAS-102, (9), p. 3148
182 VISACRO, S., et al.: Evaluation of communication tower grounding
behaviour on the incidence of lightning. 24th international conference on
Lightning protection ICLP-98, Birmingham, UK, 1998, pp. 556561
183 GUPTA, B.R., and THAPAR, B.: Impulse impedance of grounding grids,
IEEE Trans. Power Appar. Syst., 1980, PAS-99, (6), p. 2357
184 DAVIES, A.M., GRIFFITHS, H., and CHARLTON, T.: High frequency per-
formance of a vertical earth rod. 24th international conference on Lightning
protection (ICLP), Birmingham, UK, 1998, pp. 536540
185 GRIFFITHS, H., and DAVIES, A.M.: Effective length of earth electrodes
under high frequency and transient conditions. 25th international conference
on Lightning protection, Rhodes, Greece, 2000, pp. 469473
186 GRIFFITHS, H., andHADDAD, A.: Earthingsystems under lightningimpulse
conditions, in ERA Technology report no. 2002-0113, Lightning protection
2002: standards and practices, pp. 12.112.11
187 GRCEV, L.D., andHEIMBACH, M.: Computer simulationof transient ground
potential rise in large earthing systems. 23rd international conference on
Lightning protection ICLP-96, Firenza, Italy, 1996, pp. 585590
Chapter 9
Circuit breakers and interruption
H.M. Ryan
9.1 Introduction
Much literature exists relating to the historical developments worldwide associated
with switchgear related technology [129]. Acircuit breaker (CB) can be described [3]
as a device used in an electrical network to ensure the uninterrupted flowof current in
that network under normal operating conditions, and to interrupt the flowof excessive
current in a faulty network. Under some circumstances, it may also be required to
interrupt load current and to perform opencloseopen sequences (autoreclosing) on
a fault on others. The successful achievement of these, and indeed all, duties relies on
extremely careful design, development and proving tests on any circuit breaker. Good
and reliable mechanical design is required [1] to meet the demands of opening and
closing the circuit breaker contacts, and an effective electrical design is also essential
to ensure that the CBcan deal with any of the electrical stresses encountered in service
[4, chap. 7].
During the opening and closing sequences, an electric arc occurs between the
contacts of a circuit breaker and advantage is taken of this electrical discharge to
assist in the circuit interruption process, as has been discussed in detail [4, chap. 7].
In an AC electrical network, the arc is tolerated in a controlled manner until a natural
current zero of the waveform occurs when the discharge is rapidly quenched to limit
the reaction of the system to the interruption. With asymmetrical waveforms and for
DC interruption, advantage is taken of the arc resistance for damping purposes or to
generate a controlled circuit instability to produce an artificial current zero.
As discussed elsewhere, the arc control demanded by such procedures may
require gas pressurisation and flow, which in turn, make additional demands upon
the circuit breaker mechanism [1, 3] (see also chapters 7, 8, 10, 11 of Reference 4).
The reader is directed specifically to G.R. Jones [4, chap. 7] for a fuller treatment
of important system-based effects together with the principles of current interrup-
tion in HV systems. However, Figures 9.19.3, and their respective subcaptions,
current
voltage
high current
phase
dielectric
recovery
period
current
zero
period
di
dt
dV
dt
voltage
V
V
o
V
o
L R
C
l
F
1
1
2f
n
R=0
R>0
V>V
o
cos ot
t
f
n
=(2LC)
1
a
b
~
Figure 9.1 Typical circuit breaker current and voltage waveforms (after Jones
[4, chap. 7]). Voltage transients produced by short circuit faults and
short line faults are of particular practical interest. Short circuit faults
occur close to CB b, producing the most onerous fault conditions. Short
line faults occur on transmission lines a few km from the CB c, and
constitute the most onerous transient recovery voltages, typically at
current zero (dV/dt) 1020kV/s. Note: as pointed out by Jones, the
situation is made more complicated in three-phase systems because
current zero occurs at different times in each phase, implying that the
fault is interrupted at different times leading to different voltage stresses
across the interrupter units in each phase
Circuit breakers and interruption 417
V
o
l
t
a
g
e

p
.
u
.
t
line side f
L
= (2L
T
C
T
)
1
source side f
S
= (2L
S
C
S
)
1
total
0
c
1.0
L
S
C
S
V
L
T
E
CB
C
T
a voltage and current waveforms during the current interruption
process for a simple symmetrical powerfrequency (50 Hz) current
wave; here, current reduces naturally to zero once every half
cycle, at which point interruption is sought (i.e. minimum natural
rate of di/dt); for conventional power systems with inherently
inductive circuits, the contact gap is less severely stressed
transiently during both the thermal and the dielectric recovery
phases
b short line fault onerous fault current
c short line fault onerous transient recovery voltages
provide a brief indication of some of the strategic considerations covered by Jones.
The above simplified description of circuit interruption processes [3], still serves to
illustrate the complexity of the interactions involved, which are determined, on the
one hand, by the nature of the arcing and arc quenching medium and, on the other, by
the electrical network demands. It is of strategic interest to note that since Garrards
excellent reviewin 1976 [7], there has been a major swing to SF
6
as the arc quenching
medium for modern transmission and distribution GIS (gas-insulated switchgear)
[3, 4, 8], combining the excellent insulating and arc interruption characteristics of
SF
6
gas.
In passing it should be noted that there is also significant current interest in the
development of GIL (gas-insulated lines) for long distance transmission circuits, i.e.
longer than 500 m. (To date experience has been mainly limited to transmission
a
b
c
d
e
current
interruption
asymmetric
waveform
generator
fault
current
DC
interruption
arc induced
instability
prospective
fault
current
current
limitation
(domestic CB)
major loop
minor loop
i
t
Figure 9.2 Circuit breaker current waveforms (after Jones [4, chap. 7])
a shows sinusoidal current waveform with its natural current zero
b shows sinusoidal current waveform superimposed on a steady
current to form an asymmetric wave with major and minor loops
which cause different CB stresses
c related condition to b which occurs in generator faults; corresponds
to the power frequency wave superimposed on an exponentially
decaying component (consequently, the current zero crossing may
be delayed for several half-cycles)
d shows a further situation relating to the interruption of DC faults
which is achieved by introducing an oscillatory current via arc
instability in the CB so forcing the current to pass through zero
eventually
e finally, at the lower domestic voltages, current limitation can be
conveniently induced, leading to an earlier and slower approach to
zero current than occurs naturally; with the additional benefit of
reducing the energy absorption demands made on the interrupter
module
over short distances, typically less than 500 m.) Two recent CIGRE technical papers
provide valuable technical overviews of typical GIL circuits [29].
At distribution voltage ratings, vacuumcircuit breakers are nowfinding increased
application [9, 10] and are challenging the new range of evolutionary commercial
SF
6
designs [4], utilising puffer, suction, self-pressurising and rotary arc interrupters.
Hybride CB designs are sometimes preferred at distribution ratings as illustrated in
dt
a
b
fail
clear
thermal mode limit
network
response
25 A/ms
(50kA
r.m.s.
)
dV=const
i
( p)
m
dt
P
1
P
2
P
1
>P
2
u/s electrode position
dV=const
n
dt
di
dt
log
dV
dt
log
dV
5kV/s
log(di/dt)
m=13
n=14.6
Figure 9.3 Typical circuit breaker characteristics (after Jones [4, chap. 7])
a thermal recovery characteristics, network response and interrupter
characteristic; this characteristic is inthe formof a critical boundary
separating fail and clear conditions on a rate of rise of recovery
voltage (dV/dt) and rate of decay of current (di/dt) diagram, typi-
cally, the boundary condition obeys the relationship dV/dt =
constant (di/dt)
n
with n = 14.6
b thermal response characteristics; effect of pressure and geome-
try on performance, thermal response characteristic may also be
improved by increasing the CB gas pressure, the nature of the gas
or the geometry of the interrupter
Figure 9.4. Here, vacuum interrupters are used, but the designers have opted for SF
6
gas to provide insulation surrounding the interrupter. Two further designs of vacuum
switchgear are illustrated in Figures 9.5ab and Figure 9.5c shows a sectional view
of a vacuum interrupter.
log (
dV
)
dt
log (V
max
)
log (V
max
)
log (V
max
)
c
d
e
fail
clear
25A/ms log(di/dt)
network
response
dielectric
mode limit
circuit
breaker
network
clear
clear
I
TH
1
I
DIE
I
TH
L
25A/s log(di/dt)
2 breaks
1 breaks
dielectric
thermal
log (di/dt)
fail
c dielectric recovery characteristic, here, for the dielectric regime,
the characteristic is represented by the critical boundary separat-
ing successful clearance and fail on a maximum restrike voltage
(V
MAX
) and rate of decay of current (di/dt ) diagram, the dielectric
recovery performance may be improved by increasing the number
of contact gaps (interrupter units) connected in series
d,e overall CB performance superposition of thermal and dielec-
tric, note that the overall limiting curves for the CB performance
are obtained by combining the thermal and dielectric recovery
characteristics
a
b
vacuum interrupter
magnetic actuator
SF
6
gas in sealed
aluminium tank
current transformer
elastomeric bushings
with moulded-in
capacitor voltage
sensors
vacuum
recloser
control unit
and battery
Figure 9.4 GVR vacuum recloser (courtesy Whip and Bourne, after Blower [9])
a sectional view of recloser
b recloser connected to an overhead line
a
b
Figure 9.5 Examples of vacuum switchgear
a type Hadrian 12 kV vacuum switchgear (after Stewart [10],
courtesy VATECH Reyrolle)
b type WS 33 kV vacuum switchgear (after Stewart [10], courtesy
ALSTOM T&D Distribution Switchgear Ltd)
9.2 Circuit interruption characteristics, arc control and extinction
The ability to interrupt electrical power circuits represents an essential function,
especially in situations of overloads or short circuits when, as a protective measure,
immediate interruption of the current flow must take place. In the early days of
current interruption, circuits were broken merely by the separation of contacts in air
followed by drawing the resulting electric arc out to such a length that it could no
moving contact
guide grading shield sputter shield
sputter shield
mounting
glassceramic
body
moving contact
stem
bellows moving contact fixed contact fixed contact
stem
c
c sectional view of vacuum interrupter (courtesy Vacuum Inter-
rupters Ltd) and Reference 4, chapter 3, also published earlier in
chapter 6, Bradwell (Ed.) Electrical insulation (Peter Peregrinus
Ltd, 1983)
longer be maintained [1, 8]. Such a means of interruption soon became inadequate as
the voltages and current capacity of power systems grew and it became necessary to
develop special switching devices called circuit breakers.
At the outset, it should be appreciated that within the power industry the general
term switchgear is widely used and this covers the combination of (i) switching
devices and (ii) measuring devices (which monitor the system, detect faults and, as
appropriate, initiate circuit breaker operation). High reliability is essential with any
circuit breaker design as the device has to protect costly power system networks.
In service, they can remain closed for long periods until called upon to open under
even the most onerous short circuit situations.
An essential requirement in any CB design concerns its in-service operating per-
formance reliability, the effectiveness of arc control and efficient arc quenching,
removal of power dissipated in the arc (via convection, conduction and radia-
tion) which is essential to ensure rapid voltage withstand capability after current
interruption [4, chap. 7]. Jones provides a useful description of the principles of
current interruption in HV systems when considering the various phases of current
interruption.
As highlighted in an earlier comprehensive book in this series, dealing solely with
power circuit breaker theory and design [1], circuit breakers have strategic roles to
perform. They are required to control electrical power networks by:
(i) switching circuits on
(ii) carrying load
(iii) switching circuits off under manual or automatic supervision.
Flurscheim [1], observed (appropriately) that the character of their duty is unusual
and they will normally be in a closed position carrying load or in the open position
providing electrical isolation. Typically, in a service life of up to 30 years, they
are called upon to change quickly from one condition to the other (possibly only
occasionally), and to perform their special function of closing onto a faulty circuit,
or interrupting short circuit current (sometimes only on very rare occasions). Hence,
the earlier comment, reliability is essential! The circuit breaker must be reliable in
static situations and also be effective instantaneously when called upon to perform
any switching operation even after long periods without movement [1]. All circuit
breakers have designated ratings, and they must satisfy rigorous standard IEC tests,
which are set to guarantee compliance in service (see IEC 62271-100 High voltage
alternating current circuit breakers [30]), covering a wide range of conditions.
Ali [4, chap. 8] informs his readers that a circuit breaker has to be capable of
successfully:
interrupting (i) any level of current passing through its contacts from a few
amperes to its full short circuit currents, both symmetrical and asymmetrical,
at voltages specified in IEC 62271-100, and (ii) up to 25 per cent of full short
circuit making currents at twice the phase voltage
closing up to full short circuit making current (i.e. 2.5 I
sym
) at phase voltage
and 25 per cent of full making currents at twice the phase voltage
switching (making or breaking) inductive, capacitive (line, cable or capacitor
bank) and reactor currents without producing excessive overvoltages to avoid
overstressing the dielectric withstand capabilities of a system
performing opening and closing operations whenever required
carrying the normal current assigned to it without overheating any joints or
contacts.
As pointed out recently byAli [4, chap. 8], interrupting devices become more complex
as the short circuit currents and voltages are increased and, at the same time, the fault
clearance times are reduced to maintain maximum stability of the system.
The major components of a circuit-breaker are:
interrupting medium, e.g. sulphur hexafluoride (SF
6
) gas
interrupter device (see Figure 9.19)
insulators
mechanism (importance of opening energy is shown in Figure 9.18b)
(Note: the writer found that regular participation in IEC/BSI standardisation com-
mittees, together with corresponding CIGRE study committee/colloquia activities,
provided him with first class opportunities for ongoing professional empowerment
and career development.)
The interruption of short circuits represents a severe duty imposed on circuit
breakers, and short circuit ratings have progressively risen during the past 80 years to
meet the increasing needs of power networks, as voltages have increased up to 420 kV
in the UK and up to 1000 kV worldwide. Modern transmission circuit breakers are
capable of interrupting faults of up to 60 kA, and for distribution systems (at voltages
in the range 1 to about 150 kV) CB ratings up to 25 kA are typical. Over the years,
the total break time of CBs, required to interrupt short circuits, has been significantly
reduced in the interests of system stability:
1020 cycles, for early plain break oil circuit breaker (required because of long
arc durations)
68 cycles, achievable by introduction of arc controlled interrupters
<2 cycles, now readily achievable with modern SF
6
CB designs.
Acircuit breaker is an electromechanical device which is used to control and protect
distribution and transmission networks (see chapters byAli, Jones, Fletcher and Pryor
in Reference 4). It can be called on to operate and to interrupt currents as low as 10 A
up to its full short circuit rating (i.e. up to 40/50 or even 60 kA) changing the circuit
breaker from a perfect conductor to a perfect insulator within only a few milliseconds
(see chapters by Jones and Ali in Reference 4). Great care is required to ensure
that appropriate dynamic operating mechanism characteristics (i.e. speed/travel) are
designed into any circuit breaker to ensure effective and safe opening and closing
of the interrupter throughout its life and for all operating conditions [14, 8]. As Ali
[4, chap. 8], has pointed out, the number of operating sequences and the consistency
of closing and opening characteristics generally determines the performance of the
mechanism (see Figure 9.18b). He further states that:
(i) although only 2000 satisfactory type-test operations are required to achieve
compliance with IEC62271-100, the current trend (or tendency) is to carry
out 5000 extended trouble-free operations tests to demonstrate compatibility of
mechanism/CB interface with the current second generation of SF
6
interrupters
(ii) the task with third-generation SF
6
circuit breakers becomes much easier as
these devices are fitted with low energy mechanisms and lightweight moving
parts, and can satisfactorily perform 10 000 trouble-free operations without any
stresses and excessive wear and tear on the moving and fixed parts of the circuit
breaker; detailed coverage of interrupter mechanisms is outside the scope of this
brief overview chapter but the reader is directed to a major body of available
resource material [e.g. see CIGRE publications http://www.cigre.org].
It is sufficient tocomment here that a circuit breaker mayhave either a single-phase
or a three-phase mechanism [1, 4, 8], depending on network operational require-
ments or client preference. Mechanisms used with modern interrupters, are generally
hydraulic, pneumatic, or spring or a combination of these [4, chap. 8]):
pneumatic close and pneumatic open
hydraulic close and hydraulic open
spring (motor charged) close and spring open
hydraulic close and spring open
pneumatic close and spring open.
Some examples of mechanisms used with distribution switchgear are illustrated by
Ali, in chapter 11 of Reference 4, and Figure 9.4 illustrates a vacuum reclosure with
magnetic actuator [9].
Much incremental research and development of improved interrupter devices has
taken place during the past 80 years and continues to the present time. This strategic
effort has led to a greatly improved understanding of the basic principles of interrup-
tion and the development of advanced modelling and testing techniques culminating
in the design of a wide range of advanced circuit-breaker devices [110]. These
developments, sometimes involving increasing levels of sophistication, have kept
pace with the requirements of expanding power systems, deregulation and frequently
special operating situations [11].
To illustrate this fact, consider the following example. On the basis of current
technology [12], the controlled switching systems is one of a number of potential
solutions to overcome transient problems. However, the performance of modern CBs
is extremely complex and the development of effective testing methodologies, which
can be influenced by a number of considerations, is crucial. Athorough understanding
of circuit breaker characteristics, together with effective testing strategies for testing
of controlled switching systems, is complex (see section 9.6). Briefly, it must be
appreciated that a complete modern transmission substation can involve many com-
ponents, and the entire installation may be produced, delivered and installed by one
manufacturer or the components may be delivered by different producers and erected
and commissioned by the user or one of the manufacturers. Because of such diversity,
a modular testing procedure has recently been suggested and this will be considered
later (see section 9.6).
Extensive literature exists relating to the historical developments [14, 8] asso-
ciated with switchgear technology, detailing numerous and varied techniques
adopted for improving the interrupting ability of modern circuit breakers (see also
www.cigre.org).
Whatever technique is used, the basic problem is one of controlling and
quenching (or extinguishing) the high power arc which occurs at the separating
contacts of the interrupter when opening high current circuits. The reliability of
a circuit breaker is interdependent on several factors: insulation security, circuit
breaking capability, mechanical design and current carrying capacity. The range of
arc quenching media used in commercial circuit breakers has included air, com-
pressed air, oil, small oil volume, strong magnetic field devices, SF
6
and vacuum
environments [110].
It must be appreciated that circuit breakers (i.e. circuit interrupters) are essentially
mechanical devices for closing or opening electrical circuits. Historically, circuit
breakers have been classified:
(i) according to the insulating media used for arc extinction
(ii) sometimes based on special interruption features (e.g. simple autoexpansion
and magnetic rotation techniques of arc interruption in SF
6
designs).
Fuller consideration of these specialist design concepts is outside the scope of this
chapter and these aspects are discussed elsewhere [110].
Detailed information relating to the dielectric and arc interruption characterisation
of atmospheric air, compressed air, oil, vacuum, SF
6
, various other gases and gas
mixtures, and solid insulating materials, exists and is now readily available to the
switchgear designer who also has recourse to a wide range of advanced simulation
packages [4].
At transmission voltages, sulphur hexafluoride gas (SF
6
) has been used for arc
interruption and part insulation since the 1960s, and has now fully replaced com-
pressed air or small oil volume devices in new substation installations in most of
the developed countries [14, 8]. Indeed, as has already been stated, SF
6
interrupter
technology is very mature, as evidenced by the third generation devices nowavailable
commercially [4]. Despite this, much of the switchgear in service on transmission sys-
tems in the UK still uses compressed air or small oil volume interruption techniques.
At distribution voltage levels, sales of newCBs, in most developed countries, are now
exclusively centred on devices where interruption is based on vacuum and SF
6
tech-
niques which has led to more efficient and cost effective designs [9]. Although these
modern types of CB have progressively replaced older types of interrupter, using air
or oil-based interruption methodologies, here again, many of the older designs are
still in service and continue to perform effectively [1, 4, 9].
9.2.1 Principles of current interruption in HV systems
(after Jones [4, chap. 7])
All methods of interruption current, to date, rely on introducing a non-conducting
gap into a metallic conductor, achieved by mechanically separating two metallic
contacts so that the gap formed is either automatically filled by a liquid, a gas, or
even vacuum. In practice, such inherently insulating media may sustain a variety
of different electrical discharges which then prevent electrical isolation from being
achieved.
Jones considers three major facets to such electrical discharges which then prevent
electrical isolation being achieved:
(i) As contacts are separated, an arc discharge is inevitably formed across the
contact gap. The problem of current interruption then transforms into one of
quenching the discharge against the capability of the high system voltage of
sustaining a current flow through the discharge. Since this physical situation
is governed by a competition between the electric power input due to the high
voltage and the thermal losses from the electric arc, this phase of the interrup-
tion process is known as the thermal recovery phase and is typically of a few
microseconds duration.
(ii) The second facet of current interruption relates to the complete removal of the
effects of arcing which only occurs many milliseconds after arc formation even
under the most favourable conditions. The problem then is one of ensuring that
the contact geometry and materials are capable of withstanding the highest
voltage which can be generated by the system without electrical breakdown
occurring in the interrupter.
(iii) The third facet bridges the gap between the thermal recovery phase and the
breakdown withstand phase. The problem in this case is that the remnant effect
of the arcing has cleared sufficiently to ensure thermal recovery but insuffi-
ciently to avoid a reduction in dielectric strength. This is known as the dielectric
recovery phase.
Based on this understanding, circuit interruption technology is concerned on the one
hand with the control and extinction of the various discharges which may occur, and
on the other it relates to the connected system and the manner in which it produces
post-current interruption voltage waveforms and magnitudes.
9.3 Distribution switchgear systems
Historically, distribution switchgear technology was based on oil or air break circuit
breaker devices. Although good reliability was achieved with CBs based on both these
interrupting media for several decades, evolving strategic health and safety guidelines
and lifetime ownership considerations have resulted in a strategic swing in the 1970s
towards new medium voltage switchgear, i.e. 170 kV, using either vacuum or SF
6
as the interrupting media. Oilless circuit breaker designs have been developed since
the late 1970s [1, 4, 810]. Initially:
SF
6
devices offered a proven replacement technology, with excellent dielectric
and interruption characteristics
early vacuum interrupters (bottles) were difficult to manufacture and were prone
to produce voltage transients due to severe current chopping [1, 4, 8].
Consequently, SF
6
circuit breakers were deemed by users to offer a more popular
and reliable replacement interrupting media. In recent years, vacuum interrupters
have found increased acceptance and application as a result of advances in material
technology and improved manufacturing techniques. Modern vacuum interrupters
have a good performance record [4, 810] especially with respect to frequent high
fault clearance operation (e.g. see design shown in Figures 9.4a and b). SF
6
is used
almost exclusively at voltages above 36 kV. Low voltage systems (as originally
stated) (>1 kV) have in the past used air and oil CBs, but advances in moulded
case air circuit breakers have made air CBs more cost effective and this is now
the most widely used technology in this range (see Reference 9 and companion
articles).
Recent progress in distribution switchgear has been discussed extensively in
Reference 4 and was also considered in a special IEE feature covering four valu-
able articles. In his introduction to these articles, Blower chose to group distribution
switchgear into three areas of use as summarised below and detailed in his original
paper [9], namely:
(i) primary distribution
(ii) secondary distribution in cable connected networks
(iii) secondary distribution in overhead line connected networks.
Some useful technical observations relating to companion articles by other authors are
also listed in [9]. Blower illustrated a newgeneration of autoreclosures which use vac-
uum interrupters for arc interruption, driven by magnetic actuators (e.g. Figure 9.4).
He considers that distribution switchgear (often referred to as medium voltage
switchgear), which covers the voltage range above 1 kV up to about 150 kV, can
be divided into the three groups identified above. Blower also points out certain
special aspects/considerations relating to this interrupter. These include:
the use of permanent magnets in the actuator has arisen with the introduction of
new magnetic materials which are both powerful and not affected by additional
magnetic fields
the magnets provide positive on and off position for the vacuum interrupter,
and the addition of a magnetic field from a surrounding coil to change the flux
distribution is used to change the state
the vacuuminterrupter has a very long life on this type of duty and the actuator has
a very low energy consumption and can operate thousands of times from small
primary cells (e.g. lithium-type cells); a life of 10+ years, before replacement is
required, has been claimed
the vacuum device can be contained in a sealed vessel filled with SF
6
to ensure
a high reliability of insulation and independence from climatic conditions (see
Figure 9.4a).
9.4 Substation layouts and control aspects
9.4.1 Substation layouts
Typical substation switching layouts have been discussed recently by Fletcher
[4, chap. 9], Pryor [4, chap. 10] and also Ali [4, chaps 8 and 11] for transmission
networks and primary distribution substations for single and double busbar switch-
ing arrangements and also for urban and rural distribution systems. Figures 9.69.11
present a summary of some strategic switching arrangements considered, with sup-
plementary information provided in subcaptions to these figures and in the rsum
reproduced below:
Pryor has recently reported that primary distribution substations, within the UK,
typically operate at 11 kV and are utilised to supply a relatively large number of
consumers within the local area (Figure 9.6a).
The double busbar arrangement, shown in Figure 9.6b. This system is widely
used for large important supplies (i.e. as at bulk supply points), which in the UK
typically operate at 132/33 kV, or perhaps for large industrial customers, higher
security may be built into the substation design. Pryor comments that:
(i) modern busbar protection is very reliable such that the same degree of
availability can normally be achieved without the need to resort to a second
busbar system
1
2
A
busbar
bus
section
incoming
circuits
outgoing
circuits circuit breaker
Figure 9.6a Typical distribution primary substation layout: single busbar switching
arrangement (after Pryor [4, chap. 10])
a primary substation feeding a rural network is commonly of the
open terminal air-insulated design whereas a primary substation
feeding urban networks is more commonly of the indoor metalclad
design
a common switching arrangement for either substation type is of
the single busbar design but with the busbar being split into two
sections and interconnected via a bus section CB, A
there are usually two incoming circuits one feeding each section
of the busbar; there may be typically five outgoing circuits feed-
ing either multiradial networks for overhead rural systems or ring
circuits for urban connected networks
for maintenance purposes disconnectors, 2, are fitted on either side
of the CB, 1
facilities are also provided for earthing outgoing or incoming
circuits and for earthing each section of the busbar; current trans-
formers, CTs, may be fitted within the outgoing circuit for either
protection or, less commonly, for tariff metering purposes (e.g. for
large single customers)
(ii) it is also considered necessary to ensure that each section of busbar is
normally kept energised because, if a section remains deenergised for
a significant period of time, then deterioration may occur such that possible
busbar failure may result on reenergisation.
Urban distribution system switching arrangement (see Figure 9.7). This system is
fed from a primary distribution switchboard, each outgoing feeder on one section
cable or overhead
line termination
disconnectors
Figure 9.6b Typical distribution primary substation layout: double busbar switch-
ing arrangement (after Pryor [4, chap. 10])
arrangement allows the option of four separate sections of busbar to
be utilised such that if any one busbar fails, supplies could generally
be maintained via the other three busbar sections; all circuits can
be selected to either of the two main busbar sections
busbar protection is usually provided to ensure very rapid clearance
of any faulted section of busbar (such arrangements are, however,
very costly)
of the busbar feeds via an 11 kV cable network to typically 10 to 12 secondary
distribution substations which are electrically connected within the ring circuit.
This ring circuit is connected back to a feeder on the adjacent section of busbar
at the same substation [4]. Pryor goes on to describe general network principles
which operate in most industrialised countries, although some may employ radial,
as well as ring, circuits. Typical distribution voltages used elsewhere may range
from 10 kV to 20 kV. 6.6 kV systems were common at one time within the UK,
but have now been largely phased out.
A rural distribution system switching arrangement is illustrated in Figure 9.8
(see Pryor [4, chap. 10]). Here, the primary switchboard will typically comprise
two sections of busbar connected via a bus section CB, A. The number of out-
going feeders tends to be less than with an urban network with typically six
panels. Surge arrestors, commonly fitted at the cable sealing end and mounted
physically at the top of the pole at the junction between the cable and overhead
secondary substation comprising outdoor
ring main unit, 11 kV/415 V transformer,
LV multiway fuse cabinet
normally
open point
11 kV primary indoor
switchboard
possible LV
backfeed
Figure 9.7 Urban distribution system switching arrangement (after Pryor
[4, chap. 10])
this ring circuit usually has a normally open point, the purpose of
which is to minimise the number of customers affected by the faulted
section of the ring circuit
ring main switches are manually operated; hence fault location and
customer reenergisation is time consuming; a modern tendency
is for the ring main switches to be fitted with remotely operable
mechanisms such that switching times can be reduced by allowing
remote substation facilities or, with modern intelligence systems,
to allow automatic deenergisation and isolation of a faulted section
and reenergisation up to the point of the fault
the tee-off point of the ring circuit feeds via an 11 kV/415V, three-
phase transformer to an LV fuse board, typically having up to five
outgoing circuits which feed directly to large customers or groups
of customers; in most of these LV circuits, it is possible to achieve
an LV backfeed from an adjacent ring main unit, thus maximis-
ing the number of customers on supply while fault repairs are in
progress
line, reduce the probability of transient induced overvoltage (caused by lightning
strikes to the overhead line) affecting either the cable or associated switchgear.
(See Pryor [4, chap. 10], for fuller details of urban and rural distribution
networks.)
Turning now to larger substations at both transmission and distribution voltages
(see Fletcher [4, chap. 9]), where security of supply is critical, the double busbar
arrangement shown in Figure 9.9 has been extensively used in the UK. Avariation
of this is shown in Figure 9.10 which illustrates a multisection double busbar
arrangement. A third layout, reproduced in Figure 9.11, is known as a one and
11 kV
cable/OHL
junction
spur lines
spur lines
disconnector
dropout
expulsion fuses or
dropout links
switch
disconnector
indoor 11 kV switchboard
(primary)
pole-
mounted
recloser
A
Figure 9.8 Rural distribution system switching arrangement (after Pryor [4,
chap. 10])
the primary switchboard may be of open-terminal formor in the out-
door weatherproofed form; it is now more common, however, for
these primary switchboards to be enclosed within brick-type enclo-
sures, in which case, conventional indoor metalclad type switchgear
is more commonly used (most 11 kVrural primary switchboards are
of the metalclad-type, whereas at 33 kV, open-type busbar connected
switchgear is more common)
most UK rural networks feed overhead line distribution circuits and
there is usually a short connection of cable connecting from the
switchboard to the terminal pole of the overhead line
2
1
4
3
Figure 9.9 Double busbar switching arrangement (after Fletcher [4, chap. 9])
Here:
each circuit is provided with a dedicated circuit breaker CB (1) and
can be selected to connect to either of the two independent busbars
by appropriate switching of busbar selector disconnectors (2)
the busbars can be coupled by means of a bus coupler circuit breaker
CB (3), the line disconnectors (4) being normally included at trans-
mission voltages to facilitate maintenance, but this feature may be
omitted in many cases, without loss of operational functionality
service continuity can be maintained with a busbar out of service;
however, Fletcher points out that a single circuit outage is necessary
for fault repair or maintenance in the CB area
1
A
2
A
3
Figure 9.10 Multisection, double busbar switching arrangement (after Fletcher
[4, chap. 9])
Here:
bus section CBs (A) are added to the standard double busbar config-
uration, providing the ability to segregate circuits across a number
of sections of busbar separated by CBs and this is used to limit
system disruption under worst case fault conditions
Fletcher provides an example in the above figure of the use of
this configuration for the connection of large generating units to a
system
if each unit is arranged so as to feed a separate section of busbar
(1, 2, 3 in the diagram above), then normally only one unit will be
tripped for a busbar fault, leaving the remainder in service
under worst case fault conditions (a bus section or bus coupler CB
fault), a maximum of two generating units will trip
the system can thus be operated with reserve generation capacity
sufficient only to cater for the loss of these two units
a half switch meshdouble busbar switching arrangement and this arrangement
is widely used abroad at transmission voltages.
For fuller information on substation layouts, the reader should follow the activities
and readily available literature of appropriate CIGRE study committees (e.g. SC.
23, see also References 712, 15, 16, 19, 26, 27, 29) which provide strategic infor-
mation of global activities, and trends in this sector. This includes consideration of
such important issues as environment, short circuit forces on busbars, pollution, and
minimum safety clearances for personnel working within substations.
9.4.2 Intelligent networks
Before considering intelligent networks, it is pertinent to comment on certain strategic
effects of deregulation within the electricity supply sector (see also section 9.9.2).
Several workers [4, 2224], the most recent being Ziegler [26] whose personal
views are briefly reproduced below, have reported that increased competition since
deregulation has forced utilities to go into cost saving asset management with new
2
1 1
2 2
3 3
Figure 9.11 One and a half switch, double busbar switching arrangement (after
Fletcher [4, chap. 9])
Here:
this arrangement has two busbars (1) interconnected by chains of
three CBs (2)
outgoing circuits are taken off from between pairs of CBs (3)
this switching arrangement is widely used internationally at trans-
mission voltages, but is not common in the UK
when fully operational, this arrangement offers a high level of
service continuity of the outgoing circuits and is beneficial where
line availability is critical
however, this arrangement can be vulnerable to single fault outages,
during periods of busbar or CB maintenance outages
risk strategy:
plants and lines are higher loaded up to thermal and stability limits
existing plants are operated to the end of their lifetime and not replaced earlier
by higher rated types
redundancy and backup for system security are provided only with critical
industrial load
corrective event-based repair has replaced preventative maintenance
considering this changed environment, Ziegler feels that power systemprotection
and control face new technical and economical challenges [26].
A summary of Zieglers opinions, recognising his authority as a former chairman of
CIGRE, SC34, is reproduced below:
(i) protection and substation control have undergone dramatic changes since the
advent of powerful microprocessing and digital communication
(ii) smart multifunctional and communicative feeder units, so called IEDs
(intelligent electronic devices) have replaced traditional conglomerations of
mechanical and static panel instrumentation
(iii) combined protection, monitoring and control devices and LAN-based
integrated substation automation systems are now state of the art
(iv) modern communication technologies including the internet are used for remote
monitoring, setting and retrieval of load and fault data
(v) higher performance at lower cost has resulted in a fast acceptance of the new
technology
(vi) the trend of system integration will continue, driven by the cost pressure of
competition and technological progress
(vii) the ongoing development towards totally integrated substations is expected
to pickup speed with the approval of the open communications standard
IEC 61850 in the next few years.
W.J. Laycock [27] has recently written an interesting and thought provoking article
on intelligent networks, covering:
(i) the development of existing networks
(ii) the need for more intelligence
(iii) future developments and opportunities.
He discusses how engineers and managers in diverse locations are being challenged
to find solutions that will simultaneously drive down cost and encourage better utili-
sation of distribution sector assets by the development of intelligent networks. Some
strategic issues raised by Laycock will now be briefly touched on and are reproduced
below, including the application of intelligence, centres of advanced intelligence,
network services, protection intelligence and drivers for the future (see also Tables 9.1
and 9.2 and Figures 9.129.15). (The reader is directed to the original article for fuller
treatment.)
Figure 9.12 illustrates a classical MV distribution network which may include
SCADA and FPIs but the presence onsite of trained staff is still normally required.
Intelligence is defined by Laycock as whatever independent action can be taken on
receipt of information, and is apparent at the following locations:
network control centre
protection and autoreclose equipment
voltage control equipment
Table 9.1 Centres of advanced intelligence
(after Laycock)
distributed network control
automation
adaptive protection
plant condition assessment
voltage and waveform control
generation scheduling
transport and supply of electricity
network services
ownership boundaries
Table 9.2 Network services (after Laycock)
On-network Off-network
frequency control
voltage control
waveform control
reactive power supply
dynamic stability
load following
fault level control
asset management
performance measurement
plant condition monitoring
refurbishment
meter reading
billing
connection management
SCADA
ACC
infeed S/S
ACC=area control
centre
ring main unit
and transformer
circuit breaker
with protection
normally
open point
Figure 9.12 Classical open ring MV distribution network configuration (after
Laycock [27])
network arrangement is in common use worldwide, employing a
single infeed S/S and ring main units (RMUs) at load tee-off points
it is passive (i.e. has no active generation), its sole purpose being
to distribute energy to customers
may have some form of remote control (SCADA) at the infeed
point and may be equipped with fault passage indicators (FPIs),
but normally requires trained people on site to respond to any kind
of incident
network owners office for:
(i) network design
(ii) asset management
(iii) purchase and sales of electricity
(vi) performance assessment.
SCADA
ACC
infeed S/S
interconnection
Figure 9.13 Open ring with tees and interconnection (after Laycock [27])
networks commonly have multibranch tees, additional CBs and
interconnection with other networks for emerging backfeeds etc
it must be realised that all the functions relating to application of
intelligence (control/management) become more complex, often
requiring compromises between conflicting requirements
SCADA
ACC
infeed S/S
interconnection
gen
wind
waste heat
gen
gen
Figure 9.14 Open ring with embedded generation (after Laycock [27])
Laycock informs the reader that intelligence may be supplied directly by human
intelligence, by relatively simple control/relaying equipment or, on more modern
networks, by digital equipment and computers, the mix varying widely between
networks and over time as a network develops. He goes on to indicate the functions
that need the application of intelligence: (i) in the hard sense of network control and
(ii) in the soft sense of network management, pointing out that there are many links
SCADA
ACC
infeed S/S
interconnection
gen
wind
gen
gen
DVR
CHP&L
technology
park
DVR=dynamic voltage
restored
Figure 9.15 Open ring with an IPP, CHP scheme (after Laycock [27])
Laycock identifies the rapid changes post-liberalisation as more
players are attracted to this sector, creating the need for significantly
more intelligence than was necessary with the earlier passive nature
of network design
between these functions:
Application of intelligence
Control Management
loading access for work
match supply to demand asset management
security billing
short circuits new connections
supply restoration supply contracts
voltage control service contracts
dips, sags, spikes and harmonics performance measurement
Laycock also recognised that, in practice, networks are rarely as simple as
the open ring shown in Figure 9.12, since they commonly have multibranch tees,
additional CBs and interconnection with other networks for emergency feedbacks
etc. Consequently, Laycock considers that Figure 9.13 presents a more typical
network, where all the functions, as detailed above, become more complex, often
necessitating compromises between conflicting requirements. He also points out that
although Figure 9.13 is still a passive network, many previously passive networks
are now becoming active with the addition of embedded or dispersed generation
(EG). Laycock comments that some generation might come from renewable energy
sources to reduce global warming, others to provide improved security or quality of
supply or will simply be an independent power producer (IPP) who is exploiting a
business opportunity. Thus, according to Laycock, modern networks may now take
on the appearance of Figure 9.14 and, moreover, the characteristics of these modern
networks require the application of considerably more intelligence than previously.
This raises some strategic issues, which according to Laycock, include:
small stochastic generators (output dependent on primary source, e.g. wind)
operation with alternative infeeds
concept of active distribution(dominimumtoenable connectioninthe first place!)
ownership and money flows
action on loss-of-grid
generator scheduling (at transmission level).
It is recognisedthat, as a consequence of the development of privatisedenergymarkets
and the connection of more sophisticated plant and changes in perception as to own-
ership, operation and outsourcing of services, networks are being further stretched.
These concepts have been illustrated by Laycock in Figure 9.15. This figure shows
the imposition of a combined heat and power scheme (CHP) which has been designed
to operate islanded with its own load under certain conditions, but not necessarily
optimised with network operation.
As a consequence, Laycock considers that the following may be added to the
issues raised by Figure 9.14:
complex generation, supply and connection contracts
island operation and power wheeling
quality and security of supply
provision of network services
electricity supply regulation.
Laycock also points out that this opening-up or liberalisation of the electricity market
creates opportunities for increased entrepreneurial input as individuals and business
groups identify opportunities for new business ventures, which could increase util-
isation of networks (e.g. see Figure 9.15) while reducing the cost of ownership.
However, it must be recognised that although here is undoubtedly a strategic need
for a significantly increased level of intelligence with modern active distribution net-
works, it is not yet clear to Laycock and others at this time the precise form(s) that
such new intelligence systems will take. In sections 9.4.3 and 9.4.4, we will briefly
review Laycocks perspective relating to future needs/developments in this sector.
(The reader should also become familiar with recent reviews relating to intelligent
condition monitoring strategies relating to switchgear [4] and also current strategic
developments in the CIGRE sector.)
9.4.3 Need for more intelligence (after Laycock)
Laycock considers that intelligence will be applied in three primary areas:
(i) innovative solutions to network design and operation to (enable) facilitate new
business streams
(ii) use of intelligent devices on the network to improve responses and collect
information
(iii) resolution of problems as they occur including disputes over responsibilities.
Table 9.1 details anticipated centres for advanced intelligence which, according to
Laycock, will appear on future networks. In the past, responsibility for provision
was generally centred on a single entity, normally a Regional Electricity Company
(REC). However, responsibilityfor provisionof future intelligent systems will involve
several interest groups operating within the still evolving liberated energy market
sector, including possibly several utilities, manufacturers, traders etc., all with access
to data via sophisticated software and computer systems.
As Laycock and others have pointed out:
the degree of automation using intelligent devices to control the network and to
collect and analyse information will vary according to the owners philosophy,
its legacy systems and the pace of network development;
provision of outsourced network services is receiving increased attention and is
a prime area for innovative thinking
the kinds of services that may be needed are listed in Table 9.2, but Laycock
considers that this cannot be exclusive at this stage in their development.
9.4.4 Network protection and future developments (after Laycock)
Laycock provides a valuable perspective relating to the critical provision of protec-
tion as networks move from passive to active operation, pointing out that what was
previously acceptable on radial distribution circuits will fail to satisfy requirements
of dynamic operation with active generation and with loads that are susceptible to
prolonged voltage dips. Major strategic considerations raised by Laycock relating to
network protection in the future are set out in his original article [27].
Finally, looking towards possible future developments/trends, Laycock points out
that one must have a good understanding of the drivers for the future development
of intelligent networks in order to apply human intelligence to produce effective
advanced solutions. He also sets out his current thinking relating to a range of strategic
factors. Further consideration of the topic of substation control, as identified by other
CIGRE experts, is given in section 9.5.
9.5 Substation control in the system control (CIGRE. WG.39.01 [11])
Brief reference will now be made to the strategically important issue of substation
control (SCS) which now represents an integral part of the electrical power systems
system
control
centre
communication
network
communication
servers
MMI
RTU
digital and analogue I/O
RTU
MMI
I/O
mixed classical modern integrated
I/O &
protection
relay
application servers
I/O & protection relays
bay unit
level
Figure 9.16 Types of substation control system [11]
control hierarchy. A recent CIGRE working group paper [11] provides the reader
with a useful positioning overview of this theme. The paper is focused on the place-
ment, functionality and coordination of a substation control in a hierarchical system
control scheme. It presents various scenarios as to where a substation control centre
may be placed and what types of function are required from the substation control to
improve the overall control performance. Feedback from a small survey carried out
by WG.39.01 also forms part of this paper. Certain major strategic aspects discussed
in this publication, including the outcomes fromthe survey, will be extensively repro-
duced and highlighted in this section as set out in the following paragraphs, figures
and tables (see Figure 9.16). (The original source [11] is gratefully acknowledged.)
This CIGREWG.39.01paper [11] comments that the SCSmayhave different roles
to play in an electrical utility operation and it fits in different roles in the hierarchical
control centres schemes, as shown in Figure 9.16:
(i) classical monitoring, control and protection using remote terminal units (RTUs)
and local applications
(ii) the SCS existing in parallel to the classical RTU configuration
(iii) a modern SCS combining protection, monitoring and control and without use
of a RTU.
The paper reports that if intelligent electronic devices (IEDs) are used to provide inte-
grated monitoring, protection and control functionality, these have to be connected
over a coupler or gateway to the station LAN. The I/O units may be distributed
throughout the substation on the process bus, and the interface from the process
bus to the station bus depends on the bay control units. Redundant configuration
may be implemented to suit the availability requirements of the substation control.
Optical fibre is the preferred communication media rather than the hardware connec-
tions of the earlier generation. The optical-fibre-based communication is claimed to
have lower interference and allow reconfiguration at a later stage with the benefit of
lower engineering costs.
Further strategic facts detailed in [11], are:
Data engineering and management: important aspects of data engineering and
maintenance play an important role for the SCS and, as reported [11], must be
planned and coordinated to fit in the system control hierarchy.
Functional requirements which influence SCS and higher level centres.
Interfaces from substation control to high level centres have been considered as
follows (see [11]):
(i) multimaster configuration
(ii) protocols in use
(iii) multimedia possibilities.
It is possible to review experience with the SCS in system control from feedback
from a limited CIGRE survey carried out by WG.39.01 [11], based on returns from
mainly large utilities with more than 100 substations and for voltage levels of 150 kV
and above:
(i) Implementation ratio of SCS to the number of substations varied from 0
to more than 50 per cent. The majority of the substations were unmanned
and allowed remote switching. Almost all utilities owned the higher level
centre and substations, which communicated with them. The higher level
centres may have supervisory control and data acquisition (SCADA) or energy
management systems (EMS).
(ii) In the emerging unbundled and competitive market which now exists, a dis-
tinction needs to be made between the owner of the substation and the system
operator. Very often, with an independent system operator (ISO) or even a
transmission system operator (TSO), the ownership of the substation differs
and a contractual agreement between the two parties must be negotiated.
(iii) The SCS must ensure that only those commands which come from authen-
ticated higher level centres be executed. Networking and remote logins are
weak spots in the security of such a system. Firewalls and other security
schemes must be considered to inhibit unauthorised access to the SCS or SCC
database.
(iv) The data must be refreshed with the resumption of the link, and system super-
vision events from SCS should be logged or alarmed in SCC. However, most
of the utilities do not see any technical problems in achieving this. The experi-
ence from the survey [11] shows that the remote switching is a regular feature
for the substations, which have been implemented with SCS. In some cases,
the remote switching was done only when a maintenance crew was present in
the substation for testing and maintenance purposes.
(v) The performance, in terms of control actions, information updates, commu-
nication link speeds and total data transfer is of keen interest to most utilities.
Most typical response times required for reporting of a breaker tripping from
the substation to the control centre varied in the range of one to three seconds.
However, responses for less than one second and higher than three seconds
were also reported for such an event.
(vi) In comparison with the tripping events, measurements were refreshed at
a range between five and ten seconds. The SCS performed some filtering
and data selection on the measurements and indications before transferring
them to the higher level centre. More than 80 per cent of the measure-
ments and about 50 per cent of the indications, which were measured and
monitored at the SCS level, were transferred to the higher level centre.
The transfer of disturbance data is normally requested in the case of dis-
turbances, but some utilities apparently prefer to transfer these data cyclically
once a day.
(vii) Communications speeds varied between 300 b.p.s. and 9600 b.p.s. with some
links being as fast as 64 kb.p.s. Most of the stations have power line carrier
and optical fibre links but radio and leased lines are also in use. Redundant
communication links are one way to increase the total availability of the SCS
from the SCC. As the cost of communications is going down, almost all
utilities are opting for redundant links.
(viii) The availability of SCS is of crucial importance for total system control
performance. At first level, the availability of the control system at the
SCS should be high, which means that the critical control components are
redundant. In addition to the SCS availability, the availability of the com-
munication link influences the total availability of the SCS as seen from
the SCS. Most of the utilities which have been asked in the survey [11]
apparently require an availability level higher than or equal to that of the
classical RTU controlled units. Quantitatively, most of the availability fig-
ures were given as 99.5 per cent, with one value positioned between 98 and
99.5 per cent.
(ix) The protocols used for communication between SCS and SCChave in practice
been driven towards standard non-proprietary protocols. Utility communica-
tion architecture-driven ICCP and IEC870-5-101 telecontrol protocols were
the most favoured ones for the implementation. Exceptions were some substa-
tions where proprietary protocols were used in order to match the implemented
protocol in the SCC.
To summarise, this valuable WG.39.01 paper [11], representing just one of many
strategic published outputs from CIGRE/IEC organisations in recent years, has
provided valuable evidence that the present state of the substation control system
from the hierarchical control system viewpoint is slowly gaining acceptance and the
technology has been shown to be maturing. The results from the small survey indi-
cated that most of the utilities have acquired some experience with SCS systems
and are in the process of upgrading the classical RTU links to the SCS and
SCC link.
These authors [11] state that the factors that must be considered while designing
or planning a substation control system are:
the operational hierarchy of the system
the level of acceptable integration of protection and monitoring devices
whether the substations are manned or unmanned
the complexity of the substation
the data flow and communication performance.
From this evidence [11], it would seem that practically all new substations are being
considered for SCS. Apparently, some utilities prefer a mixed solution of SCS with
RTU-type links to the SCS, others install an RTU and plan to upgrade to SCS in the
near future and certain other utilities go purely for the SCS-type solution.
There is a need to differentiate between real-time and non-real-time data to the
SCC in order to leave enough capacity for fast real-time activities.
9.6 Planning specification and testing of controlled HVAC switching
systems [12]
9.6.1 Background
The uncontrolled switching of inductive and capacitive elements can create electrical
transients whichmaycause equipment damage andsystemdisturbances. This problem
has been well known for many years. CIGRE Study Committee 13 recognised that
controlled switching offered a potential solution to this problem, as restated below
and also in Sections 9.6.2 and 9.6.3:
In 1992, they appointed a task force TF13.00.1 which prepared a document sum-
marising the state of the art [1]; subsequently, Working Group WG.13.07 was
created in 1995 to study, in greater detail, the strategic issues associated with
controlled switching from both supplier and user perspectives [12].
The first publication of WG.13.07 addressed fundamental technical issues such as
the potentially achievable transient reductions for various switching conditions. It
is claimed [12] that the focus was on the most common applications of switching
of shunt reactors, shunt capacitor banks and lines. The fundamental theoretical
and scientific aspects were presented and the dominant role of the circuit breaker
characteristics was clearly identified.
On the basis of the information presented in the first document, a second document
was prepared which addresses the more practical aspects of controlled switching.
Extensive guidance is givenonhowtoapproachcontrolledswitchingprojects and,
in particular, howto study, specify and test the circuit breaker and control system.
(Note: The full text of this report can be obtained via the CIGRE website under
the title Controlled switching of HVACcircuit breakers planning, specification
and testing of controlled switching systems [12].)
Initial planningquestions are introduced, inorder toassist the decisionmakingprocess
for any application, such as, is controlled switching a technically suitable and cost
effective solution?. There is considerable interaction between the component parts
of a controlled switching system (CSS), and the hierarchy of responsibility to ensure
the overall performance is discussed in some detail [12]. Subsidiary issues are also
introduced, such as the interfaces between the controller and the auxiliary systems of
a substation.
Recommendations for the type testing of CSS components and also the integrated
system are made, including detailed guidance on difficult issues such as the determi-
nation of the circuit breaker characteristics. The document concludes with guidance
on complete system performance checks and commissioning tests.
The key aspects of this CIGRE report will now be considered and restated.
A case study prior to decision taking [12], with a schematic decision process for
capacitor bank switching, is also set out.
9.6.2 Specification of controlled switching installations [12]
This recent abridged CIGRE report [12] comments that the specification of a con-
trolled switching system must consider the interactive combination of a switching
device, a controller and a variety of associated items of auxiliary equipment:
The systemmay be assembled fromcomponents froma single source or disparate
sources, and it is important to identify which parameters or requirements are of
particular relevance to each component of the system and which to the system as
a whole.
Any application of controlled switching must be undertaken with due regard to
the characteristics, interactions and compatibilities of its component parts. The
authors [12] go on to point out that the performance of the individual components
must be well defined and understood but the performance of individual compo-
nents of the systemis of limited interest in isolation since it is the overall accuracy
requirement which should be met. The principle of interaction between the power
system and the CSS is clearly illustrated in the original paper [12].
There are likely to be numerous parties involved in a CSS application and clear
identification of responsibilities is vital. The requirement may be to construct
a total system on the basis of a comprehensive specification, but may equally
require the application of controlled switching into an existing uncontrolled sce-
nario. In this latter case, although the overall specification remains valid, certain
parameters will be predetermined (or may not be available in an accurate form)
resulting in a need to ensure that any newly installed equipment is sufficiently
robust in its capabilities to overcome such potential difficulties.
In order to take these basic conditions into account, an approach is adopted which
considers the individual components of the CSS to be discretely identifiable units.
However, considering ongoing developments in integration and digitisation of
substation protection and control facilities, the authors [12] felt that the principle
of this document should be equally applicable in all cases whether of analogue,
digital or hybrid type.
Three levels of specification requirements were advanced [12]:
(i) Performance: defines the basic functional compatibility requirements to which
the chosen CSS will be subjected.
(ii) System: defines specific technical and material requirements for the CSS itself.
(iii) Component: addresses the internal interactions of the CSS taking into account
the external requirements of the system specification.
The requirements set out and discussed in [12] are incorporated at the highest possible
level and cascade to lower levels as appropriate.
Finally, it should be emphasised that this important paper [12], which summarises
and explains strategic aspects relating to the testing of controlled switching systems
and circuit breaker testing, represents an invaluable reference source for any reader
wishing to achieve a good understanding of this subject area.
9.6.3 Concluding remarks
In their closing remarks the authors [12] state that:
controlled switching has already been widely applied by many utilities
controllers and CBs available from a variety of manufacturers have suitable
characteristics for controlled switching
they consider that the information and studies undertaken in this programme of
work [12] go some considerable way towards consolidating available knowledge
on issues of preparing studies, specifications and testing
the contents of the full report are intended to provide guidance on the subject of
controlled switching in a broad context and, as far as possible, on a scientific
level as well
the potential variety and complexity of the applications of controlled switching is
such that many issues can only be covered in general terms but the authors feel
that the guidance presented in the extended version of their paper should form
the building blocks for more specific cases to be developed
moreover, they suggest that this work would be well suited as the basis for future
standardisation activities.
The authors [12] also point out that the idle time behaviour of a CBis another issue of
major importance. Supported by experimental evidence, they point out that depending
on the period of quiescence, there may be a change in the operating time of the first
operation.
Note: This further example of recent international collaborative CIGRE work, as
restated and described above, illustrates to the reader the real benefits of participating
in such a forum (e.g. CIGRE/IEC committee, working group, task force activities, or
even merely obtaining such information directly as a CIGRE member!) as a means of
empowerment by keeping oneself abreast of current relevant strategic international
collaborative technical work in the sector.
9.7 Dielectric and global warming considerations
At this point, very brief mention can be made to the strategic role of [4,13,14]:
undertaking appropriate characterisation development studies on potential
solid/gaseous insulation systems, for a wide range of conditions
analytical techniques to enable the switchgear designer to develop efficient
designs with the robust expectation that the switchgear insulation systems for
GIS and GIL plant will achieve a long and trouble-free life in service.
For many years, the writer was extensively involved in developing numerical field
techniques which have found widespread application in the insulation design of
GIS, GIL and other switchgear [4, chap. 3]. Simple empirical and semiempirical
breakdown estimation methods by Ryan et al. [31], which are extensions of the
work by Pederson [32], together with an available experimental database obtained
from extensive Paschens Law/similarity-type high voltage studies in gaseous insu-
lants covering a wide range of electrode systems, gas pressures and temperatures,
were thoroughly developed to such a degree that minimum breakdown voltages of
practical GIS/GIL design layouts (as well as a host of SF
6
gas-gap arrangements
[4, chap. 3, 13, 14]) can be estimated to within a fewper cent, at the design stage, often
without recourse to expensive development testing. Such derived voltages are gener-
ally the minimal withstand levels attainable under practical conditions. Undoubtedly,
there is still considerable scope in the future to predict dielectric performance by
utilising advanced planning simulation tools incorporating genetic algorithms and AI
techniques, linked to various extensive databases relating to breakdown characteris-
tics of gaseous insulation, including equipment service performance data (e.g. from
surveys similar to Reference 15). Further discussion relating to dielectric issues is
presented in section 9.9.1.
Finally, it is anticipated that the continuing, and growing, environmental concerns
relating to SF
6
, which was put on the list of greenhouse gases of the Kyoto Protocol, as
discussed in a valuable CIGRE paper [16] will influence the development of the next
generation of gas-insulated switchgear. The writer recommends that the reader(s) keep
abreast of this developing theme. (This positioning paper can be downloaded from
the CIGREwebsite.) However, he is of the opinion that a commercial replacement for
100per cent SF
6
gas, for interruptionpurposes at the higher ratings, still seems remote!
9.8 Some examples of modern switchgear
9.8.1 SF
6
live-tank and dead-tank switchgear
Live-tank and dead-tank types of SF
6
circuit breaker (see Figures 9.179.19) have
been developed for ratings up to 63 kA at 525 kV. They have given satisfactory
service worldwide for more than 20 years. The particular choice of CB type for any
substationdepends onmanyfactors [4, chap. 8] whichinclude cost of switchgear, level
of atmospheric pollution, potential environmental restrictions, price and availability
of the land, individual preferences, security against third party damage etc.
Figure 9.17 Outdoor installation of 145 kV dead tank switchgear in the USA
(courtesy ALSTOM T&D Ltd)
(i) Live-tank designs: the interrupters in live-tank circuit breakers are housed in
porcelains. The interrupter heads are live and mounted on support insulators on
top of a steel structure to conform with safety clearances. The live-tank circuit
breakers are normally installed in open terminal outdoor substations.
(ii) Dead-tank designs: the interrupters in dead-tank circuit breakers are housed in
an earthed metal tank, usually aluminium, mild or stainless steel, depending on
the current rating [4, chap. 8]. GIS dead tank circuit breakers can be of either
horizontal or vertical configuration. They can be used in GIS indoor and outdoor
substations (see Figure 9.17).
Finally, it can be seen that Figure 9.18a, after Jones [4, chap. 7], provides some useful
interrupter performance characteristics, showing the effect of various parameters
on the thermal recovery characteristics of gas blast interrupters, and Figure 9.18b
shows the strategic incremental improvements of opening energy as a function of the
interrupting technique used and the interrupting capacity expressed in kA.
Turning now to modern distribution switchgear, Figure 9.4 presents an example
of a new generation of reclosure, recently discussed at greater length by Blower [9],
which uses long life vacuum interrupters for arc interruption. In this figure, particular
attention should be directed towards three design aspects:
(i) the disposition of
the vacuum interrupter
the magnetic actuator
the toroidal current transformer, mounted within an SF
6
insulated
gas-sealed aluminium enclosure
single pressure interrupter
synthetic test performance
10kA
32kA
72 kA
150kA
puffer
natural
performance
ramp circuit
two-pressure
interrupter
t
i
30
10
5
3
1
0.3
1 3 5 10
di/dt, A/s
30 50 100
0.5
50
power
circuit
c
r
i
t
i
c
a
l

