Technical Brochure: Network Modelling For Harmonic Studies

C4/B4
Technical Brochure
Network modelling for harmonic

studies
Reference: 766
April 2019
TB 766 - Network modelling for harmonic studies
Network modelling for
harmonic studies
JWG C4/B4.38
Contributing Members
M. VAL ESCUDERO, Convenor G. LIETZ, Chapter Lead DE

IE
and Chapter Lead and main Editor
Z. EMIN, Chapter Lead UK C. F. JENSEN, Chapter Lead DK
L. SHUAI, Task Lead DK K. LORENZO, Task Lead FR
F. CEJA-GOMEZ, Task Lead CA L. KOCEWIAK, Task Lead DK
N. SHORE UK K. LEONG KOO UK
T. MARTINICH CA D. O C BRASIL BR
D. ARLT DE F. BARAKOU NL
C. BUCHHAGEN DE A. SCHWOB FR
D. VUJATOVIC UK L. SOTO CANO ES
B. KHODABAKHCHIAN CA A. CASTRO LOBATO ES
O. LENNERHAG SE Y. SUN NL
M. NGUYEN TUAN FR I. VALADE IT
X. WU CN D. TING UK
Reviewers
J. GING IE M. HALPIN US
R. A. WALLING US M. BOLLEN SE
J. A. R. MONTEIRO UK R. BERES DK
I. IYODA JP P. WANG CA
C. LETH BAK DK
Copyright © 2019
“All rights to this Technical Brochure are retained by CIGRE. It is strictly prohibited to reproduce or provide this publication in any
form or by any means to any third party. Only CIGRE Collective Members companies are allowed to store their copy on their
internal intranet or other company network provided access is restricted to their own employees. No part of this publication may
be reproduced or utilized without permission from CIGRE”.
Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any responsibility, as to the
accuracy or exhaustiveness of the information. All implied warranties and conditions are excluded to the maximum extent permitted
by law”.
WG XX.XXpany network provided access is restricted to their own employees. No part of this publication may be
reproduced or utilized without permission from CIGRE”.
Disclaimer notice
“CIGRE gives no warranty or assuranceISBN : 978-2-85873-468-9
about the contents of this publication, nor does it accept any
responsibility, as to the accuracy or exhaustiveness of the information. All implied warranties and
conditions are excluded to the maximum extent permitted by law”.
ISBN : 978-2-85873-468-9
4
Executive summary
The issue of harmonic distortion in power systems is not new, with publications dating as early as 1916
dealing with distorted waveforms in transmission lines and their effects on transformers, rotating
machines and telephone interference. Significant efforts were devoted then to investigate, understand
and mitigate their negative effects. Standardisation was introduced to control and limit the amount of
harmonic distortion present in the system. The focus was on large non-linear installations, such as
Electric Arc Furnaces (EAF), smelters, industrial converters, SVCs or HVDC systems, which were
subject to harmonic limits and mitigated their emissions using filters where necessary. As a
consequence, harmonic distortion was not a source of practical problems in most transmission power
systems for many years.
This picture is changing. Power systems globally are experiencing a transition towards decarbonisation
of electricity production through large-scale deployment of renewable energy sources (RES), which are
gradually displacing conventional thermal plant. The connection of RES to the electricity network is
mostly achieved through the use of Power Electronic (PE) converters, which are sources of harmonic
distortion.
It has been observed that many power systems are experiencing an increase in harmonic distortion.
Drivers for this trend include not only the integration of RES, but also increased connection of FACTS
devices, HVDC systems, HVAC cables and general proliferation of PE converters in the demand (e.g.
electric vehicles and domestic appliances). Power quality issues associated with harmonics in power
systems are therefore becoming more pronounced and are driving a new focus towards the need to
undertake detailed analysis at the planning stages in order to ensure adherence to statutory limits.
Performing harmonics studies requires a more advanced skill set and more detailed system models and
simulation tools than those required for traditional planning studies (e.g. load flow and short-circuit
analysis). In general, there is some lack of knowledge when it comes to conducting meaningful harmonic
distortion assessments in modern power systems. Most of the issues stem from the lack of practical
information on modelling electrical plant equipment for such studies. Availability of information and
guidance for such modelling requirements are either scarce or in scattered form, mostly delegated to
appendices of various documents (examples are the short articles in Electra 164 [1] and 167 [2],
published in 1996). Both Electra articles are authoritative in nature but are becoming outdated and
inaccessible. Previous CIGRE publications have tried to bridge the gap to a certain extent but the need
for such studies has increased tremendously in most parts of the world, and hence the need for up-to-
date information on the topic of modelling. In response to this increased interest and to review state-of-
the-art modelling and simulation practices, CIGRE Study Committees C4 and B4 established the Joint
Working Group (JWG) C4/B4.38: “Network Modelling for Harmonic Studies” in late 2014.
This Technical Brochure has been compiled drawing expertise from the JWG members and provides
comprehensive guidelines for practising power system engineers when they need to perform harmonic
distortion assessments. The document covers the modelling of the most common network components
and discusses key features that need to be considered in the assessments. The focus is on practical
aspects of modelling for direct application in the planning process of connecting a new customer to the
power system, or when introducing a change to the system as part of asset replacement or system
expansion. As such, the guidelines are concentrated on frequency-domain modelling for steady-state
AC harmonic analysis in transmission and distribution networks, typically in the range from power
frequency up to the 50th harmonic (2.5 kHz in 50 Hz systems or 3 kHz in 60 Hz systems), consistent
with typical power quality assessments. The approach and modelling guidelines provided are
reasonably valid up to the 100th harmonic if specialized studies are required. These guidelines will be
valuable in the definition of harmonic performance specifications for new HVDC converters, FACTS
devices or other non-linear installations. They will also assist connectees when modelling their
installation to assess or demonstrate compliance with the emission limits provided by the System
Owner/Operator and to investigate and specify mitigation measures such as harmonic filters.
Furthermore, this Technical Brochure can also be used post-commissioning for any incident
investigation or to assist resolution of customer complaints via modelling and analysis.
This Technical Brochure has summarised the state-of-the-art and reflects best practice in modelling for
harmonics studies in power systems. Areas where further research work is required have also been
identified. These include: (i) improved load modelling with focus on PE-based load, e.g. load
characterisation and aggregation; (ii) validation of power converter models with field measurements,
e.g. methods to separate the effect of system harmonic impedance and background distortion from the
converter emission; (iii) summation of harmonic sources, i.e. gather enough information, based on field
5
measurements and theoretical analysis, to propose robust and realistic alternatives to improve the IEC
summation law currently adopted in most harmonics analysis; (iv) improved methods for accurate
aggregation of wind farm components (wind turbine generators, transformers, cables, etc) into a single
frequency dependent equivalent; and (v) develop practical methods for the estimation of background
distortion in meshed network topologies where availability of measurements is limited.
6
Table of contents
Executive summary ............................................................................................................. 5
1. Introduction.............................................................................................................. 17
1.1 Background.............................................................................................................................................. 17
1.2 Scope ........................................................................................................................................................ 18
1.3 Structure................................................................................................................................................... 19
2. Study domain and modelling aproaches ............................................................... 21

2.1 Introduction to Study Domains .............................................................................................................. 21
2.2 Frequency-Domain Methods .................................................................................................................. 21
2.2.1 Frequency Scan................................................................................................................................. 21
2.2.2 Harmonic Penetration ........................................................................................................................ 22
2.2.3 Harmonic Load Flow .......................................................................................................................... 25
2.3 Time-Domain Methods ............................................................................................................................ 26
2.3.1 Application and Limitations ................................................................................................................ 26
2.4 Hybrid Methods ....................................................................................................................................... 27
2.5 Harmonic Domain .................................................................................................................................... 27
2.6 Recommendations................................................................................................................................... 27
3. Classical network element models ......................................................................... 31

3.1 Overhead Lines ........................................................................................................................................ 31
3.1.1 Modelling Considerations .................................................................................................................. 31
3.1.2 Consideration of Skin Effect .............................................................................................................. 33
3.1.3 Representation of Line Transposition and Inter-Circuit Coupling ....................................................... 34
3.1.4 Variation of Earth Resistivity with Depth ............................................................................................ 35
3.1.5 Examples: Single 220kV Overhead Line Circuit ................................................................................ 36
3.1.6 Average Conductor Height above Ground ......................................................................................... 41
3.1.7 Summary ........................................................................................................................................... 42
3.2 Cables ....................................................................................................................................................... 44
3.2.1 Factors Influencing the Harmonic Impedance of Cables ................................................................... 44
3.2.2 Models ............................................................................................................................................... 45
3.2.3 Impedance and Admittance Formulae ............................................................................................... 46
3.2.4 Summary ........................................................................................................................................... 50
3.3 Power Transformers ................................................................................................................................ 51
3.3.1 Power Transformer Models ............................................................................................................... 52
3.3.2 Comparison of Power Transformer Models and Measurement Results ............................................. 55
3.3.3 Comparison of Power Transformer Models in a Transmission Grid................................................... 61
3.3.4 Summary ........................................................................................................................................... 63
3.4 Loads ........................................................................................................................................................ 66
3.4.1 Harmonic Load Models ...................................................................................................................... 66
3.4.2 Examples of Load Modelling Approaches in Real Power Systems .................................................... 77
3.4.3 Comparison of Load Models in a Test Grid ....................................................................................... 85
3.4.4 Summary ........................................................................................................................................... 94
3.5 Synchronous Generators........................................................................................................................ 95
3.5.1 Models ............................................................................................................................................... 95
3.5.2 Comparison of Synchronous Generator Models in a Transmission Grid ........................................... 98
3.5.3 Summary ........................................................................................................................................... 99
3.6 Shunt and Series Compensation.......................................................................................................... 100
3.6.1 Shunt Capacitors ............................................................................................................................. 100
3.6.2 Shunt Reactors ................................................................................................................................ 101
3.6.3 Passive Harmonic Filters ................................................................................................................. 102
3.6.4 Series Capacitors ............................................................................................................................ 102
3.6.5 Series Reactors ............................................................................................................................... 103
3.6.6 Frequency-Dependency of Passive Components ............................................................................ 103
3.6.7 Summary ......................................................................................................................................... 103
7
3.7 Network Equivalent for Harmonic Studies .......................................................................................... 103
4. Power electronic based network element models ................................................105

4.1 Introduction ............................................................................................................................................ 105
4.1.1 Generic Norton/Thévenin Equivalent Harmonic Model Structure .................................................... 105
4.1.2 Considerations on Norton/Thévenin Modelling Limitations .............................................................. 106
4.2 Converter-Based Generation ................................................................................................................ 107
4.2.1 Converter-based Wind Generation .................................................................................................. 107
4.2.2 Converter-based PV Generation...................................................................................................... 113
4.2.3 Other Types of Converter-based Generation ................................................................................... 114
4.2.4 Summary ......................................................................................................................................... 114
4.3 HVDC Converters................................................................................................................................... 115
4.3.1 HVDC - LCC .................................................................................................................................... 115
4.3.2 HVDC – VSC ................................................................................................................................... 124
4.3.3 Summary ......................................................................................................................................... 130
4.4 FACTS Devices ...................................................................................................................................... 131
4.4.1 Line-Commutated FACTS Devices .................................................................................................. 132
4.4.2 Self-Commutated FACTS Devices .................................................................................................. 138
4.4.3 Active Harmonic Filtering ................................................................................................................. 143
4.5 Traction Systems ................................................................................................................................... 144
4.5.1 AC Electrified Railways .................................................................................................................... 144
4.5.2 DC Electrified Railways ................................................................................................................... 145
4.6 Electric Arc Furnaces ............................................................................................................................ 146
4.6.1 General Information ......................................................................................................................... 146
4.6.2 Harmonic Behaviour of Electric Arc Furnaces ................................................................................. 146
4.6.3 Modelling of Arc Furnaces for Harmonic Studies ............................................................................. 148
4.7 Variable Speed Drives ........................................................................................................................... 150
4.7.1 General Information ......................................................................................................................... 150
4.7.2 Type 1: Classical VFD with Diode Rectifier ..................................................................................... 151
4.7.3 Type 2: VFD with PWM Front End ................................................................................................... 152
4.7.4 Type 3: PWM Current-Source Inverter Drives ................................................................................. 154
4.7.5 Type 4: Load-Commutated Inverter Drives ...................................................................................... 154
4.7.6 Type 5: Phase-Controlled Cycloconverters ..................................................................................... 156
4.8 Summary ................................................................................................................................................ 158
5. General considerations for harmonic studies ......................................................159

5.1 Introduction ............................................................................................................................................ 159
5.2 Types of Harmonic Studies .................................................................................................................. 159
5.3 Considerations for Power System Representation ............................................................................ 160
5.3.1 Extent of the Network Model ............................................................................................................ 160
5.3.2 Power System Configuration ........................................................................................................... 161
5.4 Harmonic Impedance Loci and Envelopes .......................................................................................... 164
5.4.1 Types of Harmonic Impedance Envelope ........................................................................................ 167
5.4.2 Practical Considerations for Creating Harmonic Impedance Envelopes .......................................... 168
5.5 Use of Power Frequency Short-Circuit Thévenin Equivalent ............................................................ 171
5.6 Representation of Customer installation............................................................................................. 173
5.7 Aggregation of Harmonic Sources....................................................................................................... 175
5.8 Background Harmonic Voltage Distortion .......................................................................................... 176
5.8.1 Representative Period of Measurement .......................................................................................... 176
5.8.2 Estimation of Background Distortion ................................................................................................ 184
5.9 Summary ................................................................................................................................................ 188
6. Conclusions ............................................................................................................191
7. Bibliography/references .........................................................................................195
8
Appendix A. Compoment representation for frequency-domain studies……….203

Appendix B. CIGRE 14-bus benchmark system……………………………………...207
Appendix C. CIGRE 14-bus model development examples………………………..209
Appendix D. Cable sensitivity analysis………………………………………………..215
Appendix E. Example of load models………………………………………………….229
Appendix F. Synchronous generator model validation…………………………….235
Appendix G. Harmonic state estimation in Spain……………………………………239
Figures and Illustrations