R
R
R
V
,

k
V
/
s
100
0.1
a
Figure 9.18 Typical interrupter performance characteristics (after Jones
[4, chap. 7])
a effect of various parameters on the thermal recovery characteris-
tics of gas blast interrupters
(ii) the use made of elastomeric bushings with moulded-in capacitive voltage
sensors (which provide good environmental performance and are vandalism
and mishandling-damage resistant)
(iii) signals from the current and voltage sensors can be fed to intelligent digital
relays which can be programmed for a wide range of protection requirements
e.g. autoreclosing functions based on overcurrent, earth-fault and sensitive-
earth-fault protection.
Figure 9.4b illustrates this type of reclosure pole-mounted and connected to an over-
head distribution system. This device apparently uses a low energy consumption
magnetic actuator, capable of thousands of operations. Permanent magnets using
new magnetic materials are used in the actuator energised from small primary type
cells (e.g. lithium) with a claimed life expectancy of up to ten years providing positive
on and off positions for the vacuum interrupter [9].
In passing, the reader should be aware of the exciting recent developments and
improvements in transformer technology (including improved ratings) achieved by
the application of advanced magnetic materials and highly sophisticated design tools
(e.g. [28]).
b
0
500
1000
e
n
e
r
g
y
,

J
D
C
B
A
40.50kA
25 kA
12.5 kA
12.5 kA
12.5 kA
40.50 kA
40.50 kA
25 kA
25 kA
b evolution of opening energy as a function of the interrupting tech-
niques used and the interrupting capacity expressed in kA; the
increasingly simple design and the utilisation of the arc energy
result in a reduction in switching energy
(A) autopneumatic technique
(B) autopneumatic +thermal expansion technique
(C) thermal expansion +rotating arc technique
(D) thermal expansion or rotating arc technique
9.9 Equipment life expectancy: condition monitoring strategies
Before considering modern condition monitoring strategies it is appropriate to look
first at:
(i) the knowledge available to switchgear designers/power utilities in the early
1980s, prior to the development and widespread implementation of CM
methodologies
(ii) the strategic changes in network structure/operation since the advent of the open
market.
9.9.1 Evaluation of solid/gaseous dielectric systems for use
in HV switchgear
As reported by Milne and Ryan [33], the application of synthetic resins in the design
and construction of switchgear for power equipment, originated over 100 years ago
with the introduction of phenolic resins. Since this period, a very large and ever
widening number of synthetic materials and their compounds have been developed
but the one dominating in the heavy electrical industries for the past 50 years is the
epoxy-resin-based system. The aim of good electrical insulation design is to provide
the required electrical characteristic and service life at minimum cost. In making the
opening
mono-blast arrangement
opening
partial-duo blast arrangement 1 exhaust cylinder
2 current carrying contact
3 puffer cylinder
5 arcing contact (fixed)
6 arcing contact (moving)
7 sliding contact
8 molecular sieve
9 PTFE nozzle
4 current carrying tube
opening
duo-blast arrangement
puffer
cylinder
closed position
compression stage
arc extinction
open position
a
moving
contact
interrupting
nozzles
fixed
piston
c
l
o
s
e
d

p
o
s
i
t
i
o
n
o
p
e
n
i
n
g
a
r
c

e
x
t
i
n
g
u
i
s
h
i
n
g
o
p
e
n

p
o
s
i
t
i
o
n
Figure 9.19 SF
6
puffer-type interrupters (after Ali [4, chap. 8])
a principles
puffer cylinder
fixed piston
closed
position
open
position
interrupting nozzles
b
arc
b examples of PTFE nozzles
Notes: puffer type interrupters can be classified according to the
flow of compressed SF
6
gas. They are often referred to as mono,
partial duo, and duo blast interrupters [4]. The nozzle is a vitally
important component of a puffer type interrupter. The interruption
characteristic of an interrupter is governed by nozzle geometry,
shape, size and nozzle material (see Ali [4, chap. 8])
final selection (see examples in Figure 9.19) and deciding on thickness and shape,
the designer is seldom fortunate to have adequate technical data on new alternative
materials applicable to his/her particular design. Because of this, insulation designs in
switchgear change very slowly both in type of material used, component size, shape
and electrical operating stresses. Evaluation of a new insulating material is a costly
and time consuming process, which should include testing under conditions closely
representative of those that are experienced in service.
Solid insulation in switchgear is expensive and a failure at high voltage is
often catastrophic because it is not usually self-restoring. Operating stresses are,
therefore, usually a few orders of magnitude lower than the established intrinsic
electrical strength determined under controlled laboratory conditions. The require-
ment of an insulating material is governed by the specific application and can be
as listed principally as electrical, mechanical, physical and chemical properties.
Frequently, properties other than dielectric are equally as important, such as com-
patibility, mechanical, thermal and processability [20]; consideration of these may
cause an otherwise attractive material to be rejected.
Milne and Ryan [33] reported that medium and high voltage insulation material
research in their company at that period was oriented around the evaluation of basic
material characteristics, followed by studies involving thermal ageing, long termelec-
trical stress ageing (both at power frequency and at higher frequencies, to accelerate
ageing by electrical discharges), environmental testing and multifactor stress ageing
involving in some cases full scale test rigs. This 1984 contribution was primarily
concerned with solid insulation operating in an atmosphere of air or SF
6
, as used in
modern switchgear in the voltage range 11565 kVwhich was (at that time and still is)
a technical subject which does not achieve the same prominence in the literature as
other insulating media such as gases or liquids.
Table 9.3 provides a convenient summary of insulation materials, characteristics
of interest and evaluation procedures for switchgear application, listing:
material characteristics of interest
evaluation procedures before application in switchgear (prior to the application
of approved CM techniques for subsequent in-service condition monitoring and
asset management of the switchgear)
certain electrical/mechanical properties.
Clearly, it is possible to build on such earlier physical characterisation studies
[4, 13, 14, 21] to create extensive databases which can be integrated into modern
in-service condition monitoring techniques which will be briefly considered in
section 9.9.3.
9.9.2 Open market: revised optimal network structure
During the past 12 years, there have been massive strides in the liberalisation of the
electricity supply industry in the UKand worldwide [2224]. The main driving forces
for network development before liberalisation were increase of demand, connection
of new power stations built to meet the demand and high reliability of the network.
The cost incurred to develop the network was borne by the customers [24, 4, chap. 1].
Some consequences of this practice [24] were that:
reliability and voltage quality increased continuously
the technical complexity of the networks gradually increased by the use of the
latest technology components
the economic point of view was of less importance.
Table 9.3 Insulation materials: characteristics of interest and
evaluation procedures (after Ryan and Milne)
Characteristics of interest
Electrical Mechanical
electrical strength tensile
surface flashover flexural
surface resistivity impact
volume resistivity hardness
permeability
loss angle (tan ) Chemical
tracking resistance compatibility
dry arc resistance resistance
voltage endurance environmental
stability
Thermal
heat distortion Physical
expansion density
ageing moisture absorption
flammability weathering
toxic gases fungi resistance
Evaluation procedures for switchgear applications
basic material characteristics
short term testing
life testing
thermal ageing
environmental ageing
voltage stress ageing
full scale duration testing of switchgear assemblies
Note: references to further strategic information relating to the dielectric perfor-
mance of gaseous insulation and solid support insulation, or barriers, as used in
high voltage SF
6
insulated switchgear systems are given in chapter 4 of Reference 4
After liberalisation (deregulation), the electricity industry worldwide has gone
through several changes in organisation and structure, attitude to cost and relationship
with customer [24, 4, chap. 1]. Some examples of the change of roles have recently
been given [24]:
after roles are separated with different ownership or control, the power producers are organ-
ised for competition, the transmission companies have become natural monopolies, the
consumers can shop around for cheap power and good service, and investments in new
technology must result in increased profit or reduced cost. There is a tendency to consider
the short termrather than take a longer termviewin decision making. This newenvironment
for the network owners has resulted in changed planning procedures with emphasis put on
economic performance.
It can be appreciated fromthe foregoing, and froma thorough evaluation of recent
excellent CIGRE task force reports, that:
one of the effects of market liberalisation is the unbundling of roles within the
electricity supply industry
new principal role categories are those who manage the transmission networks,
those who use the transmission networks to trade, those who facilitate energy
trading and statutory authorities
major stakeholders, identified as the demand customers, power producers system
operator, network owners and regulator, exert new and different driving forces,
and the long standing driving forces have reduced in importance
the authors also studied the effects of the new driving forces upon the plan-
ning processes and propose a methodology for accounting for them in a
pseudo-optimisation.
From the foregoing paragraphs, it can now be appreciated that the effective manage-
ment of assets to ensure that the user obtains the optimumlife for the plant is becoming
more vital as electricity transmission and distribution systems are operating to dif-
ferent criteria and are being worked harder [4, chap. 23, 2224]. Lowen [24] provides
a high level view of the electricity industry and considers the effects of reform in the
industry since privatisation from the perspective of the England and Wales transmis-
sion system operator (National Grid). He considers that the single biggest driver for
the future is the world growth in electricity generation and consumption. Based on
the levels in 1995, forecasts show that there will be a 54 per cent and 33 per cent
increase in energy and electricity consumption, respectively, in 2015 worldwide.
9.9.3 Condition monitoring strategies
Recently, in a chapter devoted to Condition Monitoring (CM) of high voltage
equipment, White [4, chap. 22] has:
(i) equated the asset value of electrical plant owned by the electricity supply
generation and distribution industry in the UK as billions of pounds sterling
(ii) estimated the replacement value of the transformers alone on the National Grid
to be about 1 billion pounds sterling.
White goes on to observe that:
to be able to buy, fit and forget such valuable assets is not an option in todays
financial environment
unless a transformer is operating or is operational, it is worth only the value of its
recyclable components
an asset can very quickly lose its value if it is not correctly managed to give
optimum life performance
management of an asset such as a transformer will almost certainly involve CM
in some form or other.
Similarly, CM techniques are now in use worldwide to monitor a variety of network
equipment, including overhead lines, GIL, cables, motors, generators and switchgear,
and are the focus of this section. Clearly, the safety of personnel is essential for all
CM studies whatever the network equipment being monitored or the particular CM
technique adopted.
To be able to effectively manage the assets of a network system, it is important to
knowwhether an itemof HVequipment continues to remain in a serviceable condition
or not. Fortunately, this can be assessed by monitoring many different quantities using
a wide variety of techniques hence the adoption of the term condition monitoring
or in-service condition monitoring. As the term CM suggests, some form of ongoing
measurement process is performed periodically and comparisons made with earlier
measurements on the same item of equipment to determine whether this item of
equipment is still in a serviceable condition. As Coventry et al. [4, chap. 22] have
recently pointed out in a valuable overview on CM of HV equipment, this topic now
attracts hundreds of papers per year and adopts methodologies ranging in complexity
fromsimple manual checks, at one extreme, to state of the art computer-based systems
at the other.
It must be recognised at the outset that a vast amount of design data have been
accrued for many years by equipment designers and network users relating to strate-
gic information concerning the design, construction, performance characteristics of
equipment and in-service performance of similar items of equipment [1317]. In addi-
tion, a recent report [19], briefly summarising the findings of a second CIGRE survey
concerned with GIS service reliability, considers that GIS technology has contributed
very effectively to increasing the reliability of new substations and to improving the
asset lifecycle of existing ones. The report provides a valuable database of GIS service
experience survey on SF
6
-insulated equipment relating to more than 13 500 circuit
breaker bays and 118 500 bay years service. Much valuable information is presented
in the report which provides an analysis of results on installations and GIS major fail-
ure report data, including general data about GIS installations, data concerning GIS
failure frequencies, major failure characteristics and consideration of life expectancy,
maintenance and environmental issues [19]. This work, typical of CIGRE surveys
in other sectors of electrical plant, provides a valuable resource and benchmark for
those involved in condition monitoring methodologies including users and manufac-
turers operating within the area of GIS substation planning, design, construction and
in-service performance.
Three broad aims of CM were recently identified by Coventry [4, chap. 22]:
(i) To prevent component/network failures which can have strategic consequences
on network safety and costs. Electrical failure of an item of equipment can
result in a short circuit fault which can result in significant damage (and some-
times even catastrophic damage). Consequently, there is an obvious benefit of
adopting CM methodologies which can provide advanced warning of potential
problems relating to an item of equipment.
(ii) To enable maintenance programmes to be planned/performed within an
appropriate timescale, according to the latest measured condition of the
equipment.
(iii) To estimate the remaining life of the item of equipment, based on its present
condition, together with knowledge on the rate of degradation. This approach
allows equipment to be used closer to the end of their life and facilitates the
implementation of an effectively planned asset replacement strategy (e.g. see
Reference 4, chaps 2123).
Some examples, originally reported by Coventry and Jones [4] are reproduced in
Figures 9.209.24 and in associated text below describe the application of a variety
of strategic condition monitoring methods based on collaborative studies in the UK
relating to intelligent CM strategies. Aspects illustrated include:
(i) Figure 9.20, which illustrates a UHVcoupler, defect recognition and decision
support approach used in GIS equipment [4, chap. 22]. Figure 9.20a shows
a schematic of a UHF coupler, perhaps the most commonly used sensor for
the detection of UHV signals in GIS:
the coupler consists of a disc-shaped electrode mounted on a hatch cover
plate of a GIS chamber and connected to the outside world by means of
a gas-tight feed-through
as a precaution, the coupler may be shunted by a resistance of a few kilo
ohms to reduce the level of the voltage induced by the incident electric
field at power frequency
in the UK, it is current practice for all new GIS at 420 kV to be equipped
with UHV couplers as standard equipment; satisfactory sensitivity is
achieved with the couplers spaced at intervals of up to 20 m.
(ii) Figure 9.20b and c provide typical discharge characteristics. Curve b
shows equivalent circuit and discharge pattern with poor contact between
conducting components (e.g. where a stress shield has become loose and this
poorly bonded component will tend to charge capacitively so that the potential
difference across the contact gap rises until the gap sparks over). This spark
reduces the potential difference across the contact gap to zero [4, chap. 22].
(iii) Figure 9.20c, shows typical discharge pattern with free conducting particles
(e.g. this pattern is accumulated over a number of cycles). A free particle
moving under the influence of the electric field may initiate breakdown if
it enters a region of critical electric field. Coventry [4, chap. 22] points out
that the behaviour of different types of defect has been studied extensively
in a library environment to assist with defect recognition and risk analysis.
Information on defects detected in service is normally maintained in a database
to assist further decisions. (Note: the reader should also be aware that such
characterisation of discharge patterns has also been known (and accrued) for
more than 30 years from first generation SF
6
-insulated GIS, and since the
1960s for 300/420 kVSF
6
-insulated instrument transformers which were first
developed by UKmanufacturers in conjunction with the then UKtransmission
utilities. GIS and free standing SF
6
instrument transformers have been very
equivalent circuit
time, ms
0 10 20
discharge pattern
dB
(voltage across contact gap
in absence of breakdown
shown dotted)
coupler
electrode
feedthrough
and connector
hatch cover plate
a
b
voltage across contact gap
Figure 9.20 UHV coupler defect recognition and decision support (after Coventry
[4, chap. 22])
a schematic of UHV coupler
b, c discharge patterns
b poor contact between conducting components
0 c 10 20
discharge pattern
time, ms
dB
power frequency
voltage
c free conducting particles
also included: defect recognition and decision support
reliable and have seen more than 30 and 40 years service, respectively, in the
UK grid and elsewhere.)
(iv) WhenUHFsignals are detected, decisions must be made onwhether corrective
action is necessary and in what timescale. Information may be obtained from
the UHF signals on which to base these decisions. Frequently, it is possible to
recognise the type of defect and the way in which it is discharging from the
point on the power frequency wave at which discharges occur.
Figure 9.20b shows an example of poor contact between conducting com-
ponents, such as where a stress shield has become loose. The poorly bonded
component will tend to charge capacitively so that the potential difference
across the contact gap rises until the gap sparks over. The spark reduces the
potential difference across the contact gap to zero.
A free conducting particle, such as a piece of swarf, lying on the floor
of the GIS enclosure becomes charged by induction under the influence of
the powerfrequency electric field. The induced charge may be sufficient to
overcome the particles weight, so that it moves under the combined influ-
ence of the electric field and gravity. Coventry comments that the particle
may return to the enclosure at any point on the powerfrequency wave. In
general, this will be a different point on the wave to that at which it lifted
p
a
r
t
i
a
l

d
i
s
c
h
a
r
g
e

d
e
t
e
c
t
i
o
n

n
e
t
w
o
r
k
p
a
r
t
i
a
l

d
i
s
c
h
a
r
g
e

d
e
t
e
c
t
i
o
n

n
e
t
w
o
r
k
o
u
t
s
t
a
t
i
o
n
m
a
s
t
e
r

s
t
a
t
i
o
n
f
u
t
u
r
e

o
u
t
s
t
a
t
i
o
n
o
u
t
s
t
a
t
i
o
n
F
i
g
u
r
e
9
.
2
1
G
I
S
c
o
n
t
i
n
u
o
u
s
m
o
n
i
t
o
r
i
n
g
s
y
s
t
e
m
a
n
d
N
G
C
p
a
r
t
i
a
l
d
i
s
c
h
a
r
g
e
m
o
n
i
t
o
r
i
n
g
s
t
r
a
t
e
g
y
(
a
f
t
e
r
C
o
v
e
n
t
r
y
[
4
,
c
h
a
p
.
2
2
)
x y
nozzle
top electrode
z
35 mm
arc
optical
fibre 1
16 mm
bottom electrode
a
0
r
arc
nozzle
photo diode 2 optical
fibre 2
photo
diode 3
optical
fibre 3
photo
diode 4
optical
fibre 4
photo diode 1
x
y
z
optical
fibre 1
optical
fibre 3
Figure 9.22 Arc optical emission probes deployment on various types of circuit
breaker (after Coventry [4, chap. 22])
a SF
6
puffer
off, and a discharge occurs as charge is exchanged between the particle and
the enclosure. The particle bounces and the sequence continues. The particle
moves comparatively slowly, typically returning to the enclosure once every
few cycles. The resulting pattern is characterised by discharges occurring at
any point on the powerfrequency wave and a repetition rate of one discharge
every few cycles. A typical pattern, accumulated over a number of cycles, is
shown in Figures 9.20c. A free conducting particle moving under the influ-
ence of the electric field may initiate breakdown if it enters a region of critical
electric field.
Several researchers have studied the behaviour of different types of defect
has been studied extensively in the laboratory to assist with defect recognition
and risk analysis. Information on defects detected in service is maintained in a
database to assist future decisions. Artificial intelligence techniques are being
introduced for defect recognition and decision making in GIS.
r
[
0
detector 1
detector 3
detector 2
arc
/6
/6
/3
/3
annular contact
with radius R
c
:=3
:=5/2
:=2
:=3/2
:=
[
=
7
/
6
[
=
4
/
3
[=3/2
[=5/6
[
=
2/3
[=5/3
[=11/6
[=0
[=
[
=
/
6
[
=
/
3
[
=
/
2
0.8
d
o
m
i
n
a
n
t

a
n
g
l
e
0.4
A
nominal radius
b
b SF
6
spinner
(v) Figure 9.21 presents a schematic of a continuous monitoring systemfor partial
discharge monitoring in GIS.
Coventry [4, chap. 22] reports that this system consists of a partial
discharge detection network which collects and transmits data from the
UHV couplers to an outstation
fixed top contact
40 mm
75 mm
72 mm
70 mm
1
4
2
3
6
5
8
7
11
10
9
12
13
14
15
16
fibre plane 1
fibre plane 2
fibre plane 3
fibre plane 4
moving lower contact
optical fibres
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
c
interruption chamber
synthetic oil mineral oil
interruption chamber
actuator
control system
c oil
current
voltage
gas pressure
particle concentration
acoustical vibration
stalk temperature
piston chamber
gas pressure
arc radiation
dielectric strength
contact travel
Figure 9.23 Location of various optical fibre sensors on a model high voltage SF
6
circuit breaker (after Jones et al. [4, chap. 22])
the outstation consists of PCperipherals and software to control the partial
discharge detection network and store the collected data
outstations at a number of substations with GIS connected by modem link
to a master station at the headquarters location
data are transferred from the outstations automatically and at least once
per week [4, chap. 22].
(vi) Coventry comments that this continuous monitoring system (Figure 9.21) is
intended to produce more effective cover than that provided by routine manual
2
0.5
0.4
0.3
0.2
0.1
0
1
0
s
i
g
n
a
l

m
a
g
n
i
t
u
d
e
,

V

l
i
g
h
t
n
e
s
s
1
2
0 0.02 0.04 0.06 0.08 0.1
time, s
2
1
0
s
i
g
n
a
l

m
a
g
n
i
t
u
d
e
,

V

1
2
0
10 2510 5010 7510 10010 12510 15010 17510 20010
0.02 0.04 0.06 0.08 0.1
time, s
0 a
b
c
14.3 31.1
arc current, kA
arc current, kA
44.1 51.9 59.5
200
190
180
170
160
150
140
0 14.3 31.1 44.1 51.9 59.5
minimum (windows 8 to 11)
average (windows 8 to 11)
maximum (windows 8 to 11)
1.00
0.98
0.96
0.94
0.92
s
a
t
u
r
a
t
i
o
n
h
u
e

a
n
g
l
e
,

d
e
g
r
e
e
s
0.90
0.88
0.86
0.84
0 14.3 31.1 44.1 51.9 59.5
0.4
0.3
0.2
0.1
0
p
o
w
e
r

s
p
e
c
t
r
a
l

d
e
n
s
i
t
y
cb25ap01 acoustic signal from puffer
31.1 kA
acoustic signals from 145 kV puffer
chromatic analysis of arcing period
cb24ap01 acoustic signal from puffer
cold flow
ft245ap3 FFT of acoustic signals
time =0.04s to 0.055s
1=14.3 KA 1=44.1 KA 1=59.5 KA
frequency, Hz
arc current, kA
Figure 9.24 Optoacoustic vibration monitoring (after Jones [4, chap. 21])
a typical time variationof acoustic signals fromanSF
6
puffer circuit
breaker without a fault current and with a 31.1 kAfault
b power spectral density curves for acoustical signals from an SF
6
puffer circuit breaker for part of the fault current arcing period
c signal discrimination in terms of time window compressed
coordinates L.H.S
surveys and to reduce its demands on engineers time. The system is able to
detect, store and display information on all defects that occur. It operates
unattended, and an engineer need only intervene when the system indicates
that captured data require their attention.
The normal mode of operation is for the system to signal when captured
data require the attention of an engineer. Such an indication is generated where
UHF signals are captured for the first time or where a change occurs in the
behaviour of UHFsignals as indicated by preset thresholds based on amplitude
or count rate. On being notified by the system of the above condition, the
engineer may retrieve and display partial discharge data on which to base
decisions on action to be taken.
Coventry comments that the system must be designed to provide the engi-
neer with sufficient information on which to base decisions, without handling
and storing excessive quantities of data. Typically, displays showing the point
on the powerfrequency wave at which discharge occurs (phase-resolved dis-
play) and displays showing changes in discharge behaviour over a period of
time (trend analysis) are provided.
(vii) Coventry states that all GIS on the NGC system are monitored for partial dis-
charge using the UHF technique. The decision to use continuous rather than
periodic monitoring is based on a costbenefit analysis. Continuous monitor-
ing systems are installed at the most strategic GISwhere the costs of unplanned
outages are higher.
(viii) Figure 9.22 illustrates some arc optical emission probes (and indicates typical
dispositions of sensor/detectors) deployed on various types of circuit breakers.
Parts a, b and c relate to SF
6
puffer, SF
6
spinner and oil-interrupters,
respectively (for further information, see Coventry [4, chap. 22]).
Since the early 1980s, the fundamentals of optical fibre systems (OFS) have been well
covered by G.R. Jones [4, chap. 21] and his research group at Liverpool University.
In numerous quality research papers, these researchers have demonstrated that optical
fibre technology has the scope of being used for a variety of purposes within the
electrical power systems sector as:
it has the advantage of inherent bandwidth availability and relative immunity to
lightning strokes and EMI and EM interference
it is becoming more attractive and increasingly economically viable year on year
as data transfer rates of local area networks (LANs) increase
research into optical fibre based parameter sensing has progressed well and Jones
considers that it is at the stage where properly engineered systems are becoming
available utilising purely optical monitoring.
Jones [4, chap. 21] describes examples of optical fibre sensors, such as for measuring
electrical, mechanical, aerodynamic/thermodynamic and chemical monitoring which
his research group has investigated for monitoring parameters such as in high voltage
equipment and systems,
As examples, consider:
(i) Figure 9.23, which illustrates the deployment and monitoring locations
of various optical fibre sensors on a model high voltage SF
6
circuit breaker.
Specific parameters monitored by Jones et al. include current, voltage, gas pres-
sure, contact stalk temperature, contact travel, particulate concentration and arc
radiation. Jones has discussed a range of different forms of optical fibre based
current transducers investigated by his research group at Liverpool University
[4, chap. 21]. Jones and his colleagues found that by winding an optical fibre
around part of the circuit breaker structure acoustical and mechanical vibrations
produced by the arcing and other mechanical effects associated with the opera-
tion of the CB can be conveniently monitored [4, chap. 21]. In a similar manner
Jones states that analysis of other features from acoustical records taken from
tests on CBS, using chromatic approaches, yields deeper levels of information
regarding the mechanical operation of the circuit breaker e.g. hydraulic drive
pressure [4, chap. 22].
(ii) Finally, Figure 9.24 summarises certain strategic information attained by Jones
et al. [4, chap. 21], using optoacoustic vibration monitoring. The reader is
directed to the original sources for fuller explanation(s) of the complex phe-
nomena represented herein, but brief consideration of subcaptions a, b and c
to this figure provides the reader with an appreciation of how interpretation of
time-dependent acoustical signals, which are representative of records obtained
during the operation of an SF
6
puffer circuit breaker under various fault arcing
current conditions, has led to effective vibrational characterisation of practical
circuit breakers. Moreover, the writer of this chapter wishes to emphasise that
the ongoing high quality fundamental research by G.R. Jones et al. at Liverpool
for over 20 years, working in association with manufacturers and utilities, has
resulted in:
incremental improvements in the knowledge of fundamental/physical
characterisation of arc interruption performance phenomena; this has
resulted in improved computer modelling and computer prediction of
CB interruption performance capabilities which can now result in the
attainment of near optimal performance levels
the meaningful and effective development of a variety of intelligent
in-service condition monitoring, CM, methodologies nowfinding increas-
ing application in practical equipment installed in modern transmission
networks.
Recent CIGRE working group reports [19, 20, 25] considers the excellent reliabil-
ity and availability of modern switching equipment which continues to get better
with the introduction of modern improved designs of CM and DT devices. The role
of condition monitoring systems (CMS) has assumed even greater importance with
the recent thrust to work network systems harder [2224]. In the future, mainte-
nance of switchgear will increasingly only be carried out when the condition of the
equipment warrants intervention [25]. This approach provides the user with bene-
fits of reduced cycle costs, improved availability due to fault prevention and the
ability to plan for any outages necessary for maintenance. Moreover, the applica-
tion of CMS together with other new technologies offers the user other advantages
such as increased functionality and performance enhancement [20, 25]. (C.J. Jones
also points out that this development creates a need for guidelines regarding the
efficient utilisation of this increased functionality, see also earlier comments in
section 9.6.)
The application of CMS and diagnostic techniques to substation equipment (when
adopting approved operational guidelines) achieves a variety of benefits for the
equipment operator:
maintenance prediction
failure prevention
improved commissioning tests
more accurate end of life assessments.
C.J. Jones, convener of W.G.13.09 CIGRE brochure 167 [20], has recently distilled
the contents of this report into a later conference paper [25]. The following aspects
are covered in this guideline brochure [20], by section:
1 definitions given for the commonly used terms in the area of monitoring and
diagnostics (necessary given the diversity of understanding that exist for the same
terms)
2 the need for monitoring is discussed, based on the objectives of the equipment
technology being considered
3 justification of the application of monitoring and diagnostics is analysed and
examples given of possible approaches (note that a single approach cannot be
prescribed for this process)
4 state of the art review of the diagnostic techniques and sensors suitable for the
application to switching equipment
5 requirements for the design and testing of the switching equipment and the
monitoring systems considered
6 the issue of dependability, monitoring and diagnostics is covered
7 management of information; strategy for how data from sensors are handled to
provide useful information
8 consideration of future of monitoring and diagnostics
9 main conclusions and report recommendations are given.
Item 9 in the above report includes the following information:
There is now clear evidence that the application of diagnostic techniques DT and
CM can bring benefits to users and suppliers of switching equipment. (These
benefits can generally be given some financial value.)
The application of diagnostics and condition monitoring will become more
commonplace in the future.
There is a trend towards the use of continuous online CM to supplement offline
or periodic monitoring.
Users require simple information in a timely manner rather than large quantities of
data which require analysis. Monitoring systems that incorporate expert systems
to provide such information will be of greater benefit to users.
Avariety of sensors, intelligent electronic devices (IEDs) and monitoring systems
already exists in substations. Unfortunately, at present, there is little standardis-
ation on the communication of data. The work that is currently ongoing in this
area will allow significant benefits to be achieved.
It is not possible to provide a single approach that will cover the justification of
the implementation of CM to switching equipment.
From the utilities point of view, in some cases it is much more important to
monitor the old, unreliable equipment than the new equipment. Unfortunately,
this task is not easy to accomplish.
An essential matter in the context of diagnostic testing and monitoring is the
interpretation of the measurements. In some cases, it is straightforward to convert
measurements to useful information about the condition of the equipment, in other
cases it is not.
It is suggested that the supplier of sensors and monitoring equipment should be
responsible for ensuring the continued availability of spare parts or interchange-
able units for at least ten years fromthe date of final manufacture of the monitoring
equipment. This includes also the tools that are necessary for maintenance and
commissioning of these spare parts.
The CIGREWorking Group.13.09 (nowdisbanded), identified the need for future
work in the area of lifecycle cost evaluation not just for circuit breakers but all
primary equipment.
C.J. Jones has recently reported that his company has had online condition monitoring
experience since 1991 and offered the following brief perspectives [25a]:
some benefits of CM are directly relevant to the manufacturer such that sensors
can really be fitted at no cost to the user
as we apply substation information systems it becomes obligatory to utilise
extensive CM and DT on switching equipment (and other items of plant)
try to keep the diagnostic technique as simple as possible (i.e. coil, current?)
a hierarchy of diagnostic techniques can be applied
high level on an online basis
low level on an as needs basis
monitoring must be matched to the equipment technology
dont fall into the trap of monitoring something because it can be monitored
processing/communications technology has developed dramatically, so think
about what is useful not what is possible, as things will become possible
changing as fast as technology are utilities, so consider training issues; systems
will need to give not data nor information but decisions.
Finally, further important aspects of CM strategies, considered and discussed by
WG.13.09 and C.J. Jones [20, 25], are reproduced in Tables 9.4 and 9.5.
9.9.4 General discussion
As mentioned earlier, the use of diagnostic techniques is not new having been widely
used and developed for more than 30 years. In the UK, a (DTI) government initiative
in the early 1980s encouraged research into intelligent switchgear etc. and, after a
slow start, great progress has been made on this theme. The current intense interest
in the application of diagnostics and condition monitoring, CM, techniques to HV
T
a
b
l
e
9
.
4
S
F
6
c
i
r
c
u
i
t
b
r
e
a
k
e
r
a
n
d
c
i
r
c
u
i
t
s
w
i
t
c
h
e
s
(
a
f
t
e
r
C
.
J
.
J
o
n
e
s
)
(
r
e
p
r
o
d
u
c
e
d
f
r
o
m
T
a
b
l
e
2
.
1
,
C
I
G
R
E
W
.
G
.
1
3
.
0
9
[
2
0
]
)
F
u
n
c
t
i
o
n
/
p
a
r
a
m
e
t
e
r
c
o
n
t
i
n
u
o
u
s
l
y
m
o
n
i
t
o
r
e
d
(
C
)
o
r
p
e
r
i
o
d
i
c
a
l
l
y
d
i
a
g
n
o
s
e
d
(
P
)
E
q
u
i
p
m
e
n
t
o
p
e
r
a
t
i
o
n
/
s
t
a
t
u
s
F
a
i
l
u
r
e
p
r
e
v
e
n
t
i
o
n
M
a
i
n
t
e
n
a
n
c
e
s
u
p
p
o
r
t
/
l
i
f
e
a
s
s
e
s
s
m
e
n
t
O
p
t
i
m
u
m
o
p
e
r
a
t
i
o
n
C
o
m
m
i
s
s
i
o
n
i
n
g
t
e
s
t
s
n
e
e
d
r
a
n
k
n
e
e
d
r
a
n
k
n
e
e
d
r
a
n
k
n
e
e
d
r
a
n
k
n
e
e
d
r
a
n
k
S
w
i
t
c
h
i
n
g
:
o
p
e
r
a
t
i
o
n
t
i
m
e
s
C
t
o
d
e
t
e
c
t
a
n
y
a
b
n
o
r
m
a
l
c
o
n
d
i
t
i
o
n
H
C
t
o
e
v
a
l
u
a
t
e
C
B
c
o
n
d
i
t
i
o
n
H
C
c
o
n
t
r
o
l
l
e
d
s
w
i
t
c
h
i
n
g
E
P
t
o
e
s
t
a
b
l
i
s
h
r
e
f
e
r
e
n
c
e
d
a
t
a
H
p
o
l
e
d
i
s
c
r
e
p
a
n
c
y
C
t
o
d
e
t
e
c
t
a
n
y
a
b
n
o
r
m
a
l
c
o
n
d
i
t
i
o
n
H
C
t
o
e
v
a
l
u
a
t
e
C
B
c
o
n
d
i
t
i
o
n
H
C
c
o
n
t
r
o
l
l
e
d
s
w
i
t
c
h
i
n
g
H
P
t
o
e
s
t
a
b
l
i
s
h
r
e
f
e
r
e
n
c
e
d
a
t
a
H
a
r
c
i
n
g
t
i
m
e
C
t
o
d
e
t
e
c
t
a
n
y
a
b
n
o
r
m
a
l
c
o
n
d
i
t
i
o
n
L
C
c
o
n
t
r
o
l
l
e
d
s
w
i
t
c
h
i
n
g
H
c
o
n
t
a
c
t
s
v
e
l
o
c
i
t
y
C
t
o
d
e
t
e
c
t
a
n
y
a
b
n
o
r
m
a
l
c
o
n
d
i
t
i
o
n
H
C
c
o
n
t
r
o
l
l
e
d
s
w
i
t
c
h
i
n
g
H
m
a
i
n
c
o
n
t
a
c
t
s
p
o
s
i
t
i
o
n
C
t
o
d
e
t
e
c
t
a
n
y
a
b
n
o
r
m
a
l
c
o
n
d
i
t
i
o
n
E
c
o
n
t
a
c
t
w
e
a
r
(
I
2
t
)
C
t
o
s
u
p
p
o
r
t
o
v
e
r
h
a
u
l
p
l
a
n
n
i
n
g
H
M
e
c
h
a
n
i
c
a
l
d
r
i
v
e
:
n
u
m
b
e
r
o
f
o
p
e
r
a
t
i
o
n
s
C
o
r
P
t
o
s
u
p
p
o
r
t
o
v
e
r
h
a
u
l
p
l
a
n
n
i
n
g
H
P
t
o
e
s
t
a
b
l
i
s
h
r
e
f
e
r
e
n
c
e
d
a
t
a
H
s
t
o
r
e
d
e
n
e
r
g
y
(
s
p
r
i
n
g
p
o
s
i
t
i
o
n
,
p
r
e
s
s
u
r
e
)
C
t
o
a
s
s
e
s
s
C
B
a
v
a
i
l
a
b
i
l
i
t
y
E
C
t
o
a
s
s
e
s
s
C
B
a
v
a
i
l
a
b
i
l
i
t
y
E
C
c
o
n
t
r
o
l
l
e
d
s
w
i
t
c
h
i
n
g
E
l
a
t
c
h
p
o
s
i
t
i
o
n
C
t
o
d
e
t
e
c
t
a
n
y
a
b
n
o
r
m
a
l
c
o
n
d
i
t
i
o
n
H
v
e
l
o
c
i
t
y
C
t
o
d
e
t
e
c
t
a
n
y
a
b
n
o
r
m
a
l
c
o
n
d
i
t
i
o
n
H
P
t
o
e
v
a
l
u
a
t
e
C
B
c
o
n
d
i
t
i
o
n
H
P
t
o
e
s
t
a
b
l
i
s
h
r
e
f
e
r
e
n
c
e
d
a
t
a
H
v
i
b
r
a
t
i
o
n
f
i
n
g
e
r
p
r
i
n
t
P
t
o
e
v
a
l
u
a
t
e
C
B
c
o
n
d
i
t
i
o
n
L
P
t
o
e
s
t
a
b
l
i
s
h
r
e
f
e
r
e
n
c
e
d
a
t
a
L
s
t
a
t
e
o
f
m
e
c
h
a
n
i
s
m
n
u
m
b
e
r
o
f
m
o
t
o
r
p
u
m
p
s
t
a
r
t
s
/
m
o
t
o
r
c
u
r
r
e
n
t
/
r
u
n
n
i
n
g
t
i
m
e
C
t
o
d
e
t
e
c
t
a
n
y
a
b
n
o
r
m
a
l
c
o
n
d
i
t
i
o
n
H
C
t
o
d
e
t
e
c
t
a
n
y
a
b
n
o
r
m
a
l
c
o
n
d
i
t
i
o
n
H
P
t
o
e
v
a
l
u
a
t
e
C
B
c
o
n
d
i
t
i
o
n
H
c
h
a
r
g
i
n
g
c
u
r
r
e
n
t
/
t
i
m
e
(
m
o
t
o
r
p
r
o
t
e
c
t
i
o
n
)
C
t
o
d
e
t
e
c
t
a
n
y
a
b
n
o
r
m
a
l
c
o
n
d
i
t
i
o
n
H
C
t
o
d
e
t
e
c
t
a
n
y
a
b
n
o
r
m
a
l
c
o
n
d
i
t
i
o
n
H
P
t
o
e
v
a
l
u
a
t
e
C
B
c
o
n
d
i
t
i
o
n
H
r
a
n
k
:
E
e
s
s
e
n
t
i
a
l
;
H
h
i
g
h
;
L
l
o
w
Table 9.5 Diagnostic techniques and sensors for testing the switching function of
circuit breakers and other types of switching equipment (reproduced
from Table 4.3, CIGRE WG.13.09 [20]) (after C.J. Jones)
Parameter Application(s) Method/sensor
position of primary
contacts
all position transducer, e.g.:
auxiliary switch, contactor C
electronic proximity sensor C
optical sensor C
contact travel
characteristics (position,
velocity, acceleration)
all dynamic position sensor with
analogue output:
resistance potentiometer C, P
magneto-resistive sensor C, P
linear variable differential
transformer
C, P
dynamic position sensor with digital
output:
optical with incremental coding C, P
optical with non-incremental
(absolute) coding
C, P
operating time all electrical recording of time to
close/open of primary circuit
P
GS, D motor running time C, P
pole discrepancy in
operating times
all electrical recording of time to
close/open of primary circuit
C, P
GS, D motor running time C, P
arcing time all combined recording of load current
profile and travel characteristics
P
arcing contact wear CB accumulated I
2
t
arc
by:
current and time measurement C
statistical estimates P
all dynamic contact resistance P
Abbreviations:
CB: SF
6
, air blast, minimum oil and oil tank circuit breakers
GS: grounding switches
D: disconnectors
C: used for continuous monitoring
P: used for periodic diagnostic testing
switchgear and switching equipment and also a wide range of other HV equipment,
including transformers, cables, overhead lines, surge arrestors, motors etc. has gained
further impetus with the increased emphasis on working network systems harder and
the strong focus on asset management (see section 9.9.2 and References 21 to 24).
As C.J. Jones [25] and others have pointed out, the primary need for CM systems
in the switchgear sector has been driven mainly by equipment manufacturers to
support a strong move from predetermined interval based maintenance to predictive
(or condition-based) maintenance.
Without doubt, we are witnessing a transitional period with regard to diagnostics
and CM strategies, because of increased utility interest in this subject area [20], as a
holistic network-wide approach is developed. Subsequently, the present writer antic-
ipates that in the future one will look at not only the benefits of increased equipment
availability, through a range of diagnostic and condition monitoring strategies, but the
increased functionality will facilitate monitoring asset management, life extension.
It is also likely that significant improvements in software support will result in the
widespread adoption of evolving data warehousing techniques which will facilitate
greatly expanded data evaluation capability to provide even better, i.e. a more com-
prehensive, diagnostic monitoring service and possible integration with SCADAand
other systems.
The topic of condition monitoring has assumed increasing strategic importance as
the electrical power sector drives relentlessly towards cost saving, asset management,
long life and improved plant performance [4, 2124]. In this chapter, it has not been
possible to critically review strategic issues associated with CM methodologies but
rather to indicate, and provide a brief flavour of the wide range of published work in
this area supported by generous referencing of strategic publications [4, 2125].
Arecent IEE event (within the HVET2002 international school) described a wide
range of monitoring systems used in the various sectors of the power industry, with
contributions fromexperts covering developments fromthe perspectives of electricity
supply industry, ESI, users, manufacturers, equipment designers, system planners,
monitoring equipment designers, consultants and academics.
A broad range of contributions were presented, with reference to earlier strate-
gic studies by these authors (including coverage of intelligent condition monitoring
systems as applied to GIS, transformers, arresters, bushings, cables etc.), who dis-
cussed many strategic techniques such as RFemissions, acoustic emissions, dissolved
gas analysis, optical fibre sensing, remote E-field monitoring etc. This event clearly
reflected howCMof electrical equipment is assuming an increased importance as the
commercial drive to improve efficiency is pursued.
9.10 Summary
This chapter has provided a general introduction to the vast subject of circuit breakers,
interruption strategies, characterisation and performance evaluation of gaseous/solid
insulation systems, the identification of strategic online condition monitoring (CM)
methods and diagnostic techniques (DT) as applied to modern switchgear systems,
before and post deregulation. Although it has only been possible to touch very
briefly on these aspects together with certain related strategic issues including sub-
station control systems (SCS), also working networks harder and environmental
matters, important strategic issues have been considered briefly and a generous list
of references, including CIGRE electronic sources is included, which will enable the
reader to explore this developing topic in greater depth.
9.11 Acknowledgements
The author wishes to thank the directors of NEI Reyrolle Ltd (now VA TECH
REYROLLE) and several other organisations for permission to publish earlier papers
which have been briefly referred to in this chapter. He has called heavily on mate-
rials originally published by CIGRE, Professor G.R. Jones, Mr C.J. Jones, Mr S.M.
Ghufran Ali, Mr W.J. Laycock and Dr P. Coventry to whom he extends grateful
thanks. He also gratefully acknowledges the assistance given and contributions made
by many of his former colleagues and students at Reyrolle, the University of Sun-
derland, Cigrean colleagues and those within the industrial, IEC, IEE Professional
Groups and academic communities worldwide and for their help and generous support
over the years.
9.12 References
1 FLURSCHEIM, C.H. (Ed.): Power circuit-breaker theory and design (Peter
Perigrinus Ltd, London, 1982)
2 RAGALLER, K.: Current interruption in high voltage networks(PlenumPress,
1978)
3 RYAN, H.M., and JONES, G.R.: SF
6
switchgear (Peter Perigrinus Ltd,
London, 1989)
4 RYAN, H.M. (Ed.): High voltage engineering and testing (IEE Publishing,
Power series no. 32, 2001, 2nd edn.)
5 WRIGHT, A. (Ed.): Arcs sparks and engineers (Reyrolle Heritage Trust Press,
Hebburn, 2001)
6 CLOTHIER, H.W.: Switchgear stages. A collection of articles written by
Henry Clothier, bound and published by G.F. Laybourne, 1933
7 GARRARD, C.J.O.: High voltage switchgear, IEE Proc., 1976, 123,
pp. 10531080
8 BROWNE, THOMAS E. Jr. (Ed.): Circuit interruption: theory and techniques
(Marcel Dekker Inc, 1984)
9 BLOWER, R.: Progress in distribution switchgear, Power Eng. J., 2000, 14,
(6), pp. 260263
10 STEWART, S.: Primary switchgear, Power Eng. J., 2000, 14, (6), pp. 264269
11 CIGRE Working Group 39.01 report: Substation control in the system control,
Electra, 2002, (200), pp. 4053 (www.cigre.org)
12 CIGRE Working Group 13.07 report: Controlled switching of HVAC circuit
breakers: planning, specification and testing of controlled switching systems,
Electra, 2001, (197), pp. 2333 (www.cigre.org)
13 RYAN, H.M., and WHISKARD, J.: Design and operation perspective of
a British UHV laboratory, IEE Proc. A, Phys. Sci. Meas. Instrum. Manage.
Educ. Rev., 1986, 133, (8), pp. 501521
14 RYAN, H.M., LIGHTLE, D., and MILNE, D.: Factors influencing dielectric
performance of SF
6
insulated GIS, IEEE Trans., 1985, PAS-104, (6),
pp. 15271535
15 CIGRE Working Group 13.09: Monitoring and diagnostic techniques for
switching equipment (convenor JONES, C. J., United Kingdom) Electra, 1999,
(184), p. 27
16 OCONNELL, P. et al.: CIGRE, WG.23.02 Paper SF
6
in the electric industry,
status 2000, Electra, 2002, (200) pp. 1625
17 JONES, C.J., HALL, W.B., JONES, G.R., FANG, M.T.C., and
WISEALL, S.S.: Recent development in theoretical modelling and mon-
itoring techniques for high voltage circuit-breakers. CIGRE 1994, paper
13-109
18 OLKEN, M.I. et al.: Restructuring and reregulation of the electric industry in
North America: challenges and opportunities, Electra, 2000, (45), pp. 4445
(special issue)
19 CIGRE Working Group 23.02, Task Force 02 report: Report on the second
international survey on high voltage gas insulated substations (GIS) service
experience, Electra, 2000, (188), p. 127
20 CIGREWorking Group 13.09: User guide for the application of monitoring and
diagnostic techniques for switching equipment for rated voltages of 72.5 kVand
above. Brochure 167
21 Online condition monitoring of substation power equipment utility needs.
CEAreport 485T1049
22 JEFFERIES, D.: Transmissiontoday: lessons froma decade of change, Electra,
1999, (187), pp. 918. Reproduced opening speech, London CIGREsymposium
Working plant and systems harder: enhancing the management and performance
of plant and power systems, 79 June 1999
23 URWIN, R.J.: Engineering challenges in a competitive electricity market.
Keynote address, ISH High voltage symposium, paper 5.366.SO, 2227
Augst 1999, London
24 LOWEN, J.: AC/DC power transmission: keynote address, Power Eng. J.,
June 2002, 16, (3), pp. 97101
25 JONES, C.J.: CIGRE Working Group 13.09 monitoring and diagnostic
techniques for switching equipment. Based on paper first published at
IEE/PES TandD conference, Atlanta, November 2001 (a also presented at IEE,
HVET 02, international High voltage Summer School, Newcastle upon Tyne,
10 July 2002)
26 ZIEGLER, R.G.: Protection and substation automation; state of art and
development trends, Electra, 2003, (206), pp. 1423
27 LAYCOCK, W.J.: Intelligent networks, Power Eng. J., 2002, pp. 2529
28 BAEHR, R.: Transformer technology: state of art and trends for future
developments, Electra, 2001, 198, (13), pp. 1319
29 CIGRE Working Group 23/21/33 report: Gas insulated lines (GIL), Electra,
2003, (206) pp. 5557 also KOCH, H., and SCHOEFFNER, G.: Gas insulated
transmission line (GIL) an overview, Electra, 2003, (211), pp. 817
30 IEC 62271-100: High voltage alternating current circuit breakers.
31 RYAN, H.M.: Prediction of alternating sparking voltages for a few simple
electrode systems by means of a general discharge law concept, IEE Proc.,
1967, 114, (11), pp. 18151821
32 PEDERSON, A.: Calculation of spark breakdown or corona starting voltages
in non-uniform fields, IEEE Trans., 1967, PAS-86, pp. 200206
33 MILNE, D., and RYAN, H.M.: The evaluation of solid dielectric systems for
use in high voltage switchgear, IEE Conf. Publ. 239, 1984, pp. 7679
Chapter 10
Polymer insulated power cable
J.C. Fothergill and R.N. Hampton
10.1 Introduction
10.1.1 Structure
The name cables is given to long current-carrying devices that carry their own insula-
tion and present an earthed outer surface. In this context, overhead lines for example,
are not considered as cables. Power cables have a coaxial structure: essentially, they
comprise a central current-carrying conductor at line voltage, an insulation surround-
ing the conductor and an outer conductor at earth potential. AC cables are generally
installed as a three-phase system and hence the outer conductor should only carry
fault and loss currents. In practice, a more sophisticated construction is adopted.
The interfaces between the metal conductors and the polymeric insulation would
tend to include protrusions and voids; features that would lead to electrical stress
enhancement and premature failure [1]. To overcome this, a polymer semicon, a con-
ductive polymeric composite, is placed at both interfaces. The inner semicon, the
insulation and the outer semicon are co-extruded to ensure the interfaces are smooth
and contaminant free. Surrounding this cable are layers to protect the cable during
installation/operation and carry the loss/fault currents. These layers also serve to keep
out water, which may lead to water treeing (section 10.4.2). Aschematic diagram of
a power cable is shown in Figure 10.1.
10.1.2 Voltage ratings
10.1.2.1 MV, HV and EHV
High voltage cables that are used for distribution and transmission purposes are
generally categorised according to the voltage rating:
medium voltage (MV) 636 kV
high voltage (HV) 36161 kV
extra high voltage (EHV) 161500 kV (or more)
jacket
or
oversheath
metal
sheath
outer
semicon
or
core
screen
or
insulation
shield
insulation
inner semicons
or
conductor
shield
or
conductor
screen
conductor
Figure 10.1 Cut-away section of a power cable
There is no international consistency on the distinction between distribution and trans-
mission. In the UK, transmission is at 275 kV and 400 kV, i.e. using EHV lines and
cables, whereas in Italy, for example, transmission starts at 150 kV. The distribu-
tors then transfer power from the transmission system across their local distribution
systems to the domestic customer.
10.1.2.2 Electrical stresses
The electrical stress within an AC cable is given by:
E =
V
x ln(R/r)
(10.1)
where V is the applied voltage, r is the radius over the inner semiconductive screen,
R is the diameter over the insulation and x is the intermediate radius (between r
and R) at which the electric stress is to be determined.
The probability of failure depends upon the electrical stress and increases with
stress. The effect of an increased stress can be estimated using the Weibull probability
function [2]:
P
f
= 1 exp
_
_
E
_
(10.2)
where P
f
is the cumulative probability of failure, i.e. the probability that the cable
will have failed if the stress is increased to a value E. The two parameters, and
are known, respectively, as the characteristic stress and the shape parameter.
Inspection of Equation (10.1) shows that the electric stress varies with the
position within the cable. There are three potentially useful stresses that can be
Polymer insulated power cable 479
considered:
maximum stress at conductor screen
mean geometric average stress for the whole insulation
minimum stress at the core screen.
The decision as to which of these to consider is an important one and is guided by
the potential modes of failure:
Maximum: the highest level of stress also corresponds to the highest probability
of instantaneous failure or, equivalently, the highest rate of electrical ageing. This
is most important if the most serious cable defects are located on or near the
conductor screen.
Mean: this is most important if the most serious defects are uniformly located
throughout the bulk of the insulation.
Minimum: this is most important if cable system reliability is determined by the
performance of accessories or if the electrical design or installation method of
accessories degrades cable performance. It is also important if the most serious
cable defects are located on or near the core screen.
10.1.3 Uses of cables
Power cables are commonly used in underground or underwater (submarine)
connections. Cables are placed at strategic points of the transmission grid to sup-
plement overhead lines or, in some cases, they can form the whole backbone.
Interconnection between networks is particularly well suited to cable solutions [1].
They may also be used in other applications. For example, overhead covered
conductors allow smaller phase clearance between the conductors on medium volt-
age overhead lines. Objects, particularly tree branches, may touch the lines without
tripping or customer outage. This has led to substantial improvements in service reli-
ability (e.g. Reference 4). In most cases, the aluminium alloy conductor is covered
with black UV-resistant crosslinked polyethylene and filled with grease, to provide
corrosion protection and longitudinal water-tightness. Arcing guides are applied at
insulator tops, to protect the line from arcing damage. Another application is the
Powerformer
TM
[5] and the related Windformer
TM
and Motorformer
TM
. These new
generators are able to supply electricity directly to the high voltage grid without the
need for a step-up transformer. They are suitable for power generation at output volt-
ages of several 100 kV. The newconcept is based on circular conductors for the stator
winding, and it is implemented by using proven high voltage cable technology. Thus,
the upper limit for the output voltage fromthe generator is only set by that of the cable.
10.1.4 AC and DC
Cables are used in both alternating current (AC) and direct current (DC) schemes.
The cable designs used in each case are outwardly very similar and have many iden-
tical design elements. However, the detailed engineering and the materials used are
very different.
AC is the globally preferred means of transferring electric power. This form of
transfer makes it straightforward to generate electricity and to transform voltages up
and down. This means of transfer accounts for more than 98 per cent of the global
power infrastructure.
In long distance transmission schemes, there are advantages in using DC over
AC. System instabilities caused by connecting regions with slightly different AC
phases and frequencies are obviated. Capacitive charging current implies that there is
a maximumuseful length of ACcable without the use of shunt reactors. DCcables are
therefore particularly useful for long distance submarine connections. Furthermore,
the lack of electromagnetic effects under DC conditions eliminates the skin effect in
which the conductor resistance can rise by up to 20 per cent at 50 Hz. The drawbacks
for a DCsolutionare that the terminal equipment forACDCconversionis more costly
and less efficient than an AC transformer. (Reviews on AC/DC power transmission
are contained in Reference 7.) Thus, the system must be sufficiently long to be
economically viable. In addition, the control of the electrical stress in a DC cable
insulation is much more difficult.
10.1.5 Cable types
There are, in general, four types of underground cable used characterised by the type
of insulation:
(i) Polymeric: low density polyethylene (LDPE), high density polyethy-
lene (HDPE), crosslinked polyethylene (XLPE) or ethylene propylene
rubber (EPR).
(ii) Self contained fluid filled (FF or LPOF): paper or paper polypropylene lami-
nated (PPL) insulated with individual metal sheaths and impregnated with low
pressure biodegradable fluid common on land >1 kV.
(iii) Mass impregnated non draining (MINDor solid): paper-insulated with individ-
ual metal sheaths and impregnated with an extremely low viscosity polybutene
compound that does not flow at working temperatures common at MV and
submarine DC.
(iv) High pressure fluid filled (pipe type or HPOF): paper-insulated and installed
in trefoil in steel pressure pipes and impregnated with high pressure non-
degradable fluid which is maintained at high pressures by pumping plants
common in the USA.
The typical electrical stresses employed in cables are shown in Table 10.1.
Up until the mid 1980s, paper-insulated cables were the system of choice at high
voltages. However, improvements in polymeric cables and accessories plus environ-
mental concerns has led to a significant reduction in the use of paper cables for land
applications. In particular, there has been a strong preference for XLPE cables over
EPR (except in Italy and the USA).
XLPE and EPR have emerged as the favoured polymeric insulations through
the 90
C continuous operating temperature that can be achieved when they are

used. This temperature matches that which can be attained when fluid-filled lapped
insulations (paper and PPL) are used. In contrast, LDPE and HDPE are limited to
Table 10.1 Average stress levels for selected
insulations
Type Stress, kV/mm EHV HV MV
fluid average at core screen 10 8 8
filled average at conductor screen 17 14 10
EPR average at core screen 4 3 2
average at conductor screen 8 5 3
XLPE average at core screen 5 3 2
average at conductor screen 11 6 3
operating temperatures of 70
C and 80
C, respectively. Table 10.2 identifies some of

the main advantages of the respective technologies.
The situation today is:
almost all new HV systems that are being installed are new build/expansion
the majority of HV cables, already installed within the existing system, are
insulated with paper (83 per cent paper and 17 per cent polymeric)
there is very little replacement of existing paper cables: most of installed capacity
is based on paper
most HV transmission is by overhead lines (OHLs)
HV cables are replacing OHLs in environmentally sensitive areas but there is
limited impact (globally) on new OHLs
XLPE cables make up two to five per cent of the total installed HVcable capacity
in the range 115161 kV
XLPE cables make up 5070 per cent of the HV cable capacity presently being
installed.
The reasons that there has been little replacement of paper cables are:
Paper cables continue to operate reliably.
The previous designs of XLPE cables were too large to fit existing rights of way
or pipes that have been designed for paper cables. This is of special importance in
the USAwhere there are many paper cables. In this area, the key task is to develop
polymeric cable designs that are flexible, small (i.e. working at high stresses) and
easy to joint, as it is essential to use the existing pipes/ducts. Present designs
address these issues, and now cables exist that match the size and performance
of paper cables [6].
10.2 The components of the polymeric cable
10.2.1 Conductor
Conductors in the USAtend to be based on theAmerican wire gauge (AWG), but in the
rest of the world are based on IEC228 and are therefore metric. Stranded conductors
Table 10.2 Advantages (+) and disadvantages () of HV cable insulations (quan-
titative comparisons given in italics)
XLPE EPR Paper
+ no risk of oil leakage + no risk of oil leakage oil leakage
+ simple design + simple design complicated design
high maintenance burden
+ very low dielectric losses
losses for a 132 kv 1000 mm
2
cable are 5.1 mA/m
high dielectric losses + low dielectric losses
losses for a 132 kV
1000 mm
2
cable are
14.7 mA/m
+ acceptable fire performance + acceptable fire
performance
fire performance in
tunnel and substation
applications is a concern
due to the copious supply
of flammable oil
complicated accessories
(possible long repair times)
complicated accessories + simple accessories
limited but growing
track record
15 years of experience
limited track record + extensive track record
>60 years of experience
most of existing system
is paper, transition joints
difficult
most of existing system
is paper, transition joints
difficult
+ most of existing system
is paper
may suffer from water
trees if no metal sheath
+ longer life in wet
conditions, metal sheath
not generally required
+ metal sheath required
to contain fluid thus
water ingress is not
possible
high conductor losses
132 kV 1000 mm
2
can transmit 250 MVA
of power
high conductor losses + low conductor losses
132 kV 1000 mm
2
can transmit 275 MVA
of power
large size
132 kV 1000 mm
2
cable has a 98 mm diameter
and a weight of 25 kg/m
large size + small size
132 kV 1000 mm
2
cable
has a 70 mm diameter
and a weight of 18 kg/m
have generally comprised concentric layers (Figure 10.2a) but, in order to make them
smoother and more compact, they are often specially shaped in rolling mills nowa-
days (Figure 10.2b). When operating at high voltages and currents, the AC current is
preferentially carried more in the outer than in the inner conductors (the skin effect).
In addition, the electromagnetic fields induce eddy currents (proximity effect). These
effects tend to increase the conductor resistance under AC above that which is seen
under DC. The AC/DC ratio (R
AC/DC
) can be as large as 1.15. This increased resis-
tance serves to increase the joule heating losses within the conductor and increases
a
b
c
Figure 10.2 Methods of laying conductors
a concentric-lay
b shaped compacted
c Milliken (an oil-filled cable) (from [3])
the temperature of the cable. Thus, special Milliken conductors, Figure 10.2c, may
be used for large conductor designs to reduce the AC/DC ratio.
Conductors are virtually all made from either copper or aluminium. Copper has
the advantage of being more conductive, and, therefore, requires less material to
carry a given current. Copper conductors, therefore, have the advantage of being
small. However, the cost of aluminium is lower than copper, and even though it has
a lower conductivity it is often used as the conductor. An additional advantage of
aluminium over copper is that the conductors are lower in weight even though more
volume of material is used. At MV, aluminium is preferred whereas, at HV and EHV,
the smaller size of copper provides the greatest advantage.
10.2.2 Semicon
Semiconductive screening materials are based on carbon black (manufactured by the
complete and controlled combustion of hydrocarbons) dispersed within a polymer
matrix. The concentration of carbon black needs to be sufficiently high to ensure
an adequate and consistent conductivity. The incorporation must be optimised to
provide a smooth interface between the conducting and insulating portions of the
cable. The smooth surface is important as it decreases the occurrence of regions of
high electrical stress [8]. To provide the correct balance of these properties, it is
essential that both the carbon black and polymer matrix be well engineered.
The same care needs to be paid to the manufacture of the matrix polymer for
semicons as for the XLPE insulation. However, the chemical nature of the polymers
is subtly different because of the need to incorporate the carbon black. The carbon
black and other essential additives (excluding crosslinking package) are compounded
into the matrix. The conveying and compounding machinery used are designed to
maintain the structure of the carbon black within a homogeneous mix. Before the
addition of the crosslinking package, filtration may be applied to further assure the
smoothness of the material toa higher standardthanthat providedbythe compounding
process.
The smoothness of the extruded cable screens is assured by extruding a sample
of the complete material in the form of a tape. The tape is optically examined for the
presence of pips or protrusions. Once detected, the height and width of these features
are estimated, thereby enabling width-segregated concentrations to be determined.
When using such a system, care needs to be exercised when examining the present
generation of extremely smooth (low feature concentration) screens, as the area of
tape examined needs to match the likely number of detected features: smooth screens
require larger areas of examination.
10.2.3 Insulation
It is clear from Table 10.2 that XLPE is the most common cable insulation currently
being installed. Insulating XLPE compounds need to fulfil a number of require-
ments. They should act as thermoplastic materials within the extruder and crosshead.
They should crosslink efficiently with the application of high temperatures and pres-
sures within the vulcanisation tube. They should be immune to thermal degradation
throughout the cable manufacture process and operation at the maximum cable tem-
perature for the life of the system. They must display an extremely low occurrence of
the features that can enhance the applied electrical stress and thereby lead to prema-
ture failure. To deliver these requirements, it is essential that the greatest care is paid
to the design and manufacture of the polymer and the engineering of the appropriate
crosslinking and stabilising packages.
The manufacturing technology employed for XLPE compounds to be used for
power applications needs to ensure the highest level of cleanliness at all points
of the production chain. The sequence comprises three main parts: base polymer
manufacture, addition of a stabilising package and addition of the crosslinking
package.
There are two types of crosslinking process that are commonly used for power
cables:
(i) Peroxide cure thermal degradation of an organic peroxide after extrusion
causes the formation of crosslinks between the molten polymer chains.
(ii) Moisture cure chemical (silane) species are inserted onto the polymer chain.
These species formcrosslinks when exposed to water; this process occurs in the
solid phase after extrusion.
The peroxide cure method is the most widely used crosslinking technology. It is used
for MV, HV and EHV. The moisture cure approach is limited to MV where it has
proven itself as a flexible and economic solution.
10.2.3.1 Peroxide cure
This first part of the production sequence starts with the polymer reactor, which
must be designed to deliver polymers with consistent properties that are free from
particulate contaminants (free from moving metal parts) and chemical species that
degrade the dielectric properties. In practice, this means that the sequence of poly-
mers manufactured in the reactor and the materials employed in its construction must
be controlled. For example, the materials used to seal the closed conveying system
must be made from thermoplastic polyethylene. Within the manufacture, it is essen-
tial that materials be conveyed in the safest way possible to ensure that extraneous
contaminants are not introduced. Experience has proved that the best solution is to use
the shortest possible distance between the reactor and the compounding equipment.
The transport from the polymerisation reactor should be affected by stainless steel
pipelines and vessels with dense phase or gravity conveying, as this significantly
reduces the formation of polymer strings and dust.
The addition of the crosslinking and stabilising packages is the next step in the
chain. These processes incorporate the active chemical species into the polymer
matrix. Before the addition of the crosslinking package, the material passes through
extremely fine filters and screens that operate in cascade. The purpose of these pro-
cesses is to ensure that the infrequently occurring contaminants do not proceed to the
packaging process. These contaminants may be within the matrix or adhering to the
surface.
At all interprocess points samples are taken for the essential quality assurance
processes. These processes are in two parts:
(i) assurance of the physical and chemical properties
(ii) measurement of the cleanliness.
The physical and chemical analyses serve to establish that the extrudability and
crosslinking characteristics are both within specification. These checks ensure the
consistency of both the polymer flow, within the extruder and the crosshead, and the
formation of the chemical crosslinks within the vulcanisation tube.
10.2.3.2 Moisture cure
The key part of the moisture cure approach is the insertion of the chemically active
species onto the polymer backbone. There are three methods for achieving this:
(i) Sioplas: a suitable silane is melt compounded with a peroxide and a polymer.
During this process, the silane becomes chemically grafted to the polymer
chain. The material is then pelletised for use in the cable extruder. The cata-
lyst for the formation of the crosslinks, and other additives, are added at the
extrusion step by way of a masterbatch.
(ii) Monosil: this process is similar to Sioplas except that all of the components
(including the catalyst) are added at the cable extruder. The silane grafting
reaction therefore takes place at the same time as the cable is being extruded.
(iii) Ethylene-vinyl silane copolymers (EVS): this process does not graft silanes
onto existing polymer chains, but inserts the silane groups into the poly-
mer chains while they are being constructed within the reactor. The polymer
exits the reactors ready for extrusion. The catalyst for the formation of the
crosslinks, and other additives, are added at the extrusion step by way of a
masterbatch.
After the core exits the extrusion line, it is solid and in a thermoplastic state.
It must be crosslinked offline. The amount of time taken for this process will
depend upon the technology used, the water or steam temperature and thickness of
insulation.
10.2.3.3 Cleanliness
The cleanliness of insulation materials (peroxide curable, Sioplas and EVS) may
be assessed by converting a representative sample of the polymer into a transpar-
ent tape and then establishing the concentration of any inhomogeneities. This is not
possible for monosil as the chemistry occurs immediately prior to the cable extru-
sion. The inhomogeneities are detected by identifying variations in the transmission
of light through the tape. To gain the required level of consistency and sensitivity,
the tape is inspected by an automated optical system (Figure 10.3). This approach
has an excellent signal to noise level and good rejection of spurious data. The data
Figure 10.3 Automatedequipment for determiningthe contaminationconcentration
of XLPE (courtesy of Borealis AB)
processing is carried out by a microcomputer, which is able to produce size segregated
concentration data for a number of selected levels of obscuration.
10.2.4 Metal sheath
For many years, lead, or lead alloys, were the main materials used for the metal sheath
layer. This is principally because the low melting temperature allows the lead to be
extruded at a temperature of approximately 200
Cover the polymeric cable. The main

disadvantages of lead are its high density (11 400 kg/m
3
) leading to a heavy product
and the tendency to creep, flow or embrittle under cyclic temperature loadings. This
latter effect has led to a number of cases where the sheath has ruptured. There are
also environmental concerns relating to the use of lead, and its use may become
restricted by EUdirectives. In 2000, the EUCommission officially adopted the waste
electrical and electronic equipment (WEEE) and Reduction of Hazardous Substances
(ROHS) proposals. The ROHS proposals required replacement of lead and various
other heavy metals from 2008. To address these problems a number of materials have
been used:
Extruded aluminium: this has excellent mechanical performance but requires
a large bending radius and can be difficult to manufacture since it requires
corrugations. It can be heavy.
Aluminium foil: this is light and easy to manufacture but small thicknesses
(0.20.5 mm) do not give mechanical protection; the strength comes from the
polymer oversheath. It relies on adhesive to make a water tight seal, and it can
suffer from corrosion.
Copper foil: this is light, easy to manufacture but not as flexible as aluminium
foil. It also relies on adhesive to make a water tight seal.
Welded copper: this is strong, robust and capable of carrying significant current
but difficult to manufacture since it requires corrugation and it is difficult to ensure
perfect longitudinal welds in practical situations.
Welded stainless steel: this is strong, robust and capable of carrying
significant current but has similar manufacturing difficulties to welded
copper.
10.2.5 Oversheath ( jacket)
In addition to the meticulous attention that must be paid to the insulation system,
care also needs to be taken with the oversheath layer. The vast majority of HV and
EHV XLPE cables are of the dry design type, which means that a metal barrier is
included. The purpose of the metal barrier is toprotect the core withinfrommechanical
damage, carry fault and loss currents and to exclude water from the construction (the
electrical ageing rate is significantly higher in the presence of moisture, see section
10.4.2). The metal barrier is a key part of the cable design and much care needs to
be taken as this significantly affects how a cable system may be installed in practice.
The metal layer is itself protected by a polymeric oversheath. Due to the critical
performance needed from the oversheath, there are a number of properties that are
required: good abrasion resistance, good processing, good barrier properties and
good stress crack resistance. Experience has shown that the material with the best
composite performance is an oversheath that is based on polyethylene.
10.3 Cable manufacture
10.3.1 Stages of cable manufacture
The stages in cable manufacture may be summarised as:
(i) Conductor manufacture: this involves:
wire drawing to reduce the diameter to that required
stranding in which many wire strands and tapes are assembled
laying up; the assembly of non-circular (Milliken) segments into a quasi-
circular construction
(ii) Core manufacture: this involves:
triple extrusion in which the core of the cable is formed comprising the
inner semicon, insulation and outer semicon
crosslinking which can be carried out directly after extrusion (peroxide
cure) or offline (moisture cure)
degassing in which peroxide crosslinking byproducts are removed by heat-
ing offline; the diffusion time depends upon temperature and insulation
thickness.
(iii) Cable manufacture: this involves:
core taping during which cushioning, protection and water exclusion
layers, inner semicon, insulation and outer semicon are applied over the
extruded core
metal sheathing; the application of a metal moisture and protection layer
armouring; the application of high strength metal components (steel) to
protect the cables; essential for submarine cables.
10.3.2 Methods of core manufacture
All of the production processes will be common to all methods of manufacture,
with the exception of the extrusion process where there are three types of peroxide
crosslinking method (crosslinking is often referred to as vulcanisation or curing):
(i) VCV: vertical continuous vulcanisation
(ii) CCV: catenary continuous vulcanisation
(iii) MDCV: Mitsubishi Dainichi continuous vulcanisation, often called long
land die.
In all of these processes (moisture cure and peroxide cure), the three layers of the cable
core are extruded around the conductor. This uncrosslinked core then passes directly
into the curing tube; this is where differences in the processes become apparent. In the
moisture cure approach, which takes place offline after extrusion, the manufacturing
process is considerably simplified as the length of the tube following extrusion only
has to be enough for the thermoplastic core to cool sufficiently to prevent distortion.
10.3.2.1 Vertical continuous vulcanisation
The curing tube is arranged vertically, and the cable is maintained within the centre
of the tube by control of the cable tension. A schematic diagram showing VCV is
given in Figure 10.4. The conductor is fed from a payoff through an accumulator.
This allows a new reel of conductor to be loaded onto the payoff when the old reel
runs out without the continuous extrusion process being stopped; the accumulator
allows time for the two conductors to be welded together. The conductor is pulled to
the top of an extrusion tower, which may be 100 mtall, over a large capstan and is fed
via a preheater through a triple extrusion head supplied by three extruders. Curing is
effectedbyheatingthe cable withhot nitrogengas. The gas is pressurisedtoensure that
gas-filled voids do not formfromdecomposition of the peroxide. The VCVtechnique
ensures concentricity of the conductor within the cable core because of the vertical
alignment of the cure tube. VCV lines are particularly effective for producing cables
with large conductors >1600 mm
2
as there is not the same difficulty of maintaining
the tension compared with the CCV technique. Cables with insulation thickness up
to approximately 35 mm can be produced with VCV lines.
Followingproduction, the cable is checkedfor concentricityandpossiblyfor gross
imperfections, such as detachment of the inner semicon, using x-rays or ultrasonic
techniques before being laid up with the outer protection layers. In land cable, the
heated
nitrogen
(several
zones)
cooling
water
triple
extrusion
conductor preheating
water seal
capstan
concentricity
and purity
testing
to cable
takeup
accumulator
from
conductor
payoff
Figure 10.4 Schematic diagram showing the principal components of a VCV line
cable is taken up on drums, but for subsea cables it is preferable to avoid jointing by
laying up in long lengths and loading directly onto a waiting ship.
10.3.2.2 Catenary continuous vulcanisation
This technique is similar to VCVexcept that the curing tube is arranged in a catenary
and the cable is maintained within the centre of the tube by very careful control of
the cable tension. This has become much easier with sophisticated automatic control
systems. Care needs to be taken to ensure that the polymer, which is molten in the
stages prior to crosslinking, does not drip or droop off the conductor under the influ-
ence of gravity. This effect becomes stronger as the thickness of insulation increases
or the ratio of the insulation thickness to conductor size increases. Techniques includ-
ing special polymers, rotation of the cable and quenching of the insulation are all
effective at reducing the drooping effect. An additional problem can arise with large
conductors (heavy cables), when it becomes difficult to both apply a large tension
(needed to keep the cable in the centre of the tube) and control the tension. This prac-
tically limits the conductors sizes to below 14001600 mm
2
. Cables with insulation
thicknesses of up to 25 mmcan be produced with CCVlines. The VCVtechnique does
not suffer from the effects of drip or droop off of the conductor under the influence
of gravity to the same extent as for the CCVtechnique. However, due to construction
costs of large vertical structures VCV lines are shorter than CCV lines: 80100 m
for VCV and 200250 m for CCV. This difference enables CCV lines to be faster as
identical cables require the same residence time.
10.3.2.3 Mitsubishi Dainichi continuous vulcanisation
In this technique, the cure tube is arranged horizontally after the extruder but, unlike
VCVand CCV, it does not employ nitrogen gas to maintain the pressure. The MDVC
technique arranges that the diameter of the die is the same size as the final cable
and thus fills the tube/die. The thermal expansion resulting from heating the polymer
to the molten state to initiate the crosslinking exerts a pressure that suppresses the
formation of voids. The MDCVprocess does not experience the same droop issues as
those faced by the CCVprocess due to being fully enclosed by the die. However, it is
important to ensure that the conductor remains centralised within the tube while the
polymer is molten before crosslinking. Centralisation may be achieved by ensuring
a very high tension (thereby essentially eliminating the catenary) or by using special
polymers with high viscosities. These special methods are normally required for
conductors above 1000 mm
2
.
10.4 Failure processes
Polymeric insulatedpower cables are designedtobe highreliabilityproducts. The fail-
ure of a major power cable is likely to have a considerable effect on the power
transmission grid and may take several days/weeks to repair. If it is under the sea, it
may take months to repair and cost well in excess of 1 million ($1.5 million). Cables
have a good service history. The majority of cable failures are caused by external
influences such as road diggers or ship anchors. Cable system failures may occur
at joints and terminations. A study of MV cable systems in France [9] has shown
that the failure rate for paper cables is 3.5 failures per 100 cable km per year which
compares with a rate of two failures per 100 cable km per year for XLPE cables. The
failure rate for XLPE has been classified in Table 10.3. This analysis shows that the
Table 10.3 20 kV XLPE system performance
(after [9])
Class Fault rate
(#/yr/100 cct km)
Total system 2.0
Third party damage 1.0
Accessories 0.9
Cable 0.1
XLPE cables form the most reliable component of the cable system. Further inspec-
tion would suggest that the XLPE cable system is 2.5 times more reliable than the
paper system when third party damage is excluded (we assume that a paper cable is
as likely to get dug up as an XLPE one!). This estimate may well overstate the case
and illustrates one of the major problems of evaluating field failure statistics; namely
the differing ages of the populations. XLPE data are based on a population which is
15 years old whereas the paper data are based on a population that is 3040 years old;
thus we would expect higher failure rates of the paper system since it contains older
devices; this would be quite independent of the intrinsic reliabilities of the systems.
Causes of insulation breakdown include:
Extrinsic defects (contaminants, protrusions or voids) caused during manufacture
or installation. These wouldnormallyleadtoelectrical treeingor direct breakdown
soon after production of the void and cable energisation.
Water treeing (wet ageing) caused by water leakage through the sheath or
inappropriate design or deployment of a cable without a water barrier. Water trees
lead to a weakening of the insulation and electrical treeing or direct breakdown.
Thermoelectric ageing: the combination of the electric field, acting synergistically
with a raised temperature, causes the insulation to weaken over time and for
breakdown to occur eventually. This process may not always be significant within
the lifetime of a well designed cable.
10.4.1 Extrinsic defects
We have already described the efforts made by cable manufacturers to exclude and
detect contaminants, protrusions and voids (CPVs) in their products. Contaminants
within the bulk of the insulation and protrusions into the insulation from the semicon
cause field intensifications that lead to premature failure of the polymer.
10.4.1.1 Contaminants and protrusions
Many studies [1012] have shown the degradation caused by large metallic
contaminants. Figure 10.5 shows the influence of size on the electrical breakdown
strength. The effect of a five-fold increase in concentration at the 100 micron level
is to further reduce the strength by 17 per cent and 14 per cent for AC and impulse,
respectively. As well as showing the reduction in the characteristic strengths, the
increasing size of contaminants changes the statistical nature of the failures, making
them less scattered or more certain.
The effects displayed in Figure 10.5 may be explained by the fact that metallic
contaminants increase the electric stress within their immediate locality such that the
local electric stress is higher than the breakdown strength. This effect is best described
in terms of a stress enhancement factor that acts as a multiplier for the Laplacian stress
(i.e. the geometrically calculated stress). The data from Figure 10.5 are in line with
the trends predicted by the theoretical estimates for the increase in local electrical
stress found at the tip of sharp metallic contaminants. It is interesting to note that it is
possible to get large stress enhancements but that the magnitude of the enhancement
0
50
100
150
200
250
300
0 50 100
contaminant size, m
c
a
b
l
e

b
r
e
a
k
d
o
w
n

s
t
r
e
n
g
t
h
,
k
V
/
m
m
,

6
3
%

p
r
o
b
a
b
i
l
i
t
y
Figure 10.5 Electrical breakdown strength (AC lower circles and impulse upper
squares) of model cables with artificially added contaminants
(concentration of 5 per kg) [12]
falls dramatically with distance from the tip; for a 5 micron radius, the field falls by
50 per cent within 1.5 radii of the tip [8]. Thus, the calculated stress enhancements
should be viewed as providing the upper limits of any assessment.
The electrical stress enhancements are not onlybasedonthe size andconcentration
but they have a significant influence from the nature (conducting, insulating, high
permittivity), the shape (sharp or blunt) and the way that they are incorporated into the
matrix. The effect of the shape of contaminants has been assessed [13] using XLPE
cups with a Rogowski profile. It was shown that contaminants with irregular surfaces
reduced the AC ramp breakdown strength by a greater degree than those with smooth
surfaces (Figure 10.6).
The local stress enhancement experienced within an insulator will have contribu-
tions fromthe size of the contaminants, their concentration and the nature (conducting
or high permittivity) of the contaminants. This is shown in Equation (10.3).
= 1
1
_
0.5 ln
+1
1

2
1
_
(10.3)
where:
= 0.5 ln
+1
1

1
+
1
(k 1) (
2
1)
k =

2
1
=
1
1 (r/a)
= stress enhancement factor
r = radius of the ellipse
0
20
40
60
80
100
iron copper glass
b
r
e
a
k
d
o
w
n

s
t
r
e
n
g
t
h
smooth surface irregular surface
Figure 10.6 The effect of shape of added contaminants on the AC ramp breakdown
strength (relative to base) of XLPE cups [13]
2a = length of the ellipse
1
= permittivity of the matrix
2
= permittivity of the defect
10.4.1.2 Voids
Voids are likely to lead to breakdown if discharges occur inside them. These dis-
charges are known as partial discharges since they are not, in themselves, a complete
breakdown or full discharge. The least problematic (and perhaps the least likely)
shaped void is the sphere. The field inside an air-filled sphere-shaped void is higher
than the field in the insulation by a factor equal to the relative permittivity of the
solid. For other shaped voids, the field in the void will be higher than this. For exam-
ple, a void inside XLPE (relative permittivity = 2.3) will have a field inside it of
at least 2.3 times that of the field in the XLPE itself. The criterion for a discharge
in a void is that the void field must exceed the threshold described by the Paschen
curve (see Reference 14, for example); this is dependent on the size of the void. The
Paschen field has a minimum for air at atmospheric pressure for a void diameter of
7.6 m. The breakdown voltage at this diameter is 327 V. This equates to a void field
(ignoring the non-linear effects) of 43 kV/mmor an applied field of 19 kV/mmwithin
the XLPE. Below this void size, the Paschen field increases rapidly; for example, a
2 m void would require an applied field of approximately 150 kV/mm to cause
discharging. It is clear then that voids of diameter exceeding a few microns are likely
to allow severe electrical damage to occur.
10.4.1.3 Partial discharges and electrical treeing
In an electrical discharge, electrons are accelerated by the electric field such that
their kinetic energy may exceed several electron-volts. With such energies, collisions
with gas molecules may cause further electrons to be released, thus strengthening
the discharge, or they may cause electroluminescence and the release of energetic
photons. The surface of the void is therefore likely to be bombarded by particles
(photons or electrons) with sufficient energy to break chemical bonds and weaken
the material. In the case of sharp protrusions or contaminants, which give rise to
high local electric fields, electrons may be emitted and very quickly acquire the
kinetic energy required to cause permanent damage to the insulation. Electrons may
accumulate i.e. they may be trapped around such defects and cause a further
increase in local electric fields. Mechanical stress, which may already be increased
due to the modulus and thermal expansion coefficient differences between the host
materials and the CPVs, may be further enhanced by electromechanically induced
stress. These effects, catalysed by CPVs in an electric field, may lead directly to
a breakdown path, but are more likely to lead first to the formation of an electrical tree.
An electrical tree is shown in Figure 10.7. For the sake of clarity, this has been
grown in a translucent epoxy resin from a needle acting as a protrusion. Abreakdown
path can be seen to be growing back through the electrical tree fromthe plane counter-
electrode. Electrical trees have a branched channel structure roughly oriented along
the field lines. Typically, the diameters of the channels are 120 m, and typically
each channel is 525 m long before it branches. There is considerable evidence that
the branching is determined by the local electric field, which is grossly distorted by
trapped electrons emanating from the discharges within the tree. There is a consid-
erable body of work on this subject (e.g. Reference 14). Trees tend to grow more
directly across the insulation if they are spindly, so-called branch trees. The other
type of tree, the bush tree, uses the energy to produce a lot more dense local treeing,
Figure 10.7 An electrical tree grown in epoxy resin
and therefore tends to take longer to bridge the insulation. Because branch trees occur
at lower voltages, there is a non-monotonic region between branch and bush growth
in which failure occurs more quickly at lower voltages. Trees grown from larger
voids can be clearly distinguished from those due to protrusions; it is noticeable that
the fields must cause the voids to discharge before the trees initiate. After an initial
period in which the tree grows quickly, the rate of growth decreases. Finally, as the
tree approaches the counter electrode, a rapid runaway process ensues.
Although the electric tree processes are now quite well understood, apparently
little can be done to prevent electrical tree propagation in polymeric insulation once
it has started. It is thought that electrical trees take relatively little time to grow and
cause breakdown in cables, in perhaps a few minutes to a few months. Once an
electrical tree has been initiated, the cable can be considered to be terminally ill.
It is therefore vital to prevent the inclusion of CPVs during cable manufacture and
installation. The triple extrusion continuous vulcanisation techniques and appropriate
cable protection and deployment techniques appear to have been successful in this
regard, and cables rarely suffer from these problems. There are more likely to be
problems at joints or terminations where high field stresses may inadvertently be
introduced or if water trees, described in the following section, are allowed to grow.
10.4.2 Wet ageing water trees
In the early days of polymer-insulated cables, it was assumed that the polymers
would be essentially immune to the deleterious effects of water that were well known
in paper cables. Consequently, the first designs of cable were installed with little or
no water precautions. Within a few years, a large number of cables started to fail in
service. Upon examination, tree-like structures were seen to have grown through the
insulation. It was assumed that they continued to grow and failure occurred when the
whole insulation was breached. This is the phenomenon of water treeing [15].
Many studies have been carried out into the phenomenon and its solution. Look-
ing back, it is clear that a number of improvements in cable design, manufacture
and materials have reduced the incidence of cable failures by water treeing. These
improvements have included:
water barriers (metal or polymeric) to exclude the water
triple extrusion (all polymer layers extruded at the same time)
semiconductive polymer screens to replace carbon paint or paper tapes
cleaner insulations
smoother semicons
internationally recognised approval methods
special long life insulations based either on additives or polymer structure.
The laboratory studies have concluded that the growth of water trees is affected by:
test voltage
test frequency
mean temperature
temperature gradient
type of material
presence of water (external and within the conductor).
There are essentially two types of water trees: vented trees that originate from the
surface of the insulation and are potentially the most dangerous and bow tie trees
that grow from contaminants or voids within the insulation and tend to grow to a
limiting size without breaching the insulation. These trees do not comprise tubules
containing water as might be surmised from the earlier description of electrical trees.
The branches of a water tree actually appear to comprise a high density of water-filled
voids of typical diameter 110 m. Such branches are therefore similar to a string of
pearls, but in practice even branches of water trees are not usually discernible. They
are simply diffuse regions of water-filled voids. If dried up, reimmersion in water
reopens the voids. Boiling stabilises the structure but probably also produces extra
small voids. There is limited evidence that a percolation network does interconnect the
voids, but the size scale of the interconnecting features is around 10 nm. Electrolyte
material accompanies the water into the voids and the ability of cationic dyes, such
as rhodamine B, to stain the trees permanently indicates that some oxidation must
have taken place. Chemical modification has also been shown using IR and FTIR
spectroscopy and by fluorescence techniques.
Water trees grow much more slowly than electrical trees. Typically, they may not
be observed at all for several years, even if the prevailing conditions for their growth
are in place. They will then grow fast initially and then very slowly. Indeed, in the
case of bow-tie water trees, there is much evidence that they stop growing completely
after a given length (dependent on prevailing conditions) and that they might not pre-
cipitate breakdown. Vented water trees may cross the insulation completely without
breakdown occurring, but they do greatly weaken the insulation. Generally, an elec-
trical tree or a breakdown path may growback through a water tree. Figure 10.8 shows
a badly degraded EPR cable containing water trees. The water trees are not as clear
as have been seen in XLPE cables, but this may be because EPR is not transparent
or even translucent as is XLPE. Also seen in this figure are breakdown paths starting
to grow back through the water-treed areas. The cable had broken down within 1 m
of this area. The general trend is for the ultimate failure of the cable to be due to the
conversion of a vented tree to an electrical tree.
There are many proposed mechanisms of water treeing and these have been criti-
cally reviewed in Reference 14. Essentially, it is likely that solvated ions are injected
at partially oxidised sites. These catalyse further oxidation by maintaining the ion con-
centration. A sequence of metalion catalysed reactions is proposed in which bonds
break and cause microvoids to develop. Alternating electromechanical stresses open
up pathways for solvated ions; these initiate new microvoids. Many tree-retardant
polymers contain ion catchers to prevent the metalion catalysis, and these have been
found to successfully delay the onset and growth rate of water trees.
The tree inception time, i.e. the time between the conditions being right for water
tree growth and the first observation of water trees, is highly dependent upon the
electrical stress. Typically, the inception time is inversely proportional to a high
Figure 10.8 Water trees and breakdown paths in an EPR cable. The black streaks
are breakdown paths starting to cross the cable. The black dots, which
are intense around these breakdown paths, are stained water trees.
EPR is more resistant to water trees than XLPE, but this cable had
been taken from service where it had been saturated in water with no
metal sheath
power (410) of electric field. For this reason, low voltage cables, which tend to
run at lower electric fields, may not have a water barrier to prevent water ingress
and hence electrical treeing. In such cases, with fields typically <4 kV/mm (see
Table 10.1), the probability of failure through electrical treeing is low and a water
barrier would make little difference. HVand EHVpolymer-insulated cables generally
use water barriers, and these become mandatory above 66 kV. Furthermore, the con-
ductor is often water blocked (a water-swellable compound or an extruded mastic)
to prevent the transport of water along the conductor. The water may enter the con-
ductor either after a cable breakdown or during installation or through an incorrectly
installed accessory.
10.4.3 Dry ageing thermoelectric ageing
The requirement for extra high voltage (EHV) underground power cables is increasing
[1622]. There is commercial pressure to push the mean electric field in the insu-
lation of such cables towards 16 kV/mm, and the most common insulation used is
crosslinkedpolyethylene (XLPE). Longtermexperience of XLPE, however, is limited
to moderately stressed cables with mean fields of 5 to 7 kV/mm. Furthermore, the
introduction of crosslinking processes has permitted the continuous operating temper-
ature of polymeric cables (XLPE and EPR) to be increased to 90
C, equalling that of
oil-filled (LPOF and HPOF) paper and polypropylene paper laminate (PPLP) cables.
The use of XLPE as the insulation for transmission cables has grown steadily since
the early 1990s. Many extruded power cables have been operating for 20 years and
are approaching the end of their 30-year design life. If robust methodologies could be
found for improving or/and evaluating the reliability of AC power cables, it may be
possible to continue to use them without compromising the reliability of the system.
Such methodologies require considerable improvements in the understanding of any
ageing or degradation mechanisms of cable insulation. They would enable XLPE
cables to be more competitive at EHV levels.
Recently, there has been considerable discussion in the literature about whether
semicrystalline polymer insulation, such as XLPE, will age naturally under combined
thermal and electric fields, without water or extrinsic damage (e.g. CPVs). Such
considerations suggest that the polymer is graduallyweakenedandthat this weakening
may eventually lead to failure. Some of the theories of ageing suggest that a minimum
field (perhaps around 20 kV/mm) is required for such ageing to occur, others assume
just that the ageing slows up considerably at lower fields. The three main current
theories are: the DissadoMontanariMazzanti (DMM) theory [23], that of Lewis
et al. [24] and the Crine theory [25]. There is not space to contrast these models
here (see, e.g. Griffiths et al. [26]) but the essential elements are described. It is
assumed that this ageing is macroscopic whereas the extrinsic and treeing degradation
is localised and leads directly to a breakdown channel. This ageing leads ultimately
to partial discharging or/and electrical treeing. The ageing process is likely to start at
nanometre size scales, and for this reason is difficult to observe directly. It is likely to
lead to the production of nano-voids which gradually increase in both size and number.
Once these approach the micron size scale, they may be capable of precipitating
breakdown through energetic mechanisms, such as hot electron damage, which can
take place close to the design field in the presence of large defects. The ageing
mechanisms proposed therefore involve the production of macrodefects (>10 m)
from microdefects (<10 nm).
The ageingmodels lead to equations describingthe life of the insulatingmaterial in
terms of the field and temperature. There has been considerable work (e.g. European
ARTEMIS
1
programme) to understand these mechanisms better in order to be able to
diagnose the state of a cable, developa replacement policybasedonsuchdiagnosis and
develop life models to be used in the resource management of cable systems. Such
an improved understanding is also useful in considering whether it is worthwhile
investing further effort in improving manufacturing processes.
The three models referred to earlier differ slightly in physical details and some-
what more in their mathematical development but their commonality can be found
in considering the different size scales of the various processes involved. The Lewis
model considers the breaking of chemical bonds as a starting point for the ageing.
The problemis stated more generally in thermodynamic terms in the other two ageing
models. They consider that moieties i.e. small regions of the polymer may exist in
either of two states. In an unaged polymer, most of the moieties are in state 1 whereas,
as ageing progresses, more and more moieties switch to state 2. An energy barrier
exists between the states and moieties may switch between states by thermal activa-
tion. In the Lewis model, these states would comprise unbroken and broken bonds; the
other two models are more general and may include molecular chain reconfigurations,
for example. The presence of a local electric field changes the dynamic equilibrium
between the states, and, according to the DMM theory, causes irreversible changes
when a threshold field is exceeded. It is assumed that changes from state 1 to state 2
will eventually strain, and, possibly, lead to nanometre and submicron sized voids
0
5
10
15
20
25
30
35
40
45
time to failure, hours
f
i
e
l
d
,

k
V
/
m
m
1E1 1E+0 1E+1 1E+2 1E+3 1E+4 1E+5
Figure 10.9 Electrical life line with model (line) and experimental data points (from
[27]) for AC polyimide data. Experimental points at 63.2% failure
probability are displayed
when a sufficient concentration of moieties have switched to state 2. Areas that are
mechanically weakened in this way will increase in size leading to super-micron-sized
voids and the more rapid degradation associated with hot electron injection, partial
discharging and electrical treeing, for example.
All three models assume that a section of the polymer can transfer between alter-
native local states by surmounting an energy barrier. The presence of an electric field
alters the energies of the two states, thereby altering the relative proportions of the
polymer in each of the alternative states at equilibrium. It also accelerates the rate
of approach to the equilibrium distribution from any arbitrary starting distribution. It
is important to note that these processes are temperature dependent. A typical result
from such modelling is shown in Figure 10.9.
10.5 Mathematical design models for cables
Cables are designed to operate at electrical stresses in the range of 311 kV/mm
(Table 10.1). However, the breakdown strengths of virgin cables are often in the
range of 80100 kV/mm. It is interesting to examine why the practical stresses are
so much lower than the strength of the material. Many studies have shown that the
electrical strength of practical cables is reduced with:
increased operating temperatures cable tests are often carried out at ambient
temperature whereas the cables are required to operate at 90
C
length of operation cables age (use up some of their performance) during oper-
ation, such that the performance capability after ten years operation is less than
at the start
length of cable and thickness of insulation the more insulation materials
employed, the larger the probability of occurrence of the harmful particles.
These effects need to be accommodated when designing the cable. The largest of these
effects and the one that is most difficult to address is the length of operation. This is
difficult because the lifetime of cables is generally very long, typically 3040 years.
This time span is longer than can be practically tested. Thus, engineers adopt the
approach of testing for 0.52 years and then extrapolating to the end of life. To enable
this extrapolation, an ageing model is required. Section 10.4.3 describes a number
of ageing models that have been developed. However, these ageing models have yet
to gain any significant acceptance amongst cable manufacturers and users, probably
because of their high complexity and the lack of good data to support them. Currently,
the inverse power law model is still widely used together with Weibull statistics to
indicate the reliability. In this section, we do not propose to describe these models
in detail, they are described elsewhere (e.g. Reference 14), or to describe the mathe-
matics of the new ageing models referenced above, we merely want to comment on
how important this area is for development.
The main inputs into mathematical design models are:
electrical ageing models which enable the extrapolation of test data out to service
lives
the experimental ageing data that serve as input to the ageing models
the breakdown characteristics of the cable systems.
However, as can be seen from Figure 10.10, even the global experience of cable
design data in this area is somewhat limited.
0
5
10
15
20
25
150 200 250 300 350 400 450 500 550
voltage, kV
o
p
e
r
a
t
i
o
n
a
l

l
i
f
e
,

y
r
s
Figure 10.10 Global operational experience of XLPE EHV cables systems as a
function of systemvoltage (test loops excluded) the size of the bubble
indicates the length of the cable circuit (courtesy of Borealis AB)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.1 1 10 100 1000 10000
time to failure, hours
p
r
o
b
a
b
i
l
i
t
y

o
f

f
a
i
l
u
r
e
50kV/mm
33kV/mm
Figure 10.11 Time to failure of XLPE cables installed in air when tested at two
different stresses
Furthermore, it can be seen from Figure 10.11 that the likelihood of cable failure
increases with time under electrical stress and that the ageing rate is very sensitive to
the applied electrical stress. The typical time to failure can be described by a number
of statistics (Gaussian, Weibull, Parameter Free, etc., see, e.g., Reference 28); the
relationship of these times to the stress and temperature is the ageing models. Once
the correct model has been established and parameterised, it may then be used to
determine the typical failure time for any desired stress and temperature. It should be
recognised that the ageing models deal with the relationship of the typical values but
do not, in general, deal with the statistical scatter. This scatter can be quite large (an
order of magnitude in the case for 33 kV/mm) and can have a profound influence on
the engineering predictions that emerge.
The importance of ageing considerations can be seen by considering one of the
many cable design approaches presently used for HV and EHV XLPE cables. It is
known that the AC voltage breakdown stress reduces with time and temperature.
Thus, the approach is based on four steps: [30]
(i) determining the breakdown performance of the virgin cables
(ii) determining how much of the initial performance will be lost due to elevated
temperature operation
(iii) determining howmuch of this performance will be consumed during the service
life of the cable system
(iv) determining an appropriate safety factor.
The following equations are then used:
E
design
=
E
fail
F
safety
F
age
F
temp
(10.4)
thickness =
V/
3
E
design
(10.5)
where:
E
design
= insulation design stress [kV/mm]
E
fail
= reference stress for cable failure; taken from breakdown curves [kV/mm]
F
safety
= safety factor
F
age
= ageing factor relating test duration to the required life of the cable
F
temp
= temperature factor relating test temperature to service temperature
V = cable voltage
thickness = thickness of insulation.
This approach has been used for the design of 275500 kVXLPEcables. By way of an
example, the thickness of a 400 kVcable may be determined for the ACcondition by:
E
design
=
40
1.1 2.3 1.1
= 14 kV/mm (10.6)
thickness =
400/
3
14
= 17 mm (10.7)
In this case, the level of ageing was assessed using the inverse power law for a design
life of 30 years. Inspection clearly shows that the largest factor that influences the
design stress and thus the thickness of a cable is the allowance made for ageing;
the 2.3 factor. If the allowance is too much, the cable will be too large and, hence,
there may be issues associated with weight, cost, length on a drum and number of
accessories. If the allowance is too little, then the cable will be too small and the
reliability will be compromised.
Thus, there can be no doubt that the correct understanding of the ageing phenom-
ena and the ability to accurately model themare absolutely critical to the development,
approval and successful operation of XLPE cables.
10.6 Direct current transmission
10.6.1 Economics
The incorporation of DC transmission into the pan-European network is not likely
to make massive savings in efficiency. However, even small savings may have very
large environmental benefits. For example, KEMAhas estimated [31] that the savings
fromDCdistribution in the ECwould be approximately 9.5 Mtonnes of CO
2
, approx-
imately one per cent of the total produced by electrical sources in the EC at the
moment.
Long distance HV cable interconnects are possible with reactive compensation
[29]. A simple calculation shows that a typical 100 MW capacity AC cable uses
approximately two per cent of its power in losses over 250 km, approximately
13 per cent over 500 km and almost 90 per cent over 1000 km. These large losses do
not occur under DC. Since high voltage DC cables are expected to be more compact
and recyclable, it may be possible to use them as longer links. This would also make
their installation less controversial as visual impact and noise would be reduced.
10.6.1.1 Economic drivers
In 1999, overall European electricity generation capacity was 575 GW with 322 GW
being conventional thermal plant. The corresponding actual energy generated for all
plant was 2533 TWh of which 1281 TWh was conventional thermal which is worth
$150 billion per annum.
KEMA, in their analysis of the potential benefits of the DC society [31], consider
that the following advantages exist in incorporating DC transmission into the current
AC grid:
decrease of transmission and distribution losses by 40 per cent
simple embedding of renewable energy sources such as photo-voltaics and
fuel cells
simple connection of energy storage systems
better use of existing transmission infrastructure due to higher energy densities
improved control of energy transport by fast power electronic devices, no reactive
power and limited short circuit currents
space saving with associated economic benefits.
Andersen et al. [33], in a 2002 review article, state, High voltage direct current
(HVDC) is often the economic means for delivering electric power over long distance
and/or for interconnecting two unsynchronised AC networks, which may be at dif-
ferent frequencies. HVDC schemes totalling 60 GW with individual scheme ratings
between 50 MW and 6300 MW have been installed worldwide, and many more
installations are being considered. The controllability of HVDC means that the
power delivered can be modulated to give improved damping to the AC transmis-
sion, sometimes allowing additional power to be transmitted safely through the AC
interconnection.
Such benefits are expected to generate a European internal market and a global
market for high voltage DC links. For polymer-insulated HVDC cable systems oper-
ating with newconverter technology, the opportunity is illustrated well by experience
to date with ABBs HVDC Light cable system which, in three years, has achieved
the penetration shown in Table 10.4.
Table 10.4 HVDC light cable projects
Project Power Voltage rating Cable length In service
(MVA) (kV) (km) (year)
Gotland/SE 50 80 140 1999
Directlink/AU 180 84 354 2000
Tjreborg/DK 8 10 8 2000
Murraylink/AV 200 150 355 2002
Cross Sound/US 330 150 82 2002
successfully commissioned, commercial operation delayed