Figure 2-1 Overview of Frequency-Domain Methods .......................................................................................... 21
Figure 2-2 Phase and Sequence Voltage in Inter-sequence Coupled Cable Radial Model (Receiving End) ....... 25
Figure 2-3 Frequency Scan of the Apparent Impedance of Phases A, B and C ................................................... 25
Figure 3-1 Overhead Line Nominal-PI Model ....................................................................................................... 31
Figure 3-2 Overhead Line Equivalent-PI Model .................................................................................................... 32
Figure 3-3 Cascaded Representation of Homogeneous Line Sections ................................................................ 34
Figure 3-4 Setup for Line Model Frequency Scan Calculation.............................................................................. 36
Figure 3-5 Comparison of Lumped and Distributed Parameter Line Model for Various Line Lengths (Blue=Lumped
Parameter Model; Red=Distributed Parameter Model) .......................................................................................... 37
Figure 3-6 Distributed vs. Cascaded Lumped Parameter Model Representation (Length = 250km) .................... 38
Figure 3-7 Skin Effect Modelling (Length = 250km) (Blue: Consideration of Skin Effect; Red=Neglection of Skin
Effect) .................................................................................................................................................................... 39
Figure 3-8 Effect of Earth Resistivity on Positive Sequence Impedance (Length = 250km) ................................. 40
Figure 3-9 Effect of Earth Resistivity on Zero Sequence Impedance (Length = 250km) ...................................... 40
Figure 3-10 Effect of Average Conductor Height on the Positive Sequence Impedance (Length = 250km) ......... 42
Figure 3-11 Comparison of Cable Models [39] ..................................................................................................... 46
Figure 3-12 Cable Internal Impedance: Approximate vs. Analytical Consideration of Skin Effect [39] ................. 47
Figure 3-13 Calculation of Cable Impedance: Comparison of Approaches [47] ................................................... 48
Figure 3-14 Consideration of Skin and Proximity Effects in MoM-SO: Effect on Positive Sequence Impedance
Magnitude [47] ....................................................................................................................................................... 49
Figure 3-15 Transformer Model (A) Physical Representation. (B) Single-Phase Equivalent Model ..................... 51
Figure 3-16 Power Transformer Model 1: Electra 167 [2] .................................................................................... 52
Figure 3-17 Power Transformer Model 2: IEEE Std. 399 [49]............................................................................. 53
Figure 3-18 Power Transformer Model 3: Electra-164 [1] ................................................................................... 54
Figure 3-19 Power Transformer Model 4: Arrillaga [3] ......................................................................................... 54
Figure 3-20 Power Transformer Model 5: Funk [50] ............................................................................................. 55
Figure 3-21 Example 1: Calculated and Measured Transformer Resistance as a Function of Frequency ........... 56
Figure 3-22 Example 1: Calculated and Measured Transformer Reactance as a Function of Frequency ............ 57
Figure 3-23 Example 1: Measured and Calculated Time Constant (L/R) ............................................................. 57
Figure 3-24 Example 2: Variation of Transformer Resistance with Frequency [51] .............................................. 58
Figure 3-25 Example 2: Variation of Transformer Inductance with Frequency [51] ............................................. 58
Figure 3-26 Example 2: Calculated and Measured Transformer Resistance as a function of Frequency ............ 59
Figure 3-28 Example 3: Calculated and Measured Transformer Resistance as a function of Frequency ............ 61
Figure 3-30 Effect of Detailed Power Transformer Models on Harmonic Impedance in a real Transmission System
.............................................................................................................................................................................. 63
Figure 3-31 Comparison of the 𝑳/𝑹(𝝎) Ratio of Different Transformer Types [51] .............................................. 65
Figure 3-32 Method 1 (CIGRE WG 36.05) Load Model (Static)............................................................................ 67
Figure 3-33 Method 2 (CIGRE WG 36.05) Load Model (Static and Rotating) ...................................................... 67
Figure 3-34 Method 3 (CIGRE WG 36.05) Load Model (CIGRE/EDF) ................................................................. 68
Figure 3-35 Method 2 (CIGRE WG CC02) “R//L” Load Model (Static) ................................................................. 69
Figure 3-36 Method 3 (CIGRE WG CC02) “Motor” Load Model ........................................................................... 69
Figure 3-37 Method 1 (CIGRE JTF 36.05.02/14.03.03) “Passive” Load Model (Domestic) .................................. 70
Figure 3-38 Method 2 (CIGRE JTF 36.05.02/14.03.03) “Motive” Load Model ...................................................... 70
Figure 3-39 Method 4 (CIGRE JTF 36.05.02/14.03.03) “Measurement-based” Load Model ................................ 71
Figure 3-40 IEEE Task Force: Summary of Load Models [55].............................................................................. 73
Figure 3-41 Large Asynchronous Load Area Equivalent Model [56] ..................................................................... 73
Figure 3-42 Downstream Harmonic Component Load Model Representation at Supply Bus GSP (Typically 132kV)
as used by National Grid TSO ............................................................................................................................... 74
9
Figure 3-43 Downstream Harmonic Component Load Model Representation at Supply Bus GSP as used by
Scottish Power TSO .............................................................................................................................................. 76
Figure 3-44 Test Case .......................................................................................................................................... 77
Figure 3-45 Modelling of 25kV Feeder with Various Load Levels ......................................................................... 79
Figure 3-46 Model Comparison: Feeder with Heavy Load ................................................................................... 80
Figure 3-47 Model Comparison: Feeder with Light Load ...................................................................................... 81
Figure 3-48 Frequency Response as seen from the Wind Farm (120kV) ............................................................. 82
Figure 3-49 Model Comparison as seen from the Wind Farm (120kV) ................................................................. 83
Figure 3-50 Effect of Load Proximity to Point of Interconnection .......................................................................... 84
Figure 3-51 Extended CIGRE 14-Bus System for Comparison of Harmonic Load Models .................................. 85
Figure 3-52 Impedance at SB4 220kV – Winter Load .......................................................................................... 86
Figure 3-53 Impedance at SB4 220kV – Autumn Load ........................................................................................ 87
Figure 3-54 Impedance at SB4 220kV – Summer Load ....................................................................................... 87
Figure 3-55 Impedance at Bus1 110kV – Winter Load ......................................................................................... 88
Figure 3-56 Impedance at Bus1 110kV – Autumn Load ....................................................................................... 88
Figure 3-57 Impedance at Bus1 110kV – Summer Load ...................................................................................... 88
Figure 3-58 Impedance at Bus2 33kV – Winter Load ........................................................................................... 89
Figure 3-59 Impedance at Bus2 33kV – Autumn Load ......................................................................................... 89
Figure 3-60 Impedance at Bus2 33kV – Summer Load ........................................................................................ 90
Figure 3-70 Synchronous Generator Harmonic Impedance ................................................................................. 95
Figure 3-71 Sample 370 MVA Synchronous Generator Harmonic Equivalent Resistance, R genh, and Reactance,
Xgenh ....................................................................................................................................................................... 96
Figure 3-72 Model 1: IEEE ................................................................................................................................... 96
Figure 3-73 Model 2: Electra 167 ......................................................................................................................... 97
Figure 3-74 Model 3: HQ ...................................................................................................................................... 97
Figure 3-75 Effect of Detailed Synchronous Generator Models on Harmonic Impedance in a Real Transmission
System (a) Close to Generation and (b) Far from Generation ............................................................................... 98
Figure 3-76 Maximum Damping Factor (with Respect to Base Case) Provided by each Synchronous Generator
Model ..................................................................................................................................................................... 99
Figure 3-77 Detuned Shunt Capacitors .............................................................................................................. 100
Figure 3-78 Effect of connecting Shunt Capacitors in a Transmission Network ................................................. 101
Figure 3-79 Examples of Typical Filter Configurations ....................................................................................... 102
Figure 4-1 Norton/Thévenin Harmonic Model of Power Electronic Converter (source [62]) ............................... 105
Figure 4-2 Doubly-Fed Induction Generator [65] ............................................................................................... 108
Figure 4-3 Example Type 4 Wind Turbine Generator ......................................................................................... 109
Figure 4-4 Example Comparison of WT Harmonic Impedance and Smoothing Reactor of Converter ............... 111
Figure 4-5 WTG Modelling Approaches ............................................................................................................. 112
Figure 4-6 Frequency Scan of a Wind Farm with different WTG Harmonic Models ........................................... 113
Figure 4-7 Typical PV System Overview ............................................................................................................ 113
Figure 4-8 AC Bank Switching Steps following HVDC LCC Converter Q Curve................................................. 116
Figure 4-9 Illustrative Variation of Characteristic Harmonics Magnitude vs Load (HVDC LCC) ......................... 117
Figure 4-10 Filter Bank, Sub-Bank and Branch Definition .................................................................................. 118
Figure 4-11 Harmonic Filter Impedance - Minimum DC Power Configuration .................................................... 119
Figure 4-12 Harmonic Filter Impedance - Triple-Tuned Filter ............................................................................. 119
Figure 4-13 Harmonic Filter Impedance - Maximum DC Power Configuration ................................................... 120
Figure 4-14 Simplified Equivalent Circuit for Harmonic Emission Assessment at PCC (HVDC LCC) ................ 122
Figure 4-15 Simplified Circuit to Assess Magnification (or Damping) of Background Voltage Distortion at PCC
(HVDC LCC) ........................................................................................................................................................ 123
Figure 4-16 HVDC LCC Model for System-Wide Studies ................................................................................... 124
Figure 4-17 a) MMC VSC Configuration, b) Half-Bridge Module (c) Full-Bridge Module .................................... 125
Figure 4-18 Simplified Equivalent Circuit for Harmonic Emission Assessment at PCC (HVDC VSC) ................ 127
Figure 4-19 Simplified Circuit to Assess Magnification (or Damping) of Background Voltage Distortion at PCC (VSC
HVDC) ................................................................................................................................................................. 128
Figure 4-20 HVDC VSC Model for System-Wide Studies ................................................................................... 129
Figure 4-21 HVDC Modelling Guidelines for Harmonic Studies .......................................................................... 130
Figure 4-22 SVC Typical Configuration .............................................................................................................. 132
Figure 4-23 Typical Harmonic Currents Generated by a TCR [87] ..................................................................... 133
Figure 4-24 Simplified circuit for Harmonic Emission Assessment at PCC (SVC) .............................................. 134
Figure 4-25 Simplified Circuit to Assess Magnification (or Damping) of Background Voltage Distortion at PCC (SVC)
............................................................................................................................................................................ 135
10
Figure 4-26 SVC Model for System-Wide Studies .............................................................................................. 136

Figure 4-27 Principal Diagram of TCSC ............................................................................................................. 136
Figure 4-28 TCSC Impedance as a Function of Firing Angle (source [89]) ........................................................ 137
Figure 4-29 Flow of Harmonic Currents in a TCSC ............................................................................................ 137
Figure 4-30 Simplified Three-Phase Diagram of the Modular Multi-Level STATCOM ........................................ 139
Figure 4-31 STATCOM Control and Oscillation Phenomena Impact Overview .................................................. 140
Figure 4-32 Principal Diagram of SSSC ............................................................................................................. 141
Figure 4-33 Principal Diagram of UPFC ............................................................................................................. 142
Figure 4-34 Principal Arrangement of Different Sub-Systems under Assessment.............................................. 144
Figure 4-35 V-I-characteristic of AC-Arcs (1 to 6 for Low Currents and 7 to 8 for High Currents) [102] and Example
of a Measured Harmonic Spectrum of the Primary Currents [103] ...................................................................... 147
Figure 4-36 Typical Electrical Diagram of a DC Arc Furnace with Classic 12-Pulse Operation and Harmonic
Spectrum (Measured) of the Primary Currents (Maximum and Average Values) ................................................ 147
Figure 4-37 Typical electrical diagram of a 3-electrode AC submerged arc furnace .......................................... 148
Figure 4-38 Thévenin Equivalent Model (Dual of the Norton Equivalent) ........................................................... 149
Figure 4-39 Classical 6-Pulse Diode Bridge Rectifier and 6-Pulse PWM Inverter .............................................. 151
Figure 4-40 Multi-Pulse VSI Model for Harmonic Studies ................................................................................... 151
Figure 4-41 Examples of Input Current Spectrums for 6-Pulse VSI-PWM Drives ([114], [115], [116]) ............... 152
Figure 4-42 Full PWM Rectifier and Inverter....................................................................................................... 153
Figure 4-43 CSI Variable Speed Drive with PWM CSR and CSI ........................................................................ 154
Figure 4-44 Topology of a 6-Pulse LCI Drive for an Induction Motor .................................................................. 155
Figure 4-45 Topology of a 6-Pulse LCI Drive for a Synchronous Motor ............................................................. 155
Figure 4-46 Input Current Harmonics for a 12-Pulse LCI Drive for a 12.47kV 14,200 hp Synchronous Motor ... 156
Figure 4-47 Topology of a 3-Phase to 3-Phase 6-Pulse Blocking Type Cycloconverter ..................................... 157
Figure 4-48 Harmonic Current Spectrum of an 18 MW Cycloconverter.............................................................. 158
Figure 5-1 Complete 400 kV Detailed French Network ...................................................................................... 161
Figure 5-2 Harmonic Voltage Distortion Measurements (95th Percentile) in a Transmission Station in the UK for
Three Levels of System Demand: Summer, Spring and Winter .......................................................................... 162
Figure 5-3 Harmonic Impedance in a Transmission Station in Ireland under Intact (N), Single Contingency (N-1)
and Double Contingency (N-2) Conditions .......................................................................................................... 163
Figure 5-4 Example of Harmonic Impedance in a Transmission Network: Locus, Loci and Envelope (5 Hz
Calculation Resolution) ........................................................................................................................................ 166
Figure 5-5 Generic Circle Envelope.................................................................................................................... 167
Figure 5-6 Generic Sector Envelope .................................................................................................................. 167
Figure 5-7 Generic Basic Polygon Envelope ...................................................................................................... 168
Figure 5-8 Example of Harmonic Impedance Envelopes .................................................................................... 170
Figure 5-9 Example of Two Impedance Envelopes for a Particular Band ........................................................... 170
Figure 5-10 Comparison of Harmonic Impedance Derived from a Power Frequency Short-Circuit Equivalent and
from a Detailed Frequency-Dependent Model ..................................................................................................... 172
Figure 5-11 Impact of the Offshore Cable and Transformer Harmonic Losses Modelling on System Resonance
Damping at Offshore Substation (source [145])................................................................................................... 174
Figure 5-12 Comparison of Two Different Weeks of 7th Harmonic Measurement in Germany ........................... 177
Figure 5-13 Comparison of Two Different Months of 7th Harmonic Measurement in Germany .......................... 178
Figure 5-14 Phase-to-Ground 95th Percentile Values of the 11th and 13th Harmonics in a 400 kV Danish Substation
over 20 Weeks..................................................................................................................................................... 179
Figure 5-15 Sorted 10 min. Aggregated Values Measured over One Week for the 11 th, 13th and 23rd Harmonics in
a Danish 400 kV Substation ................................................................................................................................ 181
Figure 5-16 Ratios between the 100th and 95th Percentile Levels Evaluated for the 11th and 13th Order Harmonics
Measured over 20 Weeks in a Danish 400 kV Substation ................................................................................... 181
Figure 5-17 Harmonic Voltage Distortion Measurements in five Substations in Ireland over a 7-Day Period. Legend:
7th harmonic, 11th harmonic, 13th harmonic ......................................................................................................... 183
Figure 5-18 5th Harmonic: Voltage vs Power ...................................................................................................... 184
Figure 5-19 7th Harmonic: Voltage vs Power ...................................................................................................... 184
Figure 5-20 Harmonic Voltage Distortion at a Wind Farm in Brazil ..................................................................... 184
Figure 5-21 Power Quality Monitoring and Analysis System by Red Eléctrica de España (REE) [165] ............. 186
Figure 5-22 Thévenin Equivalent Representation for Assessing Modification of Background Distortion due to a
Customer Connection [168] ................................................................................................................................. 187
App Figure A.1 Phase Amplification as Function of the Harmonic Order and (a) the Relative Permittivity and (b) the
Cable System Length .......................................................................................................................................... 204
App Figure A.2 Phase and Sequence Voltage in an Inter-Sequence Coupled Meshed Transmission Network . 205
App Figure A.3 Frequency Scans in Inter-Sequence Coupled Meshed Transmission Network at Two Unique
Substations .......................................................................................................................................................... 205
App Figure B.1 220/110 kV Transmission Network Topology............................................................................. 207
App Figure C.1 Sequential Steps followed in the 220 kV Network Model Development ..................................... 209
App Figure C.2 Harmonic Impedance at Bus-5 under Incremental 220kV Network Model Development .......... 211
App Figure C.3 Setup for Calculation of Harmonic Impedance at Bus 5 including 110 kV System .................... 212
App Figure C.4 Harmonic Impedance at Bus-5 with 220kV and 110kV Network Models ................................... 213
App Figure D.1 Cable Layers ............................................................................................................................. 215
11
App Figure D.2 Base Case Cable Layout ........................................................................................................... 216

App Figure D.3 Cable Cross-Bonding Configuration .......................................................................................... 216
App Figure D.4 Positive-Sequence Impedance: Cable Termination Open- or Short-Circuited ........................... 217
App Figure D.5 Zero-Sequence Impedance: Cable Termination is Open- or Short-Circuited ............................ 217
App Figure D.6 Harmonic Propagation for Long EHV Cable .............................................................................. 218
App Figure D.7 Positive-Sequence Voltage Amplification .................................................................................. 218
App Figure D.8 Zero-Sequence Voltage Amplification ....................................................................................... 218
App Figure D.9 Positive-Sequence Harmonic Impedance Comparison for Different Cable Lengths .................. 219
App Figure D.10 Zero-Sequence Harmonic Impedance Comparison for Different Cable Lengths ..................... 220
App Figure D.11 Positive-Sequence Harmonic Impedance Comparison for Variations in the Cable Conductor
Radius ................................................................................................................................................................. 221
App Figure D.12 Zero-Sequence Harmonic Impedance Comparison for Variations in the Cable Conductor Radius
............................................................................................................................................................................ 221
App Figure D.13 Positive-Sequence Harmonic Impedance Comparison for Variations in the Cable Insulation
Thickness ............................................................................................................................................................ 222
App Figure D.14 Zero-Sequence Harmonic Impedance Comparison for Variations in the Cable Insulation Thickness
............................................................................................................................................................................ 223
App Figure D.15 Examples of Cable Layout ....................................................................................................... 223
App Figure D.16 Positive-Sequence Harmonic Impedance Comparison for Different Cable Layouts ................ 224
App Figure D.17 Zero-Sequence Harmonic Impedance Comparison for Different Cable Layouts ..................... 224
App Figure D.18 Positive-Sequence Harmonic Impedance Comparison for Different Cable Bonding Configurations
............................................................................................................................................................................ 225
App Figure D.19 Zero-Sequence Harmonic Impedance Comparison for Different Cable Bonding Configurations
............................................................................................................................................................................ 226
App Figure D.20 Positive-Sequence Harmonic Impedance Comparison for Different Numbers of Major Sections
............................................................................................................................................................................ 227
App Figure D.21 Zero-Sequence Harmonic Impedance Comparison for Different Numbers of Major Sections . 227
App Figure D.22 Positive-Sequence Harmonic Impedance Comparison for Different Cable Models ................. 228
App Figure E.1 Simple Configuration for Sensitivity Analysis ............................................................................. 229
App Figure E.2 Load Representations ................................................................................................................ 230
App Figure E.3 Frequency Scan at 230kV Bus - Load Represented at a 230 kV Bus........................................ 231
App Figure E.4 Frequency Scan at 230kV Bus - Load Represented at 69 kV Bus............................................. 232
App Figure E.5 System Connection Configuration from 230 kV to 220 V ........................................................... 233
App Figure E.6 Frequency Scan at 230kV Bus - Load Represented at a 220 V with Explicit Representation of 69kV
and 13.8kV Systems............................................................................................................................................ 233
App Figure E.7 System Connection Equivalent at 69 kV .................................................................................... 234
App Figure E.8 Frequency Scan at 230kV Bus - Load Represented at a 220 V through a Reduced Equivalent
Representation of the 69kV and 13.8kV Systems ............................................................................................... 234
App Figure F.1 HQ Field Measurement Setup to Measure Z(f) Seen at 315 kV ................................................. 235
App Figure F.2 Measurement Data of Z(f) .......................................................................................................... 235
App Figure F.3 Simulated Data of Z(f) ................................................................................................................ 236
App Figure F.4 Measurement Data of a 50 MVA 13.8 kV Machine .................................................................... 236
App Figure F.5 Harmonic Impedance of a 120 MVA Hydraulic Machine ............................................................ 237
App Figure F.6 Sensitivity Analysis: Machine Losses......................................................................................... 238
App Figure F.7 Model Simplification following Validation ................................................................................... 238
App Figure G.1 Accuracy of the Estimation Method ........................................................................................... 240
App Figure G.2 Example of Results of the Harmonic State Estimation Method ................................................. 240
Tables
Table 2-1 Summary of Study Domains and their Applications ............................................................................... 28
Table 3-1 Corrections for Skin Effect in OHLs ....................................................................................................... 34
Table 3-2 Variation of Earth Resistivity with Depth ............................................................................................... 35
Table 3-3 Summary of Sensitivity Analysis: Impacts on Harmonic Impedance ..................................................... 43
Table 3-4 Summary of Sensitivity Analysis in 0: Impacts on Harmonic Impedance............................................... 50
Table 3-5 Values for Coefficients 𝐚𝟎, 𝐚𝟏 𝐚𝟐 and 𝐛 (requirement: 𝐚𝟎 + 𝐚𝟏 + 𝐚𝟐 = 𝟏) ............................................ 54
Table 3-6 Coefficients for Transformer Model 5 [50] ............................................................................................. 55
Table 3-7 Example 1: Transformer Model Parameters ......................................................................................... 56
Table 3-8 Example 2: Transformer Model Parameters ......................................................................................... 59
Table 3-9 Typical values of RLV and PFC *courtesy National Grid ....................................................................... 75
Table 3-10 Typical BLV (Power Factor Correction (PFC) + Cable Capacitance) and Equivalent XT Transformer
Reactance *courtesy National Grid ........................................................................................................................ 75
Table 3-11 Typical Cable Capacitance, BHV ......................................................................................................... 76
Table 3-12 Typical values of RLV for Different Load Categories .......................................................................... 76
Table 3-13 Composition of 25 kV Load Feeder 235 ............................................................................................. 77
Table 3-14 Residential-Commercial Assumed Load Compositions ....................................................................... 78
Table 3-15 Distribution-Level Circuits .................................................................................................................... 86
Table 3-16 Load Profiles Modelled ........................................................................................................................ 86
12
Table 4-1 Example Representation/Template of the Harmonic Voltage/Current Source ..................................... 106
Table 4-2 Example Representation/Template of the Harmonic Equivalent Impedance ....................................... 106
Table 4-3 Comparison of Harmonic Performance in Typical FACTS Devices .................................................... 131
Table 4-4 Typical Harmonic Mitigation Solutions in FACTS Devices .................................................................. 131
Table 4-5 Modelling Considerations for Line-Commutated FACTS Devices ...................................................... 138
Table 4-6 Modelling Considerations for Self-Commutated FACTS Devices ....................................................... 142
Table 5-1. IEC 61000-3-6 Summation Exponents for Harmonics [100] ............................................................... 175
Table 5-2: Comparison of Different Statistic Values of Measurements of 7 th Harmonic of Two Different Weeks in
Germany .............................................................................................................................................................. 177
Table 5-3: Comparison of Different Statistical Values of Measurements of 7 th Harmonic for Two Different Months
in Germany .......................................................................................................................................................... 178
Table 5-4 Maximum and Minimum 95th Percentile Levels of 11th and 13th Harmonics in a 400 kV Danish Substation
over 20 Weeks..................................................................................................................................................... 179
Table 5-5 Ratios of the Maximum and Minimum 95th Percentile Levels of 11 th and 13th Harmonics in a 400 kV
Danish Substation over 20 Weeks....................................................................................................................... 179
App Table B.1 220 kV OHL Parameters ............................................................................................................. 207