Although these systems are still unproven in long-life installations and with new
voltage source converter technology (VSC) which is currently limited to 150 kV, it
is expected that over the long term they will be capable of operating in excess of
150 kV and could achieve or probably exceed power ratings equivalent to current
HVAC power cable systems i.e. 5001000 MVA. The operational benefits of being
able to transmit higher powers over equivalent HVDC cables without the need for
static VAR compensation as required with AC systems, and the reducing cost of
new VSC technologies, are expected to produce both capital investment and oper-
ational savings that will lead to continual improvements in the cost benefit case.
When the environmental benefits are added, it is expected that HVDC intercon-
nections will present a compelling case for a variety of interconnection projects
worldwide.
10.7 Testing
Polymeric power cable testing recommendations are contained in IEC 60840 [34].
These are classified as:
type tests
routine tests
sample tests
electrical tests after installation.
Prequalification tests are included for EHVand MVcables, but not HV(IEC 60840).
The status of HV prequalification tests are the subject of much discussion.
10.7.1 Prequalification and development tests
Prequalification tests are usually carried out by manufacturers to satisfy themselves
before formal development testing. As such, predevelopment tests vary widely
although CIGRE Working Group 21.03 has made some suggestions [34] mainly
regarding checking for contaminants, protrusions and voids.
Prequalification tests are intended to be representative of onsite conditions. Accel-
erated ageing is usually provided by increased electrical stresses (<1.7 U
0
) and by
heating and cooling cycles
2
. In addition, a prequalification test may include lightning
and switching impulse testing and wet testing to ensure that water trees do not grow.
Typical protocols are described in References 35 and 36. Prequalification tests require
approximately 1 year at EHV and between 1 and 2 years at MV.
10.7.2 Type approval testing
There are industry standards for type, sample and routine tests [3440]. Type tests
are project based; they qualify a particular design of cable and accessories and there-
fore, in principle, only need to be carried out once. Type approval tests require
approximately 3040 days to complete.
Atypical type test includes both electrical and non-electrical testing including:
partial discharge testing
dissipation factor
switching and lightning impulse testing
withstand voltage during bending and crossbonding leads.
various mechanical properties before and after thermal ageing
ozone resistance
density measurement
carbon black content
cold elongation and cold impact testing.
10.7.3 Sample testing
To check that the cable is within specified limits, a reel of cable from each
manufacturing run is tested. Typically, this includes [41]:
conductor examination and measurement of its electrical resistance
checks of dimensional accuracy
hot set test for XLPE and EPR insulation
measurement of capacitance
measurement of density of HDPE
lightning impulse test followed by a power frequency voltage test.
10.7.4 Routine testing
Routine tests are carried out on each manufactured component to check that it meets
specifications. These take the form of voltage withstand and partial discharge tests,
Table 10.5 [41]; all delivered cables must successfully complete this test before
dispatch.
10.7.5 Future trends in testing
At present, there is no agreement about whether tests are necessary, or what formsuch
tests should take, to monitor dry (thermoelectric) ageing (section 10.4.3). Ageing, in
this context, is considered to be the change within the polymer, brought about by
electrical and thermal stresses, that leads to macroscopic degradation, thus loss of
serviceability, ultimately leading to partial discharging and electrical treeing. Experi-
mental techniques are therefore required that can be used in a complementary manner
and that can be expected to provide chemicalphysical, microstructural and electrical
characterisation. It is alsoimportant that these experimental techniques provide ageing
markers that can be used for the prognosis of the cable and the diagnosis of ageing
leading to both bulk and localised degradation, which can take place close to the
design field.
The European ARTEMIS programme, whose partners include cable
manufacturers, material suppliers, electricity distributors and a number of universities
throughout Europe, has been tackling this problem. In the programme, a variety of
Table 10.5 Typical requirements for routine tests [41]
Rated Withstand test Withstand test Partial discharge
voltage(kV) voltage (kV) duration (Min) test voltage (kV)
4547 65 30 39
6069 90 30 54
110115 160 30 96
132138 190 30 114
150161 218 30 131
220230 318 30 190
285287 400 30 240
330345 420 60 285
380400 420 120 330
500 550 120 435
investigative techniques have been applied to lengths of cables and to samples peeled
from full-sized XLPE-insulated cables. The techniques used range from those with
micron and submicron resolution, which are intended to determine submicron fea-
tures, through to bulkmeasurements such as differential scanning calorimetry (DSC)
that give information on larger characteristics, e.g. crystallinity, in addition to lamella
thickness information. Spectroscopic techniques have been used to identify the chem-
ical composition (e.g. in terms of crosslinking byproducts such as acetophenone and
cumyl alcohol) and aid the interpretation of the electroluminescence processes. The
electrical properties of the material are characterised by dielectric measurements,
conduction currents, electroluminescence, space charge accumulation and packet for-
mation. Relationships between the various electrical properties are attempted taking
into account the transit of space charge across the sample and the onset field at which
electroluminescence and space charge can be detected.
It is expected that tests based on such work will be developed by cable
manufacturers as the basis for the prognosis of the health of the cable systems [42].
10.8 Notes
1
Ageing and reliability testing and monitoring of power cables: diagnosis for
insulating systems
2
U
0
is the rated r.m.s. power frequency voltage between each conductor and screen
or sheath for which cables and accessories are designed
10.9 References
1 DENSLEY, J.: Ageing and diagnostics in extruded insulations for power
cables. 1995 IEEE 5th international conference on Conduction and breakdown
in solid dielectrics, IEEE pub. 95CH3476-9, 1995, pp. 115
2 ABERNETHY, ROBERT B.: The new Weibull handbook (Robert B.
Abernethy, North Palm Beach, Fla., 1996, 2nd edn.)
3 MOORE, G.F. (Ed.): Electrical cables handbook (Blackwell Science
Ltd, 1998)
4 VOLDHAUG, L., and ROBERTSON, C.: MVoverhead lines using XLPE cov-
ered conductors Scandinavian experience and NORWEB developments. IEE
Conf. Publ., 406, 1995, proceedings of the 2nd international conference on
the Reliability of transmission and distribution equipment, March 2931 1995,
Coventry, UK, pp. 5260
5 LEIJON, MATS: Novel concept in high voltage generation: Powerformer
TM
.
IEE Conf. Publ., 467, 1999, proceedings of the 11th international sym-
posium on High voltage engineering August 2327 1999, London, UK,
pp. 5.379.S55.382.P5
6 KARLSTRAND, J., SUNNEGARDH, P., ZENGER, W., GHAFURIAN, R.,
and BOGGIA, R.: Water-cooled 345 kVsolid-dielectric cable system. CIGR
paper 21-111, 2002
7 Power Eng. J., June 2002, 16, (3) special issue on AC/DC power transmission
8 BARTNIKAS, R., and EICHHORN, R.M.: Engineering dielectrics in Elec-
trical properties of solid insulating materials, American society for testing and
maetrials (ASTM), 1983, IIA, (STP 783), pp. 1721
9 BRINCOURT, T., and REGAUDIE, V.: Evaluation of different diagnostic
methods for the French underground MV network. JICABLE99 451
10 OHATA, K., TSUCHIYA, S., SHINAGAWA, N., FUKUNAGA, S.,
OSOZAWA, K., andYAMANOUCHI, H.: Constructionof longdistance 500kV
XLPE cable line. JICABLE99, pp. 3136
11 COPPARD, R.W., BOWMAN, J., DISSADO, L.A., ROLAND, S.M., and
RAKOWSKI, R.T.: The effect of aluminium inclusions on the dielec-
tric breakdown strength of polyethylene, J. Appl. Phys., 1990, 23,
pp. 15541561
12 ANTONISCKI, J.R., NILSSON, U., and GUBANSKI, S.M.: Influence of metal
inclusions on the breakdown strength of model cables. Conference on Electri-
cal insulation and dielectric phenomena (CEIDP), Annual Report, 1, 1997,
pp. 283286
13 HAGEN, S.T., and ILDSTAD, E.: Reduction of AC breakdown strength due to
particle inclusions in XLPE cable insulation. Conference on Power cables and
accessories 10 KV500 kV, 1993, 382, pp. 165168
14 DISSADO, L.A., and FOTHERGILL, J.C.: Electrical degradation and break-
down in polymers (Peter Peregrinus, London, 1992)
15 STEENNIS, E.F.: Water treeing the behaviour of water trees in extruded cable
insulation. ISBN 90-353-1022-5, pp. 132133
16 PESCHKE, E., SCHROTH, R., and OLSHAUSEN, R.: Extension of XLPE
cables to 500 kV based on progress in technology. 1995, JICABLE 610
17 ANDERSEN, P., DAM-ANDERSEN, M., and LORENSEN, L., et al.:
Development of a 420 kV XLPE cable system for the Metropolitan Power
Project in Copenhagen. CIGRE, 1996, pp. 21201
18 OGAWA, K., KOSUGI, T., KATO, N., and KAWAWATA, Y.: The worlds first
use of 500 kVXLPE insulated aluminiumsheathed power cables at the Shimogo
and Imaichi power stations, IEEE Trans. Power Deliv., 1990, 5, (1), pp. 2632
19 POHLER, S., BISLERI, C., NORMAN, S., PARMIGIANI, B., and
SCHROTH, R.G.: EHV XLPE cables, experience, improvements and future
aspects. CIGRE2000, pp. 21104
20 HENNINGSEN, C.D., MULLER, K.B., POLSTER, K., and SCHROTH, R.G.:
New 400 kV XLPE long distance cable systems, their first application for the
power supply of Berlin. CIGRE1998, pp. 21109
21 First 525 kV XLPE extra high voltage cables destined for Dachaoshan, Mod.
Power Syst., 2000, pp. 3941
22 400 kV XLPE cable going underground, Mod. Power Syst., 1999, pp. 3537
23 DISSADO, L.A., MAZZANTI, G., and MONTANARI, G.C.: The role of
trapped space charges in the electrical aging of insulating materials, IEEETrans.
DEI, 1997, 4, (5), pp. 496506
24 LEWIS, T.J., LLEWELLYN, P., VAN DER SLUIJS, M.J., FREESTONE, J.,
and HAMPTON, R.N.: A new model for electrical ageing and breakdown in
dielectrics. 7th DMMA, 1996, pp. 220224
25 CRINE, J.P.: A molecular model to evaluate the impact of ageing on space
charges in polymer dielectrics, IEEE Trans. DEI, 1997, 4, (5), pp. 487495
26 GRIFFITHS, C.L., BETTERIDGE, S., and HAMPTON, R.N.: Thermoelectric
ageing of cable grade XLPE in dry conditions. IEEE international conference
on Conduction and breakdown in solid dielectrics, 1998, pp. 279282
27 HIROSE, H.: A method to estimate the lifetime of solid electrical insulation,
IEEE Trans. EI, 1987, 22, pp. 745753
28 IEEE P930: Guide for the statistical analysis of electrical insulation breakdown
data. IEEE, 2002 (to be published by IEEE Standard Activities Department)
29 HALVARSSON, P., KARLSTRAND, J., LARSSON, D., LARSSON, M.,
REINHOLDSDTTIR, K., and SIGURSSON, E.: Anovel approach to long
buried AC transmission system. CIGR paper 21-201, 2002
30 GREGORY, B., GRIFFITHS, C.L., HAMPTON, R.N., and
MAINWARING, S.P.: A probabilistic design method for polymeric structures
to operate at high electrical stress. Dielectric materials, measurements and
applications 2000, IEE Conf. Publ. 473, pp. 419424
31 KEMAElectricityTechnologyRoadmap: Technologyfor a sustainable society.
April 4, 2002
32 CIGR: Recommendations for electrical tests (prequalification and develop-
ment) on extruded cables and accessories at voltages >150 (170) kV and <400
(420) kV, Electra, 1993, 151, pp. 1519
33 ANDERSEN, B.R., XU, L., HORTON, P.J., and CARTWRIGHT, P.:
Topologies for VSC transmission, Power Eng. J., 2002, 16, (3), pp. 142150
34 IEC Committee Draft 20A/407/CD: Tests for power cable systems. Cables
with extruded insulation and their accessories for rated voltage above 150 kV
(U
m
= 170 kV) up to 500 kV (U
m
= 525 kV). International Electrotechnical
Commission, Geneva, Switzerland, February 1999
35 PARPAL, J.-L.: Prequalification testing Alcatel 290/500 (525) kV extruded
cable system at IREQ. IEEE/PES Insulated Conductors Committee, minutes of
the 102nd meeting, appendix 13-A, St Pete Beach, FL, November 1997
36 PARPAL, J.L., AWAD, R., and BELEC, M. et al.: Prequalification testing of
345 kV extruded insulation cable system. CIGRE1998, pp. 21101
37 IEC Publication 60840: Tests for power cables with extruded insulation for
rated voltage above 30 (36) kV up to 150 (170) kV. IEC, Switzerland, 1993
38 AEIC Specification CS7-93: Specifications for crosslinked polyethylene insu-
lated shielded power cables rated 69 kV through 138 kV. AEIC, Birmingham,
AL, June 1993
39 IEEE Standard 48-1996: IEEE standard test procedures and requirements for
alternating-current cable terminations 2.5 kVthrough 765 kV. IEEE, NewYork,
June 1996
40 IEEE Standard 404-1993: IEEE standard for cable joints for use with extruded
dielectric cable rated 5000138000 V and cable joints for use with laminated
dielectric cable rated 2500500000 V. IEEE, New York, July 1994
41 HIVALA, L.J.: Extruded solid-dielectric power transmission cables, in
BARTNIKAS, R., and SRIVASTAVA, K.D. (Eds): Power and communication
cables (IEEE Press, 2000), chap. 4, p. 231
42 FOTHERGILL, J.C., MONTANARI, G.C., and STEVENS, G.C. et al.:
Electrical, microstructural, physical and chemical characterisation of HVXLPE
cable peelings for an electrical ageing diagnostic data base, IEEE Trans.
Dielectr. Electr. Insul., 2003, 10, (3), pp. 514527
Chapter 11
Numerical analysis of electrical fields
in high voltage equipment
A.E. Baker, T.W. Preston and J.P. Sturgess
11.1 Introduction
The accurate determination of the electrical field distribution is vitally important in
the production of safe and economic designs of high voltage electrical equipment.
With transmission voltages typically 400 kV (r.m.s.), line and test voltages of
630 kV (r.m.s.) at power frequencies and 1425 kV (peak) lightning impulse, the
problems of configuring the high voltage (HV) winding, leads, stress control rings
and bushings of an HV transformer become apparent. On the rare occasions when
something goes wrong, the consequences can be very serious. Figure 11.1 indicates
the severe damage that can be caused by an HV flashover, in this case between the
corona shield at the coil end of the HV bushing and a neutral lead. The insulation on
the lead had been severely ruptured, and as a result a costly repair was needed.
The HV transformer is only one of many examples that show the importance of
designers having analysis techniques capable of assessing or revalidating their designs
accurately under all anticipated steady-state and transient operating conditions.
Besides the quasi-electrostatic problems associated with HVequipment discussed
above, there is much interest in, and concern over, the large voltage gradients that
occur along the end windings of machines fed by inverters. The inverter voltage
waveforms, seen by the end windings, contain high harmonic components that can
lead to sparking and erosion at the insulation surface unless special measures are
taken. To alleviate the stresses, special insulating tapes have been developed whose
conductivity increases with the applied electrical stress [1, 2]. These tapes reduce the
stress concentration where the stator winding emerges from the stator slot.
As shown by the examples given, physical characteristics such as the geometric
arrangement, the material properties of the insulation and stress grading materials
Figure 11.1 Insulation rupture of the neutral lead (courtesy of ALSTOM T&D
Transformers, Stafford (TST))
and the applied voltage must be taken into consideration. Thus, as new materials are
developed and new configurations conceived, it is essential to be able to model them
at the design stage rather than to rely solely on intuition and experience.
Modern design analysis is based on numerical methods, several of which are
available to the designer. The question is, which one offers the best facilities in terms
of speed of analysis, flexibility of modelling and range of investigative tools?
11.2 Which numerical method?
11.2.1 The finite-difference method
The finite-difference method was originally developed by Southwell [3] in 1946 but
was not extensively used until the early 1960s when digital computers could be used
to improve solution speed.
The method relies on the problem area being divided into a grid structure, as
shown in Figure 11.2, from which a numerical equation for each grid intersection can
be obtained through a Taylor expansion of the differential equations describing the
solution field. From the resulting matrix the formulation potential can be evaluated
at each node. Normally the matrix of equations is solved using an iterative method.
The method has been mainly used for two-dimensional geometries but can also
be applied to three-dimensional arrangements. However, the representation of curved
or irregular volumes is poor and, in electrostatic calculations, the electrical stresses
can be misleading in these regions.
Numerical analysis of electrical fields in high voltage equipment 513
flange
bushing core
bushing profile
Figure 11.2 Example of a finite-difference mesh
11.2.2 The finite-element method
The finite-element method was originally used in the aeronautical and civil
engineering industries and was promoted extensively by Zienkiewicz and Taylor [4]
in the late 1950s and early 1960s. It was the late sixties before the method entered the
electrical engineering industry, but it has been continually developed since that time.
The method relies on transforming the problem area of complicated form over
a continuous region into one in which the region is divided into small elements
(discretised) and an approximate solution is used for the field quantity in each element
(Figure 11.3). By suitable choice of elements, complicated geometries can be repre-
sented and each element can have different physical properties, that may be non-linear
(i.e. field dependent). Transforming the governing partial differential equations into
an energy form, an algebraic expression may be created for the potential at each node
through the extremisation of the energy functional. The resulting set of equations is
sparse and there are many techniques that can be used to invert the matrix efficiently.
Although a two-dimensional problem is shown in Figure 11.3, the method is
equally adaptable to three-dimensional situations. However, the generation of the
mesh is more complex and time consuming. In excess of 1 million unknowns is not
uncommon.
11.2.3 The boundary-element method
This is another analysis technique that has been available to designers for many years.
It models the problem under consideration by representation of the surfaces of the
bushing
bushing core
flange
Figure 11.3 Example of a finite-element mesh
bushing core
bushing
flange
air region
Figure 11.4 Example of a boundary-element mesh
regions concerned rather than their interiors, in contrast with the last two methods.
This results in the problem being reduced by a dimension, i.e. a two-dimensional
problem is modelled with one-dimensional (line) elements, a three-dimensional one
with two-dimensional elements. Asignificant advantage of the method is the fact that
the air regions between solid objects do not need to be discretised. This can be of
major benefit when it comes to three-dimensional mesh generation. An example of
a boundary-element discretisation is shown in Figure 11.4.
The numerical equations are formed from the application of Greens theorem to
the partial differential equations that describe the field distribution within the region
concerned. An equation is formed at the nodes on the outer surfaces of all regions and
results in an unsymmetrical, densely packed matrix. An efficient method for inverting
the matrix is difficult to find and leads to large solution times.
The method is applicable to both two-dimensional and three-dimension problems.
However, the latter results in a very large unsymmetrical matrix to be inverted, so
restricting the complexity of the problem that can be solved.
11.2.4 Comparative summary
All the techniques describedhave their advantages anddisadvantages, socomplicating
choice. For instance, although it is easy to set up the problem, the finite-difference
technique is not easily applied to modelling irregular geometries, such as slanted and
curved surfaces. Nodal distribution can also be very inefficient. This is not so with
finite elements. Equally, the boundary-element method can model efficiently regions
in which the material properties are linear, but it is not efficient at handling non-
linear materials. Furthermore, the type of matrix formed is not amenable to efficient
solution. The finite-element method is well suited to the modelling of non-linear
materials and the sparse form of the resulting matrix makes it readily adaptable to
efficient solution, but it is necessary to discretise the free space surrounding the object
to be modelled.
It is impossible to be definitive as to which method is best different estab-
lishments will arrive at different conclusions according to their specific range of
applications and problems. However, history has shown that as a robust and flexi-
ble design method, the finite-element method is generally best suited to the needs
of designers, although in some cases there is little to choose between it and the
boundary-element method.
11.3 Formulation of the finite-element equations in two and three
dimensions
11.3.1 General
The finite-element method transforms a problemof complicated formin a continuous
region into one in which the region is subdivided into a large number of well defined
regions, known as elements. An approximate (though good) solution is obtained for
the potential at the nodes that lie at the vertices of the element. It will be evident that
complicated geometries can be represented well and that, in principle, each element
can have different properties.
In order to obtain the field distribution in a defined space, a formulation is obtained
either through an approach such as the Galerkin method [5] or directly via a functional
derived from the stored electrical energy of the region under consideration. This
functional is extremised with respect to the potential at the nodes defining the elements
and then integrated over all elements associated with the node concerned, given an
assumed polynomial variation of potential over the element (the shape function). An
equation is thus formed at each node, resulting in a large set of simultaneous equations
that replaces the governing partial differential equation.
To illustrate the process, the method will be applied to the modelling of an elec-
trostatic field. The partial differential equation that describes the voltage potential
distribution within any given region is derived as follows:
D =
v
(11.1)
where D is electrical flux density, Cm
2
,
v
is free charge density, Cm
3
, and since
D =
0
r
E then Equation (11.1) becomes:
(
0
r
E) =
v
(11.2)
where E is electrical stress, Vm
1
,
0
is 8.85419 10
12
, Fm
1
and
r
is relative
permittivity.
To formulate in terms of the voltage potential, V can be written in terms of E:
V = E (11.3)
Substituting this in Equation (11.2) results in the following equation:
{
0
r
(V)} =
v
(11.4)
Equation (11.4) can be rewritten for a homogeneous region as:
V =

v
r
(11.5)
Equation (11.5) is called Poissons equation and applies to a homogeneous medium.
If
v
is zero the equation reduces to Laplaces equation for homogeneous media. This
is shown in full by Equation (11.6) (an expansion of Equation (11.5) in Cartesian
coordinates):
2
V
x
2
+

2
V
y
2
+

2
V
z
2
= 0 (11.6)
Equation (11.6) is the partial differential equation that describes the voltage potential
distribution in high voltage situations were the medium is homogeneous and the
charge density is zero. If the medium is non-homogeneous then Equation (11.6) is
written as:
r
_
2
V
x
2
+

2
V
y
2
+

2
V
z
2
_
= 0 (11.7)
for each region. This is to allowfor the different permittivities in the different regions.
To use Equation (11.7) in a finite-element formulation requires the equation to
be transformed into an energy functional form that relates directly to the electrical
energy of the system. This equation can be derived by several methods as described
in the following section.
11.3.2 Forming the functional equation
Several approaches are possible for formulating the functional equation. One is to
use the variational method, which is based on Eulers theorem [6]. Unfortunately, it
is not always possible to deduce the functional from it, and it is not regularly used.
However, Eulers equation can always be used to check the validity of the functional
since the resulting equation reverts to the original partial differential form.
An alternative method, the energy-related functional, can often be written directly
fromelectrical stored energy considerations. This approach is, perhaps, more intuitive
for the engineer who can more easily relate to the energy of a system rather than
a strictly mathematical approach.
A third and increasingly popular technique is a particular implementation of the
weighted residual method known as the Galerkin method. The Galerkin method starts
by supposing that a trial solution potential exists at each node. Insertion of this set
of potentials into the governing equation will lead to a residual at each node, i.e.
the equation will not be perfectly satisfied since, in general, the trial values will
not be correct (there is a similarity to the finite-difference formulation here). It is
clear that an approximate solution can be obtained by adjusting the potentials to
minimise the sum of the residuals at all the nodes (this method is sometimes called
the collocation method). However, a better solution can be obtained by introducing
weighting functions at each node to try to force the sum of the local residual errors
to zero over the whole domain. This method is described as the weighted residual
method. Where the weighting functions used are the chosen shape functions for the
discretised region, the method becomes the Galerkin method.
11.3.3 The energy functional illustrated
The total electrical energy in a system of volume may be written as:
F =
1
2
_
D E d (11.8)
i.e.
F =
_
1
2
r
E
2
d (11.9)
If it is assumed that the permittivity is constant within the region concerned, then
Equation (11.9) may be used to write the energy (in a Cartesian coordinate system) as:
F =
_
r
2
_
_
V
x
_
2
+
_
V
y
_
2
+
_
V
z
_
2
_
dx dy dz (11.10)
It can be shown that this is the functional that, when differentiated with respect to
V and equated to zero, gives a distribution of V that satisfies the governing partial
differential equation. Physically, the process corresponds to minimising the stored
electrical energy i.e. the potential energy of the system for the imposed boundary
conditions. The differentiation is most conveniently carried out on the discretised
system, and this will be shown in relation to the two-dimensional case.
For element e:
F
e
=

0
r
2
_
e
_
_
V
x
_
2
+
_
V
y
_
2
_
dx dy (11.11)
where suffix e indicates integration over an element.
Hence, the contribution to the rate of change of F with V from the variation of
potential of node i in element e only,
e
, is:
e
=
F
e
V
i
=

0
r
2
_
e
V
i
_
_
V
x
_
2
+
_
V
y
_
2
_
dx dy (11.12)
e
=

0
r
2
_
e
_
2
V
x

V
i
_
V
x
_
+2
V
y

V
i
_
V
y
__
dx dy (11.13)
11.3.4 Numerical representation
To solve high voltage problems using the finite-element method requires
Equation (11.13) to be equated to zero. To represent the problem numerically, the
problem region is divided into elements and Equation (11.13) is applied at the nodes
forming the element vertices. The variation of the potential over the elemental shape
has then to be approximated by a polynomial distribution (known as the shape
function). The order of the chosen polynomial will dictate the type of element used,
for example a linear distribution would only require a simple triangular element. For
higher order shape functions, the number of nodes describing the element must be
capable of defining the order used, e.g. a quadratic shape function over a triangular
element requires nodes at the middle of each element side. Many books on the finite-
element method, e.g. Reece and Preston [7], cover the numerical derivation of the
functional, so it will not be discussed further here.
There will be contributions to the rate of change of the region functional with
respect to the potential V
i
at node i from all the elements connected to i. In the case
shown in Figure 11.5, there will be contributions from elements 1 to 6.
Hence, generally, the contribution to F/V from a change in V
i
is:
F
V
i
=
e
F
e
V
i
(11.14)
where

e
represents the summation of contributions from all elements associated
with V
i
, i.e., all the elements connected to node i.
When these derivatives are equated to zero, a group of simultaneous equations is
formed. These can be written in matrix form as:
[S
t
]{V
i
} = 0 (11.15)
where [S
t
] is a square matrix known as the stiffness matrix and is formed from the
geometric coordinates of the nodes defining the elements and the material properties.
V
i
is a column matrix containing all the nodal potentials. The coefficient stiffness
matrix will be sparse, i.e. it will contain many zeros since, from the discussion above,
i
j
k
element 6
element 5
element 4
element 3
element 2
element 1
Figure 11.5 Elements associated with node i
it will be evident that any node will only be coupled to nodes directly connected
to it by an element edge. Some of the nodes included in the equations will fall on
boundaries, and their potentials may be known, or some other boundary condition
may apply.
11.4 Variations on the basic formulation
11.4.1 General
When solving engineering problems, certain features arise where it is necessary to
modify the original formulation either to improve the accuracy of modelling or speed
of solution. Such features could be the representation of very thin metal foils that
are used in bushings to grade the potential, so avoiding high stress regions, or the
representation of partially conducting layers such as may occur on overhead lines due
to contamination from the environment. To model these types of region requires the
use of very thin elements (which can lead to numerical inaccuracy) or a modification
to the formulation to allow elements of zero thickness to be used. The section below
details several engineering implementations which have enabled a wider range of
problems to be solved.
11.4.2 Representation of foils
In the design of high voltage systems, there is a need, in particular regions, to grade the
voltage distribution or to shield regions of potentially high stress. Normally, this stress
earthed casing flange
foils
HV conductor
centre line
bushing insulation
Figure 11.6 Generator terminal bushing
busbar
centre line
rubber
insulation
foil
void
Figure 11.7 Conical bushing
grading is achieved through the use of foils which are thin metal strips. Figures 11.6
and 11.7 illustrate two examples where foils are used.
The electrical characteristic of a foil is a surface of constant voltage, whose value
is dependent upon the capacitance of the system in which it is used.
To model foils using finite elements the foil is either represented as a region of high
permittivity or as a constant voltage surface. The first approach has the disadvantage
that foils are extremely thin (0.1 mm) and to model them would result in skinny
elements having a large aspect ratio. This, combined with the very large permittivity,
makes the stiffness matrix ill conditioned with the result that it cannot be inverted
accurately. The second approach is by far the most convenient and accurate way of
modelling foils. Figure 11.8 illustrates, schematically, howthe foil can be represented
by line elements.
All the nodes connected by the line elements (marked F in the figure) are con-
strained to be at the same (unknown) potential, whose value floats to a level that is
FFF F F F F FFF
foil
Figure 11.8 Foil representation
consistent with the capacitive distribution of the problem. Afine mesh is required at
the end of the foils to ensure an accurate value of stress.
11.4.3 Directional permittivity
In high voltage equipment, there is a considerable amount of insulating material used
for support purposes and, in some cases, these structures can have permittivities that
depend upon the method of construction. For instance, a material used regularly in
high voltage transformers is made from a wood/resin mixture. The electrical grade
consists of laminates of wood bonded together with resin to form a layered board
having different permittivities in different directions. However, the mechanical grade
is formed from a homogeneous wood/resin mixture, so the bulk permittivities are the
same in all directions.
To model directional permittivities the functional equation is modified simply by
relating the directional permittivity to the appropriate coordinate as indicated in the
following equation:
F =
_
0
2
_
x
_
V
x
_
2
+
y
_
V
y
_
2
+
z
_
V
z
_
2
_
dx dy dz (11.16)
11.4.4 Modelling resistive/capacitive systems
There are numerous occurrences in industrial design where the material used has both
capacitive and resistive properties. Examples are:
(i) stress grading paints or tapes used to redistribute the stresses on the end
windings when they emerge from the stator core of machines
(ii) contamination on bushings derived from environmental effects
(iii) the coating of optical fibres, intended for use in high voltage environments,
with a semiconductor to prevent high stress due to surface defects.
The current density in a partially conducting material has two components: J
cond
, the
conduction current and J
cap
, the capacitive current:
J = J
cond
+J
cap
(11.17)
These can be related to the permittivity, , the conductivity, , and the time derivative
of the electric field, E:
J = E +
E
t
(11.18)
If the electric field is assumed to vary sinusoidally with time at a radian frequency ,
then:
J = E +jE (11.19)
or, a complex permittivity may be defined to relate D and E:
D =
_
j
_
E (11.20)
The electric energy over a volume for this systemis given, fromEquation (11.8), by:
F =
_
j
_
_
E E
2
d (11.21)
Replacing E by the derivative of the voltage gives the functional:
=
1
2
_
j
_
_
_
V
x
_
2
+
_
V
y
_
2
+
_
V
z
_
2
d (11.22)
The functional can be discretised using the finite-element method and then minimised
to give the voltage distribution. The problem can be solved for a three-dimensional
region which is divided into tetrahedral elements or reduced to two-dimensional and
solved over triangles, for example. The voltage over each tetrahedron can be described
by a first-order linear shape function or by higher order elements. Once the voltage
distribution has been found, it is possible to derive further field quantities such as
electric stress, current density and loss intensity.
11.4.5 Modelling partially conducting tapes and paints
Partially conducting regions such as stress grading tapes or contaminated surfaces are
typically only tenths of a millimetre thick, orders of magnitude smaller than typical
dimensions associated with other components of the problem concerned. In a normal
finite-element formulation, this leads to elements in the mesh with poor aspect ratios
which can, in turn, lead to numerical instability, particularly in a non-linear regime.
For this reason, a surface element formulation can be introduced to represent those
types of partially conducting region whose relative thickness is small. The surface
elements are assigned the material properties of the partially conducting layer but the
elements occupy no physical thickness in the model. This allows the energy associated
with the stress grading layer to be included without the numerical errors that skinny
elements would introduce. The formulation assumes that there is a voltage drop along
the length and width of the tape but not across its thickness.
The energy functional at the surface is similar to that for the volume:
=
1
2
_
j
_
_
_
V
p
_
2
+
_
V
q
_
2
d (11.23)
where p and q are two mutually orthogonal directions tangential to the plane of the
surface element.
11.4.6 Space charge modelling
The effects of space charge can be included via the Poisson equation:
2
V =
0
(11.24)
where is the charge density.
In general, the space charge distribution will not be known a priori and must be
calculated iteratively during the finite-element solution. This is achieved by including
an additional relationship between the voltage and the charge density. There are
many possibilities for this relationship, depending on the physics of the problem.
For example, in the case of a multielectrode valve device under space charge limited
conditions, the voltage and current at a point are related by the three-halves power law:
V
3/2
=
9
4
J
0
_
m
2e
_
1/2
x
2
(11.25)
where J is the current density at the given point, Am
2
, m is the mass of an electron,
kg, e is the charge on the electron, C, x is the distance of the point from the cathode,
m and V is the voltage at the given point, volts.
The current density is related to the charge density by:
J = u (11.26)
where u is the electron velocity, ms
1
.
In this case, the electron velocity can be obtained by equating the electron kinetic
energy and the electrical potential energy.
This gives a relation between V and which can be used in the finite-element
iterations until a consistent voltage and charge distribution is found:
= J
_
m
2eV
_
1/2
(11.27)
For other space charge problems, the physical relationshipbetweenvoltage andcharge
density may be different and alternative formulations may be required such as ray
tracing simulation or surface charge modelling Hanke et al. [8].
11.4.7 Time variation
When the applied voltage cannot be represented by a sinusoid, for example when
transient or containing harmonics, it is necessary to use a time domain solution to time
step through the waveform. At each time step, an iterative method is used to reach
the correct conductivity for the electric stress distribution based on the non-linear
characteristics of the material.
The finite-element equation for the voltage can be written in terms of the stiffness
matrix [S], the nodal voltages {V}, and the material properties as:
[S]{V} +[S]
d{V}
dt
= 0 (11.28)
The time derivative can be approximated by:
V
t
=
V
new
V
old
t
(11.29)
Then, evaluating the voltage at the mid-point of the time step gives:
[S]
1/2
({V
new
} +{V
old
}) +
[S]({V
new
} {V
old
})
t
= 0 (11.30)
At a given time step, V
new
is unknown while V
old
is known from the previous step.
Thus, the finite-element equation to be solved becomes:
[S]
_
2
+

t
_
{V
new
} +[S]
_
2

t
_
{V
old
} = 0 (11.31)
11.4.8 Open boundary problems
Of significant concern in most electrical field analysis is the representation of the
infinity of free space in which many devices are assumed to sit. If free space is
discretised along with the rest of the model, there is a question of where the mesh
should be truncated. The simple, but unhelpful, answer is that the region should be
truncated at a point where placing it any further away from the sources has no effect
on the solution. Clearly, this involves at least two solutions to the problem.
To avoid this, there have been quite a number of methods proposed that may be
characterised as:
(i) using elements with special properties
(ii) applying special boundary conditions.
In the former category come so-called ballooning elements and infinite elements.
These methods often require boundaries of predetermined shapes or discretisation,
which may be restrictive. In the latter come hybrid finite-element/boundary-element
methods (requiring special post-processing techniques) and, more recently, the charge
iteration method of Aiello et al. [9]. This last method is showing great promise, as
the boundaries of the mesh can be of arbitrary shape and placed quite close to the
sources of excitation. It is even possible to have some parts of the mesh physically
disconnected from other parts of the mesh, linked only through the special boundary
conditions.
11.5 Applications
11.5.1 General
Numerous problems arise in electrical engineering due to the presence of high
voltages. These range from insulation failure in large transformers and high voltage
generators through contamination effects on bushings and overhead lines to potential
health hazards from power lines and cables. It is not possible to illustrate all such
problems but the following examples attempt to show the versatility of the finite-
element method and how the variants of the functional are used to good effect in
solving unusual problems.
11.5.2 High voltage transmission line
There has been controversy recently concerning the effect that electric and magnetic
fields, emitted by high voltage overhead lines, might have on human beings both
psychologically and physically. Thus, there is a need to determine the magnitude
and distribution of these fields and if necessary to reduce them. Previously, this type
of investigation would be done experimentally, requiring a long length (500 m to
1000 m) of overhead line to be constructed and powered, and electromagnetic field
measurements to be made. To obtain results from such a system is time consuming
and expensive. The study below shows how this type of investigation can be done
using the finite-element method.
The problem to be solved is to determine the electric field distribution due to a set
of three-phase high voltage cables as shown in Figure 11.9. The position of the cables
is taken midway between the pylons allowing for the catenary effect due to the cable
weights. The arrangement considered is a 440 kV, twin three-phase overhead system.
(The complete cable arrangement between pylons can be modelled but would require
a three-dimensional solution that was considered unnecessary for the problem under
investigation.)
Figures 11.10 and 11.11 show the horizontal and vertical components, respec-
tively, of the electrical field distributions. The distribution shows clearly how the
cables act like aerials in emitting the field. By changing the phase sequence on one
side of the cable layout, as shown in Figure 11.12, a different distribution can be
obtained. Thus, the phase sequence can be selected to minimise the field at ground
level as shown in Figure 11.13.
Figure 11.13 indicates that a person or dwelling 45 m from the centre of the tower
(50 m along axis) will be exposed to a lower electric field with phase sequence b than
phase sequence a.
11.5.3 Foils in high voltage bushings
Figure 11.7 illustrated a conical bushing in which it was required to reduce the stress
at the junction of the insulation and conductor. Due to manufacturing tolerances, it
is almost inevitable that a void is created at such junctions between the busbar and
the insulation resulting in a high stress region with subsequent breakdown. To reduce
a b
earth
plane
earth wire
phase A
phase B
phase C
phase A
phase B
phase C
phase A
phase B
phase C
phase A
phase B
phase C
earth wire
Figure 11.9 High voltage overhead cable arrangement
a outline of pylon
b position midway between pylons
earth wire
phase A
phase B
phase C
earth plane
phase A
phase B
phase C
Figure 11.10 Variation of the horizontal component of electric stress
the high stresses in the void region a foil can be strategically placed as shown in
Figure 11.14 which also compares the electrical stresses for the cases without and
with foils.
However, care is needed in the placement of the foil since higher stresses can,
as shown also in Figure 11.14, occur at the ends of the foil. Normally, these higher
stresses are situated in a region that can withstand the higher values than in an air void.
earth wire
phase A
phase B
phase C
earth plane
phase A
phase B
phase C
Figure 11.11 Variation of the vertical component of electric stress
a b
earth wire
earth plane
earth wire
A
B
C
A
B
C
A
B
C A
B
C
Figure 11.12 Phase sequence layouts
11.5.4 Modelling the effect of contamination on an insulating system
Many insulating materials are used in a polluted environment; the sheds of high
voltage bushings, aerial optical fibres strung from power pylons and the end winding
conductors of some motors are just three examples. Atmospheric pollution settles
on the surface of the insulator in a thin layer. It contains a proportion of debris
that, although of very high resistivity when dry, can become partially conducting
when damp.
30 40 50
distance, m
phase sequence b
phase sequence a
60 70 80 90 100 10
2
4
6
e
l
e
c
t
r
i
c

f
i
e
l
d
,

V
/
m
10
3
20
Figure 11.13 Variation of the total electrical stress at ground level
< 5.0000E+02
< 1.0000E+03
< 1.5000E+03
< 2.0000E+03
< 2.5000E+03
< 3.0000E+03
< 3.5000E+03
< 4.0000E+03
< 5.0000E+02
< 1.0000E+03
< 1.5000E+03
< 2.0000E+03
< 2.5000E+03
< 3.0000E+03
< 3.5000E+03
< 4.0000E+03
high stress region
high stress region
foil
void
void
with foils
without foils
Figure 11.14 Comparison of the electric stress distributions
This thin, partially conducting layer at the surface of an insulator causes the
voltage and stress distribution at the surface to change substantially from the dry
(and as-designed) state. This is particularly true if the partially conducting layer is
absent in one location for some reason, for instance:
(i) the pollution layer has dried out locally due to a local source of heat (external
or arcing)
contamination
contamination
paper-
insulated
conductor
and foil
assembly
oil
ceramic
no contamination
Figure 11.15 HV bushing with partial contamination on the outer sheds
(ii) the pollution layer has been removed examples include an incomplete manual
cleaning of the sheds of a bushing or birds landing on aerial optical fibres,
wiping them clean with their claws
(iii) variable exposure of the insulating system to the source of pollution.
A major difficulty for the modeller is finding an appropriate value to use for the
electrical conductivity, although many problems are not strongly sensitive to the value
chosen. If critical, studies will have to be made covering the range of possible values.
Figure 11.15 shows a section through a high voltage bushing in which foils are
used to create a uniform stress distribution in the insulation.
Three studies were performed:
(i) with all the sheds free of pollution the as-designed state
(ii) with the surface of the sheds covered in a 1 mm thick layer of partially
conducting pollution
(iii) as (ii), but with one of the sheds free from pollution (having been cleaned,
perhaps).
Modelling partially conducting layers requires the use of an electrodynamic (or even
time-stepping) solver that can handle variations in both permittivity and electrical
conductivity. For this problem, a frequency domain (i.e. electrodynamic) formulation
was used.
The partially conducting layer of pollution was assigned an electrical resistivity
of 500 m. Of principal interest is the maximum stress suffered by the oil, occurring
at the interface between the oil and the paper insulation.
0
0
0.5
1
1.5
2
2.5
no pollution
full pollution
part pollution
3
3.5
r
e
l
a
t
i
v
e

s
t
r
e
s
s
0.2 0.4
relative distance
0.6 0.8 1
Figure 11.16 Variation of stress along the paper/oil interface
Figure 11.16 shows the stress variation along this surface, from bottom to top of
the bushing. The scale has been normalised so as to have a maximum stress value
of 1 for the as-designed case. The graph demonstrates that the effect of the overall
pollution layer is small, but that cleaning one shed near the bottom of the bushing can
increase the maximum stress in the oil by a factor of three.
The maximum stress will be dependent upon the electrical resistivity chosen for
the pollution layer. This is rarely known accurately and, in any case, will vary over
time with changes in atmospheric and environmental conditions. In these circum-
stances, it is appropriate to perform a sensitivity analysis, running through a range of
possible values of resistivity. With modern computing power, this is a trivial task and
gives increased confidence in the results obtained and the conclusions drawn.
Figure 11.17 shows the results of such an analysis. The maximum stress at the
oil/air interface (at about 0.25 of the device height, Figure 11.16) has been plotted
against the resistivity. For convenience, the stress has been normalised by division
by the maximum stress for the non-polluted, as-designed case. As can be seen, the
solution is relatively independent of the resistivity for values less than 5000 m.
Above this value, the maximum relative stress falls rapidly to a point where, beyond
100 000 m, it is having negligible effect on the stress.
11.5.5 Stress grading of high voltage windings
11.5.5.1 General
As machine end windings leave the stator core their surface can be subject to high
voltage gradients. The high electric stresses arising here can lead to sparking and
erosion. To alleviate the stresses, special stress grading materials are applied whose
resistivity varies with the applied stress. Since the tape is very thin in comparison with
other dimensions, a surface representation can be implemented. These materials have
10
1
0
0.5
1
1.5
2
2.5
m
a
x
i
m
u
m

r
e
l
a
t
i
v
e

s
t
r
e
s
s
3
3.5
10
2
10
3
resistivity, m
10
4
10
5
Figure 11.17 Variation of maximum stress with resistivity
been modelled both in the frequency and time domains as an aid to stress grading
system design.
11.5.5.2 Non-linearity
The stress grading materials have resistivity that varies with the applied electric
stress in a highly non-linear manner. Experimental measurements (Figure 11.18)
have shown that the resistivity varies very rapidly with the stress so that a simple
linear interpolation representation of the curve is not appropriate.
Examination of the experimental data has indicated an exponential variation of
the resistivity with the
2
3
power of the electric stress, Gully and Wheeler [10]:
s
= K exp{nE
2/3
} (11.32)
where
s
is the surface resistivity, ohm/ (ohm per square unit of material surface
area
1
), E is the surface electrical stress, kV/cm, and K and n are constants derived
from measurements. K is the resistivity constant in ohms and n is the exponential
constant. K may be thought of as the theoretical resistivity at zero stress and n
represents the degree of non-linearity the material possesses. A high conductivity
tape is denoted by a relatively low value of the resistivity constant, K, and a high
value of the exponential constant, n. As tape conductivity decreases, K increases
and n decreases
In the finite-element solution, an under-relaxation process is used to reach the
appropriate resistivity for the electric stress. At each iteration, the peak resultant
electric stress is calculated in each element. This value is used to derive a new
resistivity using Equation (11.32) and the provided parameters K and n.
0
0
1
2
3
4
5
6
r
e
s
i
s
t
i
v
i
t
y
,
m
50 100 150
electric stress, V/m 10
3
10
9
200 250 300
Figure 11.18 Variation of resistivity with electric stress in a stress grading
material (based on the paper Finite element modelling of non-linear
stress grading materials for machine end windings, by A.E. Baker,
A.M. Gully and J.C.G. Wheeler; as published on pp. 265268 of the
IEEConference Publication(no. 487) onthe International Conference
of Power Electronics, Machines and Drives in Bath, 1618th
April 2002)
The values of K and n must be measured for each stress grading tape under
investigation. In addition, the properties of the tape are known to vary with ageing
and temperature. At high temperatures, the n values of stress grading tape tend to
decrease and the tapes effectiveness reduces. An n value of zero means that the tape
has lost its non-linear properties and is only resistive.
11.5.5.3 Experimental comparison
An experimental test bar has been used as in Figure 11.19. The copper bar has a
rectangular cross section, and is wrapped with a 3 mm thick layer of insulation and
covered with an outer, 0.2 mm thick, layer of stress grading tape.
Avoltage was applied to the bar, varying sinusoidally at 50 Hz. The peak voltage
was measured at points along the tape surface. The external surface of the insulation
at the centre of the bar is earthed to simulate the earthing of the stator core in a
machine. The test measurements for a typical bar are compared with the finite-element
prediction for the bar with and without stress grading tape in Figure 11.20. Note how
the voltage gradient is much reduced by the presence of the tape. For this test, the
parameters for the tape were K = 2.8 10
10
and n = 4.3.
bare
insulation
tape
earthed layer
line of symmetry
copper bar
with insulation layer
and stress grading layer
HV
Figure 11.19 Experimental bar (based on the paper Finite element modelling
of non-linear stress grading materials for machine end windings,
by A.E. Baker, A.M. Gully and J.C.G. Wheeler; as published
on pp. 265268 of the IEE Conference Publication (no. 487) on the
International Conference of Power Electronics, Machines and Drives
in Bath, 1618th April 2002)
0
0
10
20
30
40
50
V
p
e
a
k
0.05 0.1 0.15
distance, m
end of tape
voltage along bar
no tape
FE
test
0.2 0.25 0.3
10
3
Figure 11.20 Comparison with test (based on the paper Finite element modelling
of non-linear stress grading materials for machine end windings,
by A.E. Baker, A.M. Gully and J.C.G. Wheeler; as published
on pp. 265268 of the IEE Conference Publication (no. 487) on the
International Conference of Power Electronics, Machines and Drives
in Bath, 1618th April 2002)
A frequency domain solution was used, where a single value of conductivity
applies through the time cycle (section 11.4.4).
For the case shown in Figure 11.19, there is also a layer of stress grading paint
applied after the partially conducting tape. The graph shows that for this particular
bar the taped portion was too short, as there is a sharp rise in the stress at the interface
of the tape and paint. The high impedance probe used to measure the voltage along
the bar does not give reliable results on the sample bar in the paint region because
it is not of sufficiently high impedance. This is a common problem, particularly as
higher voltages are applied to the bar to simulate overvoltage tests rather than normal
service conditions. The impedance of the probe dominates, resulting in inaccurate
measurement. The finite-element method clearly has no such difficulties and the stress
can be predicted along the whole length of the bar. The inaccuracy of measurements
is exacerbated further when non-sinusoidal voltage waveforms are applied to the bar.
11.5.5.4 Effect of parameters for stress grading materials
In selecting the best tape for an application, it is useful to look at the variation of
stress with the parameters K and n. Figure 11.21 shows the variation with voltage
along a bar for different values of K and with n fixed. Figure 11.22 shows similar
curves for n varying and with K fixed.
0
0
5
10
K = 1E10
K = 1E11
effect of parameter K (for n = 2.0)
K = 1E12
15
20
V
p
e
a
k
0.1 0.05 0.15
distance, m
10
3
Figure 11.21 Variation of voltage with K (based on the paper Finite element
modelling of non-linear stress grading materials for machine
end windings, by A.E. Baker, A.M. Gully and J.C.G. Wheeler;
as published on pp. 265268 of the IEE Conference Publication
(no. 487) on the International Conference of Power Electronics,
Machines and Drives in Bath, 1618th April 2002)
0
0
5
10
effect of parameter n (for K = 2.0E+11)
15
20
V
p
e
a
k
0.1 0.09 0.08 0.07
n=2.0
n=2.2
n=2.4
n=2.6
0.06 0.05 0.04 0.03 0.02 0.01
distance, m
10
3
Figure 11.22 Variation of voltage with n (based on the paper Finite element
modelling of non-linear stress grading materials for machine end
windings, by A.E. Baker, A.M. Gully and J.C.G. Wheeler; as
published on pp. 265268 of the IEEConference Publication (no. 487)
on the International Conference of Power Electronics, Machines and
Drives in Bath, 1618th April 2002)
In some cases, a particular stress grading system may give a good stress dis-
tribution but at the expense of increased heating in the tape. The finite-element
model can provide the current density in the tape by applying Equation (11.19) as
a post-processing function. The loss distribution can then be calculated using:
loss density =
J
2
W/m
3
(11.33)
The loss density values could be used as input to a thermal finite-element model to
calculate temperature rises. However, the loss density distribution alone gives a good
indication of hot spots on the bar and can be used to compare different stress grading
tape behaviour. Figure 11.23 shows the loss density for two tapes with different K
and n values. Note that the lower value of K denotes higher conductivity and hence
more loss.
11.5.5.5 Example of time-stepping solution
Using the time-stepping technique, a sinusoidal voltage was applied to a copper bar as
in Figure 11.19. Figure 11.24 shows the variation of voltage with time for a point in the
middle of the stress-grading layer. The applied voltage was a pure sine wave but the
voltage at the tape clearly contains other harmonic components. Figure 11.25 shows
0
0
10
20
25
l
o
s
s

d
e
n
s
i
t
y
,

W
/
m
3
5
15
30
10
6
0.05 0.1
distance, m
K = 2.8E10, n = 4.3
K = 3.1E11, n = 2.3
0.15 0.2
Figure 11.23 Effect of K on loss density (based on the paper Finite element
modelling of non-linear stress grading materials for machine
the time derivative of voltage at the same point. The shape of this curve compares
well with measured results such as those given in Wood and Hogg [11].
These curves illustrate the effects of the tape non-linearity on the distribution of
the voltage and its derivative (which is related to current). The frequency domain
model would have forced both of these curves to be pure sine waves like the applied
voltage. When the applied voltage is non-sinusoidal as in the PWM waveforms of
inverter fed machines, the need for a time domain solver is increased.
11.6 The choice of the order of the finite-element approximation
11.6.1 General
In the simplest form of the finite-element approximation to the electric field distri-
bution, the potential varies linearly over an element. This implies that the electric
stress is constant within the element volume. Such elements are known as linear, or
first-order, finite elements. It is perfectly possible, however, to assume a higher-order
variation (quadratic, cubic, etc.) of potential and electric stress
2
.
Higher-order elements are usually implemented by the introduction to the
element definition of additional nodes, located on the sides, faces or within the
5
5
0
5
10
v
o
l
t
a
g
e
,

V
10
3
10
3
10
10 15 20 25
time, s
30 35 40 45 50 0
Figure 11.24 Variation of tape voltage with time (based on the paper Finite
element modelling of non-linear stress grading materials for machine
5
5
4
3
2
1
0
1
2
d
V
/
d
t
3
4
0 10 15 20 25
time, s
30 35 40 45 50
10
3
10
6
Figure 11.25 Variation of voltage time derivative (based on the paper Finite
element modelling of non-linear stress grading materials for machine
element volume. With higher-order elements, the locations of the additional nodes
can be chosen to allowcurved element edges. These are called isoparametric elements
and permit the modelling of curved surfaces without introducing the abrupt changes
in shape that arise when the surface is modelled by a series of straight lines (flat
patches in three dimensions).
So, which should be chosen?
11.6.2 First-order elements
The main advantages of first-order elements are:
(i) Simplicity in mesh generation: nodes only occur at element vertices; con-
versely, elements are formed by a simple interconnection of nodes.
(ii) Simplicity in solution formulation: this is particularly true where non-linear
material properties are concerned (i.e. permittivities or electrical conductivities
that depend upon the value of the electric field). With first-order elements, the
electric field is constant in an element, so the permittivity and conductivity are
single-valued in each element.
(iii) Speed of construction of the finite-element equations: since the elements
material properties are constant within the element, simple, analytical expres-
sions can be written for the elements contribution to the set of finite-element
equations.
(iv) Simplicity in post-processing: displaying a single value of electric stress in
each element is a simple and rapid process. Plotting two- and three-dimensional
graphs of the stress variation in space is similarly a simple process of identifying
the elements involved.
And their disadvantages are:
(i) The boundaries between regions are formed from straight lines (two-
dimensional) or flat patches (three-dimensional). There are geometric disconti-
nuities distributed all across the model of a curved surface, each one introducing
an artificial stress concentration.
(ii) The stress varies in a step-wise fashion from one element to the next, i.e. the
stress field is not smooth.
11.6.3 Higher-order elements
The main advantages of higher elements are
(i) smooth stress variation from element to element
(ii) smooth stress variation at curved surfaces
(iii) smaller number of elements needed to model the field variation.
And the disadvantages are:
(i) more complex mesh construction and post-processing
(ii) slower construction of the finite-element equations
(iii) a numerical integration involving the variation of the material properties over
the element is required for each element
(iv) slower solution there are more nodes per element.
In simple theory, the higher-order elements will give a solution, to any given level
of accuracy, using fewer elements than a linear solution. The complexity of practi-
cal problems, however, leads to many nodes being required simply to describe the
geometry of the objects. The result is that a higher-order solution often takes longer
(mesh generation + solution + post-process) than the linear one.
For general field evaluation, the memory capacity and processor speed of modern
computers allows a good solution to be obtained using a large number of first-order
elements. Higher-order elements rarely give a significant improvement in solution
for the same total computing time. However, where the fields are changing rapidly
in space and need to be calculated to a high degree of accuracy and smoothness,
higher-order elements may be appropriate. It should be remembered, however, that
knowledge of the fields per se is often not the ultimate purpose of the investigation,
as discussed in the next section.
11.7 Assessment of electrical stress distribution
11.7.1 General
It is rare, in power engineering, that the magnitude of the voltage or the electrical
stress distribution is the ultimate aim of the analysis. What is usually required is an
assessment of whether or not there will be electrical breakdown of the insulation
system. Unfortunately, relating computed electrical stress distribution to breakdown
is far from simple, for two reasons:
(i) mathematical singularities in the numerical solution can result in very high,
localised stresses being calculated
(ii) there is no simple relationship between maximum stress and breakdown.
11.7.2 Mathematical singularities
Where there is an abrupt change in the orientation of a surface there will be a change
in the electrical stress, even if the surface is an equipotential one. As a general rule of
thumb, the finite-element mesh should be concentrated in the regions where the fields
are changing most rapidly. It would be natural, therefore, to place smaller elements
in the vicinity of a sharp corner.
It is often found, however, that as the element size gets smaller and smaller, so
the computed peak stress gets higher and higher although the stored electrical energy
for the problem shows no such trend. This reflects an attempt to capture a form of
mathematical singularity and the maximum value of the electrical stress appears to
be determined by the discretisation, rather than by the excitation.
One solution to this problem (if the value of the maximum stress shows no sign
of reaching a limit as the discretisation is increased in the vicinity) is to assess the
average stress in the local region. This average is formed using a volume weighting,
so reducing the effect of small, high stress regions. A more sophisticated average is
formed using the volume-weighted root mean squared stress. This is related to the
mean electrical stored energy in the region.
11.7.3 Relationship between stress and breakdown
The physics of electrical breakdown is complex, and most numerical models do not
consider such phenomena as corona discharge it would be an unwarranted compli-
cation in the face of other modelling approximations. Nevertheless, the designer must
make a judgement about whether a given stress is going to lead to insulation failure.
The first, and most useful, tool for this is experience. Given a restricted class
of electrical equipment, construction methods and insulation systems and given
experience of a certain number of failures(!), the engineer can build a base of expe-
rience relating calculated stress to observed breakdown (or lack of it). From this,
design rules-of-thumb may be derived.
Other, more technically-based, assessments may also be made. One such concept
is the stressed volume the volume of the insulation system that is experiencing
electrical stress above a nominated level. This volume has been related to breakdown
and is associated with the mean electrical stored energy, as discussed above.
11.8 Pre and post processor developments
11.8.1 General
The most sophisticated field solver is of limited use if it is difficult and time consuming
to enter data and to examine the results. Considerable effort has been expended in
improving the pre processing and the post processing stages and in integrating these
with the solver to permit automatic design optimisation.
11.8.2 Description of the problem geometry
Often, there will be a description of the problem geometry already in existence in
a CAD package. Most numerical analysis packages will interface directly with the
most popular CAD file formats, but the process is not totally straightforward.
The drawing is not only a description of the geometry: it also contains dimensions,
annotation and other parts not relevant to the analysis. These all need to be removed.
Adrawing does not necessarily describe a series of closed geometric shapes. Lines
that should meet exactly may, on the drawing, fall short of each other (and leave a gap)
or cross each other. Code is needed to cope with these cases.
A recent development is to make a parametric description of the object, i.e. in
terms of algebraic variables such as radius, width and angle rather than 32.1 mm
or 20
. The engineer can then study several designs by changing a few, characteristic
design variables rather than redrawing the whole object.
Once a parametric description of the problem is available, it is a short step to
varying the parameters in a systematic search for an optimal solution and this area is
a major, recent development.
For two-dimensional analysis, two-dimensional drawings are a good starting
point. This is less true for three-dimensional problems. For these, it may be most
convenient to describe objects in terms of basic, primitive shapes such as spheres,
cylinders and rectangular bricks and the intersections between them. Such solid
modellers are widely available for mechanical analysis, although not all will cope
with the problem of describing the air surrounding the solids.
Extrusion methods for three-dimensional mesh generation from a two-
dimensional starting mesh are still widely used. When coupled with sophisticated
mesh transformation and joining techniques, extrusion methods can handle most
mesh generation problems.
11.8.3 Creation of a discretisation from the problem geometry
Automatic discretisation from the geometry description is routine in two dimensions
and possibly in three. Following this automatic process, the engineer should be able
to change the discretisation by hand through some form of graphical tool. This is
useful for the cases where first analysis shows the discretisation to be inadequate in
some local region of the problem, or where a small design change is required (such
as the introduction of a small radius between two objects).
An extension of automatic mesh generation is to link it to an error evaluator that
makes an assessment of the solutions accuracy in relation to the discretisation used.
Areas showing a large solution error can then be rediscretised. By this means, the
solution can be performed to a specified degree of accuracy. It should be emphasised,
however, that although such a solution is accurate, it is not necessarily correct (the
material properties may be wrong, for instance).
11.8.4 Assigning material properties
An important point concerns the existence of accurate and appropriate material data.
Withmore powerful andsophisticatedanalysis techniques beingusedtocreate designs
stressed nearer and nearer to material limits, the accuracy of the overall solution
can become limited by the accuracy of the material properties used. As a first step,
a sensitivity analysis should be performed to find howthe solution might be influenced
by uncertainty in material properties.
11.8.5 Post processor developments
Post processing of a field solution is the means by which the engineer gets to the
object of the analysis. Normally, this involves inspecting the field solution in relation
to the problem geometry, so some form of visualisation is required.
Field plots in two- and three-dimensions are now routine. Three-dimensional
viewers can rotate the image easily, rapidly and at will to enable all facets of the
solution to be examined. Sections may be sliced from the three-dimensional volume
and examined either separately or in the context of the full three-dimensional model.
There is still a conceptual difficulty in viewing a field in a volume of space, rather
than on a surface, but this would still be the case even if a stereo view or a virtual
reality image were used (both are technically possible). One approach is to attempt
to draw equivalued surfaces as a series of semitransparent shells.
For two-dimensional slices, a whole range of analysis is on offer and easy export of
the results to other formats (particularly popular PC-based packages such as MatLab,
Mathcad and Microsoft Word and Excel) is required.
11.8.6 Design optimisation
The process of design optimisation starts with the analysis of an initial design. This
solution is post processed to see how the initial design performs in terms of the
target for improvement and to check whether any constraints on the performance or
dimensions have been violated. Typical values that might be used to characterise a
design are the capacitance between an electrode and earth or the maximum electrical
stress in a region or at a surface.
The results of this post processing operation are fed into one of many algorithms
to attempt to generate a better design and establish an iterative loop, leading to an
optimal design.
11.9 Notes
1
Surface resistivity is defined as the resistance between opposite edges of a square
area of the surface of the material. If the area of the square is increased, then the
change in the measured resistance due to the increased distance between the electrodes
will be cancelled exactly by the reduction in resistance due to the increased width
of the electrodes. Surface resistivity is, thus, independent of the area used for its
measurement [12]
2
The order of the stress variation is always one less than the order of the potential
variation, e.g. quadratic (second-order) potentials give linear (first-order) stresses
11.10 References
1 RIVENC, J.P., and LEBEY. T.: An overview of electrical properties for stress
grading optimization, IEEE Trans. Dielectr. Electr. Insul., 1999, 6, (3),
pp. 309318
2 ROBERTS, A.: Stress grading for high voltage motor and generator coils, IEEE
Electr. Insul. Mag., 1995, 11, (4), pp. 2631
3 SOUTHWELL, R.V.: Relaxation methods in theoretical physics(OUP, Oxford,
1946)
4 ZIENKIEWICZ, O.C., and TAYLOR, R.I.: The finite element method,
vols 1 & 2 (McGraw-Hill, Maidenhead, 1991, 4th edn.)
5 CHARI, M.V.K., and SALON, S.J.: Numerical methods in electromagnetism
(Academic Press, 1999)
6 FRANKLIN, P.: Methods of advanced calculus (McGraw-Hill Book Company
Inc., New York and London, 1944, 1st edn.)
7 REECE, A.B.J., and PRESTON, T.W.: Finite element methods in electrical
power engineering (Oxford University Press, 2000)
8 HANKE, K., HEISING, S., PROBERT, G., and SCRIVENS, R.: Comparison
of simulation codes for the beam dynamics of low-energy ions. Presented at
ICIS2001, Oakland, California, USA, 37 September 2001
9 AIELLO, G., ALFONZETTI, S., COCO, S., and SALERNO, N.: Placement of
the fictitious boundary in the charge iteration procedure for unbounded electrical
field problems, IEEE Trans. Magn., 1995, 31, (3), pp. 13921395
10 GULLY, A.M., and WHEELER, J.C.G.: The performance of aged stress
grading materials for use in electrical machines. Proceedings of IEE Confer-
ence on Dielectric materials, measurements and applications, No. 473, 2000,
pp. 392396
11 WOOD, J.W., and HOGG, W.K.: Endwinding stress in large generator stators.
International Seminar on High voltage, 1989
12 GOLDING E.W., and WIDDIS, F.C.: Electrical measurements and measuring
instruments (Pitman Press, Bath, 1968, 5th edn.)
Chapter 12
Optical measurements and monitoring in high
voltage environments
G.R. Jones, J.W. Spencer and J. Yan
12.1 Introduction
Optical techniques are attractive for measurements and monitoring under high voltage
conditions for several reasons. They enable measurements to be made remotely via
free space so providing a high degree of geometric isolation of the measuring equip-
ment from the high voltage environment. They also provide a means via optical fibre
transmission for penetrating into high voltage enclosures for monitoring purposes
with a high degree of inherent electrical insulation.
Phenomena associated with high voltage conditions such as electrical discharges
(corona, sparks, arcs etc.) are conveniently addressed since the latter produce both
optical emissions themselves, generate changes in the optical properties of the sur-
rounding media (e.g. pressure and temperature-induced refractive index changes etc.)
and often produce optically identifiable byproducts such as microparticulates. On the
other hand, high voltage equipment needs monitoring not only via the measurement
of electrical parameters such as voltage and current but also thermal (e.g. temperature
etc.), mechanical (e.g. vibration, linkage movement, catenary swings etc.) and chem-
ical (e.g. degradation of insulating media etc.) parameters, all of which in principle
can be optically addressed. The optical techniques are also immune to electromag-
netic interference (EMI) (which can be copiously produced by and via high voltage
systems) so leading to more reliable and accurate monitoring means.
By way of an example Figure 12.1 illustrates these two major facets of high volt-
age related optical measurements and monitoring. The free space aspect corresponds
to photographic recording of arcing flashover on the overhead high voltage power
lines of electrified railways. Such discharges are erratic (Figure 12.1a) and poten-
tially destructive in nature, requiring if they occur proper control and extinction.
Eurostar
HOCTs
PT
b
a
Figure 12.1
a Free space monitoring; photographic record of an arc flashover
on an overhead high voltage railway power line as an example of
free space monitoring
b Optical-fibre-based-monitoring; hybrid optical fibre current
transformer mounted on the high voltage end of a line side trans-
former for monitoring transient currents produced by discharges
of the forms shown in b
Their properties and behaviour need to be properly understood and free space optical
techniques are well suited for investigating these.
The use of optical fibre sensingis illustratedbya hybridoptical current transformer
(HOCT) which was mounted on the high voltage side of the line side transformer
Optical measurements and monitoring in high voltage environments 547
of the Channel Tunnel Rail Link shown on Figure 12.1b. This enabled transient
currents flowing during the occurrence of flashover of the type shown on the free
space photograph of Figure 12.1a to be monitored at locations where conventional
electric current measurement techniques could not be used because of high voltage
insulation limitations and electromagnetic interference.
Consideration is given in this chapter to some fundamental principles of optics
directly relevant to measurement and monitoring under high voltage conditions. There
is a description of various optical measurement techniques both in free space and via
optical fibre systems. Finally, some examples are given of typical experimental and
test results obtained with various optical systems.
12.2 Fundamental optical principles
12.2.1 Introduction
Traditionally, radiation from high voltage discharges has been investigated remotely
via free space transmission at various levels of sophistication ranging from simply
the spatial extent of the emitting plasma (using high speed photography), to detailed
spectral analysis (using conventional spectroscopy). The interaction of the discharge
plasma and its surroundings with probing light beams via scattering processes has
been used to provide localised property information such as gas density variations.
When such discharges are enclosed within high voltage equipment (e.g. circuit
breakers, gas insulated busbars) access for such optical measurements may be
gained with optical fibres. Such optical fibres may also be used in conjunction with
appropriate optical sensing elements for monitoring the condition of high voltage
equipment via the measurement of electrical, mechanical and thermal parameters of
the device itself.
A brief overview is therefore given of the fundamental principles which govern
free space and fibre transmission optics as well as relevant facets of spectroscopy,
light scattering and optical signal modulation.
12.2.2 Optical intensity
Two-dimensional optical intensity images provide quantitative information about the
location, size andshape of various electrical discharges. Suchinformationis important
for tracking the formation time variation and decay of such discharges in relation to the
operation of high voltage equipment e.g. insulator flashovers (Figure 12.1a), arcing in
circuit breakers etc. (Figure 12.2). Alternatively, the emission from a narrow section
across a discharge may be observed continuously in time as a streak photograph.
12.2.3 Spectroscopy
Spectroscopy involves resolving an optical signal into its different wavelength com-
ponents so that the wavelength-dependent structure becomes known. A knowledge
of such wavelength detailed structure of light from electrical discharges can provide
parasitic arcing optical fibre arc monitor
optical fibre contact travel recorder
fixed contact
PTFE nozzle
Figure 12.2 High speed framing photograph of electrical breakdown outside the
PTFE nozzle of a high voltage interrupter unit
information concerning the spatial and time variation in these discharges of the con-
centration and temperature of various ionic species and the density of electrons [13].
These are the fundamental properties which govern the nature and behaviour of the
discharge.
The use of such techniques is facilitated if the discharge plasma is in local
thermodynamic equilibrium(temperature and composition of species in equilibrium),
a condition which is not always satisfied in constricted discharges (e.g. lamps and wall
stabilised arcs etc.) or possibly in rapid time varying discharges. Spectral analysis
is also complicated by the presence of turbulence and discharge instabilities, which
may well exist in discharges occurring naturally and in high voltage equipment.
Consideration also needs to be made regarding the validity of other assumptions
[1] such as the path length over which the emitted radiation is transmitted before
being reabsorbed (optical depth). Other constraints are indicated in sections 12.2.3.1
to 12.2.3.3.
emitted
intensity
I( y
0
)
transmitted
intensity
y
y
y
I( y
0
)
y
0
I
0
I
0
x
Figure 12.3 Abel inversion of measured radial intensity variations across a cylin-
drical electrical discharge (emitted and transmitted intensities shown)
(courtesy of G.R. Jones, High pressure arcs in industrial devices: diag-
nostic and monitoring techniques, published by Cambridge University
Press, 1988)
Spectral measurements involve monitoring the intensity of radiation at a given
wavelength when viewed along a chord (y = y
0
) through a cylindrical electrical
discharge (Figure 12.3). This radiation may be either emitted by the discharge, as
in emission spectroscopy, or from an external beam passed through and absorbed
to some degree by the discharge, as in absorption spectroscopy. In both cases, the
change in intensity along the chord is governed by [1].
dI
dx
= e(x) k(x)I(x) (12.1)
where I is the intensity of the radiation propagating in the x direction, e(x) is the
spontaneous emission coefficient and k(x) the absorption coefficient, all parameters
being referred to the coordinate (x, y
0
). With emission spectroscopy, the emission
coefficient e(x) is to be determined, but with absorption spectroscopy the absorption
coefficient k(x) is to be measured.
Since the intensity measured along any chord y under conditions of cylindrical
symmetry (Figure 12.3) has contribution from discharge plasma elements at different
radii, it is necessary to deconvolve the intensity data in order to obtain true radial
profiles of emission or absorption coefficients. A number of deconvolution tech-
niques based upon a technique called Abel inversion are available [4]. These yield an
expression for the emission coefficient for an optically thin plasma at a radius r of
the form:
e(r) =
1
r
0
_
r
{d[I(y)]/dy}
(y
2
r
2
)
1/2
dy (12.2)
and a similar expression for k(r).
12.2.3.1 Temperature
The intensities of spectral lines are strong functions of temperature, some examples
of which are given for an SF
6
and copper discharge on Figure 12.4 [1].
Of the many methods for determining discharge temperatures from the intensity
of such spectral lines, the relative line intensity method offers several advantages. The
ratio of the intensities of two lines (e
Lmn1
, /e
Lmn2
) from the same species depends
upon temperature (T ) according to [1]:
e
Lmn1
e
Lmn2
=
A
mn1
g
m1
mn2
A
mn2
g
m2
mn1
exp
_
E
m1
E
m2
KT
_
(12.3)
where the suffices 1 and 2 refer to the two spectral lines, g
m
is the probability that
the atom will be in state m with energy E
m
, A
mn
is the transition probability of
spontaneous relaxation fromlevel mto n, is the wavelength of the emitted radiation,
k is Boltzmanns constant.
FII 4025.0
SII 5453.9
SII 54320.7
FI 6239.6
CuII 4555.9
CuII 2769
CuII 5051.8
SI 6410.5
CuI 5292.6
CuI 5782.1
10
8
10
6
10 20
temperature
l
i
g
h
t

i
n
t
e
n
s
i
t
y
30 10
3
Figure 12.4 Variation of selected S, F and Cu line intensities with temperature

for an SF
6
Cu plasma (courtesy of G.R. Jones, High pressure arcs in
industrial devices: diagnostic and monitoring techniques, published
by Cambridge University Press, 1988)
An alternative scheme for determining temperature is by using the intensity ratio
of ion to neutral species for a given element or the ion of one element and the neutral
of another element.
The presence of continuum radiation complicates the measurement of line inten-
sity and must be subtracted from the total measured intensity. The use of a narrow
spectral line is therefore to be preferred since the recorded continuum radiation is
a less significant fraction of the total measured intensity.
12.2.3.2 Electron density
Electron densities in the plasma columns of electrical discharges such as high pressure
arcs have been mainly determined frommeasurements of the broadening of particular
spectral lines [57]. However, there are other line broadening mechanisms which may
need to be taken into account under particular conditions
1 (see Note 12.7)
. If these can be
successfully taken into account then typically electron densities as lowas 210
22
m
3
can be measured [1].
12.2.3.3 Species concentration
The concentration of impurities in electrical discharges (e.g. entrained electrode mate-
rial) may be determined from a comparison of the intensities of the host and impurity
spectral lines according to [1]:
N
0g
(C
g
)
N
0m
(C
m
)
e
Lg
(c
g
)
e
Lm
(c
m
)
(12.4)
where C
g
, C
m
are the concentrations of the host and impurity, respectively, N
0g
,
N
0m
are the number densities of host and impurity, respectively, and e
Lg
, e
Lm
are the
line emission coefficients of the host and impurity, respectively. An example of the
relationship between the percentage copper vapour in an SF
6
arc discharge and the
intensity ratio of copper (5293 CuI, 5153 CuI) and sulphur lines (5321 SII) is shown
in Figure 12.5.
The method requires that the host and impurity species be in local thermal
equilibrium and that the temperature be accurately known.
12.2.4 Light scattering
Electrically charged particles are capable of scattering electromagnetic radiation as
a result of their interaction with the electric vector of the radiation. For instance,
a bound electron executing simple harmonic motion and driven by the oscillating
electric vector of an electromagnetic wave, reradiates the electromagnetic energy.
This constitutes the phenomenon of light scattering. Two major categories of such
scattering may be defined on the basis of the degree of coherence of the scattering
process.
12.2.4.1 Incoherent scattering
Incoherent scattering refers to the electromagnetic radiation that may be scattered in
a random direction relative to the incident radiation (e.g. by atmospheric particles).
65
55
45
5293 CuI
5321 SII
35
25
15
5
75
0.1 1 10
intensity ratio ( CuI : SII)
C
u

v
a
p
o
u
r
,

%
100
5153 CuI
5321 SII
Figure 12.5 Variation of CuI: SII line intensity ratios with percentage of copper
vapour impurity in SF
6
(T = 17000 k, P = 2 bar) (courtesy of
D.R. Airey, PhD thesis, University of Bath, 1977)
The intensity of the incoherently scattered light is considerably less than that of the
incident or coherent light scattered radiation (10
12
). Consequently, incoherent
light scattering experiments necessitate the use of high powered light sources such as
lasers. Incoherent light scattering forms the basis of powerful and versatile diagnostic
techniques. For instance, these techniques are capable of yielding highly localised
values of temperature and particle density without resorting to complex unfolding
procedures [4], which are necessary with emission and absorption spectroscopy.
Provided that a number of conditions are satisfied, the light scattered from
a discharge plasma [1] is composed of the sum of two components:
(i) One component is governed by scattering from free electrons (Thomson scat-
tering). When the light wavelength () is small compared with the Debye length
(
D
), the scattered radiation consists of a spectral line centred on the incident
radiation wavelength but broadened by the Doppler motion of the electrons.
When (/
D
) is large, the line shape changes into two narrow lines sepa-
rated by twice the electron plasma frequency symmetrically with respect to the
radiation frequency. The temperature and density values encountered in high
pressure electric arcs correspond to conditions between these extreme cases.
Also, for high density, low temperature arc plasmas the number of particles in a
Debye sphere may become small (<5), in which case the scattered spectrum is
modified [810].
(ii) The second discharge scattered component is governed by the bound elec-
trons of the ions or atoms (Rayleigh scattering), the intensity of the scattered
light decreasing as
4
. The classical Rayleigh scattering from neutral atoms
or molecules has been measured for a range of common gases by George
et al. [11].
When the incident radiation wavelength is of the order or greater than the par-
ticle size, the scattering becomes a weaker function of wavelength and is known as
Mie scattering [12]. Such scattering may be important from solid particles of size a
few microns produced within and around electrical discharges (Figure 12.6). Laser
Doppler techniques rely upon detecting light scattered from such microparticles to
determine their velocities.
The velocity with which a discharge plasma or surrounding gas flows may be
monitored by measuring the Doppler shift of the frequency (wavelength) of light
scattered by micron-sized particles moving with the flow. There are advantages in
using a differential Doppler system [13], in which the scattered light from two equal
intensity coherent beams is mixed in the test section to produce the Doppler signal
(Figure 12.7). In this case, the frequency shift is given by:
f
s
f
0
=
_
2v
0
_
sin
_
2
_
(12.5)
where
0
is the wavelength of the probing beams, v is the particle velocity and
is the angle between the two incident beams. Since the frequency shift is therefore
independent of the scattering angle ( in Figure 12.7), a large collection solid angle
may be used without the occurrence of aperture broadening.
In practice, small micron-sized particles are utilised in the flow to provide a scat-
tering signal of sufficient amplitude for detection. It is necessary to ensure that these
particles are in equilibrium with the local flow and adjust their motion sufficiently in
rapid flow accelerations or transients.
12.2.4.2 Coherent scattering
Coherent scattering is responsible for the transmission of light through a medium
being manifest in the refraction of light. The effect can be described in terms of the
polarisability of the medium, which is related to the refractive index according to [14]:
n
2
1 =
P
0
E
=
N
i
= 2 (12.6)
where P is the electric polarisation per unit volume, E is the electric field strength,
0
is the permittivity of free space, N
i
is the concentration of electric dipole constituent
i,
i
is its electrical polarisability and is the refractivity.
For a fluid of known composition, the mass density and hence N
i
varies with gas
pressure and temperature leading to corresponding changes in refractive index n. As
an example, the variation of refractivity (n 1) with temperature for air and copper
vapour is given on Figure 12.8. Consequently, the measurement of refractive index
provides a method for mapping changes of mass density in discharge heated and
pressurised fluids.
If electronic effects are important then the electron polarisability contribution,
e
,
may dominate in Equation (12.6) [14]. If in addition the wavelength of the light ()
is sufficiently short [15] and with small refractive index differences from unity the
nozzle exit
contact tip
contact tip
laser beam
luminous
particles
a
c
b
Figure 12.6 High speed video camera frames of particulates formation following
high fault current arcing in a circuit breaker
a N
2
30 kA30 ms after arc extinction (4 ms exposure)
b SF
6
30 kA22 ms after arc extinction
c SF
6
10 kA84 ms after arc extinction (luminous particles and laser
scattered light from non-luminous particles)
refractive index is given by:
n
= 1 4.46 +10
14
N
e
2
(12.7)
where N
e
is the electron concentration in cm
3
. Equations (12.6) and (12.7) indicate
that the electronic refractivity () is of opposite sign to the atomic and molecular
scattered beam ( f
s
)
probe beams ( f
0
)
[
[
v
f
s
f
0
= sin ([/2)
2v
z
0
Figure 12.7 Schematic of light scattering geometries for laser Doppler velocimetry
(courtesy of G.R. Jones, High pressure arcs in industrial devices:
diagnostic and monitoring techniques, published by Cambridge
University Press, 1988)
2.4
2.2
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
2.6
2.8
0 4 8 12 16 20 24
T (10
3
), K
p = 3 atm
z = 6941.21
air
copper gas
n
1

(
1
0
4
)
28 32
Figure 12.8 Refractivities of air and copper vapour as functions of temperature
(courtesy of F.A. Rodriguez, PhDthesis, University of Liverpool, 1974)
refractivities. For sufficiently long wavelengths (), the refractive index becomes
imaginary, implying that longer wavelength radiation is unable to penetrate the elec-
trical discharge. This provides a basis for remotely monitoring the radial extent of an
electrical discharge [16] in a confined environment.
n
0
y
I
1
I =2I
0
cos
2
[(n
0
n)lz]
I
(n
0
n)l/z =1
F
f f
K
P
test
volume
effective
source
knife
edge
f 0
knife
edge
a
b
c
Figure 12.9 Light scattering due to refractive index changes
a interferometry
b Schlieren
c shadowgraph (courtesy of G.R. Jones, High pressure arcs
in industrial devices: diagnostic and monitoring techniques,
published by Cambridge University Press, 1988)
A light ray passing through a refractive medium may be affected in one of three
ways. It may suffer a phase lag governed by a change in wavelength due to the
refractive index n; it may be deflected through an angle determined by a refractive
index gradient (dn/dy) (e.g. mirage formation); it may be displaced laterally through
a distance dz caused by the second derivative of the refractive index (dn
2
/dy
2
).
These three cases are the basis of three measurement techniques known respectively
as interferometry, Schlieren and shadowgraphy.
In interferometry, the phase difference produced by the refractive index n is
measured by recombining probing and reference beams (Figure 12.9a), yielding a
resultant intensity I [1].
I = 2I
0
cos
2
_
(n
0
n)l
_
(12.8)
where I
0
is the peak intensity, l the path length, and n
0
a reference refractive index.
Thus, a series of intensity fringes are formed which are separated by (n
0
n)l/.
The light source needs to be coherent to ensure a constant phase of the reference
beam over a distance as great as the optical path difference (a laser light source with
bandwidth = 0.01 has a coherence length of 50 cm).
With a Schlieren system, the displacement of light by a refractive index gradient
(dn/dy) may be detected as an intensity variation. By focusing a light beamon a sharp
edge, the angular deflection (Figure 12.9b) is given by [1]:
=
1
n
0
1
_
0
_
dn
dy
_
dz (12.9)
where n
0
is a reference refractive index, l is the optical path length. The brightness
pattern produced at a detector array therefore maps the refractive index gradients in
the test volume perpendicular (y) to the direction of the propagation (z). Optimum
sensitivity is obtained when D/f

= , where D is the beam diameter and f the focal
length of the focusing lens.
In shadowgraphy parallel rays passing through a refractive location where
(d
2
n/dy
2
) = 0 are displaced inhomogeneously whereby some rays converge and
others diverge. The medium, therefore, behaves as a lens increasing the intensity at
the focal point on axis while decreasing the intensity of the parallel beam elsewhere
to form a shadow (Figure 12.9c).
An example of a Schlieren photograph is given in Figure 12.10 which is a streak
photographof a shockwave passingthroughanelectrical arc (time vertical, radius hor-
izontal). The different light intensity profiles correspond to refractive index gradients
enabling the discharge and shock front to be identified.
12.2.5 Optical fibre propagation
The propagation of light through an optical fibre relies upon coherent scattering as
manifested in refractive index differences. On the other hand, incoherent scattering
can contribute to performance limitations, e.g. due to reduction in the optical power
guided. An optical fibre consists of a cylindrical core of a transparent medium of
refractive index n
1
surrounded by an outer annulus of a second transparent medium
of refractive index n
2
[17]. Provided the diameter of the core, d
1
, is greater than
the wavelength of light, light propagates through the fibre via a series of total
internal reflections (Figure 12.11). This has a number of implications regarding light
transmission:
(i) Only light rays falling on the end face of the optical fibre core within an angle
1
will be propagated by total internal refection where [18]:
NA = (n
2
1
n
2
2
)
1/2
= sin
1
(12.10)
(ii) Patterns formed by light rays launched at different angles form propagation
modes within the fibre core. The number of such modes (patterns) propagating
in a fibre is given by
N =
1
2
_
d
NA
_
(12.11)
where is the wavelength of the light. Typical multimode fibres, with core
diameters of 50200 m, propagate 1001000 modes. The significance of
multimode propagation is that different modes propagate at different velocities
so limiting the data transmission rate.
1
0
s
5mm
Figure 12.10 Streak Schlieren photography of shockwave propagating through a
cylindrical discharge (courtesy of G.R. Jones, High pressure arcs in
industrial devices: diagnostic and monitoring techniques, published
by Cambridge University Press, 1988)
(iii) Since the refractive index of materials in general varies with wavelength and
since the refractive index is inversely proportional to velocity, different wave-
lengths propagate with different velocities leading to another limitation to data
transfer rates i.e. due to pulse broadening [17]. For instance, a light emitting
diode (LED) source operating at 850 nm and with a line width of 40 nm will
give pulse spreading of some 4 ns/km.
Incoherent scattering of the Rayleigh type leads to attenuation in optical fibres
which varies as
4
so that shorter wavelengths (800 nm) are significantly sup-
pressed. At longer wavelengths (1.6 m), electron absorption effects dominate
leading to an attenuation which increases with wavelength. There is therefore an
optimum range for low attenuation propagation between these limits.
Optical fibre sensing differs from optical fibre telecommunications in that optical
elements, which are sensitive to external parameters, form component parts of the
incoherent
scattering
mode 2
mode 1
total internal
reflection
refractive
index air
d n
1
n
2
n
2
(z)
z
2
0
3
0
2
0
1
n
2
(z)
z
1
n
2
n
1
n
1
(z)
Figure 12.11 Propagation of polychromatic light in a step index optical fibre
showing:
total internal reflection
wavelength dependent refraction (
1
,
2
)
different propagation modes (1, 2)
incoherent scattering
system, whereas in telecommunications systems such external influences are delib-
erately excluded. In sensing applications, the optical signal is modulated to carry
information about the magnitude of a perturbing parameter. The fundamental light
wave properties which can be modulated to carry sensing signals are:
intensity
phase (via interferometer measurements, section 12.2.4.2)
polarisation (via rotation of the plane of polarisation)
wavelength (e.g. laser Doppler velocity measurement, section 12.2.4.1).
In addition, it is possible to encode the spectrum of polychromatic light carried by an
optical fibre (chromatic modulation [17]).
The general relationship between the output voltage (V) froma photodetector and
the optical modulation M
1
() of the optical fibre sensor is given by [17]:
V = q
__

(P()
__
1
T ()dl
_
R()M
1
()d
_
p
(12.12)
where P() is the source power, T () the fibre transmission loss, R() is the wave-
length dependent responsivity of the photodetector, is the optical wavelength. l is
the fibre length and p, q are numerical constants.
For interferometric sensors, the modulation factor M
1
() is
M
1
() cos
2
() (12.13)
where () is the phase difference between the interfering light waves of wavelength
. For a FabryPerot cavity of thickness (x/2) and refractive index n:
() =
(nx)
(12.14)
For chromatic sensors the spectral modulation M
1
() is sampled with two or
three photodectors having non-orthogonal responsivities R
n
(), n = 1, 2, 3. The
outputs from the three detectors are then processed to yield as the output optical
signal, a dominant wavelength (H), an effective bandwidth (S) and a nominal signal
strength (L) [19].
Shifts in dominant wavelength (H
d
) as small as 0.01 nm are detectable with
conventional photodiode detectors using such chromatic systems [19]. Changes in
measurand are then detectable in terms of variations in H, L and S leading to the
possible simultaneous measurement of several parameters.
12.3 Optical equipment and systems
12.3.1 High speed imaging
Highspeedimagingis concernedwiththe optical recordingof a transient or fluctuating
discharge in order to determine the spatial and temporal variation of the discharge
size, movement or light intensity. Both framing and streak images are obtainable
(Figure 12.2). The former involves recording a succession of two-dimensional images
each taken with a given exposure time, and the latter involves continuously streaking
a one-dimensional section of the image along the recording medium (Figure 12.10).
The acquisition of high speed images may be either with a conventional optome-
chanical or an optoelectronic camera, both requiring a radiation energy density of
approximately 10
8
J cm
2
for legible recording.
Optomechanical systems require the optical image to be swept along a recording
film by one of the following means [1]:
rotating prism or mirror (writing speeds of s per mm for streak photography)
movement of recording film in synchronisation with the rotating prism (6 10
3
frames per second, limited by film tearing)
rotation of the film on a drum in synchronisation with the rotating mirror
(3.3 10
4
frames per second).
Optoelectronic systems convert an optical image into electronic signals, which are
subsequently amplified and displayed photo-electrically. Additionally, amplification
may be achieved using an electronic image intensifier before finally displaying the
image. Such image converter cameras are capable of providing >10
6
f.p.s. with a few
nanosecond exposures but the recording duration is limited (12 frames).
Hybrid cameras are available whereby the primary recording is via conventional
high speed cameras with excellent spatial resolution. Thence the film images are
digitised optoelectronically. This provides the best compromise between length of
recording time and time resolution.
High speed cameras are used in conjunction with precision shuttering to prevent
overwriting on the film. Coarse shuttering is achieved with conventional electrome-
chanical shutters, and finer shuttering is achieved with electronic shutters. Such
shuttering can lead to problems of synchronisation with the event to be recorded
and image breakthrough which could lead to damage to the photo cathode of the
electronic cameras.
Substantial electronic amplification makes image converter cameras more sensi-
tive than mechanical cameras but only at the expense of modest resolution and no
spectral information (which with mechanical cameras may be retrieved with colour
film).
Alens system reduces the light flux density in the ratio [(1+P)
2
4F
2
]
1
, where
is the transmission, P the magnification and F the aperture number. The aperture
number F for mechanical cameras may be a function of the exposure time and limited
by the moving parts rather than by the objective lens. This may prove particularly
troublesome when the camera is used as part of a larger optical system such as those
used for shadowgraphy or interferometery (section 12.2.4.2).
Adetailed quantitative evaluation of photographic records requires a careful con-
sideration of the optical system, a knowledge of the recording characteristics of the
film and, in the case of image converter systems, a knowledge of the electrooptical
response of the camera. Operation needs to be on the sensitive part of the light inten-
sity: response characteristic of the film emulsion (Figure 12.12a) to avoid halation at
one extreme and lack of sensitivity at the other. Calibrated neutral density filters may
be used to ensure operation in the correct range. For image converter cameras, the
response characteristics of the photocathode (Figure 12.12b) are needed. The image
intensity is also affected by the gain of the camera tube, which may be controlled
electronically.
12.3.2 Spectrometer systems
Spectrometers for investigating electrical discharge spectra may produce spectral
dispersion via a diffraction grating or prism [3]. The wavelength resolution (/)
for a prism depends upon the prism base length, b, and the wavelength dependence
of the refractive index, n:
= b
_
dn
d
_
(12.15)
The wavelength resolution for a grating depends upon the number of lines (N
L
) on
the grating and on the order (m) of the principal maximum:
= mN
L
(12.16)
Gratings with large numbers of lines (10
4
cm
1
) produce higher resolutions than
prisms, but prisms avoid problems of overlapping of orders.
The profile of a spectral line obtained with a spectrometer is influenced by the
width of the spectrometer slit, the choice of which is governed by a compromise
between having a sufficiently narrow slit to yield a narrow spectral line width but
wide enough to provide sufficient light intensity for short exposure recording (s).
The sensitivity of a spectrometer varies with the optical wavelength.
0.1
0.01
0.001
1 10
1
10
2
relative expousure
range of
operation
o
p
t
i
c
a
l

t
r
a
n
s
m
i
s
s
i
o
n

f
a
c
t
o
r
10
3
1.0
32
28
24
20
16
8
12
4
0
q
u
a
n
t
u
m

e
f
f
i
c
i
e
n
c
y
,

%
a
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 1.0 0.9
wavelength, m b
Figure 12.12 Some parameters affecting photographic image intensities
a typical calibration curve for a polaroid film
b typical spectral response of the photocathode of an image con-
verter camera (courtesy of G.R. Jones, High pressure arcs
Atypical spectroscopic system used for arc discharge measurements is presented
on Figure 12.13a. This shows an optical system [3] utilising a grating spectrometer
with mirror optics for obtaining survey spectra along with a rapid scanning spectrome-
ter for short exposure spectra. Whereas survey spectra are recorded photographically,
the scanning spectra are detected with a photo multiplier and the output recorded
oscillographically. In the latter case, each spectral line is swept across the focal plane
rapid
scanning
spectrometer
grating
spectrometer
f
hsc
arc
rm
tm
g
pm
osc
rm = rotating mirror (150 Hz)
tm = trigger mirror
g = grating (580 lines)
pm = photomultiplier
osc = oscilloscope
f = recording film
hsc = high speed camera
argon ion laser dye laser
rotary
shutter
detectors
M
1
M
4
M
2
M
3
L
2
L
3
L
4
L
1
arc
monochromator
a
b
Figure 12.13 Spectrometric systems for arc studies [3, 39]
a framing and scanning systems (courtesy of D.R. Airey, PhD
thesis, University of Bath, 1977)
b absorption spectrometric system (courtesy of K. Ibuki, PhD
thesis, University of Liverpool, 1979)
of an exit slit by a rotating mirror. This slit samples each line in turn. The use of a half
silvered mirror allows the simultaneous recording of high speed photographs.
Streak spectra may be used to monitor continuous time changes in discharge
behaviour and properties [20].
Optoelectronic techniques are used for recording short exposure spectra. These
range from photomultiplier detection (Figure 12.13a) to vidicon image converter
cameras (whereby the spectrum from the exit of a polychromator is focused upon the
target of a vidicom camera tube). Such systems enable the electronic output to be
transferred directly to a microcomputer for displaying the spectra and for additional
processing (e.g. deconvolution along a line sight (Figure 12.3)).
Absorption spectroscopy involves the use of an auxiliary spectral source (which
may, for instance, be anargonionlaser tunedwitha dye) toprovide botha probingand
a reference beam(Figures 12.13b) [39]. Both beams are passed onto a monochromator
before being monitored photo electronically.
12.3.3 Light scattering systems
12.3.3.1 Coherent scattering
Refractive index responsive techniques require a test volume to be illuminated by light
which is then analysed. The illuminating radiation should not be strongly absorbed
by the electric discharge medium and its intensity should be sufficient to outshine
the light emitted by the discharge. The latter condition is easily fulfilled with a laser
light source, and the former condition usually restricts the wavelength domain to the
visible and near infra-red regions.
Details of optical systems which are particularly useful for investigating dis-
charges have beengiven bya number of authors [2124]. Some typical optical systems
which have proved useful for electric discharge investigations are shown in Fig-
ure 12.14. Such systems normally utilise a laser light source, light conditioning optics
and a detector. The laser source may be either pulsed (e.g. ruby) to provide extremely
short exposure (a few nanoseconds) or continuously operated (4 W argonion) to
allow long duration (10 ms) recording.
The conditioning optics are determined by the type of diagnostic (shadowgraphy,
Schlieren or interferometry). The differential interferometric and shadowgraphic sys-
tems shown on Figures 12.14a and c examine the whole test volume with a parallel
beamof laser light produced with two long focal length, concave mirrors. The second
of these mirrors also serves to focus the recording camera onto a plane in front of
the test volume. Narrowband filters matched to the laser wavelength serve to elim-
inate light emitted from the discharge. In the case of the differential interferometer,
two interfering light beams are produced by dividing equally the laser beam with a
parallel-sided glass plate (front and back surface reflections) close to the focal length
of the first concave mirror.
The distortion suffered by the laser beam in traversing the test volume may be
recorded using high speed framing or streak cameras (section 12.3.1). Care is required
to ensure that the internal shuttering and optics of the camera do not create any
additional distortion.
Analysis of the photographic records from interferometric experiments involves
measuring shifts in interference fringes at various locations followed by Abel
inversion (section 12.2.3) to determine the radial variation of the refractive index.
12.3.3.2 Incoherent scattering
Plasma light scattering
Since the intensity of incoherently scattered light is extremely small (10
12
times
smaller than the incident light intensity (section 12.2.4.1)), stringent precautions are
needed to eliminate stray scattering fromoptical windows, and microscopic particles.
A powerful laser is required to produce scattered light to outshine the discharge
camera
focus
laser
filter
camera
arc
test
section
light source
camera
focus
mirror
streak camera
slit
cut
off
framing camera
test head
lens
lens
camera
la
s
e
r
concave
mirror
concave mirror
camera focused
on this plane
narrowband filter
a
b
c
Figure 12.14 Systems for coherent light scattering measurements
a interferometer
b Schlieren
c shadowgraph (courtesy of G.R. Jones, High pressure arcs
light within the spectral domain of interest. This requirement suggests that the ion
scattered component I
i
may be easier to detect than the free electron component I
e
(section 12.2.4.1) on account of its narrower spectral width and despite its lower
intensity.
An optical system for measuring incoherent light scattered from electrical dis-
charges (Figure 12.15a) uses a rubylaser source whose beamis collimatedandfocused
onto a locality within the electrical discharge by two lenses, a pinhole and a number of
irises. The incident beamis dumped after passing through the discharge, the Brewster
angle filters avoiding back reflection. Light scattered through 90
to the incident beam

passes through a 10 optical filter before being detected via a FabryPerot etalon
and photomultiplier.
Micron particles scattering
The Doppler shift of light scattered from moving micro particles (section 12.2.4.1)
may be measured with the system shown on Figure 12.15b. This uses two beams
from an argonion laser 1.5 W, 4880 superposed within an electric discharge being
monitored. Adouble parallel plate compensator is used to reduce the beamseparation
to 0.3 mm. The scattered light is collected around a disk stop which absorbs the high
intensity main beam and determines the axial spatial resolution. Alarge aperture lens
behind the stop collects the maximum possible amount of forward scattered light.
The polarised (collimated) scattered light is detected with a photomultiplier together
with a 10 interference filter and associated polariser. AFabryPerot filter of 0.5
bandwidth further reduces spurious optical noise. Signal acquisition needs to respect
the short duration (0.2s) and small numbers of scattering events over a prolonged
observation period (e.g. 10 ms) [25].
12.3.4 Optical fibre sensing systems
The simplest formof an optical fibre sensing systemconsists of a light source coupled
to an optical fibre which addresses a measurand sensitive element, and which returns
a measurand modulated optical signal to a photodetector (Figure 12.16). If the sensing
element is distinct from the optical fibre the system is known as extrinsic; if the fibre
serves to both transmit and modulate the optical signal in response to a measurand the
system is intrinsic. Athird form of optical fibre sensing exists whereby the sensing
is non-optical but signal transmission is by optical fibre following optoelectronic or
optomechanical conversion. Such systems are known as hybrid sensing systems.
Although optical fibres offer many advantages, they need to be carefully deployed.
The advantages and difficulties may be summarised as follows:
(i) high sensitivity (but often respond to more than one parameter)
(ii) inherently electrically insulating (but are surface breakdown sensitive at extra
high voltages)
(iii) immune to EMI (but are susceptible to mechanical vibration)
(iv) geometrically flexible (but even gradual bending over of extended lengths
causes attenuation)
ruby
laser
pinhole iris arc
photomultiplier
aperture
constant temperature
FabryPerot etalon
10 filter
stop
oscilloscope
0 = /2
circular aperture
dump
dump iris
glass
filters
(Brewster
angle)
1.5 W argon
ion laser
adjustable
mirror
50% beam
splitter
thick plate
compensator
70 cm
arc chamber
50 mm f/0.98
collecting lens
50 mm f/1
collimating lens
10 bandwidth
filter (50 mm dia.)
S 20
photomultiplier
pinhole
dump
disk
stop
stop
20 cm 5 cm
5.5 cm 20 cm
7 mm
typical scattering volume dimensions: length
a
b
diameter
fringe separation
0.6mm
5mm
0.14mm
Figure 12.15 Systems for incoherent light scattering measurement
a scattering from a discharge plasma (courtesy of F.A. Rodriguez,
PhD thesis, University of Liverpool, 1974)
b scattering from moving micro particles (laser Doppler) (cour-
tesy of G.R. Jones, High pressure arcs in industrial devices:
processing
temperature
detection
V
n
oM
M(z)
T(z)
R(z)
P(z)
hv
equipment
optical sensor
(e.g. temperature)
lamp,
led
laser
optical
fibres
(10 30 m)
Figure 12.16 Structure of an optical fibre sensing system
(v) lightweight (but the need for heavier protective sheathing)
(vi) corrosion resistant (but susceptible to ionising radiation, microwave heating
and solvents).
12.3.4.1 Examples of optical fibre transducers for high voltage equipment
There are a range of optical fibre transducers which have been tested on high voltage
equipment. For example, transducers which have been deployed on high voltage
circuit breakers include transducers for current, voltage, arc radiation, solid particu-
lates, gas pressure, contact temperature and mechanical vibration. Some examples of
such transducers are discussed in the following section.
Electric current transducers
Figure 12.17 shows a range of different types of optical-fibre-based current
transducers. These are [26]:
an extrinsic magnetooptic element based upon magnetic field polarisation
modulation (f )
intrinsic sensor based on replacing the extrinsic magnetooptical element by
a polarisation preserving fibre wrapped around the current carrying busbar (e)
bulk extrinsic sensor representing an intermediate case between the purely
intrinsic and extrinsic magnetooptic sensor (d)
magnetic field concentrator based on a magnetic yoke around the current carrying
busbar concentrating the B-field into a magnetooptic element in a gap on the
concentrator (c)
current
transformer
conductor
electrical
to optical
conversion
burden
output fibre
conductor
sensing
optics
magnetic
core
output fibre input fibre
conductor
light path
sensing optics
conductor
sensing fibre
sensor
insulator
support
structure
control
house
high voltage
line
interface
unit
fibre cable
a
b c
d
f
e
conductor
sensing optics
input fibre output fibre
Figure 12.17 Various forms of optical fibre current transducer (IEEEworkinggroup
(1994)). a essential elements of OCTsystem; b conventional CTwith
optical readout; c magnetic concentrator with optical measurement;
d OCT using bulk optics (e.g. Jackson); e fibre optics based current
measurement (e.g. Rogers); f witness sensor
hybrid optical current transformer in which optical fibre transmission is used
to energise and address an electronic current transformer mounted on the high
voltage busbar (b).
The hybrid current transducer can draw its operating electrical power either by
electromagnetic induction from the current busbar or from a control room laser
via a fibre link. A paramount need is to minimise the power consumption of the
telecommunications module, which in turn dictates the form of telecommunications
encoding used, e.g. pulse frequency modulation, pulse code modulation etc. [27].
Hybrid current transducers are attractive because they rely on established telecom-
munications techniques, canbe adaptedtotransmit other measurand(e.g. temperature,
voltage) and could be coupled into existing telecommunication fibre systems on EHV
transmission networks.
The magnetooptic current transducers are based upon the Faraday rotation effect
[28], whereby the plane of polarisation of a light wave propagating through a mag-
netooptic element in a magnetic field of flux density B is rotated through an
angle:
=
_
V()Bdl (12.17)
where V() is the wavelength-dependent Verdet constant of the magnetooptic
element, and dl the optical path traversed in the element.
Since generically it is only possible to measure light intensity, a change in is
measured as the change in intensity of the light passing through a second polarising
filter so that only a resolved component of the polarised light wave is transmitted and
detected (Figure 12.18), i.e.:
P
0
= P
i
cos
2
(12.18)
Thus, the variation of with B is cyclical, so that the range of operation is governed by
the range within which (B) has linearity within the required specifications. Practical
Faraday
sensor
input
fibre
output
fibre
light
source
o
p
t
i
c
a
l
e
l
e
m
e
n
t
p
o
l
a
r
i
s
a
t
i
o
n
s
t
a
t
e
polariser sensing optics detector
D
0
[
analyser
L
Figure 12.18 Typical arrangement of optical components and polarisation compo-
nents in Faraday sensor
electrooptic
sensor
input
fibre
output
fibre
source
polarisation
state
polariser detector
D
0
analyser
L
n
x
n
y
Figure 12.19 Electrooptic sensing system
designs need to accommodate the temperature dependence of V() and the strong
dependence of on stress for high V() materials.
Various chromatic modulation forms of the concentrator OCThave been described
[28] which seek to overcome the practical limitations described above.
Electrooptic voltage transducers
Electrooptic voltage transducers sense changes in voltage via the effect of the associ-
ated electric field on the polarisation of light propagating through an electrooptic
element [17]. In such elements, light is propagated via two contrary circularly
polarised waves which when recombined constitute a linearly polarised wave
(Figure 12.19). The phase difference between the two circularly polarised waves
varies with the magnitude of the electric field, as therefore does the orientation of the
output linearly polarised wave. Thus, the output optical power varies with the phase
difference between the two circularly polarised waves, according to:
P
0
= P
i
sin
2
_
2
_
(12.19)
where
=
2
0
n
0
3T
ij
V (12.20)
0
= optical wavelength, n
0
= refractive index under zero E-field conditions, T
ij
is
the linear electrooptic coefficient of the sensing element (Pockels effect) and V the
voltage. Thus, the optical modulationis described bythe interference-type modulation
factor M
1
() from Equation (12.13).
Similar problems to magnetooptic elements exist with electrooptic elements in
having high sensitivities to temperature and to vibration. In addition, there are difficul-
ties in exposing miniature electrooptic elements to E-fields which are representative
of the voltage levels (145, 420 kV) found at distribution and transmission levels
without exposure to electrical breakdown across the electrooptic element itself.
Such electrooptic transducers may be deployed as extrinsic sensors across the low
voltage end of a high voltage capacitor divider or an intrinsic element of extended
length to provide good high voltage breakdown avoidance. Ahybrid form of voltage
transformer is also possible.
Electrode temperature
The temperature of a high voltage electrode (e.g. circuit breaker contact) can be
measured via a FabryPerot cavity interferometer [29] addressed polychromatically
(section 12.2.5) via optical fibre transmission. As such, the technique is illustrative
of how chromatic modulation defined by Equation (2.12) and interferometric sensing
defined by Equations (12.12) and (12.13) may be used combinatorially to provide an
effective sensing system for such a difficult operating environment.
A fibre addressed FabryPerot cavity consists of a silicon wafer into which
a cavity is etched and with a semireflective glass plate bonded to cover the cavity
(Figure 12.20a) [30]. Two multimode step-index fibres for delivering and receiv-
ing the polychromatic light are butted to the glass plate. The depth of the cavity
(0.31 m) is monitored via the optical interference between light reflected from
the glass plate surface and that reflected from the silicon surface on the other side of
the cavity leading to the phase difference defined by Equation (12.14). The other outer
surface of the silicon wafer is coated with aluminium to prevent infra-red radiation
from an electrical discharge being transmitted into the cavity via the silicon. The sili-
con wafer is made sufficiently thick so as not to flex under pressure and the glass plate
is arranged to have a thermal expansion coefficient which is very different from that
of the silicon. Temperature variations produce internal stresses which lead to changes
in the cavity depth and which in turn produce polychromatic optical interference. A
typical dominant wavelength:temperature calibration curve is given in Figure 12.20b
showing a low level of hysteresis between heating and cooling cycles. A resolution
of 0.1 per cent of full scale has been shown to be achievable.
Discharge radiation and particulates transducer
The chromatic modulation approach [19, 40] enables the optical emission from an
electrical discharge to be monitored along with the concentration of particulate mate-
rial formed during complex chemical reactions in the discharge plasma using a single
optical fibre probe. Particle concentration may be obtained by monitoring the change
in the spectral signature of polychromatic light [30] due to incoherent scattering.
Monitoring the output optical signal from the probe signal chromatically leads to the
arc signature being quantified by the measured dominant wavelength shift. An opti-
cal fibre probe for achieving such monitoring is shown on Figure 12.21a. It involves
transmitting a broad collimated polychromatic beamacross a monitoring gap between
two optical fibres. Light from the discharge plasma is collected by the receiving fibre
superimposed on a collimated beamfroma known optical source. The signature of the
emission is then obtained as a shift in the dominant wavelength of the superimposed
optical signals.
glass
Si
~1 mm
pressure
~ 5 mm
optical cavity
semireflector
20
810
815
820
825
830
d
o
m
i
n
a
n
t

w
a
v
e
l
e
n
g
t
h
,

n
m
835
840
845
850
855
30 40 50 60 70 80
temperature, C
90 100 110 120
heating cycle
cooting cycle
a
b
Figure 12.20 Chromatic FabryPerot temperature pressure transducer [29]
a sensor structure
b dominant wavelength:temperature calibration curve
Particle concentration monitoring may be achieved with the same optical fibre
sensor arrangement as shown in Figure 12.21a. The particle concentration is obtained
by monitoring the change in the spectral signature of polychromatic light produced
by scattering from micron-sized particles once the electrical discharge is quenched.
The wavelength of the forward scattered light is a function of the fractional volume
of micron-sized particles NR
2
P
/R
2
A
, where R
A
is the radius of the cylindrical vol-
ume being optically addressed, N, R
P
are the particle concentration and radius,
respectively.
Calibration of such a probe with known quantities of well dispersed micron-
sized particles yields the dominant wavelength:particle concentration curve of
Figure 12.21b which shows a high degree of linear correlation.
source channels
source light
modulated light
circuit
breaker
intensity
detector channels
libido
chromatic
system
intensity
500 1000
wavelength
z
D
500 1000
wavelength
z
D
particle modulation
source modulated
0
836
834
832
830
828
826
824
822
820
828
816
20 40 60 80 100
weight by volume concentration, mg/l
d
o
m
i
n
a
n
t

w
a
v
e
l
e
n
g
t
h
,

n
m
PC
400 m core
multimode
a
b
Figure 12.21 Plasma and particulate chromatic monitoring [30]
a sensor structure and location in circuit breaker
b dominant wavelength:microparticles concentration calibration
curve
Optical fibre transducers of mechanical parameters
It is also possible to monitor mechanical parameters on high voltage equipment
using optical fibre sensing. Examples of high voltage environments in which such
optomechanical transduction is important are mechanical operation of high voltage
switches (circuit breakers, tap changers etc.), gas pressure in gas-insulated equipment
(circuit breakers, transformers, GIS etc.) and vibration on overhead lines, busbars
switchgear and transformers.
Pressure measurements may be made using an optical fibre addressed FabryPerot
cavity (Figure 12.20a) with a pressure flexed membrane [30].
Mechanical vibrations may be measured with an optical fibre homodyne inter-
ferometer [18]. This is an intrinsic sensor (section 12.3.4) in which changes in the
interference between various propagation modes are produced by vibration-induced
variations in refractive index and manifest as different light intensity patterns at the
fibre output.
Mechanical movements of operating mechanisms etc., are measurable with a
fibre-addressed chromatic scale which moves past the fibre end [31] (Figure 12.2).
12.4 Examples of test results
The principles outlined in section 12.2 and the systems described in section 12.3
may be used to provide experimental and condition monitoring results relevant to
high voltage systems. Examples of such results are presented in this section and their
relevance to high voltage engineering indicated.
12.4.1 High speed photography
Examples of high speed photographs relating to high voltage electrical discharges are
given in Figures 12.2 and 12.6. Figure 12.2 is a photograph of an electrical breakdown
occurring outside the PTFE nozzle of a high voltage circuit breaker obtained with a
rotating mirror and drum type of framing camera (section 12.3.1). Figure 12.6 is a
photograph of luminous particles formed during arcing in an SF
6
HV circuit breaker
obtained with a high speed video camera.
Sequences of such images enable changes in the structure of electrical discharges
with time to be mapped. For example, Figure 12.22 shows how the luminous radii
of an arc discharge contracts during an 800 s period as the discharge current passes
through zero following a 41 kA peak current of a 100 Hz current waveform. After
the zero current, the arc column contraction is accompanied by an axial displace-
ment of the plasma due to an imposed flow of gas. The axial displacement reflects
the timescale of the axial convection losses, and the radial contraction indicates the
timescale associated with the combined influence of radial diffusion and flowentrain-
ment effects. Information fromsuch test results is important in understanding the role
of the arc discharge in switching fault currents in high voltage networks.
12.4.2 Spectroscopic results
The information that may be derived from spectroscopic measurements of electrical
discharges using spectrometer systems of the type discussed in section 12.3.2 includes
estimates of electron temperature, impurity concentration etc. (sections 12.2.3.1,
12.2.3.3).
Figure 12.23a shows a typical electron temperature map (derived from rela-
tive line intensity measurements) for a 10 kA, 10 cm long electric arc in air [20].
The temperature reduces from a maximum value of about 29 000 K on the axis in
anode
after
i =0s
+400
+300
+200
+100
0
300
400
before
i =0s
visible
boundary
200
100
0
cathode
r
v
, mm r
v
, mm
40 20 20 40
( 25s)
Figure 12.22 Variation with time of the luminous boundary of a decaying 41 kA
electric arc discharge measured from high speed framing pho-
tographs (courtesy of University of Liverpool, Arc Research Report,
ULAP-T28, I.R. Bothwell and B. Grycz, 1974)
front of the anode. Contours of copper concentration (entrained from the copper
cathode) are shown in Figure 12.23b. The constricted nature of the copper vapour
jet is apparent, particularly close to the cathode derived from the ratio of copper to
nitrogen line intensities. Temperature estimates made from the ratio of two copper
line intensities (Figure 12.23c) show that close to the cathode the copper vapour tem-
perature (1.7 10
5
K) is considerably lower than that of the host discharge plasma
(2.9 10
5
K).
Such test results provide an insight into the manner in which electrical discharges
in high voltage equipment can erode metallic members such as tank walls, switch
contacts, busbars etc.
A form of absorption spectroscopy (Figure 12.13b) is currently being used with
a free space laser beam for identifying SF
6
leakages from high voltage equipment.
The laser wavelength is tuned to one of the SF
6
absorption spectral lines and absorp-
tion maps produced which are recorded with electronic imaging. The method is
anode
cathode
anode
cathode
T, K
N
cu
+
(10
16
cm
3
)
0.2
0.5
1.0
2.0
3.0
4.0
7.0
T
m
Cu jet
T
m
NII/NI
8000
10000
12000
14000
16000
18000
20000
26000
29000
anode cathode
30000
25000
20000
15000
10000
Main arc
Jet
(a)
(b)
10 8 6 4 2 0
a b
c
K
Figure 12.23 Properties of 10 kA free burning air arcs derived from measured
spectra
a gas temperature map
b copper concentration map
c comparison of gas and copper jet temperatures (courtesy of
University of Liverpool, Arc Research Report, ULAP-T18,
D.E. Roberts, 1972)
an important contribution to the quest to reduce greenhouse gas emission into the
atmosphere, SF
6
being such a potent greenhouse effect gas.
12.4.3 Coherent scattering results
Two-dimensional images obtained with shadow, Schlieren and interferometric tech-
niques (section 12.3.3.1) can yield details of the size, structure and properties of
not only the electrical discharge but also its surroundings. The radial refractivity
variations determined from an interferometric image of a 9.7 kA electric arc are
shown in Figure 12.24. Various regions of the discharge are identifiable, namely the
2.6
2.4
2.2
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
(
n
1
)