App Table B.2 110 kV OHL Parameters ............................................................................................................. 208
App Table B.3 220/110kV Transformers Parameters ......................................................................................... 208
App Table D.1 Cable Parameters ....................................................................................................................... 215
App Table F.1 Measured Parameters of a 50 MVA Machine vs. Frequency ...................................................... 237
Equations
Equation 2.1 .......................................................................................................................................................... 22
Equation 3.1 .......................................................................................................................................................... 33
Equation 3.2 .......................................................................................................................................................... 35
Equation 3.3 .......................................................................................................................................................... 41
Equation 3.4 .......................................................................................................................................................... 52
Equation 3.5 .......................................................................................................................................................... 53
Equation 3.6 .......................................................................................................................................................... 53
Equation 3.7 .......................................................................................................................................................... 53
Equation 3.8 .......................................................................................................................................................... 53
Equation 3.9 .......................................................................................................................................................... 53
Equation 3.10 ........................................................................................................................................................ 53
Equation 3.11 ........................................................................................................................................................ 53
Equation 3.12 ........................................................................................................................................................ 53
Equation 3.13 ........................................................................................................................................................ 54
Equation 3.14 ........................................................................................................................................................ 54
Equation 3.15 ........................................................................................................................................................ 54
Equation 3.16 ........................................................................................................................................................ 54
Equation 3.17 ........................................................................................................................................................ 55
Equation 3.18 ........................................................................................................................................................ 55
Equation 3.19 ........................................................................................................................................................ 67
Equation 3.20 ........................................................................................................................................................ 67
Equation 3.21 ........................................................................................................................................................ 67
Equation 3.22 ........................................................................................................................................................ 68
Equation 3.23 ........................................................................................................................................................ 68
Equation 3.24 ........................................................................................................................................................ 68
Equation 3.25 ........................................................................................................................................................ 69
Equation 3.26 ........................................................................................................................................................ 69
Equation 3.27 ........................................................................................................................................................ 69
Equation 3.28 ........................................................................................................................................................ 69
Equation 3.29 ........................................................................................................................................................ 69
Equation 3.30 ........................................................................................................................................................ 70
Equation 3.31 ........................................................................................................................................................ 70
Equation 3.32 ........................................................................................................................................................ 70
Equation 3.33 ........................................................................................................................................................ 70
Equation 3.34 ........................................................................................................................................................ 70
Equation 3.35 ........................................................................................................................................................ 70
Equation 3.36 ........................................................................................................................................................ 70
Equation 3.37 ........................................................................................................................................................ 71
Equation 3.38 ........................................................................................................................................................ 71
Equation 3.39 ........................................................................................................................................................ 73
Equation 3.40 ........................................................................................................................................................ 73
Equation 3.41 ........................................................................................................................................................ 73
Equation 3.42 ........................................................................................................................................................ 74
Equation 3.43 ........................................................................................................................................................ 74
13
Equation 3.44 ........................................................................................................................................................ 96

Equation 3.45 ........................................................................................................................................................ 96
Equation 3.46 ........................................................................................................................................................ 96
Equation 3.47 ........................................................................................................................................................ 97
Equation 3.48 ........................................................................................................................................................ 97
Equation 3.49 ........................................................................................................................................................ 97
Equation 3.50 ........................................................................................................................................................ 97
Equation 3.51 ........................................................................................................................................................ 98
Equation 3.52 ........................................................................................................................................................ 98
Equation 3.53 ........................................................................................................................................................ 98
Equation 3.54 ........................................................................................................................................................ 98
Equation 3.55 ........................................................................................................................................................ 98
Equation 3.56 ...................................................................................................................................................... 100
Equation 4.1 ........................................................................................................................................................ 150
Equation 4.2 ........................................................................................................................................................ 151
Equation 4.3 ........................................................................................................................................................ 151
Equation 4.4 ........................................................................................................................................................ 157
Equation 4.5 ........................................................................................................................................................ 157
Equation 5.1 ........................................................................................................................................................ 172
Equation 5.2 ........................................................................................................................................................ 175
Equation 5-3 ........................................................................................................................................................ 187
App Equation A.1................................................................................................................................................. 203

App Equation A.2................................................................................................................................................. 203
App Equation D.1 ................................................................................................................................................ 216
App Equation D.2 ................................................................................................................................................ 216
Acronyms and abbreviations
ACS Average Cold Spell

ACSR Aluminium conductor steel-reinforced
AF Active Filter
AFE Active Front End
ASD Adjustable Speed Drive
CCV Cycloconverter
CSC Current Source Converter
CSI Current Source Inverter
DFIG Double-Fed Induction Generator
EAF Electric Arc Furnace
EHV Extra High Voltage
EMF Electromotive Force
FACTS Flexible AC Transmission System
FDNE Frequency-Dependent Network Equivalent
FEM Finite Element Method
GMD Geomagnetic Disturbance
GMR Geometric Mean Radius
GSP Grid Supply Point
GTO Gate Turn-off Thyristor
HD Harmonic Domain
HSE Harmonic State-Estimation
14
HV High Voltage
HVDC High Voltage Direct Current
IGBT Insulated-Gate Bipolar Transistor
IGCT Integrated Gate Commutated Thyristor
LCC Line Commutated Converter
LCI Line Commutated Inverter
LV Low Voltage
MMCC Modular Multi-level Cascaded Converter
MoM Method of Moments
MSC Mechanically Switched Capacitor
MSCDN Mechanically Switched Capacitor with Damping Network
MV Medium Voltage
OEM Original Equipment Manufacturer
OHL Overhead Line
PCC Point of Common Coupling
PoC Point of Connection
PE Power Electronics
PFC Power Factor Correction
PLL Phase-Locked Loop
PQ Power Quality
PV PhotoVoltaic
PWM Pulse Width Modulation
RES Renewable Energy Sources
SAF Submerged Arc Furnace
SC Single Core
SCR Silicon Controlled Rectifier (Thyristor)
SHE Selective Harmonic Elimination
SM Sub-Module
SRF-PLL Synchronous Rotating Frame PLL
SSC Static Shunt Capacitor
SSSC Solid State Series Compensator
STATCOM Static Synchronous Compensator
SVC Static Var Compensator
TCR Thyristor Controlled Reactor
THD Total Harmonic Distortion
TO (TSO) Transmission System Operator
TSC Thyristor Switched Capacitor
TSR Thyristor Switched Reactor
UGC UnderGround Cable
UHV Ultra High Voltage
15
UMEC Unified Magnetic Equivalent Circuit

UPFC Unified Power Flow Controller
VFD Variable Frequency Drive
VSC Voltage Source Converter
VSI Voltage Source Inverter
WECC Western Electricity Coordinating Council
WPP Wind Power Plant
WT Wind Turbine
WTG Wind Turbine Generator
XLPE Cross-linked Polyethylene
PPM Power Park Module
16
1. Introduction
1.1 Background
Harmonics have been part of the electric power network since their introduction into the system by
nonlinear devices. They occur as emissions, usually as an undesirable side-effect of the main function
of equipment, at multiple or sub-multiple integers of the system frequency. It is also possible to define
inter-harmonics as those frequencies which appear at non-integer multiples of the system frequency.
The issues caused by harmonics can be categorised under two distinct but heavily-interrelated areas:
steady-state distortion and transient effects. The first area relates to steady-state harmonic voltage
distortion, where system resonance or resonances are excited via the introduction of harmonic injections
by some active devices. Typical examples of active devices are power electronics being utilised in
converter or inverter circuits. These devices are normally under continuous operation and hence are a
source of either harmonic current or voltage which creates a steady-state distortion on the system. The
second area relates to the excitation of system resonances by harmonic sources that are transient in
nature. The most typical case is the energization of a transformer. When a transformer is energized by
uncontrolled or random closing of the energizing device, the magnetic characteristic of the transformer
core results in it being driven into partial or full saturation, and hence a current rich in harmonics is
drawn. This current could excite a system resonance thereby creating a temporary overvoltage. The
phenomenon is short in duration and in most cases the level of overvoltage reached is not excessively
high. This Technical Brochure (TB) concentrates on aspects related to steady-state harmonic distortion.
The adverse effects of harmonics have been discussed in many publications (e.g. [3], [4]) and will
therefore not be treated in detail in this introduction. However, it suffices to say that harmonics could
have detrimental effects on electrical equipment via the introduction of additional heating or the creation
of a complex environment for protection or measuring devices to function properly. Therefore,
standardisation exists in order to control and limit the amount of harmonic distortion present in the
system. This is done using limits for current or voltage distortion. These limits are normally coordinated
across different voltage levels to have a graded approach and are usually two-tier (i.e. planning and
compatibility limits) depending on which standard is utilised. Planning limits are usually classified as the
internal objectives for utilities, i.e. to plan the system such that these levels are not exceeded.
Compatibility limits on the other hand are the absolute levels that should not be breached as they define
the safe operating environment of all connected equipment.
To keep the harmonic distortion levels within planning limits, utilities need to perform extensive studies
or assessments at the planning stage of any changes being introduced to the electric power network.
The changes could be due to the expansion of the system itself, such as the installation of a new cable
route, or the integration of a new non-linear installation.
Integration of large renewable energy sources, and in particular those that involve offshore connection
requiring the use of long HVAC cables, have a twofold effect. Not only do the wind turbine generators
introduce harmonic current emissions into the system, but the integration of long HVAC cables moves
system resonance frequencies further down in the frequency spectrum and hence has an amplifying
effect on the low-order harmonics that are already present on the system. The introduction of FACTS
devices and HVDC converter technology, be it voltage or current source, also brings further harmonic
sources onto the system and hence the requirement to control them. Harmonic sources can excite local
resonances but also remote resonances in the power system, which are not detectable locally at the
injection point. Therefore, the need for performing harmonic assessments is becoming even more
pronounced as the connection of nonlinear devices and/or loads increases, driven by integration of
renewable energy sources and connection of new HVDC converters.
Adequate harmonic performance specification is therefore crucial as it impacts the subsequent design
of harmonic filters with associated costs and risks. In order to draw up proper specifications or limitations
to coordinate the emission of harmonics onto the system, the vast majority of cases require proper
modelling of the power system for this purpose. The assessment can take different forms according to
the local or international standard requirements, however in most cases network changes or connections
above 33kV will require some form of detailed system modelling.
All utilities usually have a model of their network that will replicate ordinary power flows and be able to
facilitate short-circuit calculations. In addition, they also have a dynamic model of their system in order
to perform stability calculations. However, many of them do not have a frequency-dependent system
model of their network for the purposes of harmonic distortion analysis. Traditionally, most utilities also
have some form of knowledge of the transient nature of harmonics acquired during insulation
17
coordination studies. However, this type of study approach is limited by the extent of the model, and the
knowledge and applicability has usually been confined to a select few. This used to be an acceptable
situation in the past as the need to perform this type of harmonic distortion assessment was seldom.
For example, there was previously no need to quickly perform an assessment and specify emission
limits to the connectee in a commercial environment when different projects were competing for a
connection to the same system.
In general, there is some lack of knowledge when it comes to performing meaningful harmonic distortion
assessments in modern power systems. Most of the issues stem from the lack of practical information
on modelling electrical plant equipment for such studies. Equally, the parameters that need to be
considered and the extent of this consideration have not been explained for the purposes required by
the practising power system engineer. Availability of information and guidance for such modelling
requirements are either scarce or in scattered form, mostly delegated to appendices of various
documents (examples are the short articles in Electra 164 [1] and 167 [2], published in 1996). The source
of many of these models has been lost in history, and reference can only be made to publications, some
of them by CIGRE, which contain formulae for the representation of network components but with little
or no proof of derivation. Both Electra articles are authoritative in nature but suffer from the issues
outlined above and are also becoming outdated and inaccessible. Previous CIGRE publications have
tried to bridge the gap to a certain extent but the need for such studies has increased tremendously in
most parts of the world, and hence the need for up-to-date information on the topic of modelling.
The main objective of this Technical Brochure is to provide comprehensive guidelines for practising
power system engineers when they need to perform harmonic distortion assessments. The focus is on
practical aspects of modelling for direct application in the planning process of connecting a new
customer to the transmission or distribution system, or when introducing a change to the system as part
of asset replacement or system expansion. As such, the guidelines are concentrated on frequency-
domain modelling for steady-state AC harmonic analysis in power systems, typically in the range from
power frequency up to the 50th harmonic (2.5 kHz in 50 Hz systems or 3 kHz in 60 Hz systems),
consistent with typical power quality assessments. The approach and modelling guidelines provided are
reasonably valid up to the 100th harmonic if specialized studies are required. These guidelines will be
valuable in the definition of harmonic performance specifications for new HVDC converters, FACTS
devices or other non-linear installations. They will also assist connectees when modelling their
installation to assess or demonstrate compliance with the emission limits provided by the System
Operator and to investigate and specify mitigation measures such as harmonic filters. Furthermore, this
document can also be used post-commissioning for any incident investigation or to assist resolution of
customer complaints via modelling and analysis.
1.2 Scope
The scope of this Technical Brochure, as defined in the Terms of Reference for the Working Group
(JWG C4/B4.38) concerning system harmonic modelling, is to:
1. Collate and provide all available information in the literature on modelling individual electrical plant.
2. Evaluate and suggest best practice in the use of available models to represent modern equipment.
3. Identify any shortfalls with available models and the possible need for further development in this
area.
4. Provide clear and concise guidelines on modelling existing nonlinear devices (HVDC converter
stations, wind farm generation, etc.) within the system of interest.
5. Provide guidelines on the general approach to such studies and the availability and choice of tools.
Identify any shortfalls with the available analysis tools and suggest possible developments.
The focus of this Technical Brochure is on frequency-domain modelling for steady-state AC harmonic
analysis in electric power networks, typically in the range up to the 50th harmonic (2.5 kHz in 50 Hz
systems or 3 kHz in 60 Hz systems), consistent with typical power quality assessments. The approach
and modelling guidelines provided are reasonably valid up to the 100th harmonic if specialized studies
are required. The document is not intended to cover guidelines for transient issues under harmonic
resonance although most of the modelling guidelines given will equally apply to that area also. As such,
modelling for transient harmonic performance during transformer energization or geomagnetic
disturbances (GMD) is outside the scope of this document. Similarly, the topic of harmonics on the DC
side of converters is outside the scope of this Technical Brochure.
The control systems of power electronic converters directly connected to HV and EHV grids can interact
with system resonances leading to high magnitudes of harmonics in the grid. This phenomenon is
normally referred to as harmonic stability and the root cause is the interaction of a converter controller
18
with a grid resonance. Converter controller harmonic stability requires different modelling, analysis and
mitigation methods from those described in this document. This topic is outside the scope of this
Technical Brochure. The reader is referred to the outcome of the recent CIGRE Working Group B4.67
(TB754) [5] for information on this phenomenon. Furthermore, the recently established CIGRE Working
Group C4.49 (Multi-frequency stability of converter-based modern power systems) is currently
addressing this topic in detail.
1.3 Structure
This Technical Brochure contains a further five chapters and several appendices. The appendices
include a benchmark system and further detailed analyses and examples to complement the guidelines
and recommendations provided in each of the chapters.
Chapter 2 – Study Domain and Modelling Approaches – provides an overview of analysis techniques
and solution methods for the study of harmonic distortion in power systems. This chapter examines the
strengths and limitations of each technique, including balanced and unbalanced solution methods, and
provides recommendations on applicability based on the purpose of the analysis. State-of-the-art and
future trends in solution techniques are also discussed.
Chapter 3 – Classical Network Element Models – reviews and discusses modelling options and
approaches for the accurate representation of most relevant passive network components in harmonics
studies. These include overhead lines, insulated cables, transformers, loads, synchronous generators
and shunt/series compensation devices. The relevant input data and level of detail required to represent
each component are presented and discussed. Different modelling options are assessed in a benchmark
model as well as in real system models to illustrate the effects and consequences of each type of model
in the context of system-wide studies. When available, models are validated against measurements and
recommendations are provided.
Chapter 4 – Power Electronics Based Network Element Models – reviews and discusses the
representation of non-linear devices connected to power systems acting as sources of harmonic
distortion. This chapter covers converter-based generation, HVDC converters, FACTS devices, traction
systems, electric arc furnaces and variable speed drives. The chapter reviews the mechanisms of
harmonic distortion generation for each device and provides recommendations on the modelling
structure and input data requirements. The harmonic performance of these devices is generally complex
and dependent on many factors such as converter topology, control strategy, operating point, etc. As
such, manufacturer-specific models need to be adopted. Recommendations on the structure of the
models and the features that need to be captured are provided. Emphasis is placed on the need to
capture not only the harmonic current/voltage emissions but also the harmonic impedance of the
devices.
Chapter 5 – General Considerations for Harmonic Studies – discusses practical aspects, other than the
modelling of each component, that need to be considered when performing harmonic studies related to
the connection of non-linear devices to the power grid. The chapter presents the most common types
of harmonic studies, considerations for the representation of the power system including the extent of
the model and the scenarios/contingencies to analyse, guidelines and considerations in the production
of harmonic impedance loci and envelopes, limitations of power frequency short-circuit equivalents,
practical aspects to consider in the representation of customers installations, aggregation of harmonic
sources and issues related to measurement and estimation of background distortion. The discussions
in this chapter aim to address perspectives as seen from (i) the system owner/operator; and (ii) the new
connectee, with the overall objective of minimising risk of equipment failure due to excessive harmonic
distortion as well as avoiding unnecessary investment in mitigation. While finding the right balance is
never an easy task, it is hoped that the considerations presented in this chapter will aid the engineer in
making informed decisions.
Chapter 6 – Conclusions - summarises the guidelines and recommendations of this work. Finally, areas
identified as requiring further research are highlighted. A list of technical references is included in the
bibliography.
19
20
2. Study domain and modelling aproaches

2.1 Introduction to Study Domains
Electrical signals and the characteristics of network components can be represented in either the time
domain or the frequency domain. The relationship between functions in these two domains can be
described using a Fourier series. Harmonic studies may be performed in either of these domains, or
their hybrid combination, depending on the study requirements. Such studies may include the calculation
of voltage distortion at network nodes, branch harmonic current flows and the harmonic driving point
impedance at, or mutual impedances between, specific points in the network. Due to the widespread
use of frequency-domain methods, this Technical Brochure will primarily address these. Detailed
mathematical and textual descriptions of the full breadth of frequency- and time-domain harmonic
analysis techniques can be found in [3], [4], [6], [7] and [8].
2.2 Frequency-Domain Methods

Frequency-domain methods employed in harmonics analysis, mainly frequency scans and harmonic
penetration studies, are the most commonly used in industry [6].
Formulation of harmonic analyses in the frequency domain can be very efficient and reliable for steady-
state solutions [6], [9]; the description of harmonics in this domain is simple, and the calculation time is
short [10]. However, for the accurate modelling of devices such as converters and non-linear behaviour,
as for instance transformer saturation, frequency domain models may be over-simplified [3].
Frequency domain analysis methods are briefly described in this section and an overview of them is
provided in
, originally presented in [11] (summarising [6]).
Frequency-Domain
Methods
Network Calculation of
Impedance Harmonic
Calculation Voltages and
Currents
Frequency
Scan Harmonic Harmonic
Penetration Load Flow
Balanced Unbalanced
Direct Iterative Balanced Unbalanced
Balanced Unbalanced Balanced Unbalanced
Note: Balanced implies that the three phases are equal, or in sequence component terms, that only the positive sequence
exists. Unbalanced implies unequal phase voltages or phase currents.
Figure 2-1 Overview of Frequency-Domain Methods
2.2.1 Frequency Scan