(
1
0
4
)
0.4
0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
2 6 10 14 18 22 26 30
vapour
dominated
region
electron dominated
region
thermally dominated
region
9.7 kA, 1 ms
z
=
+0.2 cm (from cathode)
copper cathode
value for air
at ambient
conditions
Figure 12.24 Radial variation of refractivity for a 9.7 kA electric arc dis-
charge determined interferometrically (courtesy of G.R. Jones, High
pressure arcs in industrial devices: diagnostic and monitoring
techniques, published by Cambridge University Press, 1988)
electrical plasma region (with the negative refractivity due to electron domination),
a surrounding copper vapour region, a hot but only low ionisation region and the
ambient air.
The time evolution of radial temperature profiles surrounding an electrical dis-
charge may be determined frominterferometer refractivity profiles of the formshown
in Figure 12.24 to yield the results in Figure 12.25. These results are for a decaying
3 kA electric arc and illustrate the severe gradient of temperature at the boundary of
the discharge ( indicates the five per cent and ten per cent accuracy limits).
Although shadowgram results do not provide detailed profile information of the
kind available from interferometric measurement, they may be used to determine the
location of the outer thermal boundary (i.e. the locations of the steep temperature
gradients of Figure 12.25) with reasonable accuracy.
Test results of this formare useful for understanding gas compression and decom-
pression effects which can be produced by transient, high current discharges in high
voltage equipment enclosures and which can lead to catastrophic fracture of the
containers.
12.4.4 Incoherent scattering results
Incoherent light scattering (section 12.2.4.1) from a discharge plasma may be mea-
sured with systems of the form described in section 12.3.3.2, plasma light scattering,
5
4
3
T

(
1
0
3
)
,

K
2
1
6
7
8
0 5 10 15
1 1.50 ms
2 2.25 ms
3 3.75 ms
4 4.50 ms
5 5.25 ms
6 6.00 ms
7 6.75 ms
8 gas flow
r, mm
5
4
3
2
1
:
=
10%
:
=
5%
Figure 12.25 Change in the radial boundary temperatures of a decaying 3 kA elec-
tric arc determined from refractivity values yielded by interferograms
(courtesy of G.R. Jones, High pressure arcs in industrial devices:
using a high power laser source. Light intensity and the change in spectral width of
the laser line can yield values of the electron density, the electron temperature and
the ion temperature. As an example, Figure 12.26 shows the radial variation of the
concentration of various ion species in a 9.7 kAelectric arc in air at an axial locality
1 cm above the cathode and 1 ms after discharge initiation [32].
An example of results from incoherent light scattering from micron-sized partic-
ulates (section 12.2.4.1) for measuring the flow of plasma and gases around electrical
discharges (laser Doppler technique (section 12.3.3.2, micron particles scattering))
is given in Figure 12.27 [33]. The dashed curve shows the position of the discharge
heated gas boundary determined shadowgraphically (section 12.3.3.1). The signifi-
cance of these types of measurement is that they show how the electrical discharge
reacts fluid-dynamically with its surroundings. In this particular case, gas entrained
into the discharge heated region is accelerated to high velocities, while a flow wake
9.7 kA
1 ms
z
=
+1 cm
NI
OI
NII
OII
CuI
CuII
CuIII
e
10
17
10
16
N

,
c
m
3
10
15
10
14
0 2 4 6 8 10 12
r, mm
Figure 12.26 Radial profiles of species concentration derived from light scattering
experiments (courtesy of F.A. Rodriguez, PhD thesis, University of
Liverpool, 1974)
is formed in front of the discharge electrode (cathode). Typically, velocities in excess
of 2 10
3
ms
1
can occur within the plasma of a 3 kAelectric arc discharge.
Aknowledge of such fluid-dynamic behaviour is important in designing the geom-
etry and structure of high voltage circuit breakers so that their dielectric recovery
function is not impeded by a redistribution of plasma and heated gases after fault
current interruption.
12.4.5 Optical fibre transducer results
Because of the inherent electrical insulating nature of optical fibre sensors coupled
to their EMI immunity, the results of tests with such sensors yield online information
about high voltage equipment which was previously difficult if not impossible to
5
4
3
2
1
0
1
2
5 4 3 2 1 0
radius, mm
1 2 3 4 5
12.8
13.5
14.3
15.0
15.8
16.6
17.3
18.1
cold flow
velocity
distance
from cathode
tip, cm
v
,

m
s
1
thermal
boundaries
taken from
shadowgraphs
scale for
profiles
local
velocity
ratio
v
v
4
3.5
3
3.5
2
1.5
1(v =v
)
c
a
t
h
o
d
e
Figure 12.27 Flow velocity profiles under arcing condition; 30 A arc in accelerat-
ing flow (courtesy of University of Liverpool, Arc Research Report,
ULAP-T13, Collings et al., 1973)
obtain. Some examples of such results obtained with a range of different optical fibre
sensors are described.
Figure 12.28 shows the fault current variation through a high voltage circuit
breaker using a hybrid optical current transformer (HOCT) [27], (section 12.3.2).
Comparison of the results with those obtained with a resistive current shunt indicates
a maximum deviation of only 0.4 per cent.
Figure 12.29 [29] shows the time variation of temperature with the cali-
bration curve (Figure 12.20b) derived from the FabryPerot temperature probe
(section 12.3.4.1, electrode temperature) embedded in the contact stalk of a high
voltage circuit breaker, 13 cm from the contact tip. The results show temperature
variations following the interruption of a number of different currents in the range
1760 kA peak. Temperature changes of up to about 11
C occur. Two half cycles of

operation at 60 kApeak produces a prolonged transient which only returns to ambient
in about 0.5 s.
30
20
10
c
u
r
r
e
n
t
,

k
A
0
10
20
30
0 5 10
time, ms
15 20
HOCT
shunt
error 100
Figure 12.28 High voltage circuit breaker current measured with a hybrid optical
current transformer (HOCT) [27] (differences between HOCT and
shunt measured currents shown by triangles)
30
28
26
24
22
20
18
16
14
12
10
0 100 200 300 400 500 600 700 800 900 1000
37 kA T245
60 kA T250
60 kA T249
17 kA
t
e
m
p
e
r
a
t
u
r
e
,
C
time, s
Figure 12.29 The variation with time temperature within the contact of a high volt-
age circuit breaker for various peak arc current and measured with
a FabryPerot optical fibre probe [29] (60 kA, T250 is for two half
cycles of current)
Figure 12.30a [18] shows the time variation of dominant wavelength of optical
emissions froma number of fault arc currents in a high voltage circuit breaker ranging
from215 kApeak, using the probe of Figure 12.21a. The arc emissions are superim-
posed upon a continuous optical fibre signal from an external polychromatic source
830
825
820
815
810
805
800
795
790
10 5 0 5 10 15 20
time, ms
d
o
m
i
n
a
n
t

w
a
v
e
l
e
n
g
t
h
,

n
m
arc period
2
5
9
15 kA
1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
0 1 2 3 4 5 6 7 8 9 10
n
o
r
m
a
l
i
s
e
d

n
o
z
z
l
e
,

w
e
a
r
event number
832
830
828
826
824
822
820
0 50 100 150
d
o
m
i
n
a
n
t

w
a
v
e
l
e
n
g
t
h
,

n
m
p
a
r
t
i
c
l
e

c
o
n
c
e
n
t
r
a
t
t
i
o
n
,

m
g
/
l
time, minutes
nozzle (net)
nozzle (data)
event number
70
60
50
40
30
20
10
0
15 kA
30 kA
45 kA
56 kA
a
c
b
Figure 12.30 Results fromdischarge emission/microparticles probe (Figure 12.20).
a time variation of dominant wavelength (section 12.2.5);
b neutral network prediction of HV circuit breaker nozzle wear
from optical emission monitoring; c microparticles concentration
as a function of HVcircuit breaker fault currents and time determined
from dominant wavelength of scattered light
leading to a convenient referenced method for plasma intensity monitoring. One
implication of such optical emission results is that they may be cross correlated with
circuit breaker nozzle ablation indicating the extent of nozzle wear. Figure 12.30b [34]
shows the prediction of nozzle wear due to radiative effects from such measurements
compared with directly measured values of the mass reduction in the nozzle.
The shift in dominant wavelength pre and post arcing on Figure 12.30a is due
to Mie scattering (section 12.2.4.1) of polychromatic light from micron-size particu-
lates formed by the arc discharge. By tracking changes in the dominant wavelength
following a series of high current arcs (1556 kA) and using the calibration curve of
Figure 12.21b, the variation in particulates concentration with time and duty of a high
voltage circuit breaker can be determined (Figure 12.30c) [35].
Examples of optomechanical fibre sensor results are shown in Figures 12.31a
and b [36]. Figure 12.31a shows the time variation of pressure in the compression
20 10 0
6
7
8
9
10
11
12
13
10 20 30 40 50
movement begins
time, ms
time, s
arc
fixed contact clears nozzle
p
r
e
s
s
u
r
e
,

b
a
r
g
2
1
0
1
2
0 0.02 0.04 0.06 0.08 0.1
s
i
g
n
a
l

m
a
g
n
i
t
u
d
e
,

V
time, s
2
1
0
1
2
0 0.02 0.04 0.06 0.08 0.1
s
i
g
n
a
l

m
a
g
n
i
t
u
d
e
,

V
15 kA fault current
a
b
Figure 12.31 Results from optical fibre probes for measuring mechanical parame-
ters on high voltage circuit breaker [30, 38]
a time variation of piston pressure in a high voltage circuit breaker
under 15 kAfault current condition measured with a FabryPerot
pressure sensor (Figure 12.20a) (Issac, 1997)
b optoacoustic fibre monitoring of mechanical vibrations from an
SF
6
puffer circuit breaker under zero and 31.1 kA fault current
conditions (Cosgrave et al., 1997)
cylinder of a puffer circuit breaker during and after the separation of contacts and
arcing at 15 kAfault current; Figure 12.31b shows the difference in the optoacoustic
signals produced by a high voltage circuit breaker when operating with no fault current
and with a fault current of 31 kApeak. The results were obtained with the distributed
optical fibre vibration sensor of section 12.3.4.1.
12.4.6 Time and wavelength response of optical fibre and free space
techniques
Although optical-fibre-based techniques offer attractive advantages over the free
space alternative (e.g. access into enclosures, reduced interference from external
effects etc.), care is required in ensuring that measurements are not compromised
by properties of the fibres themselves. An example of such a fibre limitation is in
monitoring electrical breakdown between the contacts of a disconnector switch in an
SF
6
filled gas-insulated system (GIS).
Figure 12.32 shows a simplified geometry of such a GIS mounted disconnector
with a viewing port in the tank wall and the busbar lying along the tank axis. The
rise time of the optical emission from the breakdown discharges is extremely rapid
(lower part of Figure 12.32) being of the order of 56 ns. Reflections of electric field
perturbations from the tank ends occur on timescales of 7 and 35 ns, and optical
emissions radially from the tank wall occur in about 2 ns. These timescales are
comparable with response limiting dispersion effect in optical fibres (section 12.2.5)
which are in the range 3 ns (waveguide dispersion), 4 ns (chromatic dispersion)
and 7.5 ns (intermodal dispersion). Discharge emission rise time measurements can,
therefore, be prejudiced by the transmission dispersion effects of optical fibres.
With free space techniques, time response limitations are governed by the opto
electronic detectors. A photomultiplier can provide a rise time of 23 ns, and that
for a vacuum photodiode may be as short as 0.3 ns. Consequently, free space tech-
niques can, where useable, provide better transient responses than optical-fibre-based
systems when extremely rapid events are involved.
It should also be recognised that conventional silica optical fibres have trans-
mission properties which are severely wavelength dependent in the range 300 nm to
2 m (section 12.2.5). Consequently, such wavelength degrading effects need to be
evaluated before application to a particular discharge monitoring situation. Special
purpose optical fibres may therefore be needed (e.g. quartz cored fibres for ultra-violet
transmission) with additional cost penalties.
12.5 Conclusions
The range of measurement and monitoring possibilities of optically-based techniques
for use under high voltage conditions has been described. Utilisation and deployment
of such approaches in the future are likely to be driven not only by the need of newand
conventional high voltage applications but also by the substantial developments in the
optoelectronic sector itself. The evolution of high specification, lower costs digital
imaging systems, economic white quantumsources (without thermal power wastage)
a
a
c
b
d
b
r
e
f
l
e
c
t
e
d
o
p
t
i
c
a
l
p
u
l
s
e
d
e
l
a
y
2

n
s
7

n
s
d
7

n
s
3
5

n
s
intensity
system
effects
fibre effects
detector
effects
theoretical
worse cases
rise time
2
7
56
35
7
time, ns
intermodal dispersion (spread)
chromatic dispersion (spread, over 400800 nm range)
waveguide dispersion
vacuum photodiode
photomultiplier
statistical photon effects
7.5
4
3
23
0.3
b
Figure 12.32 Transient response of optical fibres and photodetectors for recording
electrical discharge formation in GIS
and short wavelength transmitting optical fibres (polymeric fibres etc.) are all set to
drive further the usage of optical monitoring for a considerable range of industrial
applications including those involving high voltage technology. Coupled with devel-
opments in intelligent information extraction systems [40], these approaches have
considerable potential for future deployment.
120
100
80
60
40
1325K
Planck formula
ribbon lamp
20
H
0
340
320
300
0
260
without particles: (140, 180, 220, 250 A, Z=30mm)
with particles: (particle injection rate =0.63 gs
1
240 A, Z=60 140mm)
240
220
air jet
no particles
current inc.
air jet
with particles
zinc
200
180
160
140
2500K
L1
L2
0.2 0.4
S
2300K
Figure 12.33 Distinction between plasma and particulates optical emission using
chromatic mapping [37]
As an example of the evolving possibilities, Figure 12.33 shows some test results
obtained from monitoring a high voltage plasma jet for producing microparticles for
surface spraying of superconducting coatings [37]. The results are in the form of a
dominant wavelength:effective bandwidth (H-S) map (section 12.2.5) of emissions
monitored with a tristimulus chromatic system. It shows howthe conditions of plasma
and micro particles in the jet can be separated in chromatic space without recourse
to data overload or expensive and cumbersome conventional spectroscopy and pro-
vide outputs easily assimilable into the operating control system (e.g. power, gas,
particulates flow etc.) of the plasma torch system as a whole via modern information
extraction means. This is but one of several examples which are emerging.
12.6 Acknowledgements
The authors appreciate the efforts of Miss M. Burns in the preparation of this chapter.
12.7 Note
1
Line broadening mechanisms include broadening due to finite lifetime of an
excited state ( 10
5
nm), strong magnetic fields interacting with electron spins
(Zeeman shift, 0.03 nm), strong electron fields influencing the radiating atoms
(Stark broadening), the random movement of the emitting atoms (Doppler broaden-
ing, 0.01 nm), pressure broadening ( 0.02 nm), ( values are typical
for a high pressure electric arc discharge).
12.8 References
1 JONES, G.R.: High pressure arcs in industrial devices (Cambridge University
Press, Cambridge, 1988)
2 LOCHTE-HOLTGREVEN, W.: Plasma diagnostics (North-Holland,
Amsterdam, 1968)
3 AIREY, D.R.: Energy balance and transport properties of very high current SF
6
arcs. PhD thesis, University of Bath, 1977
4 STANISLAVJEVIC, M., and KONJEVIC, N.: Fizika, 1972, 4, (13)
5 GRIEM, H.: Plasma spectroscopy (McGraw-Hill, 1964)
6 WIESE, W.L.: in HUDDLESTONE, R., and LEONARD, S.L. (Eds): Plasma
diagnostics techniques (Academic Press, New York, 1965)
7 TRAVING, G.: in LOCHTE-HOLTGREVEN, W. (Ed.): Plasma diagnostics
(North Holland, Amsterdam, 1968)
8 THEIMER, O.: Phys. Lett., 1966, 20, p. 639
9 PLATISA, M.: PhD thesis, University of Liverpool, 1970
10 KATO, M.: Fluctuation spectrumof plasma with fewelectrons in Debye volume
observed by collective light-scattering, Phys. Fluids, 1972, 15, p. 460
11 GEORGE, T.V., GOLDSTEIN, L., SLAMA, L., and YOKOYAMA, M.:
Molecular scattering of ruby laser light, Phys. Rev., 1965, 137, p. 369
12 KERKER, M.: The scattering of light. Academic Press, Library of Congress
Cat. Card 69-2664, 1969
13 MAZUNDER, M.K.: NASA, CR-2031, 1970
14 DITCHBURN, R.W.: Light (Blackie and Son, 1957)
15 DYSON, J., WILLIAMS, R.V., and YOUNG, K.M.: Interferometric
measurement of electron density in high current discharge, Plasma Phys.,
1964, 6, p. 105
16 DHAR, P.K., BARRAULT, M.R., and JONES, G.R.: A multi ring radio
frequency technique for measuring arc boundary variations both at high cur-
rents and close to current zero, University of Liverpool, Arc research report,
ULAP-T64, 1979
17 JONES, G.R., JONES, R.E., and JONES, R.: Multimode optical fibre sensors,
in GRATTAN and MEGGITT (Eds): Optical fibre sensor technology (Kluwer
Academic Publishers, 2000), chap. 1
18 JONES, G.R.: Optical fibre basedmonitoringof highvoltage power equipment,
in RYAN, H.M. (Ed.): High voltage engineering and testing (The Institution
of Electrical Engineers, London, 2001, 2nd edn.), chap. 21
19 JONES, G.R., and RUSSELL, P.C.: Chromatic modulation based metrology,
Pure Appl. Opt., 1993, 2, pp. 87110
20 ROBERTS, D.E.: Spectroscopic investigation of the 10 kA free burning arc
in air, University of Liverpool, Arc research report, ULAP-T8, 1972
21 KOGELSCHATZ, U., and SCHNEIDER, W.R.: Quantitative Schlieren tech-
niques applied to high current arc investigations, Appl. Opt., 1972, 11,
pp. 18221832
22 TIEMANN, W.: IEEE Trans. Plasma Sci., 1980, PS-8, pp. 33683375
23 BLACKBURN, T.R., and JONES, G.R.: Measurement of density and
temperature in an ORIFICE arc, J. Phys. D, Appl. Phys., 1977, 10,
pp. 21892200
24 WALMSLEY, H., JONES, G.R., and BARRAULT, M.R.: Properties of the
thermal region of a gas blast a.c. arc particularly near current zero, University
of Liverpool, Arc research report, ULAP-T56, 1978
25 TODOROVIC, P.S., and JONES, G.R.: IEEE Trans. Plasma Sci., 1985, PS-13,
pp. 153162
26 IEEEWorking Group: Optical current transducers for power systems: a review,
IEEE Trans. Power Deliv., 1994, (4), pp. 17781788
27 PATE, A., HUMPHRIES, J.E., GIBSON, J.R., and JONES, G.R.: The mea-
surement of fault arc currents using a hydrid opto electronic current transformer
(HOCT). Proceedings of XIV international conference on Gas discharges and
their applications, Liverpool, 2002, pp. 180183
28 JONES, G.R., LI, G., SPENCER, J.W., ASPEY, R.A., and KONG, M.G.:
Faraday current sensing employing chromatic modulation, Opt. Commun.,
1998, 145, pp. 203212
29 MESSENT, D.N., SINGH, P.T., and HUMPHRIES, J.E., et al.: Optical fibre
measurement of contact stalk temperature in a SF
6
circuit-breaker following fault
current arcing. Proceedings of 12th international conference on Gas discharges
and their applications, Greiffswald, 1998, 1, pp. 543546
30 ISAAC, L.T.: Puffer circuit breaker diagnostics using novel optical fibre
sensors. PhD thesis, University of Liverpool, 1997
31 ISAAC, L.T., SPENCER, J.W., JONES, G.R., HALL, W.B., and TAYLOR, B.:
Live monitoring of contact travel on EHVcircuit breakers using a novel optical
fibre technique. Proceedings of 11th international conference on Gas discharges
and their applications, Tokyo, 1995, 1, pp. 238241
32 RODRIGUEZ, F.A.: PhD thesis, University of Liverpool, 1974
33 COLLINGS, N., BLACKBURN, M.R., JONES, G.R., and BARRAULT, T.R.:
Laser Doppler velocimetry of flow fields surrounding low and high current
arcs, University of Liverpool, Arc research report, ULAP-T13, 1973
34 COSGRAVE, J.: Private communication, 1998
35 ISAAC, L.T., JONES, G.R., HUMPHRIES, J.E., SPENCER, J.W., and HALL,
W.B.: Monitoring particle concentrations produced by arcing in SF
6
circuit-
breaker using chromatic modulation probe, IEE Proc., Sci. Meas. Technol.,
1999, 146, (4), pp. 199204
36 RUSSELL, P.C., COSGROVE, J., TOMPTIS, D., VOURDAS, A.,
STERGIOULAS, L., and JONES, G.R.: Extraction of information from
acoustic vibration signals using Gabor transform type devices, Meas. Sci.
Technol., 1998, pp. 12821290
37 RUSSELL, P.C., DJAKOV, B.E., ENIKOV, R., OLIVER, D.H., WEN, Y., and
JONES, G.R.: Monitoringplasma jets containingmicroparticles withchromatic
techniques, Sens. Rev., 2003, 23, (1), pp. 6065 (Emerald Group Publishing
Limited)
38 COSGRAVE, J., HUMPHRIES, J.E., and SPENCER, J.W., et al.: Chro-
matic characterisation of optoacoustic signals from fault current arcs in high
voltage circuit-breakers. Proceedings of XIIth international conference on
Gas discharges and their applications, Greifswald, 1997
39 IBUKI, K.: Spectroscopic study of the high current vacuum arc. PhD thesis,
University of Liverpool, 1979
40 JONES, G.R., RUSSELL, P.C., VOURDAS, A., COSGROVE, J.,
STERGIOULAS, L., and HABER, R.: The gabor transform basis of chromatic
monitoring, Meas. Sci. Technol., 2000, II, pp. 489498
Chapter 13
Pulsed power principles and applications
J.E. Dolan
13.1 Introduction
Pulsed power deals with the generation of extremely high power, short duration
impulses. Peak powers typically range frommegawatt (MW) to terawatt (TW) levels,
and pulse durations from nanoseconds to milliseconds. The aim of this review is to
indicate the development of pulsed power as a discipline, and to discuss some of the
key elements in pulsed power systems. The range of applications of pulsed power is
also outlined.
Pulsed power began in the 1920s when Erwin Marx at the Technical University
of Braunschweig devised a novel form of high voltage impulse generator for light-
ning testing of high voltage power transmission equipment. In the Marx generator
(Figure 13.1) a set of n capacitors are charged in parallel at moderate voltages (typi-
cally 10100 kV). The capacitors are then switched into series connection to achieve
an output impulse voltage which is a multiple n of the charging voltage. The series
connection arises due to the sequential over-volting and breakdown of the spark gaps
once the first gap breaks down. Output voltages up to 2 MVare readily achieved. The
spark gap closure maintains the capacitors in series for the duration of the ensuing
current discharge. In more recent years, the development of high energy and fast
rise-time Marx banks has been driven by the requirements of flash radiography, ion
beam generation, and plasma fusion drivers for MJ energy levels to be delivered in
sub-microsecond timescales.
The first pulsed power systems were probably the power modulators used to
drive radar magnetrons at 100 kW level in the Second World War era. The mag-
netron requires a repetitive pulse driver developing typically 1050 kV at 110 A,
and running at 110 kHz repetition rates. The necessary pulse width is of the order of
microseconds. The problem lay in the peak pulse power and the fact that this was at
a high repetition rate. Spark gaps could not conduct the required current levels, and
still recover in the available timescale.
S
4
S
3
+
+
+
+
S
2
charging resistors
O/P
S
1
+V
Figure 13.1 Four-stage Marx generator with positive polarity output
The solution was found in magnetic pulse compression [1]. A series of saturable
magnetic stages are driven by the primary closing switch and energy is transferred
between stages in progressively shorter timescales. The primary switch both benefits
from turn-on snubber action and does not handle the peak magnetron power and
current. Although rotary spark gaps were often used to drive the magnetic pulse
compressors, these would only handle a moderate current and so the recovery time
would be adequate. Magnetic pulse compression fell into disuse post war as the power
handling capabilities of various tubes and the hydrogen thyratron in particular were
raised enormously by development. However, the technique has enjoyed a steady
renaissance fromthe early 1980s through to the present day due to novel requirements
in high power lasers, ultrawideband radars, corona reactors etc.
Since the 1940s, pulsed power has gone through a series of distinct phases
as applications have developed. The modern day pulsed power collective is very
much centred on the US, and this is reflected in the preeminence of the biannual
IEEE International Pulsed Power conference. The shape of pulsed power has been
moulded largely by the requirements of nuclear weapons, inertial confinement fusion
(ICF) and directed energy weapons (DEW) programmes. All of these have bene-
fited from extensive support in the United States and former Soviet Union and, to
a lesser extent, in Japan and Western Europe. Nonetheless, industrial applications
have always been an important aspect of pulsed power. By the 1970s, moderately
high power pulse techniques were becoming well established for a range of techni-
cal/industrial applications flash photography, metal forming, pulsed lasers, electric
Pulsed power principles and applications 593
fences, pulsed electrostatic precipitators. This is covered in detail in the series of
books by Frngel [2]. In something of a paradigm shift, the spread of applications
and pulse techniques that Frngel describes is in present day terminology largely sub-
sumed under industrial electronics. Meanwhile pulsed power has shifted strongly
towards the requirements of high energy physics.
The first high energy pulsed power area to develop was that of flash radiography.
Much pioneering work was carried out in the UK at the Atomic Weapons Research
Establishment, largely due to the individual genius of J.C. Charlie Martin. Flash
radiography is the production of intense submicrosecond duration x-ray beams of
extremely short duration typically of the order of 10100 nanoseconds. The par-
ticular application was and remains today the x-ray imaging of nuclear warhead test
detonations from outside the metal casing. The pulsed power requirement is for sub-
microsecond pulses of 110 MV and 1 MAlevels to drive the electron beam which
produces the x-rays by the Bremsstrahlung mechanism. Rise times of 50100 ns or
less are also required. This led to the development both of fast rise time high energy
Marx banks in oil tanks, and of transmission line techniques for shaping of pulses
and channelling the energy to the e-beam load. J.C. Martin established many of the
ground rules which distinguish pulsed power from general high voltage technology.
In particular, the realisation that dielectrics can be operated considerably above their
DC breakdown strength for very short impulse durations. This led to the much used
large area breakdown formulae for water and oil as transmission line dielectric
liquids. The Martin equation for coaxial geometry is:
V
b
= kA
1/10
dt
1/3
(MV) (13.1)
where V
b
is the applied voltage in MV, k is a value in the range 0.30.6 depending
on whether oil or water is used and on the polarity, d is the radial separation of the
plates in cm and t is the time to breakdown in s.
Martin also established formulae for spark gap behaviour which are still in use
today, and developed many other pulsed power techniques. Ahighly readable account
of this work may be found in Charlie Martin on pulsed power [3]. As a consequence
of the Aldermastons capabilities, very strong links developed between the Aldermas-
ton group and US groups including Los Alamos, Sandia National Laboratories and
the Naval Research Laboratories.
The second major area to develop related to inertial confinement fusion (ICF)
programmes based on laser and on particle beams, respectively. The pulsed power
requirement is to energise either intense krypton fluoride (KrF) lasers or to produce
particle beams, which in turn compress and heat the fusion pellet target. Useful
outlines may be found for the Titania KrF scheme [4], the Sandia National Labora-
tories PBFA-II (Particle Beam Fusion Accelerator-II) scheme [5], and the ongoing
Megajoule project in France [6].
Directed energy weapons (DEW) programmes have run continuously since the
Star Wars Strategic Defense Initiative (SDI) under President Reagan in the mid
1980s. The term directed energy weapons refers to a whole range of weapons with
a variety of functions. High power lasers are envisaged as a means of disabling or
directly destroying items such as ballistic missiles or sensor components, in particular
by interception in space or near space. High power microwave devices and other radio
frequency sources would principally be intended to disrupt enemy communications,
computing or control systems.
The decline in funding for nuclear weapons programmes since the end of the Cold
War has led to a certain degree of realignment of pulsed power activities towards
commercially viable industrial applications. The principal areas of industrial pulsed
power application today can largely be grouped under the following headings: envi-
ronmental cleanup, materials treatment, biotreatments, food treatment, rock breaking
and particle generation. The general rationale for the application of high power pulsed
techniques rather than forms of continuous DCor ACexcitation is that it enables oper-
ating regions to be accessed which are not accessible in the steady state. For example,
higher values of pulsed fields may be applied to a corona-generating assembly with-
out electrical breakdown occurring than for steady state fields. If electric fields are
to be applied in water, the conductivity of the water means that only pulsed fields
may realistically be applied at the kV/cm level and above. Some systems may have
a non-linear response which can only be driven effectively in a pulsed mode.
Environmental cleanup is applied principally in the areas of flue and exhaust gas
cleanup, and of destruction of volatile organic compounds (VOCs). The chemistry
can be broadly understood as the stimulation of reactions by means of ionised parti-
cles or free radicals. The injection of ionised particles can be by means of an electrical
discharge or by ion beam. Controlled electrical discharges can be achieved effectively
by means of pulsed corona, which results in the generation of microstreamers. Various
other plasma techniques may also be used. Principal among these is the use of cold
(non-thermal) plasmas established using the dielectric barrier (DBD) technique. The
bigissue is not somuchwhether gas cleanupcanbe achieved, but whether schemes can
be devised which give cost effective cleanup in terms of the joules per cubic centimetre
of treated substance (J/cc). Which approaches are likely to be cost effective remains a
veryopenquestion. For example, Penetrante et al. [7] present experimental data show-
ing that for a given rate of NO
x
removal, ion beam injection requires only 20 per cent
of the input energyrequiredbypulsedcorona. However, the capital cost of anionbeam
source compared with a pulsed corona source is likely to be high. Ageneral problem
with commercial applications of pulsed power is that solutions based upon submi-
crosecond pulses inherently tend to require relatively complex, costly, systems, as the
simplest spark-gap-based approaches are not capable of long time operation between
parts replacement. This is typified by the abandonment by the 1950s of static and rotat-
ing spark gaps in favour of more stable and controllable gas discharge tubes for radar
modulators. Perhaps some of the most promising future applications areas for pulsed
power are in fact in very high technology areas with large payback, such as extreme
ultra-violet/soft x-ray lithography for the next generation of digital semiconductors.
13.2 Pulsers and topologies
Any formal definition of pulsed power will include the concepts of pulse compression
and pulse forming. Pulse compression involves the discharge of energy froma storage
charging resistor
capacitor
charging
supply
switch load
Figure 13.2 Capacitor discharge circuit
device at a faster rate than it has been charged. This is exemplified by capacitor
discharge systems, although there are various other forms. Pulse forming is the use of
circuit, switching and transmission line techniques to achieve the pulse shape required
bythe load. Adiscussionof generic pulse compressionandshapingtechniques is given
by Pai and Qi Zhang [8]. Agood up-to-date review of circuit topologies and analysis
techniques is given by Smith [9].
13.2.1 Capacitive discharge
Capacitor discharge circuits are certainly the most widely used pulsed power topology.
Figure 13.2 indicates a generalised discharge circuit. A suitable charging system is
used to charge a capacitor to the required voltage and energy level. The charging
process may take minutes for a set of F capacitors being charged to 50 or 100 kV in
a high energy, single-shot system. On the other hand, it can be as short as milliseconds
for a 1 nF capacitor in a high repetition rate system. When the capacitor is charged to
the required level, a closing switch connects the capacitor to the load. The resulting
discharge has a pulse width defined by the load time constant = CR
load
. In pulsed
power applications, the discharge time is typically of the order nss. The pulse
rise time into the load may be of more significance than the pulse width. The rise
time is likely to be affected principally by the closing switch characteristics and by
the inductance of the load circuit. One means of obtaining a more rapid voltage
rise onto the load is by use of a sharpening gap, as indicated in Figure 13.3. The
circuit has a second capacitor intermediate between the source capacitor and the
load. This peaking capacitor is of much lower value than the source capacitor.
When the primary switch is closed, this intermediate capacitor charges at a rate
determined by the closing switch circuit. The sharpening gap is designed to self-break
as the intermediate capacitor reaches a suitably high voltage, and thereby connects
the intermediate capacitor across the load. The load voltage rise time may then be
very rapid since the inductance of the peaking capacitor load loop can be made very
low. Elements of the capacitive circuit topology are briefly outlined below.
13.2.2 Charging supplies
Resistive charging has the benefit of simplicity and low cost. The DC source could
typically be any of the following: mains transformer with rectifier, CockcroftWalton
multiplier, switched-mode high voltage DC supply. However, an inherent disadvan-
tage is that when a capacitor is charged from a fixed voltage source, as indicated in
charging resistor sharpening switch
charging
supply
discharge
capacitor
peaking
capacitor
switch load
Figure 13.3 Capacitor discharge circuit with peaking capacitor and sharpening
switch
E C
S R
Figure 13.4 Resistive charging circuit for capacitor
Figure 13.4, an energy equal to that stored in the capacitor is dissipated in the charg-
ing resistance. For this reason, the use of switched-mode power electronic charging
units, generally termed capacitor chargers, is becoming increasingly common. This is
particularly the case in rep-rate applications. Charger units range from the miniature
systems found in camera flash systems to multi-kW systems with outputs of up to
50100 kV. The benefit of specialised capacitor charging systems is that the instan-
taneous output voltage during charging is generally controlled to maintain a constant
charging current into the capacitor, giving very high efficiency. Companies producing
capacitor chargers include General Atomics, Lambda EMI and ALE.
Although mains supplies will generally be used, battery storage is often employed
for mobile systems, for systems with extremely high current demands, or where
EMI/EMC concerns can otherwise cause problems. Low voltage ultracapacitors
(typically rated at several farads and a few volts) may provide an interface between
batteries and faster pulsed circuits, as has been done in some electric vehicle appli-
cations. However, these techniques are in the realm of power systems operation
and essentially fall outside the pulsed power area. Railgun and coil gun systems
will require MWGW of prime power for several milliseconds pulse duration, and
this generally implies the use of specialised pulsed alternators such as compulsators
[10, 11].
In passing, Electrical interference and protection by E. Thornton [12] should be
mentioned as a near unique, invaluable practical guide to the design of pulsed circuits
in terms of the necessary shielding, filtering, screening and grounding.
13.2.3 Capacitors
In order of pulse discharge speed, the main types of capacitor employed in pulsed
power workare electrolytic, woundpaper andfoil, metallisedplastic film, andceramic
dielectric. Despite their ubiquity in power electronics rectifier and smoothing appli-
cations, electrolytic types are only suitable for millisecond timescale discharges due
to their high equivalent series resistance (ESR).
Capacitors based on wound plastic and aluminiumfoil layers or metallised plastic
are suitable for nFF ranges and have been developed extensively for high volt-
age (i.e. 1100 kV) and fast pulse discharge characteristics. Foil-wound capacitors
can be used for discharges with timescales as short as a few hundred nanoseconds.
The 1960s technologies of Kraft paper impregnated with oil were effective for s
timescale discharges, but have been largely superseded by the more recent plastics
technologies. Ceramic capacitors based on ferroelectrics, e.g. BaTiO
3
, are capa-
ble of nanosecond timescale discharges. Ceramic capacitor types are commercially
available up to 50 kV in values between 100 pF and a few nF. A general feature
of all types of capacitor used in pulse work is that the lifetime is highly depen-
dent on the degree of voltage reversal imposed as well as on the actual voltage.
The highest voltage rating can be achieved if it is possible to ensure no voltage
reversal during pulse operation. The greater the voltage reversal, the lower the volt-
age rating. Reversal characteristics are therefore standard data sheet items for pulse
capacitors.
13.2.4 Voltage multiplication: the Marx bank
Limitations imposed by DCcorona, DCinsulation, power electronics and transformer
design mean that generally 50100 kVis the maximumfeasible static capacitor design
voltage. Many pulsed power requirements are for voltages in the range 100 kV2 MV
and higher. One solution to this is provided by the Marx generator; this was shown
in Figure 13.1. A set of capacitors is charged in parallel to a voltage V via a set
of charging resistors. Closure of the switches typically spark gaps reconfigures
the capacitors in series and a high voltage impulse of the order of microseconds
duration can be delivered to the load. Very often, the load will be a transmission
line which is resonantly charged by the Marx bank. Generally, the lowest (first)
switch in the Marx bank will be actively triggered, and subsequent switches will
close due to a combination of ultra-violet (UV) illumination and overvolting. Many
high energy Marx banks will use additional spark gaps with mid-plane gaps which
can be triggered, either directly or by use of interstage coupling. Reliable operation
of Marx banks with high numbers of stages (i.e. 2030) generally depends upon the
use of some triggering linkages between stages and careful optimisation. Ideally, the
voltage produced by an n-stage Marx will be nV. Loading by the resistive charging
stacks and voltage drops across the switches will reduce this somewhat. Nonetheless,
voltages within a fewper cent of the ideal can be achieved in practice. Variations on the
Marx topology include bipolar charging, which enables doubling of the stage voltage
within the 100 kVcharging voltage constraint; use of more than one charging stack
in parallel to give faster charge time constants; use of inductive charging for low
dissipation in rep-rated applications, etc. Note that a typical high energy Marx bank
may have a series

LC time constant of a few s. This is determined by the total
series inductance, due to the combination of the geometrical layout and the internal
inductance of the capacitors and switching elements. Pai and Qi Zhang [13] provide
a detailed review of Marx bank technology.
13.2.5 Compact, fast rise time Marx banks
The development of compact high voltage sources is of self-evident importance for
applications of all types outside the laboratory.
The present day development of Marx bank techniques therefore focuses on com-
pact and high voltage systems. For high voltage, low energy systems (500 kV1 MV
and 10100 J/shot) the use of ceramic capacitors from manufacturers such as TDK
and Murata has become the de facto standard [14]. Such devices enable 50 kV or
100 kV charge voltages, as with high energy Marx banks. Up to MV outputs can
therefore be achieved with reasonable numbers of stages (i.e. 430). The capacitors
have low series inductance contributing to fast load rise times of 10100 ns. The
spark gaps may also be designed for low inductance. Hydrogen is typically used as
the switching gas as it has both high voltage hold-off capability and fast recovery rate,
so that repetition rates of 1 kHz are readily achieved. Pressurised SF
6
is commonly
used as an insulant dielectric gas around the Marx components [15].
Interest in the use of ultrafast Marx generators to provide nanosecond rise time
pulses relates to potential applications in ultrawideband (UWB) radar or directed
energy (DEW) applications [16, 17]. Achieving this level of rise time typically relies
on the Marx being given a layout which enables effective wave propagation between
input and output. As with avalanche semiconductor switching generators and DBD
lines, a degree of pulse sharpening can occur as the output pulse propagates and builds
up between successive spark gaps along the Marx stack.
13.2.6 Pulse compression
Energy may be transferred resonantly between two capacitors by a circuit with a
suitable closing switch and inductor. If a second LC loop is added, energy can then
be transferred into the third capacitor via the second closing switch. If the

LC
resonant period of the second loop is made shorter, the energy transfer occurs in a
shorter period and with a higher associated peak current. Figures 13.5 and 13.6 show
the generalised LCcircuit and waveforms, respectively. With reference to Figure 13.5,
the time constant of the first loop is given by:
1
=
_
L
1
C
0
C
1
C
0
+C
1
(13.2)
C
0
C
1
C
2
S
1
S
2
L
1
L
2
V
C
0
V
C
1
V
C
2
Figure 13.5 Single-stage LC pulse compression circuit
current
voltage
V
C
1
V
C
2
i
L
1
i
L
2
time
time
Figure 13.6 Idealised current and voltage waveforms in the pulse compression
circuit of Figure 13.5
C
0
S
ms
0
ms
1
C
1
C
2
C
3
load
ms
2
ms
3
Figure 13.7 Three-stage magnetic pulse compressor circuit
and the time constant of the second loop is given by:
2
=
_
L
2
C
1
C
2
C
1
+C
2
(13.3)
This process is termed pulse compression. The switching of the second switch will
usually be made to depend upon the state of the circuit. For example, a self-breaking
spark gap could be used. An inherent problem with the second and further switches
in pulse compressors is that pulse compression inherently means higher power levels
and a more arduous switching duty.
13.2.7 The Melville line magnetic pulse compressor
One way to make the pulse compression switching process occur automatically was
devised by Melville in the 1940s [1]. The inductors in the magnetic pulse compression
circuit (Figure 13.7) have saturating soft magnetic cores, typically of NiFe ferrite or
amorphous metal. These give the inductors a magnetic switching characteristic. While
the core is unsaturated (
0
r
), the inductor presents a high impedance to current flow.
When the core is driven into saturation, the inductor presents a low impedance (
0
),
and current is able to flow according to the value of L
sat
. The transition between
unsaturated and saturated states can occur in 10100 ns timescales, which means that
the inductor exhibits a quite rapid magnetic switching characteristic. The two main
parameters for the magnetic switch are the voltsecond hold-off and the saturated
inductance. The voltsecond hold-off is equivalent to the flux swing in the inductor
when it is driven from its initial bias point to saturation:
(B
s
B
r
)AN =
_
V dt (13.4)
where B
s
is the value of saturated magnetic core flux density, B
r
is the value of reset
magnetic core flux density, A is the magnetic core cross-sectional area and N is the
number of turns.
The operation of the magnetic pulse compression circuit of Figure 13.7 is as fol-
lows. MS
0
is initially in a high inductance state. After the primary switch S typically
a thyristor is closed, MS
0
is driven into saturation and the same process repeats
through MS
2
and MS
3
. C
0
resonantly transfers its charge to C
1
. Inductor MS
1
is
designed to remain in a high inductance until the point when C
1
is fully charged. MS
1
then saturates, and resonant current flows between C
1
and C
2
. Meanwhile, reverse
current between C
1
and C
0
is blocked by the fact that MS
0
requires to be driven into
negative saturation before it is closed for current in the reverse direction. Only mag-
netisation current flows during the magnetisation reversal phase. Further pulse com-
pression stages may be added. Between one and four compression stages are usual.
The magnetic pulse compressor (MPC) has gained considerable application in
repetitively pulsed applications [18]. These include driver circuits for TEA CO
2
gas
lasers [19], excimer lasers, copper vapour lasers [20], corona discharge reactors and
ultra-wideband (UWB) radar systems.
13.2.8 Transmission line circuits
The high voltage aspect of most pulsed power circuits imposes generally quite large
dimensions and spacings on circuit elements in order to achieve workable electric
stress levels. When electrical energy is required to be delivered in the form of short
pulses, the physical separation between circuit elements often makes it necessary for
these pulses to be delivered along some form of transmission line. The transmission
lines will be two-conductor lines e.g. coaxial or parallel-plate, supporting TEM-
mode waveforms as the principal mode. Helical slow wave lines are also used. The
transmission line transmits power according to the telegraphers wave equations. It
can support forward and backward travelling waves.
13.2.9 Charge lines
Of particular significance in pulsed power applications is the use of transmission line
cables as energy stores and pulse sources. A single charge line circuit is shown in
Figure 13.8 and a coaxial geometry in Figure 13.9. Alength of cable or transmission
line is charged to a DC voltage level V by a suitable arrangement. When the switch
connecting the transmission line to the load is closed, the load begins to discharge the
transmission line. The load pulse rise time is equivalent to the switching time. The
values of voltage and current in the load may be found by solving the transmission
line equations. In the case of a matched load, the load voltage is equal to half the
charging voltage. The load pulse then also has duration of 2T , where T is the cable
charging
supply
switch
transmission line load
Figure 13.8 Single charge line circuit
charging
supply
charge line spark gap load
Figure 13.9 Schematic arrangement of coaxial charge line with spark-gap
switched load
Marx bank
charge line compression
gap
sharpening
gap
load
Figure 13.10 Transmission line pulse compression and pulse sharpening
arrangement
transit time, given by:
T =
l
r
c
0
(13.5)
where l is the cable length, c
0
= 3 10
8
m/s is the free-space velocity of EM waves
and
r
is the relative dielectric constant of the cable.
In the case of the matched load, 2T is the time required for the transmission
line to become fully discharged. The load pulse waveform is essentially rectangular,
with rise and fall times defined by the turn-on time of the switch and the inductance
of the line-load connection. Note that the length of transmission line required for a
given pulse width depends upon the dielectric constant of the medium. The relative
dielectric constants for air/vacuum, oil and water are 1.0, 2.1 and 81, respectively.
A1 m line length therefore corresponds to a double transit time 2T (i.e. pulse width)
of 6.6 ns in air/vacuum, 10 ns in oil and 90 ns in water.
The charge line configuration is ideally suited to pulse compression and pulse
sharpening, as were outlined previously in the circuit context. Figure 13.10 shows
the general arrangement. The charge line is resonantly charged by a Marx generator in
the order of 12 s. The compression gap initiates a transmission line pulse discharge
with a pulse width of typically 100400 ns. This represents the pulse compression
phase. The sharpening gap holds off the initial voltage rise until near peak voltage,
then closes. The rise time through the sharpening switch is typically 1020 ns, and
this then propagates forward to the load.
In a variation of the charge line configuration, the charge line is switched to
ground by the closing switch. In principle, this applies the same waveform to the
load, but has the benefit that the closing switch is not floating. One disadvantage
is that the charge line outer conductor floats instead. A second disadvantage is that
the switching impulse travels along the charge line before it reaches the load, so that
the rise time may be degraded (lengthened) by transmission line losses. This type of
system is more suited to relatively low voltage, power and energy levels.
13.2.10 The Blumlein circuit
A variation on the charge line circuit is the Blumlein circuit, invented by
A.D. Blumlein. This uses two charge lines with the load interposed, as shown in
Figure 13.11. For matched operation, the impedance of each of the charge lines
should be one-half of the load impedance. The two lines are initially charged by
some suitable charger; note that the right-hand line charging path is through the load
impedance. The closing switch is between the two plates (conductors) of the left-
hand line. Closing the switch then launches a travelling wave into the left-hand line,
and when this wave reaches the load, the load voltage is developed. Analysis shows
that, for matched operation, the magnitude of the load voltage impulse is equal to the
charge voltage V. The duration of the load impulse is 2T , where T is the transit
time of each of the two charge lines. The benefit of the Blumlein is that the load pulse
voltage is equal to the full charge voltage, rather than one-half of the charge voltage
as in the single charge line configuration discussed above. Acoaxial implementation
of the Blumlein is shown in Figure 13.12.
13.2.11 Inductive voltage adders
Inductive voltage adders for pulsed power arose as a means of achieving multi-MV
pulses for radiographic and gamma-ray generators. A prime example is the 20 MV
Hermes system [21]. The basic principle is that of a series-coupled set of 1:1 trans-
formers (Figure 13.13). Each primary is fed by a separate pulse generator, which may
charging
supply
switch
transmission line 1 transmission line 2
load
Figure 13.11 Blumlein charge line circuit
charging
supply
closing switch
triaxial charge line
load
Figure 13.12 Coaxial Blumlein arrangement; the Blumlein outer conductor is
usually grounded
1:1
1:1
1:1
Figure 13.13 Three-stage inductive adder circuit using 1:1 transformers
be referenced to ground, and all of the secondaries are coupled in series. The output
voltage is the summation of the secondary outputs. In the transmission line embodi-
ment of this method (Figure 13.14), the secondary forms the centre conductor (stalk)
of a coaxial line which leads to the anode or cathode of the load, typically an e-beam
generator. The magnetic cores are located around the outside of the coaxial line
and provide the inductive coupling between the individual primaries with the cen-
tral coaxial stalk. In the Hermes system, there are twenty 1 MV primaries driven by
1 MV Marx banks, giving the 1920 MV output voltage. However, systems are not
necessarily this large. In MOSFET-switched pulsers manufactured by Kentech Instru-
ments, the outputs of 25 boards, each producing 400 V pulses, are added inductively
to form a 10 kV output into 50 . The output rise time is near equivalent to that of
a single switching MOSFET element at 23 ns. Two-stage inductive adders are used
in the RHEPP systems at Sandia. Inductive adders are currently being developed for
the next generation radiographic machine (LINX) at AWE Aldermaston in the UK
[22, 23].
pulse feeds
inductive cores
central stalk
Figure 13.14 Showing first three stages of a transmission line inductive adder
13.2.12 Inductive energy storage
Although capacitive storage has the merit of simplicity, inductive storage potentially
offers energy densities around 1001000 higher. This is based, respectively, on an
assumed electric field strength in the capacitive store of 25 MV/m and magnetic
flux density in the inductive store of 10T [24]. The caveat lies in the technology: for
inductive energy to be extracted, it is necessary to open a switch in order to divert the
inductor current into the load. Current interruption is inherently extremely difficult.
In the pulsed power context, the two major opening switch types are the plasma ero-
sion opening switch (PEOS) and, increasingly, semiconductor switches. The PEOS
addresses the MAcurrent range, whereas semiconductor device types such as MOS-
FETs and GTOs will open against 0.15 kA. Anew semiconductor approach due to
the Ioffe Institute, St Petersburg, Russia is based on exploiting the reverse recovery
spike in pn diodes. When a diode current reverses, the diode initially continues to
carry current while the pn junction is still in a conductive state. However, as the
pn junction recovers into a blocking state, the high impedance produces a reverse
recovery voltage spike due to the current which is still flowing. If not properly con-
trolled, this can cause serious damage and device failure in conventional circuits.
However, if the spike voltage can be connected across a load, it will give a useful
pulse. Work in Russia has focused on optimising pn diodes to maximise this effect.
The diode is driven by a relatively short (100200 ns) resonant current waveform.
Reverse recovery pulse rise times of a few ns can be obtained, at voltages of sev-
eral kV, and devices may be stacked to obtain up to 100 kVtotal pulse voltage. The
device type is referred to as a drift step recovery diode (DSRD) or silicon opening
switch (SOS). Most work and exploitation to date has been carried out in Russia, and
pulser units are commercially available from Russian companies such as Megapulse
and FID Technologies. Some amount of technology transfer out of Russia has also
secondary winding
HV end LV end
primary winding
Figure 13.15 Concentric primary and secondary spiral windings for the Tesla
transformer
occurred. Teramoto et al. [25] report a system using pulse compression from 60 Hz
mains via magnetic pulse compression and voltage transformation circuit, with final
pulse sharpening by the SOS resulting in an output pulse with 140 kV amplitude and
70 ns FWHM.
13.2.13 The Tesla transformer
The open-cored Tesla transformer is generally seen as an exotic curiosity due to its
unsuitability for utility power distribution applications. However, amateur enthusi-
asts construct versions with extreme turns ratios and lowcoupling coefficients. When
driven by primary spark gaps, these can produce radio frequency outputs with ampli-
tudes of up to several MV and capable of producing lengthy ionised channels in air.
The former Soviet Union (FSU) has treated the Tesla transformer more seriously and
developed the technology considerably [26]. Although the magnetisation current is
high, the coupling coefficient is nonetheless generally in the range 0.50.8 and pulsed
output voltages up to 1 MV can readily be achieved. Figure 13.15 shows a typical
Tesla transformer coil arrangement. Primary windings of a few turns are used, which
may be axially spaced coils or simply several turns of copper foil wound concentri-
cally. Turns ratios of the order of 1:1200 are typical. The principle of operation is that
of a pulse transformer: a thyristor or other primary switch is triggered in the primary
circuit and this applies a capacitor voltage to the transformer. Rise times of the order
of s are produced on the secondary winding. Typically, a transmission line or charge
line is capacitively charged by the Tesla coil output. Closure of a self-breaking or trig-
gered high pressure spark gap then initiates pulse discharge of the transmission line
into the load with a pulse width defined by the charge line length, and subnanosec-
ond rise times can be achieved. Typical loads include UWB emitters and high power
X-band oscillators (10 GHz at 150 MW pulse power) with impedances in the range
20150 .
13.3 Semiconductor switching
13.3.1 Introduction
The prospect of long lifetime, low jitter electronic switches in pulsed power has been
a tantalising one over the past decades. The take-up of commercial power devices has
been limited, because available device characteristics have only mapped onto certain
quite specific and limited areas. Nonetheless, as semiconductor power switching
devices have generally developed toward multi-kV, multi-kA voltage and current
ratings, combined with submicrosecond switching speeds, the ability to match power
devices with applications has steadily broadened. The thyristor has long been applied
to microsecond and longer timescale pulse requirements in the 1100 kArange. The
ability of the thyristor to carry large pulse currents is due to the large die areas
of up to tens of square cm. Gate turn-off (GTO) and similar thyristor types with
heavily interdigitated gate structures are used, due to their inherently rapid switching
times, which are of the order of 0.11 s. The bipolar transistor when operated in
avalanche mode is capable of subnanosecond switching times, for which there are
certain specialised applications. As voltage and current ratings have increased, power
MOSFETs and IGBTs are increasingly being used for rapid switching applications.
These devices are generally driven within their standard operating envelope. In terms
of circuit topology, the two configurations of principal interest are the series/parallel
stack and the semiconductor Marx bank. Finally, a range of novel pulsed power
semiconductor devices operating at thyristor power levels combined with nanosecond
switching speeds has emerged from work at the Ioffe Institute in St Petersburg.
13.3.2 Thyristor
The thyristor is an ideal single shot switch in that as a minority carrier device it is
inherently capable of carrying very large impulse currents relative to its continuous
current rating. The trigger requirements are also very modest relative to the switched
power. Note that the active turn-off capability will generally not be exploited in
pulsed power, so that the heavy reverse gate currents required by GTO thyristor
devices when switching off large currents are not usually an issue. The Marx bank
principle is readily applied to thyristors. Gregory et al. [27] describe a four-stage
thyristor Marx. The general circuit is shown in Figure 13.16. Diodes are used to
clamp the individual thyristor voltages, and each thyristor is actively triggered by
resistive coupling to the gate. The use of active triggering to all devices is a general
feature of semiconductor Marx circuits, and contrasts with the classic Marx, which
relies on sequential overvolting of the spark gaps in the stack. The system described
by Gregory et al. is relatively modest, producing 2.2 kV, 250 A, 2.5 s pulses into
a 10 load, but it illustrates the thyristor Marx concept very well. Development
work has shown that 1015 stage thyristor Marxes are feasible, using capacitive
links between each stage and the next thyristor gate. Repetitive operation is practical
except for the efficiency problems posed by resistive charging. Repetitive Marxes
have been applied for application in corona flue gas treatment. Again, parallel diodes
provide voltage-limiting control on the thyristor devices during the Marx erection
phase [28].
In Japan, there is also strong interest in the use of static induction thyristors
[29, 30]. These are relatively novel devices with MOSFET-type gate characteristics,
with a rapid turn-on time of 100 ns, and capable of carrying 30 kAat 15 V forward
drop. Pulse widths of 100500 ns can be switched efficiently. It is envisaged that
D R
T
+
+
+
+V
+
R
T
R
T
D
D
trigger
pulse
charging
resistors
Figure 13.16 Four-stage thyristor Marx with resistive triggering of upper stages
this switching capability could be exploited effectively in rapid resonant charging
of corona discharge capacitance for exhaust gas processing. To date, a three-device
stack switching 10 kV, 10 kArepetitively has been developed and demonstrated.
13.3.3 Bipolar transistor avalanche mode switching
Baker provides a useful outline of avalanche switching in bipolar junction transistors
(BJTs) [31]. A distinction is drawn between slow avalanche processes and current
mode second breakdown. The latter is characterised by much higher currents and
more complete voltage collapse, but also device destruction if the pulse energy is
not curtailed. The subnanosecond switching of BJTs is due to current mode second
breakdown, although it is generally referred to as avalanche mode. The successful
application of devices in this mode, therefore, depends upon switching only a modest
energy per pulse.
Transistors operated in avalanche mode have proved to be reliable and stable
switching elements for operation in various forms of Marx bank circuit, switching
typically 12 kV in <1 ns. Applications are typically micro channel plate systems,
pockels cells, and providing the beam sweep in streak cameras. Streak cameras are
used to image x-ray emissions, typically over 12 ns timescales. The x-rays are
directed onto a photo-emissive plate, and the intensity of the emitted photo-electron
beam varies over time with the x-ray intensity. The photo-electron beam is then
directed between the time axis sweep plates onto photographic film. The requirement
on the sweep plate drive voltage is typically for a pulse rise time of 12 kV in the
order of 1 ns. Circuits for streak camera application are given by a number of authors
[32, 33].
Grouping the avalanche transistors into series stacks can give benefits in reducing
the effect of the transistor junction capacitance, which acts as a parasitic element on
the Marx circuit. Rai et al. [34] use five stacks with three series transistors per stack.
Each transistor stack may be treated as a single switching element.
Transmission line layouts for avalanche mode switching circuits have been devel-
oped at the Lawrence Livermore National Laboratories in the US [35, 36]. Since the
inherent switching time of transistors in avalanche mode can be 50100 ps, the lim-
itation on circuit rise time is generally due to the circuit inductance capacitance and
physical size. Design of the Marx layout and ground plane on transmission line prin-
ciples enables the Marx voltage to propagate from stage to stage in the form of a
travelling wave as it develops. Rise times as low as 100 ps have been achieved, with
output voltages of up to 8 kV.
13.3.4 MOSFET Marx
The Marx bank configuration has also been applied to power MOSFETs [37]. The
achievable rise time is limited by the MOSFET switching speed, which at min-
imum is typically of the order of 3 ns. This limitation is essentially due to the
minimum feasible gate capacitance charging times. As with all semiconductors, it
is imperative to trigger the gates of all devices. This may be done by providing
each device with its own gate drive, or alternatively by providing suitable cou-
pling capacitances to the gate of each device. The second approach can give equally
fast switching times and may give a circuit with a much lower total element count.
A disadvantage of the second approach is that it requires a more complex design
analysis.
13.3.5 MOSFET switching stacks
MOSFETs can readily be stacked in series to form high voltage stacks, just as thyris-
tors are used in HVDC conversion schemes. Conventional drive arrangements to
each floating gate gives the most reliable arrangement. A basic system is described
by Continetti et al. [38] which was for applying potential difference ion beam exper-
iments. However, the stack switching time in this work was very slow at around
250 ns. Baker and Johnson [39] also describe the design of MOSFET stacks. They
reiterate the need for careful gate drive design to ensure that the maximum MOSFET
V
gs
rating of typically 20 V is not exceeded, as this will result in oxide layer punch
through. They use passive coupling to drive the gates of devices higher in the stack.
This is generally highly effective for stacks at turn-on, as each gate is driven by the
switching on of the device beneath it. Clearly, passive coupling arrangements to each
device gate generally have cost and simplicity benefits. However, passive coupling
is less effective when stacks are required to turn off fast, and this results in an asym-
metry between turn-on times of a few nanoseconds, and turn-off times of hundreds of
nanoseconds.
Stacked MOSFET techniques are increasingly being adopted by industry. Fully
packaged MOSFET stack modules with integral triggering circuits handle voltages
up to 60 kV and currents up to 800 A [40]. IGBTs are employed for higher currents
of up to 4800 A, albeit at lower switched voltages of 015 kV. Applications for
such switches are wide ranging and include EMC test equipment, klystron drivers,
EMC burst generators, nanosecond pulse generators, pockels cell drivers, thyratron
triggering, deflection grid drivers etc. However, when passive coupling triggering
techniques are used, this results in slow turn-off characteristics. The mean power
ratings for these devices are also relatively low.
A more recent approach now being adopted widely is series connection of com-
mercial MOSFET devices to form 2050 kV stacks capable of handling 100200 A
[41]. Each MOSFET has a fully active gate driver circuit, and so the stack can be
switched off as well as on in the order of 1020 ns. The series stack essentially
functions as a direct replacement for hard tube modulators. This is one reason for
engineering a series stack rather than a paralleled switch systemdriving a pulse trans-
former. Applications include magnetron modulators, x-ray drivers and very high duty
modulators with burst mode pulse repetition frequencies of up to 750 kHz. Atypical
example of a series/paralleled stack of MOSFETs achieves 55 kV, 200 Aswitching at
repetition rates up to 1333 Hz [42]. The stack volume is 0.1 m
3
and it weighs 91 kg.
Forced oil cooling is employed, and the on time may be varied between 0.56 s at
duty cycles up to 0.1 per cent.
Kicker pulsers are required for precise ion beam steering and marker applications
in linear accelerator installations. A s length beam can be cleaved into sections by
using kicker magnets to deflect segments of the beam as it passes by. It is necessary
to energise the kicker magnets with very square pulse waveforms. The pulse spec-
ifications are of the order of 10 ns rise and fall times, and 20200 ns pulse width
at 1015 kV and 200300 A, i.e. the kicker magnet coils are designed to present a
50 impedance. The requirement has previously been addressed with either pla-
nar triode or thyratron switching, but is now increasingly achievable using solid
state MOSFET stack approaches which can readily achieve the rise and fall time
specifications [43].
13.3.6 MOSFETs with inductive coupling
One approach to high voltage pulse generation is the use of a pulse transformer with a
single low voltage primary winding and single high voltage secondary winding. The
design of this type of pulse transformer is in fact quite difficult due to the large voltage
ratio, high numbers of turns and high voltage insulation requirements. In particular,
achieving fast rise times becomes difficult. An alternative approach is to couple the
secondaries of a number of 1:1 transformers in series. Each primary can be driven by
a grounded pulse circuit, as indicated in Figure 13.13. In the form of a set of primary
boards coupling to a transmission line inductive adder it lends itself to highly modular
design. Rise times below 3 ns can be achieved.
13.3.7 General semiconductor switching design issues
The attraction of semiconductor switching is the promise of very good lifetimes and
high controllability compared with most gas discharge switching devices. However,
as with all electronic circuit design, in order to achieve reliable circuit operation it
is necessary to ensure that the ratings for device voltage, current and dissipation are
adhered to absolutely. In the presence of very large switching currents, this is not
necessarily a trivial task. A particular problem with semiconductor switching is the
relative sensitivity of lowvoltage trigger circuits to noise. EMCdesign and screening
of the trigger circuits takes on at least as much significance as it does in measurements
and diagnostics. Circuits need careful engineering to ensure that false or self-induced
triggering does not occur with consequent runaway failures.
13.3.8 Novel semiconductor devices
Over the past two decades, the group led by KardoSysoev and Grekhov at the Ioffe
Institute in St Petersburg has devised at least four novel device types. The group has
approached the pulsed power semiconductors issue froma rather different perspective
than the US and other western nations. Essentially, they have set out to develop novel
devices suitable for pulsed power applications, i.e. having both rapid switching and
high power capabilities. This contrasts with the approach taken in the west, which is
to apply standard commercial semiconductor devices. The principal limitation of the
western approach is the powerspeed product. Although thyristor kAand kV ratings
are high, the switching times even for high speed GTO devices are broadly in the s
range. Most systems switched by thyristor will therefore require pulse compression or
pulse sharpening elements. Semiconductors optimised for general power electronic
application may well also have internal features different to what would be ideal for
pulsed applications. The approach taken by KardoSysoev and Grekhov has been to
revisit generic device types including the pn diode, thyristor, and bipolar transistor and
reengineer both device topologies and the junction design specifically to enhance high
speed capabilities. Most work has been carried out using silicon as the semiconductor
material. The four device types that have arisen are the reverse switching dynistor
(RSD), field ionisation dynistor (FID), drift step recovery diode (DSRD) and the
delayed breakdown diode (DBD). The RSD and FID devices are essentially variants
of the thyristor structure, but with switching times of the order of 0.55 ns [44]. The
RSD is a two-terminal device and is switched by a reverse pumping current followed
by a relatively rapidly applied forward voltage. The FID is a three-terminal device
and is more conventionally triggered by a gate pulse of the order of 50100 V. Both
FID and RSD device types are inherently thyristor-like, large die area devices, and
can be designed to carry up to several kAand kV.
The drift step recovery diode (DSRD) is a pn junction diode designed for a
very abrupt reverse recovery characteristic. Figure 13.17 shows a drive circuit.
V
C
V
load
load
DSRD
C
S
i
L
Figure 13.17 Drift step recovery diode (DSRD) pulse generating circuit
V
load
i
D
time
time
Figure 13.18 Diode current (i
D
) and load voltage (V
load
) waveforms in the DSRD
circuit
Figure 13.18 shows device waveforms. The device is driven by a resonant current
pulse, initially in the forward direction. As the resonant current reverses, the junction
initially conducts, but as the junction regains its blocking state, the current is forced
to flow into the high impedance load [45]. Voltages of several kV per device may
be generated, and devices may be stacked. This device is often referred to in the
literature as the silicon opening switch (SOS).
Finally, the delayed breakdown device (DBD) is a small pn junction device which
is driven into reverse breakdown by a suitably high voltage and fast rise time pulse
in the 110 kV range [46]. If the overvoltage can be applied fast enough, i.e. in a
nanosecond timescale, the avalanche breakdown across the pn junction takes the form
of a plasma shock wave, and the current rise time through the device is much faster
than that of the applied voltage. The device acts as a pulse sharpener.
13.3.9 Applications of novel semiconductors
A typical example of the application of reverse switch-on dynistor (RSD) is driv-
ing a pulsed electrostatic precipitator [47]. In this case, the RSD circuit produces
26 kV, 2.5 s pulses into a 50 pF load. The pulse voltage is superimposed onto a
steady 50 kV DC component on the precipitator. The pulsed voltage improves par-
ticle charging effectiveness and inhibits the breakdown that would occur for higher
values of DC voltage. Energy recuperation is a key capability since most of the load
capacitance energy is recoverable. This is done via magnetic pulse compression cir-
cuits. The dynistor is a 1.5 kV operating voltage, 6 kAdevice, i.e. ratings equivalent
to a typical GTO thyristor. The key features are that the RSD device is triggered
by applied circuit voltages and current reversal. From a conventional viewpoint, the
dynistor circuit operation may appear relatively complicated. However, the benefits
are that the rapid switching characteristics of the RSD enable good energy transfer
efficiency between source and load. Also, a fully solid state solution between the
command triggering thyristor input stage and the load is achieved.
Rukin et al. [48] describe a system using SOS devices to drive a pulsed corona
streamer. A thyristor-triggered magnetic pulse compressor (MPC) is used to drive a
series stack of ten paralleled pairs of SOS devices. The output pulse parameters are
voltage 1.2 MV, current 4 kA, rise time 10 ns and duration 4060 ns.
13.3.10 Electro-optic switching
Photoconductive switching (PCSS) involves placing a semiconductor slab into a con-
ducting state by means of laser light illumination. Primary applications of the PCSS
include ground-probing radar, precision gas switch triggering, short pulse ion accel-
erators and pockels cell drivers. Pockels cells typically require 0.32 kV and 20 A
to switch polarisation state. Benefits of PCSS are the low jitter of 10 ps, rapid
turn-on down to 50100 ps and low switch inductance. Essentially, there are three
operating modes for photoconductive semiconductor switching (PCSS): linear mode,
non-linear mode and lock-on. The linear mode requires a high ratio of laser power to
the switched electrical power, since one photon drives one electron into conduction
mode. The non-linear mode requires high on-state fields 1030 kV/cm, and hence
electrical switch losses are high. In the lock-on mode, avalanching occurs which is
promising for low on-state loss. In principle, using avalanche mode photoconductive
switching in bulk semiconductors, it is possible to control gigawatt levels of peak
power with <1 ns rise time [49]. However, current filamentation may lead to low
device lifetimes.
The generic issues for efficient optical switching are therefore lock-on and current
filamentation. Lock-on is the required condition where the device remains switched
on, i.e. in a conductive state, until the current falls below a holding value. This is
analogous to the latching effect when a thyristor is triggered. However, lock-on is
generally associated with filamentation. Filamentation leads to hot spots and early
failure. Resolving these issues is proving to be a major technological challenge.
13.3.11 Conclusions on semiconductor switching
It is clear that successful semiconductor applications in pulsed power require
extremely careful circuit engineering. This is in order to limit and control fault modes
on the one hand, and to ensure that devices are not stressed above their voltage,
current and dissipation ratings on the other. However, as with the progressive appli-
cation of power electronics to industrial drives and traction applications, it is clear
that the developing range of device capabilities available is resulting in increased
penetration of the pulsed power applications sector. Various applications match the
principal device types; MOSFETs, IGBTs and thyristors (GTO, asymmetric and field-
controlled). IGBT and MOSFET stacks are being developed as drop-in replacements
for thyratrons. Stacked thyristors have been used for some time in kicker magnet
drivers, for laser pulsers and also as coil and rail gun switches. The less well known,
fast-switching Russian devices are making inroads into novel repetitive applications
including corona discharge reactors and ultrawideband sources. Electrooptic switches
are being developed for subnanoseconds applications.
13.4 Non-linear transmission lines
The use of non-linear transmission lines (NLTLs) as a means to pulse sharpening
was first outlined in detail by Katayev in 1966 [50]. Electromagnetic shock waves
may develop in non-linear transmission lines where the point-on-wave velocity varies
[51]. The small-signal phase velocity is generally defined as:
c
ph
=
c
0
r
(13.6)
where c
0
is the phase velocity of electromagnetic waves in free space, and
r
and
r
are the relative permittivity and permeability, respectively. If the permeability varies
with the applied H-field:
r
= f (H) (13.7)
then the phase velocity clearly also varies with H. Adecreasing or saturating charac-
teristic for
r
with increasing H results in steepening of the waveform rising edge,
and stretching of the falling edge. The effect is analogous to generation of shock
waves in air. Clearly, the same mechanism operates for non-linear dielectric, and
non-linear transmission lines may be based on both saturable magnetic and non-linear
capacitance effects. Equivalent circuits for discrete element NLTLs are indicated in
Figures 13.19 and 13.20. Because of the varying point-on-wave velocity, the lead-
ing edge rise time of the waveform in a continuous transmission line (Figure 13.21)
becomes steeper until it is limited by the magnetic or dielectric frequency response.
Shock lines based on ferrite cores are the most widespread type of high power NLTL.
These are generally operated in magnetisation reversal mode [52, 53]. In reversal
mode the rate of change of magnetisation is governed by the expression [54]:
dM
dt
=
0
HM
s
1 +
2
_
1
M
2
M
2
s
_
(13.8)
where M is the ferrite magnetisation, M
s
is the saturation magnetisation, H is the
pulse field in the core in A/m, = 1.610
11
(rads/s/Tesla) is the gyromagnetic ratio
and is the damping factor, typically in the range 0.11.0.
Impedances are typically50, but lower values have alsobeendemonstrated[55].
It is possible to combine several shock line sections to obtain very high repetition
rates in a burst mode [56]. Since the early 1990s, application of an axial magnetic
source
load
Figure 13.19 Discrete-element NLTL based on non-linear magnetic elements
source
load
Figure 13.20 Discrete element NLTL based upon non-linear capacitances
ferrite outer conductor
inner conductor dielectric insulator
Figure 13.21 Ferrite-loaded coaxial NLTL
field bias has been used to increase switching speed by nearly an order of magnitude.
This type of shock line has been demonstrated to be capable of producing sub-100 ps
rise times at up to 100 kVinto 50 systems [5759]. Analysis and design of the non-
linear transmission line is difficult [60, 61], and lines need to be operated at relatively
high electric stress levels in order to achieve high enough non-linearity in the pulsed
fields. Nonetheless, the ability to achieve subnanosecond rise times at 10100 kV
levels and at high repetition rates is near-unique and creates a niche for non-linear
pulse sharpening lines.
Non-linear capacitance has also received considerable attention as a means for
obtainingpulse sharpening[62, 63]. However, the degree of non-linearity(i.e. the ratio
of initial to final permittivity at peak field) is far lower for paraelectric or ferroelectric
materials such as barium titanate than for ferrites. The high permittivity of most
non-linear dielectrics also makes it difficult to design systems for reasonably high
impedance. The typical relative dielectric constant for barium titanate of 10002000
gives rise to fractional ohm systems, whereas most applications at nanosecond rise
time level require 2550 .
In a discrete element transmission line with periodic non-linear elements such as
varactor diodes or capacitors [64, 65], the rise time may be limited by the non-linear
Figure 13.22 Parallel plate line with periodic dielectric slab loading for soliton
generation
DBD DBD DBD DBD
Figure 13.23 DBD-based pulse sharpening transmission line circuit
relaxation characteristic. However, it may also be limited by the

LC time con-
stant of each section if this is slower. In the latter case, it is also possible to obtain
soliton-like oscillatory waveforms. Interest in solitons as a means of obtaining fast
rise time pulses has enjoyed recurrent interest. Soliton formation requires a medium
which is both non-linear and dispersive. The non-linear material should also be low
loss. Slabs of non-linear dielectric material may be used to form a periodic struc-
ture (Figure 13.22), resulting in the generation of recurrent waveforms at 100s MHz to
a fewGHz. The lowimpedance typical of these systems enables GWpower levels tobe
achieved [66].
Another form of pulse sharpening is an extension of the sharpening gap approach
whereby a switching device transiently holds off an applied rising voltage wave-
form before switching in a shorter timescale. Aseries of delayed breakdown devices
(DBD) may be used in a variation of the shock line circuit to provide pulse sharp-
ening (Figure 13.23). Output pulses into 50 of 3.8 kV with 100 ps rise time are
measured [67].
13.5 Pulsed power applications
13.5.1 Introduction
Industrial applications are driven by a varying combination of technology push and
technology pull. The push is exerted by organisations with established pulsed power
capabilities searching for new applications or markets. This is exemplified by the
Atomic Research Centre (KernforschungszentrumKarlsruhe) at Karlsruhe, Germany.
In the early 1980s, an inertial confinement fusion (ICF) programme based on ion
beams was established. The pulsed power requirement was for a 1.7 MV, 1.5 TW,
50 ns beam. However, the ICF programme was then dropped, and only industrial
applications of pulsed power are now being pursued. In practice, this means find-
ing uses for the two existing ion beam sources. The types of application being
addressed include surface treatments and materials properties characterisation by ion
beam impact driven shock wave [68]. The pull is due to applications requirements
demanding solutions. Typical examples are in the exhaust gas cleanup and food pro-
cessing areas. In this section, a variety of industrial and other applications areas for
pulsed power techniques will be discussed.
13.5.2 Ion beam materials treatment
The two electron beam facilities at Karlsruhe are capable of 150 keV and 400 keV,
respectively. The impact power of these intense beams is of the order of MW/cm
2
over tens of square cm, high enough to melt the surface adiabatically to 2030 mand
100 m, respectively. The rapid melting and subsequent quenching of a thin surface
layer is an effective means of achieving required heat treatment effects, e.g. surface
hardening of gas turbine blades. Afurther issue in this application is that the high inlet
temperatures that are required for high efficiency operation of gas turbines mean that
the turbine blades require a ceramic thermal barrier coating on top of the conventional
metallic oxidation protection layer. Typical difficulties result from the grain structure
of the protection layer and adhesion of the ceramic layer. Remelting of the surface
with an electron beam is found to lead to much improved performance. Lifetimes of
10 000 hours at 950
C have been demonstrated.