The frequency scan is a relatively simple analysis technique performed with the aim of showing the
frequency response of the network as seen from a specific bus or node. The impedance seen from the
selected bus is plotted against frequency. A recalculation of the network admittance matrix is performed
at each frequency step within the range of interest, and a 1 pu (or 1 A) current injection is applied to
obtain the corresponding bus voltages. It is represented mathematically as:
21
[𝑌ℎ ][𝑉ℎ ] = [𝐼ℎ ] Equation 2.1

where [𝑌ℎ ] is the admittance matrix and [𝑉ℎ ] and [𝐼ℎ ] are the nodal voltage and nodal current vectors,
respectively. The value of the resulting voltage (in pu or V, depending on the injected current) at the
busbar of injection corresponds to the driving point impedance and the voltage at the nth busbar
represents the harmonic transfer impedance (or mutual impedance) between the busbar of injection and
the nth busbar. The current injected can either be positive, negative or zero sequence resulting in the
positive, negative or zero sequence driving point or transfer impedances. Impedance peaks in the scan
imply a parallel resonance whereas a trough implies a series resonance.
It should be noted that the 1 pu (or 1 A) current injection is applied to a single entry in [𝐼ℎ ] (i.e. only one
harmonic) while all other entries are zero. It is a simple, yet mature and highly-effective method to detect
resonance conditions. For these reasons it is often one of the first analyses performed in a harmonic
study [8] and is widely employed in filter design.
The frequency scan method ignores system non-linearities, meaning that the dependence of [𝑌ℎ ] on
system voltages and currents is not considered. When all components in the network model are linear,
the actual value of the injected current has no impact on the result at all. A current of 1 A will give the
same impedance as an injected current of 1000 A. For almost all network components this is the case.
However, the reader must be aware that results will not be accurate for non-linear components such as
surge-arresters and magnetic cores (power transformers, VTs, etc). For most common applications in
harmonics studies, these components either operate in their linear region or the non-linear effects can
be neglected, therefore Frequency Scan results are adequate. For more detailed analysis of non-
linearities (outside the scope of this TB), time-domain or hybrid methods are required.
2.2.2 Harmonic Penetration

Harmonic penetration is a simple linear solution method and one of the most commonly-used forms of
harmonic analysis. It is described in detail in [4], [6] and [12].
Harmonic penetration solution methods based on a fundamental frequency load flow are available in
most commercial simulation packages used for harmonics studies. These methods are often referred to
as “harmonic load flow”; however the reader should be aware that harmonic penetration methods do not
carry out a true load flow solution at harmonic orders (here “true load flow” refers to the ordinary iterative
load flow performed for every harmonic frequency catered for in the analysis).
Two approaches are described next: the direct method and the iterative method. Each method can be
implemented with a balanced or unbalanced solution.
2.2.2.1 Direct Method

In the direct method, the network admittance matrix is reformulated at each harmonic order, then
considering any harmonic current injections in the network at that order, the linear system given by (Eqn.
2.1) is directly solved (i.e. non-iteratively). This means that the influence of the terminal fundamental
and harmonic voltages on the harmonic injection or harmonic impedance of the source is not considered.
This is often referred to as the ‘direct’ method and such a formulation suffices for most applications as
discussed in Section 2.6.
Quantities such as root-mean-square values for voltages and currents are summed using the
fundamental frequency load flow results plus the additional quantities obtained from the solution (Eqn.
2.1) at each harmonic order.
2.2.2.2 Iterative Method

An extension of the direct method is the iterative method which iteratively recalculates the harmonic
current injections considering their dependency on the harmonic voltage [13], [14].
2.2.2.3 Applications and Limitations

Harmonic penetration studies are primarily used in the following contexts to ascertain the following [15]:
 Predicting network harmonic distortion levels at various nodes, and harmonic currents in various
network components;
 Verification of harmonic compliance;
 Filter performance and rating design.
22
Harmonic penetration is highly applicable to large system studies and is typically utilised by system
operators for planning purposes. This method provides a good overview of the harmonic level and
general system harmonic performance under steady-state operation.
Harmonic penetration solution methods provide a relatively simple means of analysis in which it is
assumed that there is no harmonic interaction between the network and non-linear devices. This means
that phenomena such as cross-coupling between the AC and DC sides of the converter and the
subsequent coupling between frequencies introduced by converters are ignored. It also means that the
harmonic injection as well as the converter harmonic impedance is assumed to be independent of (i)
any change in the operating point of the converter; or (ii) changes that may occur on the power system
such as the background distortion level.
There are cases in which the dependence of the converter on the operating point needs to be considered
in the frequency domain for both frequency scan and penetration studies. This can be achieved via the
use of several look-up tables (for instance those provided by the converter vendor) that change in
discrete steps as the operating point of the converter changes. This is a crude approach and quickly
results in the necessity to handle many look-up tables if the harmonic impedance and/or emission of the
converter is highly dependent on operating point. In practice, the highest emission level is normally
chosen while the converter and grid impedance are varied over all credible operating points, which leads
to conservative results.
Non-linearities in network components or power electronic devices are usually ignored in the frequency
domain. For studies where these are relevant, analysis should be performed in either the time or
harmonic domain.
2.2.2.4 Balanced Harmonic Penetration

In a balanced harmonic penetration study, it may be assumed that the network is balanced at all
harmonic frequencies of interest and injection from harmonic sources appears in purely the positive
sequence. Because of these two assumptions, a single-phase network representation is sufficient, and
the model implementation and the formulation and solution of the system’s equations becomes simpler.
This method is adequate for calculating the performance and rating of filters for a single installation, e.g.
HVDC, where the station components, filters, etc., and emission are symmetrical and with negligible
inter-sequence coupling.
2.2.2.5 Unbalanced Harmonic Penetration

The power system cannot always be adequately or accurately represented using a balanced
representation. According to [3], field tests show that at harmonic frequencies, unbalanced power
system operation is the norm rather than the exception. Unbalanced harmonic penetration is the
extension of the balanced system formulation into an unbalanced (or phase-wise) framework, as
described in [7]. The user can define harmonics in positive, negative or zero sequence components or
as a combination of the three. The system admittance matrix is formulated and solved as a three-phase
system representation and harmonic propagation in each phase can be analysed. This allows
asymmetry caused by different mutual coupling between phases, transposition etc., cable sheath
system asymmetry etc. to be considered. Two common analysis configurations are described below:
 Unbalanced Harmonic Penetration based on a Balanced Fundamental Load Flow

A balanced, positive sequence fundamental load flow forms the basis for the analysis. Load flow positive
sequence voltages and currents are transformed into phase quantities and are used in the harmonic
penetration study. The sequence characteristics of user-defined harmonic injections may be predefined
based on harmonic order. Unbalanced (phase-wise) harmonic voltages and current injections can be
modelled. Results are available as phase and sequence quantities.
 Unbalanced Harmonic Penetration based on an Unbalanced Fundamental Load Flow

An unbalanced fundamental load flow forms the basis for the analysis. The load flow phase voltages
and currents are used in the harmonic penetration study. The sequence characteristics of user-defined
harmonic injections may not be predefined based on harmonic order. Unbalanced (phase-wise)
harmonic voltages and current injections can be modelled. Results are available as phase and sequence
quantities. This provides the greatest level of detail of the harmonic penetration methods presented
here.
23
Multi-phase harmonic analysis is preferred or required under certain network conditions or topological
arrangements. In unbalanced (and balanced) systems for example, non-zero-sequence triplen
harmonics can penetrate regardless of transformer connection [3]. For the analysis of large networks,
the use of a multi-phase formulation may involve long computation times and large memory
requirements [16].
2.2.2.6 Selection of Balanced or Unbalanced Penetration Method

When beginning a new study, the engineer must make a choice between setting up a balanced or
unbalanced system model and conducting a balanced or unbalanced penetration study. The choice
depends not only on the nature of the specific network or component under study, but also on the data
available.
When using a balanced approach, the assumption is that the power system is perfectly symmetrical
(meaning equal self- and mutual impedances and shunt admittances at all frequencies of interest for all
components in the system). This assumption is often invalid in practice due to the influence of e.g.
asymmetrical transmission lines. The evaluation of choice of method for the specific study will be highly
network dependent and the onus is on the engineer to make the correct choice. When analysing
asymmetrical transmission lines near the location of interest , if a resonance peak is present in the
associated frequency range of interest, an unbalanced study type will give more accurate results [17],
[18].
Example: Difference between a balanced and an unbalanced propagation study on a

high voltage cable system
The relative difference between the apparent phase impedances (Z h=Uh/Ih) is strongly frequency
dependent due to frequency dependent mutual coupling. Therefore, at some frequencies the inter-
sequence coupling is stronger than at others. This is illustrated in Figure 2-2 where the harmonic
voltage as a function of the harmonic order is displayed at the receiving end of an unloaded 45 km
220 kV cross-bonded cable system laid in flat formation, directly in the ground. The cable is energized
by a 1 pu balanced harmonic voltage (positive sequence) at all harmonics from the 2 nd to the 20th.
If a symmetrical line model (decoupled model) was used, all harmonic voltages at the receiving open
end would be purely positive sequence. Instead it is seen in the lower figure that significant inter-
sequence coupling occurs at several harmonics. For instance, at the 4 th harmonic, the negative
sequence component has a value of 33% of the positive sequence component. At the 17th and 20th
harmonic, the negative sequence components are almost equal in magnitude to the positive
sequence component. As indicated by the results in Figure 2-2, for many harmonic frequencies the
sequence impedance matrix is diagonal dominant, and the decoupled approximation can be used
with reasonable accuracy. At other frequencies, the off-diagonal values can be significant compared
to the diagonal values, and large errors are introduced if a decoupled approximation is used. This is
for instance the case at the 20th harmonic where the highest phase voltage is 96% higher than the
positive sequence voltage. A frequency scan showing the magnitude of the phase impedances seen
from the sending end of the cable system is presented in Figure 2-3.
The frequency scan illustrates that inter-sequence coupling is strongest at frequencies where the
phase impedances differ between each other (compare Figure 2-2 to Figure 2-3). The effect is
particularly pronounced at and around resonance where the relative phase differences can be very
large. The three phases of the flat formation arrangement have naturally different mutual impedances
due to the physical placement. This effect is seen in the phase impedance from around the 3 rd to 4th
harmonic and again from the 14th harmonic and to the end of the displayed part of the spectrum. Had
the cable system been in a symmetrical trefoil formation, the three phase impedances would have
been equal and no inter-sequence coupling would occur.
As indicated in this example, a lightly-loaded long cable system placed in an unsymmetrical
arrangement is prone to inter-sequence coupling due to unequal mutual coupling at resonance
conditions. This can be very significant if for instance harmonic limits are issued based on calculated
harmonic levels at the end of such a cable.
24
Figure 2-2 Phase and Sequence Voltage in Inter-sequence Coupled Cable Radial Model (Receiving
End)
Figure 2-3 Frequency Scan of the Apparent Impedance of Phases A, B and C
This material is further addressed in Appendix A.
2.2.3 Harmonic Load Flow

There is no strict definition of what a harmonic load flow entails. Most harmonic “load flow” algorithms
available in commercial software packages are actually harmonic penetration methods. To avoid
confusion, linear solution methods described in Section 2.2.2 are referred as “harmonic penetration
studies” in this document, whereas true load flow solution methods for all harmonic orders are referred
to as “harmonic load flow”.
Harmonic load flow analysis performs the repeated solution of the power system admittance matrix at
each defined harmonic order; i.e. it is a reformulation of the standard 50/60Hz iterative load flow problem
normally used in planning studies (e.g. with Newton-Raphson or Gauss-Seidel solution methods) [6].
Using this method, the voltage or current dependence of network devices can be included. However,
this requires detailed knowledge of the devices in the system.
Full Newton-Raphson load flows (i.e. a non-linear problem) calculate the mismatches at each iteration
(until the convergence criteria has been met), and the Jacobian matrix is reformulated accordingly.
Newton-Raphson harmonic load flow formulations are available [3], and offer highly detailed
consideration of network behaviour, but are not widely used.
25
Harmonic load flow solutions can be formulated in the frequency domain, time domain, or a hybrid
combination of both, and can be applied on a single-phase or multi-phase basis [8]. An overview of
various methods including detailed mathematical descriptions is provided in [4].
2.3 Time-Domain Methods

Time-domain methods are characterised by the representation of system behaviour using differential
equations.
A time-domain simulation is carried out until the system reaches steady-state. A brute force solution of
the power system can be obtained by integrating the system differential equations after the transient
response has diminished [4].
Time-domain methods offer the advantage of accurate consideration of nonlinear devices and their
controllers, however according to [3], it may be difficult to model frequency-dependent parameters and
to obtain network impedance envelopes.
2.3.1 Application and Limitations

For cases where advanced modelling is required (for example, detailed analysis of devices operating in
their nonlinear region), the time-domain method can provide enhanced capabilities such as
consideration of system nonlinearities and complex control functions [3], [8].
Time-domain methods may be required for specialised studies involving controlled power electronics,
depending on the purpose. An example of this is the control of power converters. The operating point of
the converter adjusts due to the combination of external commands and measurement feedback loops,
and in doing so, alters the harmonic impedance of the device. While this could also be captured in the
frequency domain, the converter model must be modified for different operating conditions as mentioned
earlier for example via a look-up table. In general, the time domain solution will allow the inclusion of an
accurate model of DC controls.
During the initial design stage, when the detailed converter controls are not available, both frequency-
domain and time-domain methods could be used as an initial approximation for system level studies,
but with awareness of the potential errors. For more specific converter design studies, the harmonic flow
across the converter (involving a change of harmonic order due to modulation) could be of interest.
Time-domain analysis would be more appropriate in these cases. Coupling of harmonic frequencies
also requires detailed time-domain simulation. However, such frequency coupling effects are minor and
can usually be ignored.
It should be noted that for voltage source converters, low order harmonics are inaccurately represented
by time-domain modes if the switching element dead-time is not considered in the model. Many model
providers chose to neglect the dead-time as it requires a very small time-step. Hence there is a trade-
off between simulation speed and accuracy for steady-state harmonics.
Time-domain methods can provide more detailed modelling of non-linear electronic devices and
saturation of magnetic devices than frequency domain methods. However, it should be noted that their
accuracy is dependent on the type of analysis and the available input data. The literature seems to be
in consensus that time-domain methods are often inefficient at reaching a steady-state solution [3], [4],
[6] and [9]. In terms of computational efficiency, time-domain methods may only be appropriate for
systems with sufficient damping [4]. References to the use of acceleration techniques ([19], [20]) for the
solution process are provided in [8]. However, the general effectiveness of such techniques for purely
time-domain solutions is questioned in [3], in particular for the modelling of large systems. Hence, as
noted in [3], [4] and [9], time domain simulation can be very slow to initialise some steady-state models
(e.g. synchronous machines) due to large time constants or if the network under consideration is large
and contains short line segments. Some commercial packages offer the possibility of steady-state
initialisation. This is usually incompatible with user models (black-box models) such as OEM (Original
Equipment Manufacturer) HVDC or wind turbines models. Studies where such models are involved are
often one of the motivating factors for application of the time domain, however the issue of a long
initialisation time remains.
26
2.4 Hybrid Methods

Combinations of frequency-domain and time-domain solutions methods are referred to as hybrid
methods [3], [6]. These are the most powerful and provide great flexibility in terms of modelling
complexity [8] but are generally not available in commercial software packages.
Hybrid methods are similar in formulation to frequency-domain methods, but allow the representation
of, for example, non-linear loads by differential equations [8]. They are required for the accurate
modelling of non-linear components. According to [3], these methods may benefit significantly in terms
of performance from the use of acceleration techniques.
Hybrid methods are generally formulated such that the frequency- and time domains are interfaced via
one of the following means [21]:
1. Full solution of the network in the frequency domain followed by conversion of the frequency-domain
variables to the time domain;
2. Iteration between the frequency domain and the time domain; or
3. Simultaneous solution in the time domain and frequency domain, thereby considering the
frequency- and time-dependence of the variables.
2.5 Harmonic Domain

The harmonic domain provides a general framework in the steady-state which models the coupling
between phases and between harmonics [22]. The harmonic domain technique is a full Newton method
which offers the explicit representation and iterative solution of nodes, phases, phase unbalance, linear,
nonlinear and time-varying components, harmonics and harmonic cross-coupling effects [4].
Linearization around an operating point yields a Norton harmonic equivalent which represents the phase
unbalance and harmonic cross-coupling effects.
According to [4], the harmonic-domain solution method is numerically robust with good convergence
and [4] provides an exhaustive list of references to harmonic-domain power system models.
Representation of converters in the harmonic domain allows the consideration of the modulation of
switching instants and converter control functions, caused by AC voltage and DC current distortion.
The representation of a component in the harmonic domain requires considerable skill and effort, and
the associated program required to calculate the non-sinusoidal periodic steady-state may be highly
complex and difficult in terms of use and maintenance [3]. However, this method provides a valuable
way to assess the harmonic interaction between an AC system and large power converters.
2.6 Recommendations
The frequency domain solution methods provide adequate results for many applications, along with
straight-forward modelling of harmonic injections and frequency dependence, making this study domain
accessible and attractive to many engineers. It is generally numerically-robust and the computational
overheads of executing analyses in this domain are usually minimal and many study cases may be
considered with ease. For applications such as the calculation of AC system wide harmonics, evaluation
of AC filter ratings, impedance scans to evaluate potential resonance issues, modelling of network
harmonic impedance envelopes, DC-side harmonic profiles and filter design (amongst others), the
frequency domain is adequate and highly-recommended. However, if the nonlinear region of device
operation and any associated interaction should be modelled, the frequency domain may not provide
the engineer with sufficient accuracy.
The time domain provides a highly-accurate and powerful means to model the nonlinear time-varying
behaviour of devices. It can assist in identifying harmonic instability and other nonlinear interaction and
in performing analyses of AC/DC harmonic interaction for HVDC schemes (for control tuning and DC-
side harmonic performance validation). Significant effort and experience may be required on the part of
the engineer to incorporate all nonlinear models (e.g. transformer saturation, converter controllers
operating in their non-linear region, etc.) and to accurately represent frequency-dependent parameters.
Time-domain studies require validation that the system has reached steady-state before the parameters
of interest can be extracted (e.g. voltages and currents). For accurate representation of the low order
harmonic emission of a voltage-source converter, the dead-time of the switching elements must be
considered in the model. The computational overheads of time-domain simulations are larger than of
those performed in the frequency domain. These can be improved with automated execution,
exploitation of parallelised calculations and dedicated hardware. It should be mentioned that some
27
commercial programs offer steady-state initialization routines that can significantly reduce simulation
time. However, such routines are however often not applicable for custom models (for instance OEM
HVDC models).
The combination of the advantages of both the frequency- and time domains was the driving force
behind the development of the hybrid domain. At the time of writing, the use of this domain in commercial
software packages generally requires the manual interfacing of frequency- and time-domain
functionality, requiring experience on the part of the engineer. It may present difficulties in cases where
models are supplied by different vendors, due to complex model interfaces.
The advantages, disadvantages and possible applications of the three domains are detailed in Table
2-1.
Table 2-1 Summary of Study Domains and their Applications
Advantages Disadvantages Applications