In the US, ion beam systems are also being developed for surface treatments.
One repetitive system at Sandia National Laboratories is driven by a magnetic pulse
compressor [69, 70]. The pulsed power system is a thyratron-controlled five-stage
MPC feeding into a two-stage linear induction adder to generate 400 kV 150 ns
pulses at a rate of 10 pulses per second. At the time of reporting, the system had been
commercially operational for 18 months with only two four-hour down times due to
minor component failures.
13.5.3 Air treatment and pollution control
There is a general requirement for cost-effective means of reducing NO
x
and SO
x
emissions from vehicles and power stations, as well as reducing solid particulate
emissions. The cleanup of diesel fumes emitted from vehicle engines is a particular
issue. The problemlies inburningoff the diesel particulates, andthis mayrelypartially
on some type of ceramic reactor complex. The general chemistry for diesel particulate
cleanup is well established. Active corona reactors, surface discharges and beds of
high dielectric constant beads (which create partial discharges in the interstices) all
provide means to burn diesel particulates. However, the energy efficiency of such
processes is not well established. The issues appear to be very much the technical
questions of how to provide effective insulation, given that carbon is a moderately
good conductor, and how to provide the free radicals or ozone at a reasonable cost in
energy/power.
Gas discharge reactors are being investigated widely for use in diesel exhaust
remediation [71] and other applications. The principal problem lies in achieving a
cost-effective treatment process, i.e. achieving adequate design robustness, reliability
and lifetime at a reasonable cost and efficiency [72]. Ageneral difficulty with exhaust
remediation is that the pulsed corona scheme alone is not necessarily energy efficient
[73]. This is attributed to thermalisation of the plasma at >10 ns timescales. It is likely
that a combination of corona reactors with catalytic techniques will be required [74].
Given the need for particle collection, e.g. in diesel exhaust remediation, the idea
of combining corona precipitators with a corona reactor is attractive. A brief outline
of the corona precipitator follows.
13.5.4 Pulsed corona precipitators
The basic principle in electrostatic precipitators is that dust particles become charged
by the corona charge cloud around the high voltage electrode. The electric field then
provides a drift force, which results in the particle being attracted to the ground
electrode.
Once a particle has acquired charge, the electric field required to produce an ade-
quate drift field is quite moderate. However, the magnitude of electric field required
to ionise the region around the high voltage electrode effectively and to charge the
particles needs to be much higher. If applied continuously, this will generally result
in a transition from partial discharge and corona to arcing and breakdown. This
problem is avoided by providing a separate DC bias and pulsed corona energisation
fields. To date, pulsed fields have generally been of the order of 100500 s width.
This is effective for larger particles such as are found in coal burning power station
flues, and pulsed voltages may be switched by thyristor technology. However, it
appears likely that for sub 2.5 m particles effective corona discharge precipitators
will require faster pulses. This will be an area for investigation, e.g. in diesel exhaust
cleanup.
One of the classic problems with pulsed corona reactors in which the reaction
plates are charged as capacitors is the need for a circuit to remove the energy from
the circuit. This is inherently difficult. Resonant pulser circuits could be constructed
in order to enable reverse energy flow, for example based on MOSFET stacks with
antiparallel diodes. Alternatively, the magnetic pulse compressor (MPC) is inherently
a bidirectional circuit and may also be used to transfer energy back towards a supply.
The general viewseems to be that an electrical scheme alone is not efficient. Ther-
malisation of pulsed corona plasmas for pulse durations >10 ns is a major problem.
The combination of a suitable catalytic converter with the pulsed corona approach is
likely to provide a solution.
13.5.5 Biological applications
Biological applications cover a number of areas ranging from sterilisation to the
processing of foodstuffs. There is particular interest in the application of pulsed
electric fields to kill microorganisms such as eschericia coli by electroporation. The
bacterial cell wall is perforated by values of electric field of the order of 5 kV/mm
[75]. For adequate pulse magnitude, duration and repetition rate, this causes cell
death. The pulse repetition rate is found to be of particular significance since there
are cell repair or recovery mechanisms which operate over mss timescales. If a pulse
repetition rate of 100 Hz or above can be sustained, then the treatment effectiveness
is much enhanced [76]. It is possible that different values of field level may be
required with radio frequency exposure, but this is not yet a well understood or well
quantified area. Alternatively, irradiation with gamma rays also has potential for the
sterilisation of food surfaces. The benefit of a pulsed power approach to gamma-ray
generation based on e-beam technology is that no radioactive materials would be
required.
13.5.6 Biofouling and ballast water treatment
Discharge of ships ballast water can lead to non-indigenous species being spread
around the globe with undesirable consequences. The decontamination of ballast
waters of large ships is of increasing importance given present-day environmental
concerns [77]. On the basis of practical demonstrations, it is claimed that a suitable
spark discharge would allow decontamination by five to six orders of magnitude at
an energy cost of 27 kWh per 1000 litres of water [78]. This could provide a viable
alternative to chemical dosing or direct thermal (heating) approaches.
Biological fouling is generally a major current issue. The use of tributyl-tin (TBT)
as an antifouling coating on ships is now banned in the US with the consequence that
most US vessels are simply treated in overseas waters, i.e. the problem is shifted
to countries with less stringent regulations. There is therefore a genuine need for a
non-polluting replacement.
Experimental work shows that biological fouling of oceanographic sensors
(e.g. for salinity) can be remedied by application of 7 kV/mm, 100 ns duration
pulses to water as it is drawn through the sensor [79].
13.5.7 Food processing
Other areas in which technology push is exemplified are the food processing indus-
tries. In the production of citrus juices, Fishback and Nubbe [80] have recently found
that the application of pulsed discharges can lead to the inactivation of enzymes which
initiate degradation and limit shelf life. Typical pulse widths are 100 ns5 s, and
only moderate pulse energies are required per unit volume of juice (0.10.4 J/ml).
This is in contrast to the much higher energies required to have biocidal (sterilising)
effects. The system is driven by an AC mains-fed magnetic pulse compressor with
thyristor switching of the first stage. The system is straightforward and reliability
is good.
Juice extraction from diverse fruits and vegetables can also be enhanced by
electrically-driven breakdown of the cell walls. A benefit of this method is that the
temperature rise is extremely low. For example, in sugar processing, the applica-
tion of Marx-driven 300 kV 20 A discharges at 20 Hz p.r.f. is being found to
make juice extraction from beet more effective than alternative means [81]. Specific
energy consumption is 3 kJ/kg. The electroporation pretreatment then allows mechan-
ical extraction of juice to take place at reduced temperatures, which saves energy
and facilitates purification of the juice. Demonstrator plants to date have operated
at a few 100 kg/hour. This compares with typical throughput in large plants of
1000 tonnes/hour.
The sterilisation of liquid foods such as juices and milks by means of pulsed elec-
tric field (PEF) is an area of investigation [82]. Advantages over thermal techniques
are claimed to be that the taste is not adversely affected.
13.5.8 Water purification
Water purification relates to both bacterial decontamination and the removal of heavy
metals, organic and inorganic materials such as PCBs, hydrocarbons etc. This is
generally a multifactor process with synergies at work. Application of an electric
field followed by a pulsed discharge gives rise to direct electric field effects (electro-
poration, cell rupture, death), which are followed by production of ultra-violet, ozone
production, other active particles or radicals and possibly shock waves in the fluid.
These can give rise to chemistry effects on pollutants as well as on biological con-
taminants. The combined effect of all these is far greater than that due to electric
field alone [83]. Electrical breakdown can be achieved at moderate applied fields by
sparging of the liquid with air, nitrogen or other gases. The electric breakdown occurs
in the gas bubbles due to field enhancement effects. Another approach is the use of a
packed-bed reactor containing particles with high dielectric constant [84].
Due to industrial pollution, the need for techniques for water purification for
potable use is extreme in western Siberia [85]. The principal requirement is for reduc-
tion of heavy metals, hydrocarbons and phenols. Pulsed dielectric barrier discharge
(DBD) systems are currently being used at 40 installations [86]. This type of sys-
tem principally produces photochemical oxidation via ultra-violet, H
2
O
2
hydrogen
peroxide and ozone (O
3
), and the water is sprayed to maximise surface area. The
DBDprocess also has a bactericidal effect. The DBDis driven by a thyristor-triggered
magnetic pulse compressor, and after the DBD process, filtering is employed to
remove resulting particulates.
13.5.9 Mechanical applications of spark discharges
The shock waves produced by electric spark discharges can be exploited for a range
of applications. Much of the work to date has taken place in Russia, although there
is increasing interest in its application in Germany [87] and Japan [88]. Spark dis-
charges can be used for the direct fragmentation of reinforced concrete in recycling
applications. Concrete is well suited to shock-wave-induced fragmentation since it
is weak in tension and shear stress conditions. This potentially offers considerable
environmental benefits, given that the production and use of concrete contributes
ten per cent of greenhouse gas emissions [89]. Spark discharge techniques are also
being applied at pilot plant level at the ash treatment plant in BielefeldHerford,
Germany. Aseven-stage Marx produces 350 kV with a rise time of 150 ns. This is
rapid enough that the water around the ash acts as an insulator and effectively break-
down is channelled through the solid components. Spark discharge enables heavy
metal concentrations Zn and Pb in particular to be reduced below statutory levels.
However, the materials are still in solution and have to be removed from the water.
Energy consumption is 3 kW/h 2.5 tonnes/hour and 0.5 m
3
water per tonnes of
treated ash; the treated ash can be used in road construction, for example. Drilling,
boring and fragmentation of rocks by pulsed power spark discharge are also feasible.
Afluid is required in order to transfer the shock energy; from an electrical efficiency
perspective the fluid should be insulating oil, but from an environmental perspec-
tive water is preferable. Livitsyn et al. [90] review the processes of electric pulse
internal breakdown, electrohydraulics, surface flashover, plasma blasting and the
electric explosion of wires. Their experimental work exploits the lower bulk break-
down and surface flashover strength of rock relative to water. A 100200 kV Marx
generator is used in single-shot, both with direct connection of Marx to load, and in
a pulse compression mode with inductive storage of Marx output with opening foil
switch.
Granulation and crushing of ores are claimed to be areas in which spark discharge
(electroshock) technologies can outperform mechanical crushing. This can enable
valuable materials such as gold or diamond to be more effectively recovered from
ores inwhichthe concentrationmay be less thanone per cent. It is claimedbya number
of authors that the mechanical damage to the recovered ore and contamination of the
recovered product is very low [91].
Treatment of sewage sludge is an area in which use of electroshock techniques
accelerates the decomposition of organisms. In one test plant [92], a 10 kV voltage
is applied to the electrodes. After breakdown, a 600 A peak pulse current flows.
The repetition rate is 100 Hz. Although the treatment requires significant amounts of
energy at 1000 kJ/kg, nonetheless this compares with 10 kJ/kg for ultrasonics or high
pressure homogeneous crushers.
The principle exploding is being investigated as a means of substitution for high
explosive TNT in rock blasting applications. Ahigh current discharge of the order of
100 kAthrough fine aluminium wire contained in a cartridge can be used to produce
0.51 kg high explosive equivalent [93]. Electrical explosion of fine wires can also be
used in the production of nanopowders, i.e. powders with particle sizes in the range
50 nm5 m.
Spark discharges can also be used as a means of providing intense ultra-violet
illumination for surface treatment applications. Voronov et al. [94] describe the use
of a 300 kA 200 s wide pulse with 50 MW arc power to double the hardness of a
steel surface layer over a depth of 40 m.
13.5.10 Medical applications
One medical application of pulsed power lies in the use of pulsed magnetic fields
to provide electric field stimulation within the human body, for example for the
stimulation of kidney peristalsis following lithotripsy. A 15 kV, 15 kA coil design is
described by the Pulsed Power Group at Loughborough [95] which produces a pulsed
electric field at the kidney of 350 V/m with a FWHM of 1 ms. The possibilities of
using pulsed electric fields to induce drug delivery via cell wall electroporation effects
are currently being investigated [96]. Use of low intensity millimetre microwaves
(MMW) as a general therapeutic technique has been developed since the early 1980s
in Eastern Europe, China and Russia [97].
13.5.11 Ultrawideband and HPM applications
There is considerable interest in the effects of high power RF radiation on electronic
and electrooptic systems of all types. On the one hand, armed forces are interested
in extending their electronic warfare capabilities. On the other, there is a need to
guard both civil and military targets against potential terrorist threats based on RF
techniques. Since the development of the atomic bomb, nuclear electromagnetic
pulse (NEMP) hardening has been mandatory for military equipment of all types. The
NEMPwaveformhas double exponential formand is essentially similar to a lightning
waveform but with a much faster rise time of a few nanoseconds and a fall of a few
microseconds. However, the high power RF threat is much broader. It encompasses a
variety of waveforms including not only double exponential but also impulse and short
pulse sinusoidal waveforms with a range of envelope characteristics and frequencies
between a few MHz and tens of GHz. High power short pulse waveforms radiated
above 1 GHz are generally referred to as high power microwaves (HPM). Impulse
waveforms with rise times of the order 0.11 ns are generally referred to as ultra-
wideband (UWB) due to their very broad frequency spectrum. A generally accepted
definition of an ultrawideband signal is when the half-power bandwidth f
2
f
1
is
greater than 25 per cent of the centre frequency f
c
of the signal spectrum i.e.:
f
2
f
1
f
c
25% (13.9)
The reality is that effective application of RF weapons is a sophisticated task
requiring detailed knowledge of the target system, coupling mechanisms, genera-
tor technologies and so on. Although there has been military and defence industry
activity in the area for a number of years, programmes of university research into
pulsed RF techniques are also taking shape. In the US, multidisciplinary university
researchinitiative (MURI) programmes provide for collaborationbetweenthe Depart-
ment of Defense and the university sector. The types of effect being investigated
include the effects on electronic circuits and systems of pulse shape, modulation,
timing, clock states, pulse repetition rates, multiple failure mode synergies and
induced internal arcing and breakdown. Thus, many individual aspects of the RF
circuit threat are being brought together that have been studied separately over the
years.
Published ultrawideband research focuses largely on source technologies. These
include subnanosecond spark gap switching [98, 99], solid state switching and light-
activated switching systems driving into parabolic antennae or specialised types such
as TEM horns which can radiate impulse waveforms particularly effectively [100].
Source pulsed powers may be of the order of 10 MW to 1 GW and pulse repetition
rates up to several hundred Hz. The requirement is probably to induce fields of several
kV/m at a target location.
13.5.12 X-ray simulators
Ware et al. [101] provide an excellent review of the historical development of x-ray
simulators in the US, under the aegis of the present Defense Threat ReductionAgency
(DTRA), between the 1960s and the late 1990s. The main cost of modern simulators
in fact lies in the power conditioning and flow arrangements between the capacitor
Marx bank and the actual plasma or ion beam load. Early systems made a transition
from oil to water dielectric for transmission lines, enabling much lower impedance
and higher power flow. All such systems exploit the discovery by J.C. Martin that for
short, single-shot pulses, water dielectric can be operated at well above its breakdown
strength, because the breakdown streamer channel impedance is high enough that
relatively little energy is lost. The 1970s saw much basic physics established: diode
physics beam transport, pinch and plasma radiation sources. The 1980s and 1990s
saw stressing of all components to minimise size. The late 1990s have seen the
discovery of implosion of radiative plasmas in wire arrays (the Z machine) as a
means of producing intense x-rays. The size of the Z machine at Sandia is awesome.
It contains around 1.7 million litres of Diala insulating oil and 2.7 million litres of
deionised water. For compactness, future directions for x-ray generators are likely to
involve ultrafast Marxes, plasma opening switches and non-water systems.
Most x-raysimulator systems are modest incomparisonwithmachines suchas Z.
A good description of a modern day, compact, moderate energy x-ray generator for
simulator studies is provided by Miller et al. [102].
The current direction of ICF fusion research is summarised by Cook [103]. In
19981999, the Sandia Z machine was built up to achieve 1.81.9 MJ of x-rays
in a 6 ns burst. The energy conversion efficiency from Marx bank to x-ray yield is
surprisingly high at 15 per cent.
13.6 Conclusions
Pulsed power originated with the invention of the Marx bank in the early 1920s,
enabling the simulation of lightning strikes and switching operations on power sys-
temcomponents. Between the 1960s and the beginning of the twenty-first century, the
demands of high energy pulsed power applications have stretched Marx bank tech-
niques enormously. Refinements of triggering and switching techniques, and control
and exploitation of the stray capacitances in the systems, enable 100200 ns rise
times to MV levels and above with MJ of energy being stored and delivered. The
requirements of exotic radiographic, e-beam and plasma pinch loads have demanded
the use of pulse forming networks or lines interposed between the Marx and the load
to provide the requisite pulse shaping and power levels. Meanwhile the rise time
capability of low energy Marxes has been reduced to nanosecond levels as a result
of developments in capacitor and switching technologies. The latest direction is now
for the development of ultrafast high energy Marxes with adequately low inductance
for direct feed of compact e-beam loads.
Although the largest and most spectacular pulsed power systems at present remain
the single shot radiographic and fusion research facilities, there is an ever increasing
requirement for repetitive systems. These range from ultrawideband radar pulsers
through corona reactors to ion beam material processing facilities, and typically may
have mean power levels in the kWMWrange. The mean power and lifetime require-
ments give rise to a whole new set of design problems and constraints. Magnetic
pulse compressors were first developed in the 1940s for radar magnetron modula-
tors. The notable high power pulse compression qualities of these systems have been
revived and developed enormously for a range of new applications, including high
power gas laser drivers, e-beam drivers, UWB pulsers and pulsed corona reactors
and precipitators. The benefits of magnetic pulse compression include high energy
efficiency, rep-ratability and the ability to recover energy from reactive loads. The
development of magnetic switch technology also led to the development of inductive
adders, which use the same form of magnetic core. Based on amorphous alloy mate-
rials, these enable multi-MV pulse generation by summing the output voltages of a
number of Marx or other modules.
Pulsed power has principally developed in response to high energy physics and
weapons programme requirements; plasma drivers, x-ray drivers, magnetron drivers,
laser drivers, ion beam steering and acceleration. However, the industrial applica-
tion of pulsed power is increasing. Examples include the use of high energy electron
beams in plasma reactors for flue gas cleanup, and in ion beam surface treatments,
typically of high technology turbine blades for operation under high temperatures.
High current (100 kA) electric discharges may be used to fragment valuable min-
eral ores. Fragmentation of structural materials in recycling situations is also being
developed. Ageneral feature in pulsed power is the increasing insertion of power elec-
tronic devices into areas which even ten years ago were reserved for spark gaps and
other gas discharge devices. This is due first to the increasing power and switch-
ing capabilities of well established commercial device types such as MOSFETs,
IGBTs and thyristors. Second, there has been the recent discovery and development of
nanosecond timescale switching capabilities in a range of non-standard device types,
including the field ionisation dynistor (FID), the drift step recovery diode/silicon
opening switch (DSRD/SOS) and the delayed breakdown device (DBD). The repet-
itive and long lifetime switching of voltage and power levels in the 10100 kV
and 2200 MW range using semiconductors will therefore increasingly become
the norm.
For further reading see References 2, 3, 8, 9, 12 and 104.
13.7 References
1 MELVILLE, W.S.: The use of saturable reactors as discharge devices for
pulse generators, IEE Proc., Part 3 (Radio and Communication), 1951, 98,
pp. 185207
2 FRNGEL, F.B.A.: High speed pulse technology, (Academic Press), vol. I:
Capacitor discharges, magnetohydrodynamics, x rays, ultrasonics, 1965;
vol. II: Optical pulses, laser measurement techniques, 1965; vol. III: Capa-
citor discharge engineering, 1976; vol. IV: Sparks and laser pulses, 1980
3 MARTIN, T.H., GUENTHER, A.H., and KRISTIANSEN, M. (Eds):
J.C. Martin on pulsed power, series: advances in pulsed power technology
(Plenum Press, 1996)
4 KIDD, A.K., ANGOOD, S.M., BAILLY-SALINS, R., CARR, P.S.,
HIRST, G.J., and SHAW, M.J.: Pulsed power system for the Titania KrF
laser module. Proceedings of 9th IEEE Pulsed power conference, 1993,
pp. 718722
5 MARTIN, T.H., TURMAN, B.N., and GOLDSTEIN, S.A., et al.: The pulsed
power characterisation phase. Proceedings of IEEE Pulsed power conference,
1987, pp. 225232
6 RUBIN DE CERVENS, D., and MARRET, J.P.: Energy bank of laser mega-
joule CEA-Project. Proceedings of IEEE Pulsed power conference, 1999,
pp. 356358
7 PENETRANTE, B.M., HSIAO, M.C., and BARDSLEY, J.N., et al.: Electron
beamand pulsed corona processing of volatile organic compounds and nitrogen
oxides. Proceedings of IEEE Pulsed power conference, 1995, pp. 144149
8 PAI, S.T., and QI ZHANG: Introduction to high power pulse technology,
advanced series in electrical and computer engineering vol. 10 (World
Scientific, 1995), pp. 15
9 SMITH, P.W.: Transient electronics(Pulsed Circuit Technology, Wiley, 2002)
10 DRIGA, M.D.: Compulsators: advanced pulsed power supplies for hyper-
velocity accelerators. Proceedings of 12th IEEE Pulsed power conference,
1999, pp. 324328
11 SCOTT, D.J., CALFO, R.M., and SCHWENK, H.E.: Development of high
power density pulsed ac generators. Proceedings of 8th IEEE Pulsed power
conference, 1991, pp. 549552
12 THORNTON, E.: Electrical interference and protection (Ellis Horwood,
London, 1991)
13 PAI, S.T., and QI ZHANG: Introduction to high power pulse technology,
14 GOERZ, D.A., FERREIRA, T.J., NELSON, D.H., SPEER, R.D., and
WILSON, M.J.: An ultra-compact Marx HV generator. Proceedings of 13th
IEEE Pulsed power conference, 2001, pp. 628631
15 DRAGT, A.J., and ELIZONDO, J.M.: Compact, battery powered, 400 kV
40 joule portable Marx generator. Proceedings of 13th IEEE Pulsed power
16 MAYES, J.R., CAREY, W.J., NUNNALLY, W.C., and ALTGILBERS, L.:
The Marx generator as an ultrawideband source. Proceedings of 13th Pulsed
power conference 2001, pp. 16651668
17 MAYES, J.R., CAREY, W.J., NUNNALLY, W.C., and ALTGILBERS, L.:
The Gatling Marx generator system. Proceedings of 13th Pulsed power
18 HARJES, H.C., PENN, K.J., and REED, K.W., et al.: Status
of the repetitive high energy pulsed power project. Proceedings of
8th IEEE Pulsed power conference, San Diego, 1619 June 1991,
pp. 543548
19 SWART, P.H., VONBERGMANN, H.M., and PRETORIUS, J.H.C.: All solid
state switched pulser for multikilowatt, multi-kilohertz excimer and CO2 TEA
lasers. Proceedings of 8th IEEE Pulsed power conference, 1719 June 1991,
pp. 743749
20 COOK, E.G., BALL, D.G., and BIRX, D.L., et al.: High average power
magnetic modulator for copper lasers. Proceedings of 8th IEEE Pulsed power
21 RAMIREZ, J.J., PRESTWICH, K.R., and BURGESS, E.L., et al.: The
Hermes-III program. Proceedings of 6thIEEEPulsedpower conference, 1987,
pp. 294303
22 THOMAS, K.J., WILLIAMSON, M.C., and PHILLIPS, M.J.: LINX and
future IVA machines. IEE Pulsed power symposium, 1999, digest 99/030,
pp. 2/12/4
23 THOMAS, K.J., WILLIAMSON, M.C., CLOUGH, S.G., and Phillips, M.J.:
Prototype IVA module (PIM). Proceedings of 13th IEEE Pulsed power
24 PAI, S.T., and QI ZHANG: Introduction to High Power Pulse Technology,
25 TERAMOTO, Y., DEGUCHI, D., KATSUKI, S., NAMIHIRA, T.,
AKIYAMA, H., and LISITSYN, I.V.: All solid state trigger-less repet-
itive pulsed power generator utilising semiconductor opening switch.
Proceedings of 13th Pulsed power conference, Las Vegas, 2001,
pp. 540543
26 GUBANOV, V.P., KOROVIN, S.D., PEGEL, I.V., ROITMAN, A.M.,
ROSTOV, V.V., and STEPCHENKO, A.S.: Compact 1000pps high volt-
age nanosecond pulse generator, IEEE Trans. Plasma Sci., 1997, 25, (2),
pp. 258265
27 GREGORY, K., STEVENSON, P., and BURKE, R.: Four stage Marx
generator using thyristors, Rev. Sci. Instrum., 1998, 69, (11), pp. 39963997
28 MAEYAMA, M., and YOSHIDA, M.: Improvement of a high repetitive rate
static impulse voltage generator. Proceedings of 13th IEEE Pulsed power
29 NISHIKAWA, K., OKINO, A., WATANABE, M., HOTTA, E., KO, K-C., and
SHIMIZU, N.: High rep rate operation of pulsed power modulator using high
voltage static induction thyristors. Proceedings of 13th IEEE Pulsed power
30 TSUNODA, K., MAEYAMA, M., HOTTAM, E., and SHIMIZU, N.: Switch-
ing properties of series connected static induction thyristor stack. Proceedings
of 13th IEEE Pulsed power conference, 2001, pp. 17861789
31 BAKER, R.J.: High voltage pulse generation using current mode second
breakdown in a bipolar junction transistor, Rev. Sci. Instrum., 1991, 62, (4),
pp. 10311036
32 BAKER, R.J., and JOHNSON, B.P.: Sweep circuit design for a picosecond
streak camera, Meas. Sci. Technol., 1994, 5, pp. 408411
33 LIU, J.Y., SHAN, B., and CHANG, Z.H.: High voltage fast ramp pulse
generation using avalanche transistor, Rev. Sci. Instrum., 1998, 69, (8),
pp. 30663067
34 RAI, V.N., SHUKLA, M., and KHARDEKAR, R.K.: A transistorised Marx
bank circuit providing sub-nanosecond high voltage pulses, Meas. Sci.
Technol., 1994, 5, pp. 447449
35 FULKERSON, E.S., and BOOTH, R.: Design of reliable high volt-
age avalanche transistor pulsers. Proceedings of IEEE Power modulator
symposium, June 1994
36 FULKERSON, E.S., NORMAN, D.C., and BOOTH, R.: Driving pockels cells
using avalanche transistor pulsers. Proceedings of IEEE Power modulator
symposium, 1997
37 BAKER, R.J., and JOHNSON, B.P.: Applying the Marx bank circuit
configuration to power MOSFETs, Electron. Lett., 1993, 29, (1), pp. 5657
38 CONTINETTI, R.E., CYR, D.R., and NEUMARK, D.M.: Fast 8 kV metal-
oxide semiconductor field-effect transistor switch, Rev. Sci. Instrum., 1992,
63, (2), pp. 18401841
39 BAKER, R.J., and JOHNSON, B.P.: Stacking power MOSFETs for use in
high speed instrumentation, Rev. Sci. Instrum., 1992, 63, (12), pp. 57995801
40 Behlke Instruments GmbH catalogue, Germany
41 RICHARDSON, R., RUSH, R., ISKANDER, M., GOOCH, P., and DEAL, K.:
High voltage direct switching solid state MOSFET modulators for driving RF
sources. Proceedings of IEE Pulsed power symposium, April 1999, paper 13,
pp. 1415
42 RICHARDSON, R., RUSH, R.J., and ISKANDER, S.M., et al.: Compact
12.5 MW, 55 kV solid state modulator. Proceedings of 13th IEEE Pulsed
power conference, 2001, pp. 636639
43 DE HOPE, W.J., CHEN, Y.J., COOK, E.G., DAVIS, B.A., and YEN, B.:
Recent advances in kicker pulser technology for linear induction accel-
erators. Proceedings of 12th IEEE Pulsed power conference, 1999,
pp. 416419
44 EFANOV, V.M., and KRIKLENKO, A.V.: Small size solid state nano- and
picosecond pulsers on the basis of fast ionization devices. Proceedings of 13th
Pulsed power conference, Las Vegas, 1722 June 2001, pp. 479481
45 GREKHOV, I.V., and MESYATS, G.A.: Physical basis for high power semi-
conductor nanosecond opening switches. Proceedings of 12th IEEE Pulsed
46 KARDO-SYSOEV, A.F., EFANOV, V.M., and CHASHNIKOV, I.G.: Fast
power switches frompicosecond to nanosecond timescale and their application
to pulsed power. Proceedings of 10th IEEE Pulsed power conference, 1995,
pp. 342347
47 GREKHOV, I.V., KOROTKOV, S.V., KOZLOV, A.K., and
STEPANYANTS, A.L.: A high-power high-voltage pulse generator based
on reverse switch-on dynistors for an electrostatic precipitator power supply,
Instrum. Exp. Tech., 1997, 40, (5), pp. 702704
48 RUKIN, S.N., MESYATS, G.A., PONOMAREV, A.V., SLOVIKOVSKY, B.G.,
TIMOSHENKOV, S.P., and BUSHLYAKOV, A.I.: Megavolt repetitive SOS
based generator. Proceedings of 13th IEEE Pulsed power conference, 2001,
pp. 12721275
49 ZUTAVERN, F.J., LOUBRIEL, G.M., and MAR, A., et al.: Photoconductive
semiconductor switch technology. Proceedings of 12th IEEE Pulsed power
50 KATAYEV, I.G.: Electromagnetic shock waves (Iliffe Books, 1966)
51 OSTROVSKII, L.A.: Formation and development of electromagnetic shock
waves in transmission lines containing unsaturated ferrite, Sov. Phys.-Tech.
Phys., 1964, 8, (9), pp. 805813
52 WEINER, M., and SILBER, L.: Pulse sharpening effects in ferrites, IEEE
Trans. Magn., 1981, MAG-17, pp. 14721477
53 SEDDON, N., and THORNTON, E.: A high-voltage, short rise-time pulse
generator based on a ferrite pulse sharpener, Rev. Sci. Instrum., 1988, 59,
pp. 24972498
54 GYORGY, E.M.: Rotational model of flux reversal in square loop ferrites,
J. Appl. Phys., 1957, 28, (9), pp. 10111015
55 SHVETS, V.A.: Nonlinear multichannel pulse-sharpening line containing
ferrite rings with nonrectangular hysteresis loop, Instrum. Exp. Tech., 1982,
pp. 908912 (Pribory i Tekhnika Eksperimenta, 1982, (4), pp. 116119)
56 IKEZI, H., DE GRASSIE, J.S., LIN-LIU, R., and DRAKE, J.: High power
pulse burst generation by magnetically segmented transmission lines, Rev. Sci.
Instrum., 1991, 62, (12), pp. 29162922
57 BENSON, T.M., POULADIAN-KARI, R., and SHAPLAND, A.J.: Novel
operation of ferrite loaded coaxial lines for pulse sharpening applications,
Electron. Lett., 1991, 27, (10), pp. 861863
58 POULADIAN-KARI, R., SHAPLAND, A.J., and BENSON, T.M.: Develop-
ment of ferrite line pulse sharpeners for repetitive high power applications,
IEE Proc. H, Microw. Antennas Propag., 1991, 138, (6), pp. 504512
59 DOLAN, J.E.: Effect of transient demagnetisation fields on coherent mag-
netic switching in ferrites, IEE Proc. A, Sci. Meas. Technol., 1993, 140, (4),
pp. 294298
60 DOLAN, J.E.: Simulation of ferrite loaded coaxial lines, Electron. Lett.,
1993, 29, (9), pp. 762763
61 DOLAN, J.E., and BOLTON, H.R.: Alength equation for ferrite-loaded high
voltage pulse sharpening lines, Electron. Lett., 1998, 34, (13), pp. 12991300
62 BAKER, R.J., HODDER, D.J., JOHNSON, B.P., SUBEDI, P.C., and
WILLIAMS, D.C.: Generation of kilovolt subnanosecond pulses using a
nonlinear transmission line, Meas. Sci. Technol., 1993, 4, pp. 193895
63 WILSON, C.R., TURNER, M.M., and SMITH, P.W.: Electromagnetic
shock-wave generation in a lumped element delay line containing nonlinear
ferroelectric capacitors, Appl. Phys. Lett., 1990, 56, (24), pp. 24712473
64 RODWELL, M.J.W., et al.: Active and nonlinear wave propagation devices
in ultrafast electronics and optoelectronics, Proc. IEEE, 1994, 82, (7),
pp. 10371059
65 BAKER, R.J., HODDER, D.J., JOHNSON, B.P., SUBEDI, P.C., and
WILLIAMS, D.C.: Generation of kilovolt-subnanosecond pulses using a
nonlinear transmission line, Meas. Sci. Technol., 1993, 4, pp. 893895
66 IKEZI, H., deGRASSIE, J.S., and DRAKE, J.: Soliton generation at 10 MW
level in the very high frequency band, Appl. Phys. Lett., 1988, 58, (9),
pp. 986987
67 FROST, C.A., MARTIN, T.H., FOCIA, R.J., and SCHOENBERG, J.S.H.:
Ultra low jitter repetitive solid state picosecond switching. Proceedings of
12th IEEE Pulsed power conference, 1999, pp. 291294
68 BLUHM, H., BAUMUNG, K., and BOEHME, R., et al.: Pulsed power science
and technology at Forschungszentrum Karlsruhe. Paper A.04, proceedings of
international conference on Pulsed power applications, University of Applied
Sciences, Gelsenkirchen, 2729 March 2001
69 BARRETT, D.M., COCKREHAM, B.D., and PRAGT, A.J., et al.: A pulsed
power modulator system for commercial high power ion beam surface treat-
ment applications. Proceedings of IEEE Pulsed power conference, 1999,
pp. 173176
70 STINNETT, R.W., BUCHHEIT, R.G., and NEAU, E.L., et al.: Ion beam
surface treatment: a new technique for thermally modifying surfaces using
intense, pulsed ion beams. Proceedings of IEEE Pulsed power conference,
1995, pp. 4655
71 PUCHKAREV, V., KHARKOV, A., GUNDERSEN, M., and ROTH, G.:
Application of pulsed corona discharge to diesel exhaust remediation.
Proceedings of 12th IEEE Pulsed power conference, 1999, pp. 511514
72 WEKHOF, A.: Multi-parameter non-thermal corona laboratory system with a
build-in process reactor. Paper B.09, proceedings of international conference
on Pulsed power applications, University of Applied Sciences, Gelsenkirchen,
2729 March 2001
73 TAKAKI, K., JANI, M.A., and FUJIWARA, T.: Oxidation and reduction of
NO
x
in diesel engine exhaust by dielectric barrier discharge. Proceedings of
IEEE 12th Pulsed power conference, 1999, pp. 14801483
74 GROTHAUS, M.G., KHAIR, M.K., PAUL, P., FANICK, E.R., and
BANNON, D.R.: A synergistic approach for the removal of NO
x
and PM
from diesel engine exhaust. Proceedings of IEEE Pulsed power conference,
1999, pp. 506510
75 HEINZ, V., ANGERSBACH, A., and KNORR, D.: Membrane permeabiliza-
tion in biological materials in response to pulsed electric fields a highly
efficient process for biotechnology. Paper B.03, proceedings of international
conference on Pulsed power applications, University of Applied Sciences,
Gelsenkirchen, 2729 March 2001
76 ALY, R.E., JOSHI, R.P., STARK, R.H., SCHOENBACH, K.H., and
BEEBE, S.J.: The effect of multiple, microsecond electrical pulses
on bacteria. Proceedings of IEEE Pulsed power conference, 2001,
pp. 11141117
77 RUIZ, G.M., RAWLINGS, T.K., and DOBBS, F.C., et al.: Global spread of
microorganisms by ships, Nature, 2000, 408, p. 49
78 ABOU-GHAZALA, A., KATSUKI, S., SCHOENBACH, K.H., DOBBS, F.C.,
and MOREIRA, K.R.: Bacterial decontamination of water by means of pulsed
corona discharges. Proceedings of 13th IEEE Pulsed power conference, 2001,
pp. 612615
79 BLOCK, R., LEIPOLD, F., and LEBAHN, K., et al.: Pulsed electric field
based antifouling method for salinometers. Proceedings of 13th IEEE Pulsed
80 FISHBACK, J.R., and NUBBE, M.E.: Preliminary investigation of
enzyme inactivation in citrus juices by submerged electric discharge.
Paper B.08, proceedings of international conference on Pulsed power appli-
cations, University of Applied Sciences, Gelsenkirchen, 2729 March
2001
82 GAUDREAU, M.P.J., HAWKEY, T., PETRY, J., and KEMPKES, M.A.:
A solid state pulsed power system for food processing. Proceedings of 13th
IEEE Pulsed power conference, 2001, pp. 11741177
83 ESPIE, S., MARSILI, L., MACGREGOR, S.J., and ANDERSON, J.G.:
Pulsed power inactivation of liquid-borne microorganisms. Paper B.07,
proceedings of international conference on Pulsed power applica-
tions, University of Applied Sciences, Gelsenkirchen, 2729 March
2001
84 BYSTRITSKII, V, YANKELEVICH, Y., and WOOD, T., et al.: Pulsed
discharge in the fluidized packed bed reactor for toxic water remediation.
Proceedings of IEEE Pulsed power conference, 1999, pp. 464467
85 YAVOROVSKY, N.A., and BOEV, S.: Electropulse water treatment.
Proceedings of IEEE Pulsed power conference, 1999, pp. 181184
86 YAVOROVSKY, N.A., PELTSMANN, S.S., KHASKELBERG, M.B., and
KORNEV, J.I.: Pulsed barrier discharge application for water treatment and
disinfection. Paper B.02, proceedings of international conference on Pulsed
power applications, University of Applied Sciences, Gelsenkirchen, 2729
March 2001
87 HOFMANN, J., and WEISE, Th. H.G.G.: Material disintegration and material
breakdown with electrically generated shock waves. Paper E.01, proceed-
ings of international conference on Pulsed power applications, University of
Applied Sciences, Gelsenkirchen, 2729 March 2001
88 USHAKOV, V.Ya., andDULZON, A.A.: Performance capabilityof technolog-
ical installations using the electrical discharge of conductor electrical explosion
energy. Paper A.03, proceedings of international conference on Pulsed power
applications, University of Applied Sciences, Gelsenkirchen, 2729 March
2001
90 LIVITSYN, I.V., INOUE, H., KATSUKI, S., AKIYAMA, H., and
NISHIZAWA, I.: Drilling and demolition of rocks by pulsed power.
Proceedings of 12th IEEE Pulsed power conference, pp. 169172
91 DULZON, A.A., KURETZ, V.I., and TARAKANOVSKI, E.N., et al.:
Electric pulse selective disintegration of mountain breeds, of techni-
cal rubber of workpieces and extraction of substances from vegetative
raw material. Paper E.05, proceedings of international conference on
Pulsed power applications, University of Applied Sciences, Gelsenkirchen,
2729 March 2001
92 JUNG, M., WEISE, T.H.G.G., WITTENBERG, M., MUELLER, J.A., and
DICHTL, N.: Sludge disintegration by electrical shock waves. Paper B.01,
proceedings of international conference on Pulsed power applications,
University of Applied Sciences, Gelsenkirchen, 2729 March 2001
93 VORONOV, A., WEISE TH., and BERG, M.: Pulsed power in explosions
driven electrically. Paper E.03, proceedings of international conference on
Pulsed power applications, University of Applied Sciences, Gelsenkirchen,
2729 March 2001
94 VORONOV, A., WEISE TH., and BERG, M.: Pulsed power discharge for
producing an ultra-thin hardened layer. Paper E.04, proceedings of inter-
national conference on Pulsed power applications, University of Applied
95 YOUNG, A.J., NOVAC, B.M., SMITH, I.R., LYNN, B., and MILLER, R.A.:
A pulsed-power system for producing high-intensity magnetic and electric
fields for medical applications. Paper C.01, proceedings of international
96 SCHOENBACH, K.H., BARKER, R.J., and LIU, S.: Nonthermal
medical/biological treatments using electromagnetic fields and ionized
gases. Proceedings of 12th IEEE Pulsed power conference, 1999,
pp. 497501
97 PAKHOMOV, A., and MURPHY, M.: Low intensity millimetre microwaves
as a novel therapeutic modality. Proceedings of 12th IEEE Pulsed power
98 PRATHER, W.D., BAUM, C.E., and LEHR, J.M., et al.: Ultrawideband
source research. Proceedings of 12th IEEE Pulsed power conference, 1999,
pp. 185188
99 ANDREEV, Y.A., BYANOV, Y.I., and EFREMOV, A.M., et al.: Gigawatt
power level ultrawideband radiation generator. Proceedings of 12th IEEE
Pulsed power conference, 1999, pp. 13371340
100 JUNG, M., WEISE, TH. H.G.G., BRAUNSBERGER, U., and SABATH, F.:
High power compact UWBsystems. Paper G.01, proceedings of international
101 WARE, K., GULLICKSON, R., PIERRE, J., SCHNEIDER, R., and
VITKOVITSKY, I.: Technologies for development of more affordable large
X-ray simulators. Proceedings of 12th IEEE Pulsed power conference, 1999,
pp. 264268
102 MILLER, A.R., GILBERT, C., RAUCH, J., RIX, W., PALKUTI, L., and
WARE, K.: Pulsed power design for a compact X-ray simulator. Proceedings
103 COOK, D.: Z, ZX, and X-l: a realistic path to high fusion yield. Proceedings
104 SARJEANT, W.J., and DOLLINGER, R.E.:High power electronics (TAB
Books, 1989)
Index
0, , model 3213, 326, 335
61.8 per cent rule 3745
Abel inversion 549
AC/DC ratio 4823
accelerated ageing tests 21213
accidental earth circuit 3859
accidental path 391
acoustic detection of partial discharges
16770
acoustic velocity 169
active gap arresters 193
ageing 496503
agricultural pollution 296
AI diagnostic techniques 723
air 1456
density 28, 30, 111
refractivity 555
relative dielectric constant 601
air breakdown 13
atmospheric effects 289
breakdown processes 23
field configuration effects 258
see also sparkover
air circuit breakers 428
air density 28, 30, 111
air insulation 310
air termination 104, 11314
air treatment 61617
aircraft, lightning strikes 12830
apparent resistivity 360, 3624
arc control 427
arc optical emission probes 4624, 467
arc quenching 417
arc quenching media 4267
arc suppression coil earthing 354
arresters: see surge arresters, ZnO surge
arresters
Arrhenius equation 214
artificial neural networks 723
ash treatment 61920
asset management 304, 434, 4567
attachment 2, 4, 14, 38
attachment coefficient 45, 8, 39
attraction area 1045
attraction radius 104, 1078, 118
autoreclosures 429
avalanche, critical 812, 41
avalanche development 47, 1011, 41
air density effects 28
avalanche length 7
avalanche mode switching 6078
avalanche number 4, 8
avalanche properties 68
avalanche radius 67
backflashover 314
ballast water treatment 618
barium titanate
barrel insulators 2713
basic insulation level 45, 105
biological fouling 618
bipolar junction transistors 6078
Blumlein circuit 6023
bond scission 140, 142, 147
booster sheds 2967
borehole data 358
boundary-element method 367, 51315
breakdown voltage 42
busbars 343
634 Index
bushings 169, 520, 5256, 52831
partial discharge measurement
techniques 179
stress 5301
cable design models 5003
cable end boxes
techniques 179
cable failure 491500
contaminants 4924
dry ageing 498500
electrical treeing 4946
voids 494
wet ageing 4968
cable insulations 4802
breakdown 492500
cable manufacture 48891
cable mapping 17982
cable testing 5057
prequalification 505
routine 5067
sample 506
type 5056
cables 47781
AC 47780
DC 480, 5035
distribution 4778
electrical stresses 4789, 481
high pressure fluid filled 480
mass impregnated non draining 480
overhead 479
paper-insulated 4801
techniques 17982
polymer insulated 4808
self contained fluid filled 480
simulation 326
structure 4778
testing 5057
types 4801
underground 353, 36970, 47980
underwater 479
uses 479
voltage ratings 4778
see also polymeric cables
CAD 540
cameras 170, 5602, 6078
cap and pin insulator 2618
AC energisation 2648
aerodynamic design 261
anti-fog design 261
cracks 2648
DC energisation 2624
electrolytic corrosion 2634
figure of merit 2801
glass shattering 2623
transient overvoltage 28990
withstand stress 291
capacitor charger 596
capacitor coupler 1512
capacitor discharge circuit 5956
sharpening gap 5956
capacitors 169, 5957
techniques 179
resistive charging circuit 596
catenary continuous vulcanisation 48991
cellular phone systems 380
centres of advanced intelligence 436
ceramic
surface wettability 2734
chain impedance 352, 3679
chain resistance 369
charge iteration method 524
charge line circuits 6003
chemical partial discharge detection
1708
chemical stress 1459
chromatic mapping 587
chromatic modulation approach 572
Church Nomograph method 174
circuit breaker boxes
techniques 179
circuit breaker restrikes 228
circuit breaker testing 472
circuit breakers 288, 41527
break time 425
characteristics 41920
components 424
condition monitoring 45672
current waveforms 416, 418
mechanisms 425
nozzle wear 5834
optical fibre measurements 5824
reliability 4234
simulation 326
vacuum 41819, 4213
voltage waveforms 41617
see also air circuit breakers, SF
6
circuit
breakers, vacuum interrupters
circuit interruption 415, 417, 424, 428
circuit switches 471
clean fog test 2834, 288
coastal pollution 296
Index 635
cold plasmas 594
collection area 104, 118
collection cone 1223
collection distance 104, 119
collocation method 517
compact transmission lines 303, 311,
3404
busbars 343
electric field profile 3434
insulation 3412
magnetic field profile 3434
compulsators 596
concentrated earth electrodes 398
concrete fragmentation 619
arc optical emission probes 4624, 467
continuous 461, 463, 465
optical fibre sensors 465, 4678
optoacoustic vibration monitoring 466,
468
partial discharge 461, 463, 4657
conductor-plane gap 27
conductor-rod gap 27
continuous operating voltage 2345,
2389
controlled switching systems 4457
copper vapour
refractivity 555
core bolts 169
corona 1315, 701
DC 1456
low air density 30
negative 20
corona precipitator 617
corona reactors 61617
corona stabilised breakdown 469
creepage distance 293
critical avalanche 812
critical field 8
critical reduced field strength 40
critical volume 811
crosslinked polyethylene: see XLPE
crosslinking 4857, 489
cumulative probability function 31011
curing 4856, 48991
current chopping 315
current interruption 415, 4223, 4278
dielectric recovery phase 428
thermal recovery phase 427
Dakin-Arrhenius life plot 21415
data engineering 443
daylight corona camera 170
DC transmission 5035
dead-tank designs 449
Debye length 67
defect recognition and decision support
45860, 462
delayed breakdown device 61011, 615
Dellera and Garbagnati model 1078
depolymerisation 143
deregulation 434, 4546
desert conditions 258, 296, 301
design optimisation 542
detachment 5, 10
detachment coefficient 5
diagnostic techniques 46870, 4723
dielectric barrier discharge systems 619
dielectric barrier technique 594
dielectric performance 448
dielectric strength 389, 51
dielectric systems 4515
diesel particulate cleanup 61617
diffusion theory 6
digital imaging systems 585
directed energy weapons programmes
5934, 598
directional permittivity 521
discretisation, automatic 541
dissolved gas analysis 1717
dissolved gas extraction and measurement
1712
distribution circuit breaker box 170
distribution network 43640
distribution switchgear systems 42834
distribution system
cables 4778
earthing 3523
drift step recovery diode 604, 61011
dry ageing 498500
dry band 2745
dry salt layer method 283
Duvals triangle 174, 178
dynistor 61012
earth electrode potential 349
earth electrode systems 3968
soil ionisation effects 3968
earth electrodes 367
earth fault current 349, 352, 371
earth grid 38990, 4002
earth grid current 389
earth impedance 349, 3523, 36770
background noise 3769
measurement 3729
reducing 38990
636 Index
earth models 3626
1D model 3656
2D model 3656
3D model 3656
multilayer 365
threelayer 363
twolayer 364
earth potential 373
earth potential rise 34950
earth resistance 350, 367
measurement 3734
earth resistivity 35566, 38990
measurements 35861
moisture effects 356
salt effects 356
temperature effects 356
earth resistivity maps 3613
earth return current 371
earth wire 3512, 367, 389
earthing 34951
system 3535
transient 3945
earthing systems 34950
dimensioning 38991
distribution system 3523
impulse performance 394402
models 398400
performance simulations 4002
power frequency performance 36580
safety 38994
standards 3656
testing 3778
transmission system 3512
electric wind 140
electrical breakdown 53940
electrical discharge measurement 57581
electrical discharges
electron density 551
species concentration 5512
electrical field distribution 51112
electrical power systems 31112
electrical stress 149, 4789, 481, 5268
electrical stress distribution 53940
electrical stress enhancement 4923
electrical treeing 4946
electricity consumption 456
electricity market 4345, 440
electricity supply industry 4546
electrocution hazards 3809
limiting 38991
electrode temperature 5723
electromagnetic compatibility 127
electromagnetic model 400
electron density 551
electron temperature map 575, 577
electron-ion interactions 141
electron-molecule interactions 1401
electronegative gas 5
electronic equipment protection 1268
protector devices 128
electronic refractivity 5534
electro-optic switching 612
electro-optic voltage transducer 5712
electroporation 61720
electroshock 620
electrostatic field modelling 516
embedded generation 440
end of life criterion 21315
energy absorption capability 239
energy management systems 443
environmental cleanup 594, 61617
environmental factors 3401
EPDM insulators
figure of merit 281
epoxy resin 1479
chemical bond energies 142
EPR: see ethylene propylene rubber
equivalent salt deposit density 279,
2867, 295
Eriksson model 1078
ethylene propylene rubber 4802
insulator figure of merit 281
ethylene-vinyl silane copolymers 486
Eulers theorem 517
excitation 2
exhaust fume remediation 61617
explosion, electrical 620
Fabry-Perot cavity 559
Fabry-Perot cavity interferometer 5723,
575
Fabry-Perot optical fibre probe 5812
Fabry-Perot pressure sensor 573, 584
fall-of-potential technique 3735
Faraday sensor 5701
fault clearance time 352
fault currents 3702
distribution 3712
fault gas detectors 172
fault gas ratios 1727
feet earth resistance 387
Ferranti effect 316, 318
ferroresonance 31617
fibre-addressed chromatic scale 575
field ionisation dynistor 610
Index 637
filamentation 612
finite-difference method 51213, 515
finite-element method 51319
applications 52537
first-order 536, 538
foils 51921
functional equation 517, 5213
higher-order 536, 5389
open boundary problems 524
singularities 53940
time variation 524
finite-element/boundary-element methods
524
flash radiography 593
flashover 126, 257, 259
ice and snow 2789
polluted insulators 2729
see also pollution flashover
flashover voltage 277
floating electrode 712
foils 51921, 5256, 528
food processing 61819
footwear resistance 386
Franklin, Benjamin 778
free conducting particles 52
free electrons 2
free metallic particle 701
free space monitoring 5456
free space techniques 585
furans 177
fusion 593, 615, 622
Galerkin method 515, 517
gamma ray irradiation 618
gap factor 258
gas blast interrupters 44951
gas breakdown 142
gas cleanup 594
gas discharge reactors 616
gas ratios 1727
gas-insulated lines 65, 41718, 448
gas-insulated substations: see GIS
gas-insulated switchgear 417, 4489
generators 159
techniques 179
geological maps 358
GIS 423, 45, 512
leader breakdown 4950
partial discharge diagnostic techniques
527, 1813
partial discharge monitoring 461, 463,
4657
particle-initiated breakdown 501
streamer-controlled breakdown 489
UHF method 569, 6574, 1823
UHF signals 5965
glass 259
compressive strength 259
resistivity 2623
tensile strength 259
global warming 440, 448
see also greenhouse gases
greenhouse gases 448, 577, 619
grid mesh density 390
ground fault neutraliser earthing 354
ground faults 316
ground flash density 7981
ground penetrating radar 358
ground potential rise 349
ground strikes 1023, 125
ground termination 10911
heart current factor 381, 385
heart fibrillation 3845
Hermes system 6023
high density polyethylene 480
high power microwaves 621
high speed imaging 548, 558, 5602
high speed photography 548, 554, 5756
hollow core insulator 2723
hot line washing 300
hot zone 350
human body resistance 3867
humidity 289, 111, 113
HVDC 5035
hybrid camera 560
hybrid optical current transformer 5467,
5812
hydrophilic insulators 2868
hydrophilic surface 258, 2748
hydrophobic surface 25860, 278, 283
ice 258, 278, 2912, 300
IEC Code 599 173, 176
IEC Code 60599 173, 1778
impulse tests 21619
independent power producer 440
independent system operator 443
inductive energy storage 6045
inductive voltage adder 6024
industrial pollution 296
inertial confinement fusion 593, 615, 622
inland pollution 296
638 Index
insulating crossarms 3412
insulating materials 25960
see also cable insulations
insulating system, contaminated
modelling 52731
insulation breakdown 30910
cables 492500
insulation coordination 30911, 33340
risk of failure 3356, 338, 340
withstand voltage 33940
insulation monitoring 524
see also partial discharge detection, UHF
method
insulation withstand voltage 311
insulator coatings 2989
insulator selection 2926
insulator washing 300
live 300
insulators, outdoor 2578
AC versus DC 28990
audible noise 302
barrels 2713
cap and pin 2618
cold switch-on 3001
costs 3034
electric field distributions 3023
electrical characteristics 27992
electromagnetic compatibility 302
flashover mechanisms 2729
iced 2912
longrods 2689
materials 25960
pollution 2734
pollution tests 27988
posts 26872
role 2589
selection 2926
semiconducting glaze 301
snow 292
thermal lag 301
withstand stress 291
intelligent electronic device 435, 442
intelligent information extraction systems
5867
intelligent networks 43441
control 439
management 439
interferometry 556, 559, 5645
results 5779
interphase spacers 303, 341
interrupter head porcelains 2889
interrupters
gas blast 44951
oil 464
SF
6
puffer-type 4523
vacuum 419, 423, 4289, 44950
interruption: see circuit interruption, current
interruption
ion beam materials treatment 616
ionisation 2, 45
SF
6
3840
ionisation coefficient 45, 3940
juice extraction 61819
kicker pulsers 609
Kyoto Protocol 448
L1 limits 1723
Laplaces equation 516
large area breakdown formulae 593
laser Doppler velocimetry 553, 555
laser driver circuits 600
LC pulse compression circuit 5989
leader breakdown 4750
leaders 1520
average electric field 19
charge injection 19
development 1518, 48
humidity effects 29
negative 21
velocity 18, 21
leak resistors 315
leakage current 257
leakage current analysis 2335
Lichtenberg figure technique 31
light scattering 551
coherent 5538
incoherent 5515
light scattering systems 5647
coherent 5645
incoherent 564, 5667
lightning 312
positive 11417
lightning conductors 11314
lightning discharge 210, 212
lightning electromagnetic pulses 86, 126
lightning flash
current shape 845
electric fields 857
electromagnetic fields 867
electrostatic field changes 86
incidence 7982
Index 639
mean duration 83
multiple strokes 83
peak current 834
polarity 82
positive 11417
side flashes to a mast 1223
spatial development 878
upward 7980
lightning impulse voltage 223
lightning leader propagation 97100
charge density 979
current 979
return stroke current 99100
return stroke velocity 99
velocity 979
lightning leaders 8897
channel temperature 923
downward 109
energy dissipation and storage 912
gradient 8994
mechanisms 889
negative space leaders 947
spark leaders 89
upward 1089
velocity 93
Lightning Location System 84
lightning outages 119
lightning overvoltages 118, 125, 31214
Lightning Position and Tracking System
80
lightning protection 113, 1212, 395
electronic equipment 1268
overhead power lines 1256
lightning risk 11625
electrogeometric calculation 11920
negative flash 1205
risk factor 11625
tall structures 1235
lightning rod 778, 104
lightning standards 1012, 104
lightning strikes
aircraft 12830
space vehicles 130
lightning termination at ground 100
see also striking distance
LINX 603
liquid earthing resistor 354
live insulator washing 300
live line working 3012
live-tank designs 448
load rejection 316, 318
lock-on 612
long land die 489, 491
long sparks 8897
longrod insulators 2689
low density polyethylene 480
magnetic pulse compression 592,
599600
magnetic pulse compressor 599600, 612,
61617, 623
magnetic storms 314
magnetisation reversal mode 613
magnetooptic current transducer 570
magnetron 591
main earth grid 351
marine pollution 280
market liberalisation 4546
Martin, J. C. Charlie 593
Marx, Erwin 591
Marx bank 591, 593, 5978
avalanche transistors 6078
MOSFET 608
rise time 622
thyristor 6067
Marx generator 5912, 597
Maxwell-Wagner model 223
mechanical stress 1445
medical applications of pulsed power 620
medium voltage switchgear 429
Melville line 599600
mesh voltage 3823
metal oxide varistor 128
mica 1467
micro particles scattering 5667
Mie scattering 5534, 584
Milliken conductors 4834
Mitsubishi Dainichi continuous
vulcanisation 489, 491
mobile communication base stations 350
modal model 324, 326
moisture cure 4856, 489
monitoring
circuit breakers 45672
GIS 461, 463, 4657
insulation 524
monosil 486
MOSFETs 60810
inductive coupling 60910
switching stacks 6089
Motorformer 479
motors 159
techniques 179
640 Index
multicolumn surge arresters 2412
mutual coupling effect 3756
National Lightning Detection Network 80
negative discharges 201
neon tube method 219
network analysis 399400
network protection 441
network reduction 327, 329
network services 437
networks, intelligent: see intelligent
networks
neutral 3534
neutral lead 51112
nominal discharge current 236, 239, 241
non-linear capacitance 614
non-linear transmission lines 61315
discrete element 61315
non-nuclear electromagnetic pulses 314
non-soluble deposit density 279, 2867,
295
non-uniform field breakdown
SF
6
458
nuclear electromagnetic pulse hardening
621
nuclear electromagnetic pulses 314
nuclear weapons 5924
numerical analysis of electrical fields
51112, 5402
oil
oil interrupter 464
oil-filled equipment 1712
open ring network 43740
optical fibre homodyne interferometer
575
optical fibre sensing 5467, 55860, 566,
56875
chromatic 560
extrinsic 566, 568
hybrid 566, 570
interferometric 559
intrinsic 566, 568, 575
optical fibre sensors 465, 4678
transient response 5856
wavelength dependence 585
optical fibre transducers 56875
discharge radiation 572, 574, 5824
electric current 56871, 5812
mechanical parameters 5745
particulates 5724, 5824
results 5806
temperature 572
voltage 5712
optical fibres 5579
multimode propagation 557
optical intensity images 5478
optical measurements 5457
optical monitoring 5857
Optical Transient Detector system 812
optoacoustic fibre monitoring 5845
optoacoustic vibration monitoring 466,
468
optoelectronic camera 5602
optomechanical camera 5602
optomechanical fibre sensors
results 5845
optomechanical transduction 574
ore treatments 620
overhead busbars
techniques 179
overhead line shielding 1056, 119, 1256
overhead lines 2434, 2589, 281, 372,
481, 5258
compact 303
electric field modelling 5258
lightning 31314
overvoltage 30918
classification 313
external 31214
internal 31418
lightning 309, 31214
non-nuclear electromagnetic pulses 314
nuclear electromagnetic pulses 314
statistical 338
steady state 318
temporary 309, 31618
see also switching overvoltages
overvoltage protection 1925
ozone 1456
paper-insulated cables 4802
partial discharge 54, 139, 1834
discharge current 157
discharge energy 156
discharge intensity 157
discharge phase extinction voltage 156
discharge phase inception voltage 156
discharge power 157
quadratic rate 157
-q-n pattern 1557
partial discharge data display 701
Index 641
partial discharge degradation mechanisms
13950
chemical stress 1459
electrical stress 149
particle impact stress 1402
thermal stress 1434
partial discharge detection 623
in GIS 6574
partial discharge diagnostic techniques
528
acoustic emission 55, 57
chemical byproducts 545, 57
electrical methods 567
light output 54
UHF method 569
partial discharge measurement 15083
acoustic 16770
antenna techniques 1645
capacitive probe techniques 1657
chemical 1708
electrical 15067
loss measurements 1635
noise suppression 1613
pulse measurement 1508
thermography 170
see also partial discharge pulse
measurement
partial discharge monitor system 69
AI diagnostic techniques 723
service experience 734
partial discharge pulse measurement 1508
calibration 15961
capacitor coupler 1512
current transformers 1501
pattern interpretation 1537
transducers 1512
partial discharge pulse sequence analysis
158
partial discharge signal classification
723
partially conducting layers
modelling 5279
partially conducting regions 5223
particle control 52
particle impact stress 1402
particle-initiated breakdown 501
Paschen field 494
Paschens law 30, 42, 154
PD: see partial discharge
peroxide cure 4856, 489
Petersen coil earthing 354
Petrov and Waters model 10813
phase model 31920, 326
photochemical oxidation 619
photoconductive switching 612
photodetachment 5, 11, 13
photodetectors, transient response 5856
photoelectrons 10, 31
photographic image intensity 5612
photoionisation 11, 141
photomultiplier detection 563
photon-ion energy transfer 1412
photon-molecule energy transfer 141
piezoelectric polymers 167
piezoelectric sensors 1678
plasma erosion opening switch 604
plasma jet monitoring 587
plasma light scattering 564, 5667
Pockels cell 612
Pockels effect 220, 571
Poissons equation 516, 523
polarisability 553
pollution 2578, 280, 296, 527
pollution control 61617
pollution deposit density 2956
pollution flashover 260, 2729, 289, 296
artificial pollution 2838
interrupter head porcelains 2889
natural pollution 27983
selection of insulators 2926
pollution levels 293
pollution tests 27988
artificial pollution 27980, 2838
natural pollution 27983
polyethylenes 143, 149
polymeric cables 480
conductors 4814
DC 5035
electrical breakdown strength 4923
EPR 4802, 4978
failure processes 491500
HDPE 480
LDPE 480
manufacture 48891
mathematical design models 5003
metal sheath 4878
oversheath 488
semicon 484
testing 5057
XLPE 4802, 4847, 4934, 4989,
5013
polymers 259
glass transition temperature 144
642 Index
polymers (Contd.)
thermal oxidation 144
thermal stress 1434
polymethylmethacrylate 143
polyvinyl chloride 143
porcelain 259, 2648
porcelain insulator
salt fog test 2834
positive lightning 11417
post insulators 26872
cracks 26870
multiple cone 26870
post processing 5412
potential ramps 390
power cables 47781
see also cables
power modulator 591
power systems 31112
Powerformer 479
Preece 78
pressure relief 237, 239
problem geometry 5401
protection area 1045
protection distance 104
protective earth 349
protective levels 2378
lightning impulse current 237
steep current 2378
switching impulse 237
proximity effect 482
puffer-type interrupters 4523
pulse compression 5945, 598602
magnetic 592, 599600
pulse forming 595
pulse sharpening 5956, 6012, 611,
61315
pulse transformer 60910
pulsed corona 594
pulsed corona precipitator 617
pulsed corona reactors 61617
pulsed power 5914, 6223
applications 5924, 61522
pulsed power circuits 6005
charge lines 6002
transmission line 6001
radio interference 302
radio interference measurement 1501
radio interference noise 164
radiographic machines 603
radomes 129
railways 258, 376, 378
electrification systems 350
rain 257, 274, 278, 281
ramp rings 3901
rated voltage 2356, 2389, 241
Rayleigh scattering 552, 558
reactance earthing 354
recombination 2
recycling 619
reduced electric field 3
refractive index 5535
regeneration 23
relative air density 289
relative line intensity method 550
remote terminal units 442, 445
renewable energy sources 440
resin-bonded glass fibre 25960
brittle fracture 260
resistance earthing 3545
resistance to earth of standing point 387
resistive charging circuit 5956
resistive/capacitive systems 5212
resonant earthing 3545
restricted land area 350
reverse switching dynistor 61012
RF weapons 621
right of way 303
rise of earth potential 349
riser-pole arresters 194
risk factor: see lightning risk
risk management
earthing safety 3914
risk of failure 311, 3356, 338
risk of flashover 118
Rizk model 108
rock treatments 620
rocks
permittivity 357
resistivity 3557
rod earths 4001
rod gaps 192
rod-plane gap 237
SF
6
46
rod-rod gap 13, 23, 27
Rogowski coil 1523
rolling sphere method 1013, 111
roughness factor 434
rural distribution system 4313
Index 643
safe voltages 350
safety 38994
risk management 3914
salt fog test 258, 2834, 286
saturation effects 316
scattering
coherent 5779
incoherent 57881
Schering bridge 164, 233
Schlieren photography 902, 5568,
5645
Schlumberger-Palmer method 361
Schonland, Basil 78
seismic activity 271
seismic surveying 358
semicon 4778, 484
semiconducting glazed insulators 301
semiconductor switching 60413
design 610
sewage treatment 620
SF
6
additives 51
greenhouse gases 448
ionisation 3840
vapour pressure 37
SF
6
circuit breakers 425, 4278, 471
dead-tank 449
live-tank 448
puffer 462
spinner 463
SF
6
insulation systems 378
breakdown mechanisms in low
divergence fields 405
coaxial electrodes 423
improvements 512
non-uniform field breakdown 459
streamer breakdown 41
surface roughness effects 435
SF
6
leakage detection 5767
SF
6
puffer circuit breaker 462
SF
6
puffer-type interrupters 4523
SF
6
spinner circuit breaker 463
shadowgram 5789, 581
shadowgraphy 5567, 5645
shape function 518
sheath voltage limiter arresters 194
shed extender 2978
shed protector 2989
shielding angle 1046, 1212
ships 618
shock duration 3845
shock line 61315
short circuit ratings 4245
shunt reactors 315, 318
shuttering 560
SiC material
EJ characteristics 221
silicon opening switch 6045, 61112
silicone rubber 25960
high temperature vulcanised 259
hydrophobicity 25960
pollution 279
room temperature vulcanised 260
silicone rubber coating 2834
RTV 299
silicone rubber insulators 2846
clean fog test 284
figure of merit 2812
hydrophobicity transfer 2846
performance 304
Sioplas 486
site pollution severity 295
skin effect 65, 482
slope method 3745
slow front impulse voltage 22, 245
snow 258, 2789, 292, 300
soil ionisation 3958
critical electric field intensity 397
mechanisms 3978
soils
permittivity 357
resistivity 3557
solid earthing 3545
solid layer method 283
soliton generation 615
space charge field 78
space charge modelling 523
space stem 11, 201
space vehicles
lightning strikes 130
spark discharges 61920
spark gaps 192, 5934
sparkover
atmospheric effects 289
gap factor 258
insulator surfaces 312
lightning impulse voltage 223
slow front impulse voltage 245
spectrometers 5614
arc studies 563
spectroscopy 54751
absorption 549, 5634, 5767
emission 549
644 Index
spectroscopy (Contd.)
results 5757
temperature dependence 5501
static induction thyristor 6067
statistical breakdown voltage 339
statistical coordination factor 338, 340
statistical overvoltage 338
statistical withstand voltage 338, 340
steady state overvoltage 318
step potential 3803, 390
sterilisation 61719
stiffness matrix 518
streak cameras 6078
streak photograph 547, 5578
streak spectra 563
streamer constant 41
streamer trail 14
streamers 1013, 1516, 41
air density effects 29
humidity effects 29
negative 11, 201
retrograde 21
velocity 13, 18
stress grading materials 5307
loss density 5356
resistivity 5312
stress grading tapes 5223
stressed volume 540
strikes to a mast 111
striking distance 78, 100
altitude effects 11113
electrogeometric models 100, 1026
generic models 100, 10714
geometric models 1003
positive lightning 11417
side flashes 1223
to a mast 11112, 1202
to a plane ground 10912
strioscopy 90
substation control system 4415
availability 444
substation layout 42935
double busbar 42935
primary distribution 42930
rural distribution system 4313
urban distribution system 4302
substations
fences 390
voltage hazards 390
sulphur hexafluoride: see SF
6
supervisory control and data acquisition
443
surface element formulation 5223
surface treatments 620
surge arrester classification 241
surge arrester protection 2424
distance effects 2423
protective zones 2434
separation distance 2423
surge arrester ratings 241
surge arresters 1914, 315
active gaps 193
compact transmission lines 341
metal oxide 1934
polymeric 217, 21920
VI curve 219
see also ZnO surge arresters
surge mitigation 127
suspension unit 259
switchgear 423, 44851
dead-tank 449
dielectric system evaluation 4515
distribution 4289
live-tank 448
switching equipment testing 472
switching overvoltages 309, 31416
computed 32735
cumulative density distribution 3323,
336
density distribution 3323, 336
earth resistivity effects 32935
mean values 337
pole closing time effects 32935
trapped charge effects 32936
switching resistors 315
switching stacks 6089
switching surges 20911, 31416
system control scheme 442
system earths 349, 3535
tan 1634
TE modes 61
electric field 62
television interference 302
telluric currents 359
TEM modes 614
electric field 62
temporary overvoltage 21011, 31618
Tesla transformer 605
testing
cables 5057
circuit breakers 472
thermal dissipation factor 202
Index 645
thermal imaging camera 170
thermal runaway 2048
critical condition 2056
glass insulators 2623
thermal stress 1434
thermally simulated currents method 207
thermoelectric ageing 498500
Thomson scattering 552
thyristor 6067
time-stepping technique 5357
TM modes 61
electric field 62
tolerable body currents 3845
tolerable touch voltages 3878
tolerable voltages 384
touch potential 3806
limiting 3901
tolerable 3878
touching shock 386
tower footing resistance 3689
tower lines 3679, 372
effective length 367
towers 380
Townsend 3
transfer potential 3801, 3879, 391
transformers 159, 167, 169
techniques 179
simulation 3278
transient control level 128
transient earth voltage 1656
transient earthing 3945
models 398400
performance 4002
standards 3945
transient electromagnetic technique 361
transient overvoltage 28990, 31416
transmission cables 4778
transmission line circuits 6001
transmission line models 31826
0, , model 3213, 326
bundled conductors 325
corona losses 3245
earth resistivity 3201
line parameters 319
modal model 324, 326
phase model 31920, 326
trapped charge 326
transmission line segments 399400
transmission lines 5258
compact 3404
non-linear 61315
transmission networks 30911
simulation 31827
transmission system earthing 3512
U-curve 246
UHF couplers 659
calibration system 667
disc 678
external 669, 73
internal 66, 73
sensitivity 668
UHF method 569, 6574
couplers 659
online system 73
partial discharge data 701
UHF signals in GIS 5965
attenuation 645
excitation 59
extraction 601
propagation 60
waveguide modes 614
UHV coupler 4589
UK Lightning Flash Location System 80
ultracapacitors 596
ultrawideband radar 598, 623
ultrawideband signals 621
underground cables 353, 36970, 47980
earthing 36970
EHV 498
types 480
underwater cables 479
uniform field breakdown
SF
6
414
urban distribution system 4302
urban encroachment 350
UV solar blind imager 170
vacuum circuit breakers 41819, 421
vacuum interrupters 419, 423, 4289,
44950
magnetic actuators 429
vacuum recloser 421, 44950
vacuum switchgear 419, 4213
valve-type arresters 192
varistor degradation 207
varistor stability 207
ventricular fibrillation 3845
Verdet constant 5701
vertical continuous vulcanisation 48990
vertical earth electrode 399
vertical mast 1035, 11112
voltage distribution measurement 21820
646 Index
voltage multiplication 597
voltage source converter technology 505
vulcanisation 48991
water
water purification 619
water treeing 477, 4968
waveguide modes
coaxial waveguides 614
wavelet analysis 1613
wayleave 303
Weibull probability functions 3379, 478
weighted residual method 517
Wenner method 35860
wet ageing 4968
wettability 2724
white quantum sources 585
wind farms 350
Windformer 479
windings 169, 5307
withstand voltage 33940
statistical 338, 340
x-ray simulators 622
XLPE 4802, 4847
cleanliness 487
moisture cure 4856
peroxide cure 4856
relative permittivity 494
Young-Dupre equation 273
Z machine 622
zinc oxide: see ZnO
ZnO 1959
additive effects 1956
electrical properties 196
microstructure 1979
VI curve 224
ZnO material
dielectric constant 222
dissipation factor 222, 224
EJ characteristics 221
equivalent circuit 2302
frequency effects 2224
resistivity 223
temperature effects 222
ZnO surge arrester characterisation
21520
AC tests 21517
DC tests 21516
impulse tests 21619
voltage distribution measurement
21820
ZnO surge arrester characteristics 22132
AC + DC 2267
AC + impulse 2289
combined stress response 2259
continuous current 235
continuous operating voltage 2345,
239
crossover of DC and AC characteristics
2267
DC + impulse 228
discharge current 235
energy absorption capability 239
equivalent circuit 2302
frequency response 2214
high current impulse 236
impulse response 2245
multiple impulse response 229
nominal discharge current 236, 239,
241
pollution withstand 238
power frequency withstand voltage
versus time 238
protective characteristic 2378, 2424
protective levels 2378
rated voltage 2356, 239, 241
residual velocity 236
switching current impulse 236
voltage overshoot 2246
ZnO surge arrester selection 23942
ZnO surge arresters
accelerated ageing tests 21213
ageing 204, 2089, 21214
applications 194
classification 2367
degradation 204, 20712
design improvements 1945
destruction mechanism 21011
dynamic stability 206
fabrication 1957
heat dissipation capability 2023
identification 2389
life estimation 21215
line discharge class 2367
monitoring 2335
multicolumn 2412
Index 647
pollution effects 209, 218
pressure relief class 237
reference current 238
reference voltage 238
standard tests 238
standards 2346
surge absorption capability 20910
thermal model 2067
thermal runaway 2048
thermal stability 2013
see also ZnO surge arrester
characteristics
ZnO varistors 193, 1956
conduction mechanism 199201
high conduction model 2001
zone of protection 78
-q-n plot 1557, 180, 183
Doug Warne graduated from Imperial College with an
apprenticeship at AEI Heavy Plant Division in Rugby. At the
Electrical Research Association he took responsibility for the
short-circuit testing station and the HV laboratory, and also the
development and management of ERAs wind energy research.
Since 1990 Doug has acted as an independent consultant.
He also runs the independent NETCORDE syndicate for better
industryuniversity collaboration in electrical engineering and
he co-ordinates the ESR (Electricity Supply Research) Network.
Dr A. Haddad obtained the degree of Ingnieur dEtat in
Electrical Engineering, followed by a Ph.D. in High Voltage
Engineering in 1990. In 1995 he was appointed as a Lecturer
and then Senior Lecturer in electrical engineering at Cardiff
University, with responsibility for the High Voltage Research
Group and lecturing duties in high voltage engineering, power
systems, electromagnetics and electrical circuits.
Advances in High
Voltage Engineering
It is several years since any book has attempted to cover the
range of issues in high voltage research and development.
The area continues to be full of challenge and scope and this
book focuses on developments in experimental methods,
theory, modelling and HV technology through the past decade.
The coverage includes advances in basic understanding and
capability, for instance in earthing, numerical analysis, optical
methods, the physics of air breakdown and partial discharge.
It also addresses technological developments in key areas
such as SF
6
insulation systems, polymeric insulators and power
cables, ZnO surge arrestors, and pulsed power. The unique
blend of reputable contributors and comprehensive subject
coverage makes this book an ideal reference source for
engineers and researchers in the field for many years to come.
www.theiet.org
0 85296 158 8
978-0-85296-158-2

High Voltage Engineering PDF

Uploaded by

Copyright:

Available Formats

High Voltage Engineering PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

High Voltage Engineering PDF

Uploaded by

Copyright:

Available Formats

IET Power and Energy Series 40

, with resulting changes in

This process competes with that of collisional ionisation, by which an electron

components to their electric fields, and TM modes have E

corresponds to the positive-going zero of the power

to the positive peak, and so on.

N or S, and recent orbital observations suggest that 90 per cent of

cones of attraction are retained only for structures

, the striking distance dependence on the mast height and lightning

(Equation 3.66) does not depend on

whose apex is at the mast

the electron continues on with a lower velocity, v

either emits the extra energy as a photon:

forming a negative ion

142 Advances in high voltage engineering

(iii) splitting a negative ion into neutral and charged species:

(iv) splitting a positive ion into neutral and charged species:

the left-hand side of the reaction continues:

C whereas polymer (2), which

+SiOSi SiO +Si +M

is the energetic particle, or bond scission by cumulative localised heating:

154 Advances in high voltage engineering

), discharge magnitude (usually in pC) and number

is the zero crossing point of

C < T < 300

C < T < 700

C < T < 700

provided the plant item is air-cooled (i.e. has access vents)

provided the plant item is oil filled

C. This heat treatment has

C/h at the end of which

heat loss at ambient temperature T

power (heat) dissipation at voltage V

power (heat) dissipation at voltage V

C and m is the mass of the

C if the arrester is operated at an alternating voltage which produces

which has a new operating point.

C for one day degrades

C) degrades these arresters at a faster rate. For an applied

C, the main additive Bi

C). The current I or

C were significantly lower than those expected by extrapolation of

Figure 5.17 Permittivity dependence on frequency in ZnO material. (Courtesy

before the crest volt-

for ceramic and about 100

to the vertical; angled is about 45

C. In such tests, the

to the vertical in lieu of

are related by the lossless line equations [23]:

are obtained in a form suitable for computation as [25]:

These constraints are satisfied by an interconnection of the 0, , networks at the

and travel times T

and a minimum value of 1.2 p.u.

. Curve b shows the effect of a trapped charge represented by an initial

with the minimum value being 1.0 p.u. at = 180

100 20.0 25.0 23.3 1 1 1

includes surge arrester at line termination

includes surge arresters at line termination

circuit model computer model