 Numerically robust.  Limited modelling of  Studies aimed at allocating harmonic
 Generally good nonlinear devices and emission limits to new customer
convergence. feedback functions such connections to the power grid.
 Simple definition of as converter controls or  Calculation of AC system harmonics
harmonics. active filters. profiles.
 Intrinsic modelling of  Evaluation of AC filter ratings due to
Frequency Domain
frequency installations of nearby harmonics

dependence. load; e.g. new HVDC.
 Usually provides  Impedance scan of load bus; e.g. to
sufficient accuracy and evaluate a potential resonance issue
efficiency for devices between the AC system and installed
operating in linear capacitors by customers.
region.  Calculate DC-side harmonic profiles
 Time-effective method  Harmonic propagation studies.
allowing to cover many  Deterministic and stochastic
study cases (e.g. set harmonic summation and
points, outages, cancellation studies including fixed or
contingency cases). random phase angle.
 Modelling of nonlinear  Slow initialisation and  AC/DC harmonics cross-modulation
devices and time- long execution time to study of a HVDC scheme (for control
varying properties. achieve harmonic tuning and AC performance
 Can accurately steady state response. validation)- this will enable accurate
integrate power Often impractical for modelling of converter control and
electronic converters larger systems. transformer saturation.
with their control.  Requires effort to  System-level studies where detailed
 Can be run in real-time incorporate all non- modelling of power electronic
Time Domain
on specific and linear models correctly converters (including saturated,

dedicated hardware. (e.g. transformer passive components and/or
saturation and power controller non-linearity) is needed
electronic converters).  Active filtering implementation and
 Difficult to represent studies due to e.g. highly non-linear
frequency-dependent controller.
parameters. It requires  Real-time simulations in a realistic
advanced techniques. and discrete representation of
converters.
 Incident investigation after a
harmonic problem has been
observed in the system.
28
 Incorporates non-  Time consuming  Currently, applications of this method

linearities with the implementation in large are limited to academic research.
efficiency of frequency power systems.
domain methods.  Difficult to implement in
Hybrid Domain
 Benefit from use of cases where models are

acceleration delivered from different
techniques [3]. vendors due to complex
model interfaces.
 Risk of combining the
disadvantages of both
frequency and time
domains.
 It is not established in
industry.
29
30
3. Classical network element models

3.1 Overhead Lines
This section begins with a discussion of basic overhead line (OHL) theory, followed by calculated
examples which demonstrate the sensitivity to numerous factors, both for individual lines and for
analysis of a complete network.
3.1.1 Modelling Considerations

The modelling of OHLs for harmonic propagation studies is very well documented in technical literature.
These models consider long-line effects, frequency dependency and line imbalance. A brief synopsis of
these models is included in this section, with sufficient background information to understand the most
critical aspects of OHL modelling above power frequency. For more detail, the reader is encouraged to
consult references [23], [24], [25].
An overhead line comprises series parameters (resistance and reactance) and shunt parameters
(conductance and susceptance), which are distributed over the entire length of the circuit and are
affected by frequency. These electrical parameters can be derived from the line geometry and the
conductor data for each harmonic frequency. Higher frequencies increase the electrical distance of the
circuits, therefore long-line effects must be included when modelling harmonic distortion. This requires
cascading several lumped-parameter (nominal-PI) sections or the use of distributed-parameter models
(equivalent-PI) derived from wave propagation equations.
In early studies, the frequency dependency of the line parameters tended to be neglected [26] as a first
approximation motivated by the lack of computing power. The lines were simply represented using PI-
models calculated at power frequency, either as single sections (short lines) or as a series connection
of multiple sections (long lines). Later work ([2] and [27]) recognised the importance of including
frequency dependency in the line models as well as improving the representation of long-line effects
(i.e. distributed parameters) by using hyperbolic functions. To reduce computing power requirements in
cases where zero sequence harmonic penetration and damping at resonant frequencies were not
critical, a compromise could be made by neglecting frequency dependency and simply computing the
impedance and admittance matrices at the dominant frequency of interest. Nowadays, computing power
is rarely an issue for typical applications and most harmonics analysis tools can represent single- and
multi-phase circuits, capturing both the frequency dependency and the distributed nature of the electrical
parameters.
A simple representation of an OHL, approximated as lumped parameters, is shown in Figure 3-1
(nominal PI), where [Z(ω)] and [Y(ω)] are the shunt impedance and admittance matrices calculated at
the angular frequency, ω. The lumped parameter approximation is adequate for short line lengths and
low frequencies. However, when the line length or the frequency of interest increase, the lumped
parameter representation becomes very inaccurate as voltage and currents are affected by standing
wave effects and a distributed parameter model is needed. This effect is illustrated in section 3.1.5.1,
where Figure 3-5 shows the errors introduced with the simple lumped parameter model as a function of
frequency and circuit length. The reader is warned that the range of applicability of the lumped parameter
model is limited to low frequencies and short circuit lengths. As a general guideline, distributed
parameters should be used for line lengths exceeding 240/h (km), where h is the harmonic order of
interest [27].
Figure 3-1 Overhead Line Nominal-PI Model

The main points for consideration when choosing the most appropriate representation of an OHL for
harmonics studies are:
1. Selection of single or multi-phase model: A single-phase representation based on positive
sequence data is adequate to represent perfectly balanced systems, with the benefit of providing a
significant reduction in computational burden. However, most typical OHL constructions are highly
asymmetrical, presenting different inter-phase impedances and different resonant frequencies on
31
each phase, which can result in a large unbalance in the voltages and currents at certain harmonic
frequencies (see Section 2.2.2.6 for an illustrative example). When telephone interference is of
concern, zero sequence parameters are important. Considering the standard computing capabilities
available nowadays, the use of multi-phase models is normally the default option. Circuit
transpositions should be represented in detail, as they can influence unbalance at certain harmonic
frequencies. Likewise, any significant inter-circuit coupling should be captured, for instance parallel
circuits sharing right of way or close to metallic telephone lines. The size of the [Z] and [Y] matrices
will increase from 3x31 (single circuit) to 6x6 (double circuit) or higher for complex rights of way.
2. Capture the frequency dependency of the line parameters: The main effects contributing to
frequency-dependency are the conductor skin effect and the earth-return paths for zero sequence
currents.
3. Choice of nominal-PI (lumped parameters) or equivalent-PI (distributed parameters/long-line
effects). As discussed above, lumped parameter representations are only adequate for short lines
and low frequencies. However, when the line length approaches the wavelength of the frequency of
interest, the errors introduced are significant and a distributed parameter representation is required
(highlighted in section 3.1.5.1). Cascading nominal-PI sections is an alternative technique to
represent long lines. The more sections used, the closer the model approaches the distributed
nature of the parameters. In theory, increasing the number of cascaded sections to infinity will turn
the lumped parameter model into the distributed parameter model (demonstrated in section 3.1.5.2).
On the other hand, significant errors can be introduced if the sections are too large in relation to the
circuit length and the frequency range of interest. These errors increase with frequency. A practical
balance must be reached between accuracy and computational burden caused by the increased
number of intermediate nodes along the line. Section 3.1.5.2 discusses this topic in detail. Note that
in some cases the use of the cascading technique might be necessary; for example, when a
harmonic voltage/current profile along the line is needed.
The process of developing an OHL model for harmonic propagation studies can be divided into two
steps:
Step 1: Calculate the lumped electrical parameters from the circuit geometry and conductor physical
characteristics, capturing its frequency dependency. This requires calculation of the shunt
impedance and admittance matrices for each frequency within a selected range.
Step 2: Introduction of long-line effects.
The above functionality is typically built into commercial harmonic analysis software tools. The user only
needs to input the geometrical layout and conductor characteristics and specify the type of model
required for the study: lumped (nominal-PI) or distributed (equivalent-PI).
The distributed parameter (equivalent-PI) model is shown in Figure 3-2 and the reader is referred to [3]
for definitions of the parameters.
Figure 3-2 Overhead Line Equivalent-PI Model
1
Under the assumption that ground wire reduction does not introduce significant error for the frequency range
of interest and type of study being undertaken [28]
32
3.1.2 Consideration of Skin Effect

The skin effect is an AC, frequency-dependent phenomenon resulting in an increased current density
on the conductor surface with increasing frequency. The current mainly flows in the "skin" of the
conductor, which is the region between the conductor’s outer surface and a level called the “skin depth”.
The practical implication of this is an increase in the effective resistance of the conductor at higher
frequencies due to a reduction in the conductor effective cross-section, and a decreased internal
inductance.
Due to the relatively small value of the conductor resistance compared to its reactance for typical
transmission circuits, the skin effect is only noticeable at (or near) resonance conditions, where the line
impedance is dominated by the resistive component. At resonance, changes in the resistive component
of the line have a significant impact on the calculated voltage peak. Therefore, it is important to represent
the skin effect when dealing with harmonic resonances.
Key Point:
 Skin effect should always be modelled when calculating the series impedance for OHLs in harmonic
studies.
Assuming homogeneous non-ferrous conductors of tubular shape (as an approximation for ACSR
conductors), the internal impedance can be formulated using equations based on Bessel functions [3].
Some utilities instead model the skin effect using simpler correction factors. An approach suggested in
[2] and [27] is shown below.
𝑅(𝜔) 0.035 ∙ 𝑀2 + 0.938 𝑀 < 2.4

={
𝑅𝑑𝑐 0.35 ∙ 𝑀 + 0.3 𝑀 ≥ 2.4
Equation 3.1
𝑓 ∙ 𝜇𝑟
𝑀 = 0.05012 ∙ √
𝑅𝑑𝑐
where µr is the relative permeability of the cylindrical conductor, 𝑓 is the frequency in Hz, and 𝑅𝑑𝑐 is the
conductor DC resistance, in Ω/km. Correction factors developed in France and the UK are presented in
Table 3-1 ([3], [23]).
33
Table 3-1 Corrections for Skin Effect in OHLs
Company Voltage (kV) Harmonic Order Resistance
400, 275 3.45ℎ2

𝑅1 (1 + )
Based on 0.4 sq.in. steel-core 1 = ℎ ≤ 4.21 192 + 2.77ℎ2
al. conductors
NGC 4.21 < ℎ ≤ 7.76 𝑅1 (0.806 + 0.105ℎ)
ℎ > 7 ⋅ 76 𝑅1 (0.267 + 0.485√ℎ)

0.646ℎ2
132 𝑅1 (1 + )
192 + 0.518ℎ2
3.45ℎ2
400, 225 1=ℎ≤4 𝑅1 (1 + )
192 + 2.77ℎ2
4<ℎ<8 𝑅1 (0.864 − 0.024√ℎ

RTE + 0.105ℎ)
8<ℎ 𝑅1 (0.267 + 0.485√ℎ)

0.646ℎ2
150, 90 𝑅1 (1 + )
192 + 0.518ℎ2
3.1.3 Representation of Line Transposition and Inter-Circuit Coupling

Line transposition effectively reduces the circuit unbalance at the fundamental frequency. However, it
has been shown [29] that the effectiveness of transpositions deteriorates substantially with increasing
frequencies for electrical distances above 1/8th of their respective wavelengths. Therefore, the presence
of transposition can exacerbate voltage and current unbalance at certain harmonic frequencies, with
significant differences in the location of resonant frequencies for each phase. High voltage amplification
has been observed for electrical distances equal to 1/4 of the wavelength. To capture this effect, the
circuit needs to be represented as a cascaded connection of homogeneous un-transposed sections,
using the relevant PI model (nominal or equivalent). Similarly, any coupling with parallel circuits needs
to be captured in the homogeneous circuit sub-division.
An example is shown in Figure 3-3 illustrating two transposed circuits and their phase designations with
parallel coupling between some sections of their lengths. The circuit sub-division caters for all changes
in circuit geometry.
S-1 S-2 S-3 S-4 S-5 S-6
A B C
Circuit 1 B C A
C A B
B C A
C A B
A B C
Circuit 2
Section 1 Section 2 Section 3 Section 4 Section 5 Section 6

Circuit 1 A-B-C A-B-C B-C-A B-C-A C-A-B C-A-B
Circuit 2 B-C-A B-C-A C-A-B C-A-B A-B-C
Figure 3-3 Cascaded Representation of Homogeneous Line Sections
34
3.1.4 Variation of Earth Resistivity with Depth

The term ‘earth resistivity’ is used here rather than ‘soil resistivity’. The latter term is commonly used in
the context of the design of earthing systems, where the interest lies in the electrical behaviour of the
top layer of the earth’s surface, which is usually soil. However, for the study of harmonic impedances of
transmission lines, the characteristics of the earth at much greater depths become an important
consideration. Earth resistivity is a necessary parameter for the development of a line model. Depending
on the model used, users can either section the model and use different values for each section or
assume an average value to represent the entire circuit length. This parameter is dependent on the local
ground conditions/composition and can vary significantly along the circuit length. Earth resistivity is
affected by weather conditions (temperature and humidity).
Typically, the geological structure of the earth will consist of several distinct strata of diverse types of
soil and rock, each with distinctly different resistivities. The resistivity of these strata, and an estimate of
their depth and thickness, can be measured in the field using one of several well-known methods (e.g.
Wenner [30], Schlumberger [31]). The significance of the variation of resistivity with depth becomes
clear with reference to the commonly-used Dubanton equations [32] for the calculation of self- and
mutual impedance at harmonic frequencies, and which contain the frequency-dependent complex
depth, 𝑝, where
1 Equation 3.2
𝑝= 𝑗𝜔𝜇0
[m]
√
𝜌
and represents an imaginary mirroring plane below the earth’s surface. Clearly there is a circular
relationship: the depth of the mirror plane varies with the resistivity, but the resistivity varies with depth.
Reasonably accurate solutions to this problem have been developed, normally simplifying the earth
strata to a more manageable equivalent of three different layers.
Such detailed calculation is unrealistic during a typical study involving many transmission lines of
considerable length. While the usual practice is to take a typical average earth resistivity for the area,
the engineer should be aware of the substantial errors which may be introduced; particularly in
calculations involving zero-sequence parameters or mutual impedance.
Illustration 1
Using Equation 3.2 with homogenous earth resistivity of 200 Ohm-m and frequency 550 Hz, the
depth of the mirror plane would be 214 m. However, for 1000 Ohm-m the depth would be 480 m.
Applying the Dubanton equations [32], the impact of these two resistivities on the mutual impedance
at 550 Hz between two conductors at say 50 m separation and 18 m height would be approximately
83%, while the impact on the self-impedance of one conductor would be around 13%.
Illustration 2
The result of an actual field measurement of earth resistivity is shown in Table 3-2 as an illustration
of the wide variations of resistivity with depth which occur in reality. Simply using an average value
of resistivity for line calculations is a very substantial approximation.
Table 3-2 Variation of Earth Resistivity with Depth
Thickness of layer (m) Resistivity (Ωm)

2.2 218
4.2 776
33.5 20
450 291
1547 16
Below 232
35
3.1.5 Examples: Single 220kV Overhead Line Circuit

This section provides guidelines for the modelling of OHLs as single network components. For the
purposes of this section the calculation results are presented in the form of short-circuit impedances.
These harmonic impedances are the result of frequency scans injecting 1 A (at each frequency within
the range of interest) at the sending end with the receiving end short-circuited as shown in Figure 3-4.
The measured voltage (in V) corresponds to the impedance (in Ω) at each frequency. The 220kV circuit
data is presented in 0.
Figure 3-4 Setup for Line Model Frequency Scan Calculation
3.1.5.1 Line Models as a of the Circuit Length

This example compares the response of two OHL models for increasing circuit lengths:
 Lumped parameter model (nominal-PI representation) with frequency dependency of parameters
only;
 Distributed parameter model (equivalent-PI representation) with frequency dependency and long
line effects.
In this example the magnitude and angle of the circuit impedance have been computed using the set-
up shown in Figure 3-4. A frequency scan has been carried out with each line model and increasing
circuit lengths. The calculation results are presented in Figure 3-5.
The following observations can be made:
 The lumped parameter representation is only capable of representing the first resonant point. The
frequency of this resonance is always slightly below the first resonant frequency obtained with the
accurate distributed parameter model. However, as the circuit length increases, the difference
between the two resonant points diminishes. Long line effects causing subsequent resonant points
are not able to be represented with the lumped parameter model.
 The amplitude of the resonant peak obtained with the lumped parameter model is very similar (albeit
slightly higher) than the equivalent first resonant point obtained by the distributed parameter model.
 The accuracy of the lumped parameter model is only adequate for harmonics studies in which the
main frequency of interest is well below the first resonant point of the circuit (a factor of 5 can be
used as a rule of thumb). It has been shown in Figure 3-5 that this frequency can be very low for
high circuit lengths – i.e. at the 5th harmonic order for a 300 km line. Therefore, the use of a
distributed line parameter model (including long line effects) is recommended as the default option.
The following additional observations can be made about the frequency response observed with the
accurate distributed parameter model:
 The circuit harmonic impedance exhibits an infinite series of resonant points (series and parallel). It
can be demonstrated that these points are uniformly distributed over the frequency spectrum at
intervals of one quarter of its wavelength at fundamental frequency (λ 50/4) [3]. In this example, the
circuit wavelength is λ 50 = 5850 km and the observed intervals between “peaks” and “valleys” are
exactly as predicted: h=29.3 for 50 km length, h=14.8 for 100 km length, etc.
 The frequency of the first resonant point drops as the circuit length increases. This is because the
parallel resonances occur at the half-wavelength frequencies of the line, considering on its length.
 The amplitude of the resonant peaks drops as the frequency increases. This upper asymptotic
behaviour of the decreasing amplitudes is dependent upon the characteristic impedance of the line,
and the hyperbolic cotangent of the attenuation constant and the line length [3].
36
Figure 3-5 Comparison of Lumped and Distributed Parameter Line Model for Various Line Lengths
(Blue=Lumped Parameter Model; Red=Distributed Parameter Model)
37
Key Point:
 A frequency-dependent, distributed parameter model capturing long line effects should be default
option for modelling OHL circuits in harmonics studies, for all but the very shortest lines.
3.1.5.2 Cascading Multiple Lumped Parameter Line Sections

This example illustrates the effect of splitting a lumped parameter representation into multiple cascade
sections.
A 250 km long circuit has been selected for this example. The OHL circuit has been represented using
the accurate distributed parameter line model (as reference) and by cascading lumped parameter line
sections of uniform length. Two cascading cases have been considered: (i) 5 x 50km sections and (ii)
10 x 25 km sections. The results of the comparative frequency scan analysis are presented in Figure
3-6.
Figure 3-6 Distributed vs. Cascaded Lumped Parameter Model Representation (Length = 250km)

 Multiple series and parallel resonance points are obtained with the cascading approach, in contrast
with the single lumped section (Figure 3-5) which can only approximate the first resonant peak.
 Both cascading cases can accurately reproduce the first resonant peak. In both cases, frequency
and amplitude of the first peak match the results obtained with the accurate distributed parameter
model. This is a significant improvement from the results presented in Figure 3-5 (250 km case) for
one single lumped section.
 However, as the frequency increases, the response of the cascading representations deviates from
the distributed parameter model. The response of the 10-section case is acceptable up to the third
resonant peak, while the 5-section representation starts introducing high errors earlier in the
frequency spectrum.
Key Point:
 Cascading sections improves the accuracy of the lumped parameter model up to a certain
frequency, depending on the circuit length and the number of sections. Increasing the number of
sections improves the frequency range for which the model is acceptable.
3.1.5.3 Skin Effect

This example illustrates the influence of skin effect in the frequency response of the overhead line model.
A 250 km long circuit has been selected for this example. The OHL circuit has been represented using
the accurate distributed parameter line model with skin effect (using Bessel functions) and without skin
effect. The results of the comparative frequency scan analysis are presented in Figure 3-7.
38
Figure 3-7 Skin Effect Modelling (Length = 250km) (Blue: Consideration of Skin Effect; Red=Neglection of
Skin Effect)
 The skin effect is only noticeable at the resonant points, providing increased damping. In this
example, the impedance of the first resonant peak calculated without skin effect is 45% higher than
with skin effect.
 There is an attenuation of amplitude and phase angle when the skin effect is included. This means
that if the skin effect is neglected, the error thereby introduced increases with frequency.
 Skin effect also results in a slight upward shift of resonant frequencies. This effect is only noticeable
at high frequencies (see the third resonant peak in Figure 3-7) and it is caused by an effective
reduction in the internal inductance of the phase conductors.
Key Point:
 Neglecting skin effect leads to underestimation of circuit damping at resonant frequencies.
3.1.5.4 Earth Resistivity

This example illustrates the effect of different earth resistivity values on the positive 2- and zero-sequence
harmonic impedance of the circuit. The positive-sequence impedance has been calculated as shown in
Figure 3-4, with a balanced three-phase current source. The zero-sequence impedance was calculated
using a similar setup but changing the current injection to three single-phase current sources with the
same phase angle.
A 250 km long circuit was selected for this example. The OHL circuit is represented using the accurate
distributed-parameter line model with a variable earth resistivity in the range of 50 Ω.m to 1000 Ω.m.
The results of the frequency scan are shown in Figure 3-8 and Figure 3-9 for the positive- and zero-
sequence impedance, respectively.
2
Note that this section refers to positive sequence only for simplicity. The reader is advised that conclusions
related to positive sequence can be extended to negative sequence for the purposes of OHL modelling.
39
Figure 3-8 Effect of Earth Resistivity on Positive Sequence Impedance (Length = 250km)
Figure 3-9 Effect of Earth Resistivity on Zero Sequence Impedance (Length = 250km)
40

 Earth resistivity does not have a significant impact on the positive-sequence harmonic impedance
for the range of frequencies analysed. The positive sequence resistance should remain reasonably
constant until the frequency at which the skin effect causes it to increase [33]. The only noticeable
differences among the four curves presented in Figure 3-8 occur at the resonance points. Higher
earth resistivity results in higher impedance at resonance (i.e. lower damping). The differences
increase with frequency. This is due to the influence of the Carson correction terms (described in
[33]) in the calculation of the self- and mutual-impedances of the series impedance. These account
for the earth return effect and are frequency dependent. In this example, the maximum difference in
amplitude is just 3.6%, observed at the third harmonic resonant peak between 50 Ω.m and 1000
Ω.m.
 The effect of earth resistivity on the zero-sequence harmonic impedance is illustrated in Figure 3-9,
which shows that significant differences in peak amplitude and resonant frequencies can be
observed even in the low frequency range. The zero-sequence resistance and inductance are highly
frequency dependent, due to a greater dependency on the Carson correction terms than in the
positive sequence. These terms tend to zero for low values of earth resistivity. Therefore,
assumptions of a high earth resistivity lead to lower damping through the entire frequency spectrum.
The location of the resonant peaks is also highly dependent on the assumption of earth resistivity
and can shift by several harmonic orders.
Key Point:
 Accurate estimation of earth resistivity is not necessary if only positive-sequence magnitudes are of
interest. If unbalanced conditions need to be assessed, an accurate estimation of earth resistivity is
needed to derive a correct zero-sequence model. For long circuits crossing areas with very different
earth conditions, it may be necessary to split the model into sections calculated with representative
earth resistivities.
3.1.6 Average Conductor Height above Ground

One of the input parameters of any OHL model is the average height of each conductor above ground.
This is commonly given as [33]:
1 Equation 3.3
ℎ = ℎ𝑒𝑖𝑔ℎ𝑡 𝑎𝑡 𝑚𝑖𝑑𝑠𝑝𝑎𝑛 + ∙ 𝑠𝑎𝑔 [m]
3
However, this value is dependent on several physical factors such as terrain, river/motorways crossings,
span length, conductor tension, ambient temperature, etc. and electrical factors such as circuit loading.
Typically, an average figure is assumed to account for these uncertainties in the practical development
of a model. This example illustrates the sensitivity of the harmonic impedance to variations in conductor
height of ±10% of the assumed average value (as per Equation 3.3). A 250 km long circuit was selected
for this example. The OHL circuit was represented using the accurate distributed parameter line model.
Three cases were compared:
(i) Low average conductor height = 0.9 x 16.4 = 14.76 m
(ii) Medium average conductor height = 16.4 m
(iii) High average conductor height = 1.1 x 16.4 = 18.04 m
The results of the frequency scan analysis are presented in Figure 3-10.
41
Figure 3-10 Effect of Average Conductor Height on the Positive Sequence Impedance (Length = 250km)

 Small variations of conductor height above ground do not have a significant impact on the
harmonic impedance for the range of frequencies analysed. The only noticeable difference
between the three curves presented in Figure 3-10 is a minor shift in resonant frequencies,
which increases with harmonic orders. Lower conductor heights result in slightly lower resonant
frequencies.
 The maximum resonant frequency shift observed in this example was at the third resonant peak,
with just 2.1 Hz difference for a height variation of 3.28m.
Key Point:
 Small variations in the average conductor height above ground (within typical ranges) do not have
a significant impact on the harmonic impedance of the circuit.
3.1.7 Summary
Various aspects of harmonic modelling of OHLs were discussed in this section and supporting examples
provided clear illustrations of the most relevant aspects. A summary of the analyses presented is given
in Table 3-3.
A comparison of line models as a function of the circuit length was presented, highlighting the
importance of using frequency-dependent, distributed parameter models when performing harmonic
studies. Modelling the frequency dependence of the series resistance (due to the skin effect) was shown
to have a significant impact on simulation results. Cascading PI sections was demonstrated to improve
the accuracy of the lumped parameter model up to a certain frequency, depending on the circuit length
and the number of sections modelled.
Values for the earth resistivity should be as accurate as possible to avoid significant approximation
errors, which significantly affect the zero-sequence harmonic impedance. This is relevant if unbalanced
conditions need to be assessed. The impact of small variations in the average conductor height above
ground was shown to be minor.
42
Table 3-3 Summary of Sensitivity Analysis: Impacts on Harmonic Impedance
Parameter Sensitivity Level of Impact Effect

Analysis
Skin Effect Consider/Neglect High Skin effect must be
considered to accurately
estimate damping.
Earth Resistivity Consider/Neglect High Should be considered if
accurate zero sequence
(only for
model required.
unbalanced
analysis)
Model Representation Lumped High Only able to represent first
resonance frequency.
(nominal-PI)
Model Representation Distributed High Able to represent multiple
(equivalent-PI) resonance frequencies and
long-line effects.
Model Type Lumped High Able to represent multiple
resonance frequencies;
(cascaded PI)
increasing the number of
cascaded sections improves
the frequency range for
which the model is
acceptable.
Average Conductor Height Changed within Low Very minor shift in resonant
above Ground typical range frequencies, with increasing
harmonic order.
43
3.2 Cables
The network harmonic impedance is highly dependent upon the impedances of OHLs and cables.
Cables are more likely than OHLs to cause resonance at lower frequencies, which are of interest as
damping tends to be lowest at the lower end of the frequency spectrum. Cables typically have a lower
series impedance and higher shunt capacitance than OHLs. Both should be accurately modelled in
harmonic studies [34], [35]. The accurate modelling of submarine cable losses, for example, is difficult
and this impacts upon the damping of harmonics [36]. Achieving the required accuracy can prove difficult
due to studies being carried out during the planning stage of a project, at which time manufacturer-
provided design parameters may not be available and studies are instead based on generic data.
Deviations between data sheet values and physical values often exist for installed cables. The laying
and bonding configuration of cables play a key role in the local harmonic behaviour of the system and,
for older systems, these details might not be known. Other aspects influencing accuracy can be
simplifications or approximations made for reasons of computational efficiency in commercial power
system simulation packages. Simplistic formulae to model skin and proximity effects suffice at power
frequency but may lead to inaccuracies in terms of losses and damping at harmonic frequencies.
3.2.1 Factors Influencing the Harmonic Impedance of Cables

A thorough sensitivity analysis of the numerous factors that influence the harmonic impedance of cables
is provided in 0. For convenience, these factors are briefly described here, and the results of the
sensitivity analysis summarised.
3.2.1.1 Cable Length

The length of the cable is the parameter that has the greatest effect on both the magnitude and the
frequency of the resonance peaks. During planning, there is little certainty about the length of the circuit,
as the exact route may not be known. Some studies may be carried out when the length has been
established; others may use a very approximate estimate of the length. Circuit length deviations of ±30%
are not uncommon between planning and installation phases, meaning large uncertainties in the location
of the resonances.
Key Point:
 Cable length plays a significant role in the frequency, magnitude and number of harmonic resonance
points within the frequency range of interest in both the positive- and zero-sequence harmonic
impedance. Increases in length mean a downward shift in the locations of the resonances in both
the positive- and zero sequence.
3.2.1.2 Cable Design Characteristics

The thicknesses of the core conductor and main insulation affect the capacitance of the cable and are
therefore of relevance to harmonic studies. There are uncertainties in the thicknesses of the cable layers
provided by the manufacturer. Sometimes the choice of conductor material is up to the producer, so if
aluminium is chosen over copper for example, the conductor radius will be larger. This is not known until
after the cable tender period.
Key Point:
 Conductor radius influences the frequencies at which resonances occur in the positive sequence
and has a minor impact on the zero-sequence impedance magnitude.
 Insulation thickness affects resonances in the positive-sequence in terms of frequency and
magnitude, and in the zero-sequence only the magnitude is affected.
3.2.1.3 Cable Layout

The cable laying formation is highly dependent on the available space and any digging activities, so the
impact of likely possibilities should be investigated. Land cables could be laid in excavated trenches
which are then backfilled, or they could be pulled through cable ducts in concrete duct banks where the
joints are accessible via concrete vaults (sometimes a combination of the two). There may be more
than one set of cables occupying a duct bank. Submarine cables, which have an armour to protect them
from mechanical damage, could either be directly laid on the sea bed or they could be laid in trenches
that have been ploughed into the sea bed.
44
Key Point:
 Cable layout significantly affects the frequency, magnitude and number of resonances in the
positive-sequence harmonic impedance.
3.2.1.4 Sheath Bonding

The type of cable bonding used as well as the number of cross-bonding joints have a significant impact
on the harmonic impedance. Bonding introduces a non-continuous impedance along the cable.
Key Point:
 Bonding configuration significantly affects the positive-sequence harmonic impedance but has only
a minor effect on the zero-sequence harmonic impedances.
 The number of major sections affects the frequency, magnitude and number of resonant peaks in
the positive sequence.
3.2.1.5 Cable Model

The choice of model (lumped or distributed) used to represent the cable impacts significantly on how
accurately the harmonic impedance will be represented in the study. Typically, with the currently-
available computational power, the distributed model is the first choice for most harmonic studies.
Key Point:
 Model type should be selected carefully depending on the requirements of study. In the case of the
PI model, the number of PI sections used dictates the number of resonance points able to be
modelled. For longer cables (i.e. >2-5 km), the distributed model should be used.
3.2.1.6 Pipe Thickness

For harmonic studies, the influence of the pipe thickness is generally small. The reader is referred to
[37] for further details.
3.2.2 Models
Both cables and OHLs can be represented by a pi-circuit using either lumped or distributed parameters
as described in 3.1.5.1. The primary difference between models used for cables and OHLs is the
calculation of the impedances and admittances. This is due to:
 Inherent design differences (a cable has a metal sheath, insulating layers, etc.);
 An OHL is installed above ground, whereas cables are usually buried which affects the earth
impedance.
For a thorough discussion of the differences between the electrical characteristics of cables and OHLs,
the reader is referred to [34].
3.2.2.1 PI Models
The modelling of single-core (SC) cables has been well addressed in the literature [33], [38]. Models of
cables for several types of studies are discussed in [35], and recommendations provided. All models
are based on series impedances and shunt admittances, which are derived from the cable system
geometrical and physical characteristics. Like OHLs, cables can be modelled using either a nominal-PI
or an equivalent-PI representation. These are illustrated in Figure 3-1 and Figure 3-2, respectively.
A comparison of the cascaded nominal-PI and the equivalent-PI model is shown in Figure 3-11. As
shown, the number of nominal-PI sections that must be cascaded to maintain a certain accuracy
increases with frequency [3].
45
Figure 3-11 Comparison of Cable Models [39]
A comparison of nominal- and equivalent-PI models for harmonic studies was made in [40]. For long
cables connected to a strong grid, the nominal-PI model was found to introduce significant error.
Conversely, the nominal-PI and equivalent-PI models yielded comparable results for the computed
resonance frequencies for short cables connected to a weak grid. Small errors in resonance frequencies
were found to lead to large errors in harmonic levels. With the computational power readily available at
the time of writing, the benefit of the nominal or cascaded nominal-PI section is limited.
Key Point:
 Nominal-PI model: Should only be used to get an approximate idea of the frequency range in which
resonances can be expected. Unless used in cascade (rarely done in practice), this model can only
represent one resonant frequency. The nominal-PI model should be used with caution when
modelling long cables connected to a strong grid.
 Equivalent-PI model: Should always be used for improved accuracy in harmonic studies.
3.2.3 Impedance and Admittance Formulae

Various formulae can be used to calculate the impedances and admittances that are required as inputs
to the selected model. The accuracy of the formulae used depends on consideration of frequency
dependence and phenomena such as the skin effect and proximity effects. Classical formulae used to
calculate the impedances and admittances of SC cables are well established and described in detail in
the literature [34], [38], [41] and [42]. Cables laid in a pipe are usually modelled using the well-known
analytic formulae for pipe-type cables. Newer methods, such as the recently introduced MoM-SO
(Method of Moments – Surface admittance Operator) can provide high accuracy with small
computational overhead [43]-[45] and valuable insights into the frequency-dependent behaviour of multi-
core cables. This method is available to be used with some commercial power systems analysis
software applications and is discussed further in 3.2.
3.2.3.1 Skin Effect

For a description of the skin effect, the reader is referred to Section 3.1.2. A comparison of the use of
approximate formulae and analytical methods for the calculation of the cable impedance is illustrated in
Figure 3-12.
46
Figure 3-12 Cable Internal Impedance: Approximate vs. Analytical Consideration of Skin Effect [39]
Key Point:
 The skin effect should be considered when calculating the harmonic impedance of either SC or
multi-core cables. Analytical methods which employ Bessel functions provide the highest
accuracy whereas approximate, truncated formulae can be inaccurate above a certain frequency.
Skin effect correction factors can be inaccurate at harmonic frequencies and may therefore lead to
incorrect harmonic losses and damping.
3.2.3.2 Proximity Effects

The proximity effect is a frequency-dependent phenomenon which can occur in the presence of closely-
spaced, parallel conductors [44]. This results in a non-uniform current distribution on the core conductors
and armour (or pipe), yielding an increase in the conductor AC resistance [44]. The proximity effect
varies with frequency and has an impact at lower harmonic frequencies. According to [45], significant
proximity effects may be observed in multi-core, or closely-spaced SC cables, and ignoring these effects
when modelling cables may lead to an underestimation of the cable losses at power frequency. For
these reasons, when performing harmonic studies of cables with closely-spaced cores, the accurate
modelling of proximity effects should be considered.
Consideration of the proximity effect (core-core and core-armour; to varying levels of accuracy) can be
achieved via several means:
(i) Use of proximity effect factors, such as those according to IEC 60287-1-1 [34] and [46].
These factors yield the worst-case harmonic losses but can be erroneous in terms of
harmonic emissions.
(ii) Consideration of the (core-armour/pipe) proximity effect in the calculation of the harmonic
impedance [44]. This is considered when using standard formulae [38].
(iii) Finite element methods (FEM): These can be highly accurate but usually at the cost of
computational expense due to the fineness of the mesh required [43].
(iv) The state-of-the-art MoM-SO method [43], [45] which can perform up to 100 times faster
than FEM methods [43] and can provide highly-accurate results in harmonics analysis [47].
47
This method can consider core-core, core-armour proximity effects and account for
individual strands of stranded armours. However, in order that the model be accurate,
detailed knowledge of the physical and electrical parameters of the cable is required [36].
(v) The subdivision of metallic components into sub-conductors. This shows good agreement
with FEM methods and is less computationally complex. The interested reader is referred
to [48] for further details.
A comparison of some of these methods is illustrated in Figure 3-13.
16,00
MoM-SO with S & P Effects (Stranded Armour)
MoM-SO with Skin and Proximity Effect
14,00 MoM-SO with Skin Effect
IEC 60287-1-1:2006
EDF Approximation
12,00
NGC Approximation
10,00
R1 [pu]
8,00
6,00
4,00
2,00
0,00
f [Hz]
Figure 3-13 Calculation of Cable Impedance: Comparison of Approaches [47]
A recent study [47] demonstrates that the inclusion of accurate modelling of frequency-dependent skin
and proximity effects in cables can be critically important and may yield large reductions in resonant
peaks in some cases. Inaccurate modelling of the cable can lead to incorrect evaluation of harmonic
damping and stability margins. For studies such as Power Park Module (PPM) design, this can result
in excessive harmonic levels in the PPM, or a costly over-design with unnecessarily reduced harmonic
levels. Similarly, the stability margins can become too low, thereby compromising PPM operation. The
positive sequence resistance calculated using various levels of accuracy is depicted in Figure 3-13 and
a frequency scan showing the positive sequence magnitude (taken from [47]) is provided in Figure 3-14.
It should be noted that in [47], to illustrate the effect on damping, the positive sequence inductance is
equal for each of the cases shown, and only the frequency-dependent, positive sequence resistance
characteristic is changed. This ensures that the resonances are aligned and only the effect of damping
can be seen.
48
10000
Ideal (No Harmonic Losses Modelling)
MoM-SO with Skin Effect
MoM-SO with Skin and Proximity Effects
MoM-SO with Skin and Proximity Effects (Stranded Armour)
1000
|Z1| [Ω]
100
10
1
100 600 1100 1600 2100
f [Hz]
Figure 3-14 Consideration of Skin and Proximity Effects in MoM-SO: Effect on Positive Sequence
Impedance Magnitude [47]
Key Point:
 The proximity effect should ideally be considered when calculating the harmonic impedance of either
multi-core cables or cables placed with cores in close proximity. Engineers should be aware that
proximity effect correction factors can be inaccurate at harmonic frequencies. Advanced methods
such as MoM-SO considering both skin and proximity effects (although currently only of limited
availability for use with commercial software) are recommended for highest accuracy when
calculating the harmonic impedance.
49
3.2.4 Summary
A sensitivity analysis highlighted the most important parameters to capture accurately when modelling
cables for harmonic studies. The results are summarised in Table 3-4.
The impact of selection of model was discussed and the limitations of the nominal-PI model (including
cascaded nominal-PI sections) stated.
Formulae for the calculation of the cable impedance and admittance were referenced and discussed.
Phenomena such as the skin effect and proximity effects can be considered via various means from
simple correction factors to methods such as FEM and MoM-SO. When selecting the accuracy with
which to model these effects, the engineer should be aware of the possibility of incorrect estimation of
harmonic damping. To avoid this, the most accurate modelling available should be used.
Table 3-4 Summary of Sensitivity Analysis in 0: Impacts on Harmonic Impedance
Parameter Sensitivity Level of Resonant Frequencies Magnitude

Analysis Impact
Pos. Seq. Zero Seq. Pos. Seq. Zero Seq.
Length Increase length Significant Decrease Decrease Decreases Decreases

Conductor Increase Minor Minor Minor No change Minor
Radius conductor decrease decrease decrease
radius (all other
layers fixed)
Insulation Increase Moderate Minor No change Minor Minor
Thickness insulation increase increase increase
thickness (all
other layers
fixed)
Cable Flat formation Significant Decrease No change Decrease No change
Layout
Cable Triangle Significant Decrease No change Decrease No change
Layout
Cable Touching trefoil Significant Increase No change Decrease No change
Layout
Sheath Every major Significant Increase Negligible Decrease No change
section → solid change
Cross-
bonding
Bonding
Sheath Increase Significant Increase Negligible Increase No change
number of major change
Cross-
sections
Bonding
Model Type Nominal-PI Significant Increase - No change -
→
Equivalent-PI
50
3.3 Power Transformers

Power transformers are one of the most common components in power systems. They present an
inductive behaviour along the frequency range of interest in this Technical Brochure and can therefore
interact with capacitive elements, such as cables or capacitor banks, resulting in parallel or series
resonances. Therefore, accurate representation of this type of equipment is paramount for correct
estimation of resonant frequencies and, more generally, for performing meaningful harmonic studies in
power systems.
The overall structure of the basic transformer model is the same as that used in power frequency studies,
which can be found in many text books. Figure 3-15 shows the physical representation of a three-phase
transformer and the single-phase equivalent model. Some modelling aspects that require special
attention for harmonic studies are discussed next.
(A)
HV LV
(B)
Figure 3-15 Transformer Model (A) Physical Representation. (B) Single-Phase Equivalent Model
Transformer stray capacitances have no material impact in the range of frequencies of interest for typical
harmonics studies, and therefore do not need to be included in the model. The frequency threshold
above which transformer capacitances start to have some effect depends on the type and size of the
transformer unit. As guidance, a value of 4 kHz can be used to start considering the capacitances for
most transformers [27].
The magnetising branch (Lm and RFE) can be ignored under normal conditions, as it is generally assumed
that the transformer core operates in its linear region on a continuous basis. However, if transformer
saturation becomes a primary concern (e.g. operation at elevated voltage levels for extended periods
of time), this effect can be captured by including a harmonic current source in place of Lm and RFE. The
value of the current source should be determined from the flux-current curve and the operating voltage
[27].
Based on the above considerations, for typical harmonic studies within the scope of this Technical
Brochure, the transformer model shown in Figure 3-15 can be reduced to an impedance connected in
series between the two transformer terminals.
The position of the tap changer affects the leakage reactance of the transformer and the transfer
impedance from either side of the transformer (which follows the square of the turns ratio). The variation
in leakage reactance can produce shifts in resonant frequencies while the transfer impedance can affect
the level of damping provided between the two voltage levels. It is therefore recommended to capture
the effect of tap changers (range and step size) in transformer models for harmonic studies.
Transformer vector groups can introduce phase shifts to harmonic voltages and currents depending on
the harmonic order, the sequence and the transformer connections [27]. Therefore, explicit
representation of the winding connections is necessary for most harmonic studies. As an example, the
51
parallel connection of wye and delta transformer windings is used to combine two 6-pulse rectifiers into
a 12-pulse one. In this application, the 30° phase shift between a star-connected winding and a delta-
connected winding introduces cancellation of the 6n harmonic orders.
Transformer winding connections also affect the flow of zero-sequence harmonic currents. In general,
triplen harmonic currents tend to be trapped in delta winding connections (assuming perfectly balanced
conditions) while they add up in grounded transformer neutrals. This may result in excessive currents in
the transformer neutral leading to operation of protection relays. Explicit representation of winding and
neutral connections is recommended.
One of the key aspects of transformer modelling, and normally neglected, is the correct representation
of losses. While power transformers are specified to minimise resistive losses at fundamental frequency,
their presence at higher frequencies provides beneficial damping of harmonic resonances. Correct
estimation of this damping can reduce, or even negate, the need for harmonic mitigation measures. This
needs to be accurately captured in the series impedance components of the transformer model.
The following sub-sections concentrate on the representation of the frequency dependent behaviour of
the series impedance components of the transformer model. An overview of power transformer models
that are frequently used in industry is provided first. Then, the performance of the transformer models is
assessed and compared in a larger model representing a real power system and recommendations are
provided. Finally, comparisons of simulation results using the transformer models against real
measurements provide the basis for further recommendations.
It should be noted that the models presented in this section are based on 2-winding transformers, and
that significantly higher harmonic losses should be expected in 3-winding transformers where a tertiary
is involved. Models used for 2-winding transformers may be extended for application to 3-winding
transformers, however this should be done with caution. Furthermore, the models presented here
assume positive sequence impedances. A similar approach could be adopted for representation of zero-
sequence impedances, provided that winding and neutral connections are correctly represented in the
zero-sequence equivalent network.
3.3.1 Power Transformer Models

The most common representations for power transformers in harmonics analysis are presented in this
section. All these models are based on data that is typically available to the study engineer (or easy to
provide reasonable assumptions) such as rated power, leakage reactance or transformer resistance.
3.3.1.1 Model 1: Electra No. 167 (CIGRE WG CC02) [2]

This model represents the power transformer by a lumped impedance Zh made up from a resistance R s
in series with an assembly consisting of a reactance Xh in parallel with a resistance R p [2], as seen in
Figure 3-16.
Figure 3-16 Power Transformer Model 1: Electra 167 [2]
The input data requirements for this model are the transformer leakage reactance at fundamental
frequency (X1 ) in Ohm and the rated nominal power of the transformer (Sr ) in MVA.
The model assumes that the leakage inductance 𝐿σ is constant over the range of frequencies of interest,
therefore Xh increases linearly with frequency as per Equation 3.4 and Equation 3.5.
𝑋1 = 2𝜋𝑓𝑛 𝐿𝜎 Equation 3.4
52
𝑋ℎ = ℎ𝑋1
Equation 3.5
The resistances R s and R p are constant at all frequencies and their values (in Ohm) are given by
Equation 3.6 and Equation 3.7, based on the rated nominal power of the transformer (in MVA), as per
Equation 3.8 (if better knowledge is not available).
𝑋1
𝑅𝑆 = Equation 3.6
𝑡𝑎𝑛( 𝜓1 )
𝑅𝑝 = 10 ⋅ 𝑋1 ⋅ 𝑡𝑎𝑛( 𝜓1 )
Equation 3.7
0.693+0.796 𝑙𝑛 𝑆𝑟 −0.0421(𝑙𝑛 𝑆𝑟 )2
𝑡𝑎𝑛 𝜓1 = 𝑒
Equation 3.8
The equivalent impedance of the transformer Zh is given by Equation 3.9 below, which can be easily
implemented in a software tool as a series combination of R(h) and X(h) with frequency dependant
correction factors.
𝑗𝑋ℎ 𝑅𝑝 ℎ2 𝑋12 𝑅𝑝 ℎ𝑋1 𝑅𝑝2
𝑍ℎ = 𝑅(ℎ) + 𝑗𝑋(ℎ) = 𝑅𝑠 + = (𝑅𝑠 + 2 2 ) + 𝑗 ( ) Equation 3.9
𝑅𝑝 + 𝑗𝑋ℎ 𝑅𝑝 + ℎ2 𝑋1 𝑅𝑝2 + ℎ2 𝑋12
Equation 3.9 shows that the resistive component of Zh is a function of the frequency, demonstrating that
skin effect is captured in this model. Furthermore, even though X(h) presents a complex dependency
with frequency, experience shows that a linear approximation is adequate for the range of frequencies
of interest in harmonic analysis. Therefore, 𝑋(ℎ) ≅ ℎ𝑋1 is usually adopted as a valid approximation.
3.3.1.2 Model 2: IEEE Std. 399 [49]

This model represents the power transformer as a series impedance Zh as illustrated in Figure 3-17 [49].
s h
Figure 3-17 Power Transformer Model 2: IEEE Std. 399 [49]
The input data requirements for this model are the transformer DC resistance R DC (in Ohm) and the
leakage reactance at fundamental frequency X1 (in Ohm).
𝑋ℎ = ℎ𝑋1 Equation 3.11
The resistance modelling accounts for skin effect. The frequency dependency of the resistance is
calculated using Equation 3.12. The recommended values for A and B are 0.1 and 1.5, respectively [49].
𝑅S (ℎ) = 𝑅DC (1 + 𝐴ℎ𝐵 ) Equation 3.12
3.3.1.3 Model 3: Electra No. 164 (CIGRE JTF 36.05.02/14.03.03) [1]

53
s h
Figure 3-18 Power Transformer Model 3: Electra-164 [1]
The input data requirements for this model are the short-circuit resistance R t (in Ohm) and the leakage
reactance at fundamental frequency X1 (in Ohm).
𝑋ℎ = ℎ𝑋1 Equation 3.14
The resistance modelling accounts for skin effect. The frequency dependency of the resistance is
calculated using Equation 3.15. The recommended values for the coefficients are given in Table 3-5
with the condition that 𝒂𝟎 + 𝒂𝟏 + 𝒂𝟐 = 𝟏 [1].
𝑅s (ℎ) = 𝑅t ∙ (𝑎0 + 𝑎1 ℎ𝑏 + 𝑎2 ℎ2 ) Equation 3.15
Table 3-5 Values for Coefficients 𝐚𝟎 , 𝐚𝟏 𝐚𝟐 and 𝐛 (requirement: 𝐚𝟎 + 𝐚𝟏 + 𝐚𝟐 = 𝟏)
𝒂𝟎 𝒂𝟏 𝒂𝟐 𝒃
Small system transformer 0.85 – 0.90 0.05 – 0.08 0.05 – 0.08 0.9 – 1.4
Large system transformer 0.75 – 0.80 0.10 – 0.13 0.10 – 0.13 0.9 – 1.4
Usually, it can be assumed that the coefficients for the small system transformer should be used if its
power rating is below 100 MVA. Otherwise the coefficients for a large system transformer should be
used.
3.3.1.4 Model 4: Arrillaga [24]

s h
Figure 3-19 Power Transformer Model 4: Arrillaga [3]
The input data requirements for this model are the transformer resistance at fundamental frequency 𝑅
(in Ohm) and the leakage reactance at fundamental frequency X1 (in Ohm).
The leakage inductance 𝐿σ is not assumed to be constant and a typical L-f characteristic is included in
[24] showing a decay in inductance with increasing frequency.
The harmonic impedance is calculated using Equation 3.16 below.
𝑍T (ℎ) = 𝑅 √ℎ + jℎ𝑋1
Equation 3.16
54
3.3.1.5 Model 5: Funk [50]

s h
Figure 3-20 Power Transformer Model 5: Funk [50]
The input data requirements for this model are the transformer resistance at fundamental frequency R fn
(in Ohm) and the leakage inductance at fundamental frequency Lfn (in Henry).
The frequency dependency of the resistance and inductance of the transformer are calculated using
Equation 3.17 and Equation 3.18 below. The recommended values for the coefficients are given in Table
3-6 [50].
BR
f
R s = R fn (1 + AR ( − 1) ) Equation 3.17
fn
f BL
L = L fn A L ( ) Equation 3.18
fn
Table 3-6 Coefficients for Transformer Model 5 [50]
𝑹(𝒇) 𝑳(𝒇)
Transformer type
𝑨𝐑 𝑩𝐑 𝑨𝐋 𝑩𝐋
20 kV / 0.4 kV, 250 kVA 0.2 1.5 1 -0.03
108 kV / 10.5 kV, 40 MVA 0.2 1.4 1 -0.02
220 kV / 110 kV, 200 MVA 0.2 1.6 1 ≈0
3.3.2 Comparison of Power Transformer Models and Measurement Results

A comparative assessment of each transformer model option against measurements from three different
transformers is presented next.
3.3.2.1 Example 1
Single-phase two winding transformer unit with the following technical parameters:
 Rating: 100 MVA (single phase)
 Primary Voltage: 735 kV
 Secondary Voltage: 26 kV
 Nominal frequency: 60 Hz
 Short Circuit impedance (Zk): 12.92%
 X/R at nominal frequency: 68
Frequency response measurements were performed by white noise injection at the 26 kV transformer
terminals.
The response of each transformer model option is compared against the measured data for this
transformer unit in Figure 3-21 (resistance) and Figure 3-22 (reactance). The magnitudes plotted in
these graphs are expressed in per unit, based on the rated MVA of the transformer. The model
parameters selected for each assessment are included in Table 3-7.
Figure 3-21 (a) shows the calculated resistance values using the recommended default model
parameters. It can be observed that Model 1 results in an over-estimation of the transformer resistance
55
as frequency increases. Similarly, Model 2, Model 4 and Model 5 result in an under-estimation of the
transformer resistance. In this example, Model 3 yields the closest match to the measured values.
Further model tuning was performed in an attempt to find a better fit with the measurements - Figure
3-21 (b). The best overall result was obtained with Model 1 by using the measured X/R ratio (at 60Hz)
instead of the calculated tan 𝛹. Model 2 and Model 5 also gave reasonable results when the parameters
were optimised to fit the measured values.
Figure 3-22 shows the calculated and measured reactance values as a function of frequency. All models
produce very similar results and are in good agreement with the measurements. The reactance can be
considered to vary linearly with frequency for the range of interest in harmonics studies. Above 2 – 3
kHz this behaviour starts to deviate as the effective winding capacitances start having a noticeable
impact on the transformer response.
The measured and calculated values of the time constant L/R(f) are plotted in Figure 3-23 for
completeness.
Table 3-7 Example 1: Transformer Model Parameters
Default Parameters Modified Parameters

𝑋
Model 1: Electra-167 [2] tan 𝛹 = 47.63 tan 𝛹 = (@60𝐻𝑧) = 68
𝑅
Model 2: IEEE-399 [49] A = 0.1, B = 1.5 A = 0.585, B = 1.538
Model 3: Electra-164 [1] a0 = 0.775, a1 = 0.115, a2 = 0.11, b= 1.15 N/A
Model 4: Arrillaga [24] N/A N/A
Model 5: Funk [50] AR = 0.2, BR = 1.6, AL =1, BL = 0 AR = 0.2, BR = 1.85
0,50
(a) Default Model Parameters
0,40
R(f) [pu]
0,30
0,20
0,10
0,00 f [Hz]
0 500 1000 1500 2000 2500
Measurement Model 1 Model 2 Model 3 Model 4 Model 5
0,40
0,35 (b) Modified Model Parameters
0,30
R(f) [pu]
0,25
0,20
0,15
0,10
0,05
0,00 f [Hz]
0 500 1000 1500 2000 2500
Measurement Model 1 (modified) Model 2 (modified) Model 3 Model 5 (modified)
Figure 3-21 Example 1: Calculated and Measured Transformer Resistance as a Function of Frequency
56
6,0
5,0
4,0
X(f) [pu]
3,0
2,0
1,0
0,0 f [Hz]
0 500 1000 1500 2000 2500
Figure 3-22 Example 1: Calculated and Measured Transformer Reactance as a Function of Frequency
1000
100
L/R(f) [ms]
10
0,1 f [Hz]
60 600 6000
Measurement Model 1 (modified) Model 2 (modified) Model 3 Model 5 (modified)
Figure 3-23 Example 1: Measured and Calculated Time Constant (L/R)

Key Points:
For the transformer assessed in Example 1:
 Model 1 (Electra-167), using an X/R ratio measured at power frequency, produces the closest
results to the measured resistance data. This ratio is normally available in factory acceptance tests
reports.
 Model 3 (Electra-164), using default parameters, produces reasonable results.
 The transformer reactance can be assumed to increase linearly for the frequency range of interest
in harmonics studies.
3.3.2.2 Example 2
Three-phase wind farm transformer with the following parameters:
 Rating: 90 MVA
 Primary Voltage: 132 kV
 Secondary Voltage: 33 kV
 Short circuit impedance (Zk): 11%
 X/R at nominal frequency: 51.7
The frequency response measurement method is detailed in [51]. Figure 3-24 and Figure 3-25 below
show the variation with frequency of the resistance and inductance for this transformer, referred to the
datasheet values (at nominal frequency). A fast increase in the resistance value can be observed within
the range of frequencies of interest for harmonic analysis. However, the transformer inductance changes
very slowly with frequency. In practical terms, a representation based on a constant inductance appears
reasonable for this transformer. Therefore, this section will concentrate on assessing the representation
of the resistive component only.
57
Figure 3-24 Example 2: Variation of Transformer Resistance with Frequency [51]
Figure 3-25 Example 2: Variation of Transformer Inductance with Frequency [51]
The response of each model option is compared to the measured resistance data for this transformer in
Figure 3-26. The resistance values plotted in this figure are expressed in per unit, based on the rated
58
MVA of the transformer. The model parameters selected for each assessment are provided in Table
3-8.
Figure 3-26 (a) shows the calculated resistance values using the recommended default model
parameters. It can be observed that Model 1 and Model 3 result in an over-estimation of the transformer
resistance as frequency increases. Similarly, Model 2, Model 4 and Model 5 result in an under-estimation
of the transformer resistance. These responses match the observations made in Example 1, except for
Model 3 which gives a reasonable fit in Example 1 but overestimates the resistance in Example 2.
Further model tuning was performed to find a better fit with the measurements and the results are shown
in Figure 3-26 (b). The improved parameters gave very close results to the measured values for Model
1, Model 2, Model 3 and Model 5. Essentially, most modelling input data can be tuned to reproduce
measurements. It should be noted that, unlike Example 1, the use of the power frequency X/R ratio with
Model 1 did not produce a good match in this transformer and a higher value had to be selected.
completeness. This graph is almost identical to the one presented for Example 1 (Figure 3-23) and can
be used as reference for this type of transformer.
Table 3-8 Example 2: Transformer Model Parameters
Default Parameters Modified Parameters
Model 1: Electra-167 [2] tan 𝛹 = 30.64 tan 𝛹 = 65
Model 2: IEEE-399 [49] A = 0.1, B = 1.5 A = 1, B = 1.95
a0 = 0.87, a1 = 0.05,
Model 3: Electra-164 [1] a0 = 0.775, a1 = 0.115, a2 = 0.11, b= 1.15
a2 = 0.08, b= 1.2
Model 4: Arrillaga [24] N/A N/A
Model 5: Funk [50] AR = 0.2, BR = 1.6 AR = 0.25, BR = 1.72
1,00
(a) Default Model Parameters
0,80
R(f) [pu]
0,60
0,40
0,20
0,00 f [Hz]
0 500 1000 1500 2000 2500
0,50
(b) Modified Model Parameters
0,40
R(f) [pu]
0,30
0,20
0,10
0,00 f [Hz]
0 500 1000 1500 2000 2500
Measurement Model 1 (modified) Model 2 (modified) Model 3 (modified) Model 5 (modified)
Figure 3-26 Example 2: Calculated and Measured Transformer Resistance as a function of Frequency
59
1000
100
L/R(f) [ms]
10
0,1 f [Hz]
60 600 6000
Measurement Model 1 (modified) Model 2 (modified) Model 3 (modified) Model 5 (modified)
Key Points:
For the transformer assessed in Example 2:
 None of the transformer model options gave a reasonable match to the measured transformer
resistance. Model 1 and Model 3 over-estimated the resistance values while Model 2, Model 4 and
Model 5 under-estimated the resistance.
 The response of all transformer model options is very sensitive to the selected parameters. In all
cases it is possible to reproduce measurements by back-calculating the parameters.
 The transformer inductance can be assumed to be constant for the range of frequencies of interest.
Under this assumption, the transformer reactance increases linearly with frequency.
3.3.2.3 Example 3
Wind turbine generator transformer with the following technical parameters:
 Rating: 8.2 MVA
 Primary Voltage: 33 kV (delta connection)
 Secondary Voltage: 0.69 kV
 Resistance at 50 Hz: 3.32 Ω (delta connection)
 Reactance at 50 Hz: 31.07 Ω (delta connection)
 X/R at nominal frequency: 9.35
The response of each model option is compared against the measured resistance data for this
transformer in Figure 3-28. The resistance values plotted in this figure are expressed in per unit, based
on the rated MVA of the transformer.
Figure 3-28 shows the calculated resistance values using the recommended default model parameters.
It can be observed that Model 1 and Model 3 result in an over-estimation of the transformer resistance
as frequency increases. Similarly, Model 4 underestimates the transformer resistance. For this
transformer, Model 2 and Model 5 gave reasonable matches to the measurements.
completeness.
60
1,50
R(f) [pu]
1,00
0,50
0,00 f [Hz]
0 500 1000 1500 2000 2500
Figure 3-28 Example 3: Calculated and Measured Transformer Resistance as a function of Frequency
1000
100
L/R(f) [ms]
10
0,1 f [Hz]
60 600 6000
Measurement Model 2 Model 5
Key Points:
For the small transformer assessed in Example 3:
 Model 2 (IEEE-399) and Model 5 (Funk) gave the best matches to the available measurements.
3.3.3 Comparison of Power Transformer Models in a Transmission Grid

The effect of transformer modelling (impedance and loss at harmonic frequencies) may be insignificant
in the presence of other network components (overhead lines, insulated cables, generators, loads, etc.)
and their location in relation to each other. A case study which illustrates these modelling aspects is
presented here.
Three of the five power transformer models described in section 3.3.1 were compared in a full system
model representing the transmission grid in Ireland. The selected transformer models represent the
highest losses (Model 1, section 3.3.1.1), lowest losses (Model 4, section 3.3.1.4) and middle range
losses (Model 5, section 3.3.1.5) from all model options presented in section 3.3.1. The base-case
reflects a model without frequency dependency in the resistive components of the transformers (i.e.
constant resistance).
Frequency scans were performed at a range of transmission nodes scattered throughout the system,
giving a representative overview of the performance of the different power transformer models at various
locations in the grid. Four characteristic cases are shown in Figure 3-30. The graphs on the left present
the harmonic impedance at each selected node for the full range of frequencies of interest. The graphs
on the right show the harmonic impedance around the main resonance points to better appreciate the
effect of the different transformer model options. A commentary on each Node follows.
 Node #1 is a switching station, remote from large transmission transformers. The harmonic
impedance is dominated by overhead lines and insulated cables in the area. Therefore, the
representation of frequency dependent transformer losses has no material impact on the harmonic
impedance at that point.
 Node #2 is a distribution station serving a small amount of load. This station is one node away from
two large transmission stations with 2 x 250MVA 220/110kV transformers on each one. These
61
transformers have a material impact on the harmonic impedance in the frequency range from 2 kHz
to 2.4 kHz. Within this range, Model 1 (Electra 167) provides the highest damping with a 47%
reduction in the maximum harmonic impedance at resonance. Model 5 (Funk) provides a 25%
reduction in impedance with respect to base case and Model 4 (Arrillaga) provides the lowest
damping with just a 2% reduction in harmonic impedance.
 Node #3 is a transmission collector station for a cluster of wind farms. The station has 2 x 250MVA
220/110kV transformers and several 110kV insulated cable connections. These power
transformers have a material impact on the harmonic impedance in the frequency range from 250
Hz to 450 Hz. In this case, the impedance reduction observed with each transformer model is 24%,
13% and 8% for Model 1, Model 5 and Model 4, respectively.
 Node #4 has been fictitiously created to introduce a parallel resonance at 1220 Hz between a
250MVA 220/110kV transformer and a 110kV radial cable connection. In this case, the impedance
reduction observed with each transformer model is 84%, 71% and 23% for Model 1, Model 5 and
Model 4, respectively.
Key Points:
 The choice of transformer model is of key importance for nodes where the harmonic impedance is
dominated or influenced by power transformers. It has been shown that different transformer
representations result in large deviations in harmonic impedance at major resonance points.
 For nodes that are electrically distant from power transformers, the selection of a transformer model
(or constant resistant) does not have material impact on the calculation results.
 All transformer models considered in this document introduce some level of frequency dependent
resistive losses. The effect of these losses is only material at parallel resonant points in the power
system, where the harmonic resistance reaches a peak. All models exhibit similar behaviour outside
the resonant frequencies.
 The quantitative effect on damping introduced by each model depends on the location of resonant
frequencies for the node of interest. All models provide increasing damping with frequency;
therefore, higher differences in performance are observed when the main resonances appear at the
higher frequency range.
 All modelling options described in this document assume that the transformer leakage inductance
is constant for the range of frequencies of interest. Therefore, the reactance increases linearly with
frequency in all model options. This means that the selection of a transformer model does not affect
the location of the resonant frequencies, just the amplitude of the harmonic impedance. This can be
used as a sensitivity test to check whether the harmonic impedance of a node is dominated by a
transformer or by other network components.
62
Z [ohm] Node #1 Z [ohm] Node #1

900 900
800 800
700 700
600 600
500 500
400 400
300 300
200 200
100 100
0 0
0 500 1000 1500 2000 2500 950 1000 1050 1100 1150
f [Hz] f [Hz]
Constant Resistance Model 1 Model 4 Model 5 Constant Resistance Model 1 Model 4 Model 5

3000 3000
2500 2500
2000 2000
1500 1500
1000 1000
500
500
0
0
0 500 1000 1500 2000 2500
2000 2050 2100 2150 2200 2250 2300 2350 2400
f [Hz]
Constant Resistance Model 1 Model 4 Model 5 Constant Resistance Model 1 Model 4 Model 5 f [Hz]

200 200
150 150
100 100
50 50
0 0
0 500 1000 1500 2000 2500 250 300 350 400 450
f [Hz] f [Hz]

20000 20000
15000 15000
10000 10000
5000 5000
0 0
0 500 1000 1500 2000 2500 1150 1170 1190 1210 1230 1250 1270 1290 f [Hz]
f [Hz]
Figure 3-30 Effect of Detailed Power Transformer Models on Harmonic Impedance in a real Transmission
System
3.3.4 Summary
In general, a transformer can be represented by a series combination of resistance and inductive
reactance. Both components are frequency dependant and this section deals with different approaches
to capture this feature.
Five models were described in this section and case studies presented highlighting various important
aspects of power transformer modelling for harmonics studies. The reader is reminded that, for
simplicity, the models presented here are based on 2-winding transformers and positive-sequence
parameters. The models, however, can be extended to represent higher number of windings and zero-
63
sequence behaviour in line with common power frequency modelling approaches, if data describing
their frequency behaviour is available. .
The data required for each model can be easily obtained from the transformer data sheet, or preferably,
from the Factory Acceptance Test (FAT) report. Some of the models require additional factors for which
default values have been provided. However, best calculation results are achieved if the manufacturer
of a transformer provides a frequency-dependent model or if an appropriate model is developed by the
operator based on their own measurement data.
Frequency dependency of the transformer resistance is an area where caution is advised as certain
models used may introduce optimistic damping.
All model options capture the frequency dependency of the resistive losses with various levels of
complexity. This effect can be implemented in most commercial software tools by a correction function
applied to the 50/60Hz transformer resistance. The transformer reactance can be assumed to increase
linearly with frequency for the typical range of interest in harmonics studies.
In general, it is important to consider the position of the tap changer as it affects the leakage reactance
of the transformer and the transfer impedance from either side of the transformer (which follows the
square of the turns ratio). Transformer stray capacitances do not have material impact in the range of
frequencies of interest for most harmonics studies, and therefore do not need to be included in the
model. As a general rule, transformer capacitances can be ignored for frequencies up to 4 kHz in most
cases. The transformer winding connections should be represented to take account of the phase-shifting
effect on harmonic currents. The magnetising branch can be ignored under normal conditions, but it
should be represented as a harmonic source if transformer saturation becomes a primary concern.
The five transformer model options presented in this document were assessed against measurements
from three power transformers. This assessment highlighted significant differences between all the
options and their high sensitivity to the selected model parameters. It was observed that, in general,
Model 1 (Electra No. 167 [2]) tend to overestimate resistive losses if the default expression for tan 𝛹
was used, however reasonable results were observed when the 50/60Hz datasheet X/R value was used
instead. Model 4 (Arrillaga [24]) always gave the lowest resistive losses, significantly below the
measured values.
A comparison of the models via frequency scans using a representation of the Irish transmission grid
suggests that the selection of a transformer model only has a noticeable effect at the parallel resonant
points and at nodes whose harmonic impedance is dominated or influenced by power transformers.
If high accuracy is needed in the studies, it is recommended to obtain frequency dependent
characteristics for R and X from the transformer manufacturer. In the absence of measurements to
validate models, the analysis presented in this section suggest the use of Model 5 (Funk [50]) with
default parameters or Model 1 (Electra No. 167 [2]) with 50/60Hz datasheet X/R value as options with
reasonable results.
Alternatively, measurements from a range of transformers presented in [51] and [52] may be used. In
this recent work ([51]) the R(ω)/L ratio of different transformers types and sizes is compared (see Figure
3-31). The measurements were obtained using frequency response analysis (FRA) devices. The reader
is advised that the accuracy of the FRA measurements can be uncertain for smaller transformers within
the low frequency range, i.e. close to the fundamental component. Therefore, it is recommended to
confirm with the supplier whether the measurements are reliable and/or use a hybrid approach including
fundamental frequency short-circuit test values to define the transformer frequency-dependent
characteristic behaviour within the low frequency range [53].
64
Figure 3-31 Comparison of the 𝑳/𝑹(𝝎) Ratio of Different Transformer Types [51]
65
3.4 Loads
Loads have a considerable influence on the harmonic characteristics of the network in terms of location
of resonances and level of damping, and their appropriate modelling is a very important consideration.
This is becoming more important due to the increasing number of HVDC interconnectors, offshore and
onshore wind power plants, PV-based generation and large-scale integration of power electronic based
distributed generation. This section provides an overview of load models that are frequently used in
industry and reviews and discusses reported new harmonic load models as well as novel methods to
tackle the changing load composition due to the increase in power electronic type devices. Comparisons
of simulation results using the most common load models provide the basis upon which
recommendations are provided.
Load behaviour is usually classified as either linear or nonlinear. Linear loads draw current in a
sinusoidal manner and can be grouped into static loads (lighting, resistive heating, etc.) and rotating
loads (asynchronous induction and synchronous motors). While static loads are generally well
represented by their active and reactive power demand at fundamental frequency, the same does not
apply to rotating loads which require more sophisticated models.
Nonlinear loads include power electronic-type loads such as PVs, variable speed drives, battery storage
systems, electric plug-in vehicles, data centres, etc. These loads draw current in a non-sinusoidal
manner. Accurate harmonic modelling of nonlinear loads often presents difficulties due to them being
sources of harmonics, compounded by the fact that their variable R, L, C configuration and nonlinear
characteristics cannot be represented using the linear harmonic equivalent model [3]. Consideration of
nonlinear loads in harmonic studies may need to include (i) the type and topology of the nonlinear loads;
(ii) the interaction with the AC system impedance; and (iii) the variation in the harmonic spectra [16]. In
addition, power electronics-based loads exhibit a more complex damping characteristic than that
exhibited by linear loads, and the effect of controllers on load behaviour should be considered.
The general load composition and its harmonic behaviour is normally one of the biggest unknowns when
performing harmonic assessments in power systems. On the one hand the loads are highly variable in
time and on the other hand, the active power absorbed by rotating machines does not exactly
correspond to a damping value [26]. A common approach is to aggregate the effect of the general load
and distribution networks into a single reduced model. The aggregate harmonic load model refers to an
equivalent representation of a group of loads at a particular voltage level (i.e. LV or MV) and will include
elements of a transmission and distribution network. This means that network elements such as grid
transformers at Grid Supply Points (GSPs), OHLs, cables and shunt compensation devices would also
be part of the model. Therefore, the loads can affect the harmonic response in the network through
damping and the introduction of both parallel and series resonances. If the load is not accurately
represented in harmonic studies, results will likely be unrealistic.
The derivation of the aggregate harmonic load model can be either component-based (bottom-up or
knowledge-based) or measurement-based (top-down or behaviour-based), as described in detail in [54].
The important difference is that a component-based approach can be applied to load consisting of
multiple components, while a measurement-based approach can be applied to load consisting of single
and multiple components.
3.4.1 Harmonic Load Models
3.4.1.1 CIGRE WG 36.05

The CIGRE WG 36.05 [26] proposed the following three methods of load representation with increasing
degrees of sophistication and accuracy:
Method 1 (CIGRE WG 36.05):
The equivalent reactance to the load is neglected; i.e. the general load is assumed to be
purely resistive, as shown in Figure 3-32. The equivalent resistance is calculated using
Equation 3.19, where P1 is the active power of the equivalent load. This method captures
frequency dependency by increasing the value of R for higher harmonics. For example, the
value is 1.3R for the 5th harmonic and 2R for the 10th harmonic.
66
Figure 3-32 Method 1 (CIGRE WG 36.05) Load Model (Static)
𝑈2
𝑅= Equation 3.19
𝑃1
This method is recommended only when the motor part is very small; i.e. for commercial
and domestic loads in which the motor component is so partitioned that the resistive effect
is dominant [8].

To capture motor contribution, the load is represented as a parallel combination of a
resistance and a reactance as shown in Figure 3-33. The resistance represents the static
load component and is calculated as in “Method 1 (CIGRE WG 36.05)” above. The
reactance represents the rotating load component and is evaluated using an estimation of
the number of the motors in service, their installed unitary power (not demand) and their
negative sequence reactance, multiplied by the harmonic order.
R jhX
Figure 3-33 Method 2 (CIGRE WG 36.05) Load Model (Static and Rotating)
As precise information on the number of motors is not generally available, estimation

factors are proposed in [8] as follows:
𝑈2
𝑅= (𝑠𝑡𝑎𝑡𝑖𝑐 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡) Equation 3.20
(1 − 𝐾) ∙ 𝑃1
𝑈2
𝑋= (𝑟𝑜𝑡𝑎𝑡𝑖𝑛𝑔 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡) Equation 3.21
1.2 ∙ (𝐾 + 𝐾𝐸 ) ∙ 𝐾1 ∙ 𝑃1
Where P1 is the total MW demand, K is the motor fraction of the total MW load, K E is the
electronic controlled load fraction of the total MW load and K1 represents the severity of the
motor starting condition (to estimate the equivalent negative sequence inductance).
According to [8], the following values can be assumed when no better information is
available:
 K can assume a value of 0.8 for industrial loads and 0.15 for commercial and domestic
loads.
 K1 assumes normally a value between 4 and 7.
 KE can assume values around 0.
67
An additional parallel resistor representing the motor damping can be included as R 1 = L/K2
where K2 is a fraction of the negative sequence or locked-rotor inductances. Typical values
of K2 are around 0.2.
It should be noted that more research is required to evaluate the applicability of these
assumptions to modern load compositions.
This method is based on experimental results at MV distribution networks in France by
EDF. The loads are represented by a reactance, Xs, in series with resistance R (static
branch); this assembly being connected in parallel with a reactance X p (rotating branch),
as shown in Figure 3-34. This model is applicable over a frequency range approximately
between the 5th and the 20th harmonic orders [26].
jXS
jXP
Figure 3-34 Method 3 (CIGRE WG 36.05) Load Model (CIGRE/EDF)
The model parameters are calculated from both the active and reactive power components
of the load at power frequency and empirical constants, as follows:
𝑈2
𝑅= Equation 3.22
(1 − 𝐾) ∙ 𝑃1
𝑋𝑠 = 0.073 ℎ𝑅 Equation 3.23
ℎ𝑅
𝑋𝑝 = Equation 3.24
𝐾 ∙ 6.7 tan ∅1 − 0.74
Where U is the nominal voltage of the network, P1 is the active power of the load (at the
fundamental frequency), K is the motor fraction of the total MW load, 𝑡𝑎𝑛 ∅1 = 𝑄/𝑃 and Q
is reactive power of the load (at the fundamental frequency). Compensation capacitors and
cables are not captured in this model and must be represented externally.
3.4.1.2 CIGRE WG CC02

The CIGRE WG CC02 [2] proposed three types of aggregate load models to cover most cases in
harmonic studies. In all cases, the electronic load component must be subtracted from the total load to
avoid overestimation of damping.
Method 1 (CIGRE WG CC02):
This model, also referred to as the “CIGRE” model, is the same as “Method 3 (CIGRE WG
36.05)” in section 3.4.1.1 (Figure 3-34).

This model, also referred to as the “R//L” model, represents the load as a parallel
combination of a resistance and inductance. This model does not capture motor load and,
therefore, is only suitable when the fraction of motor load is very small. The model is shown
in Figure 3-35 and the calculation of its parameters follows:
68
R jhX
Figure 3-35 Method 2 (CIGRE WG CC02) “R//L” Load Model (Static)
𝑈2
𝑅= Equation 3.25
𝑃
𝑈2
𝑋= Equation 3.26
𝑄

This model, also referred as the “Motor” model, represents the rotating load as a series
combination of a resistance and inductance. The model is shown in Figure 3-36 and the
calculation of its parameters follows:
jhX
Figure 3-36 Method 3 (CIGRE WG CC02) “Motor” Load Model
𝑋 = 𝑈 2 /𝑆𝑠𝑡𝑎𝑟𝑡 (locked rotor) Equation 3.27

𝑅1 = 𝑋/3 (corresponding to 𝑐𝑜𝑠 𝛷𝑠𝑡𝑎𝑟𝑡 = 0.32) Equation 3.28
𝑅 = √ℎ 𝑅1(skin effect considered) Equation 3.29
3.4.1.3 CIGRE JTF 36.05.02/14.03.03

CIGRE JTF 36.05.02/14.03.03 [1] acknowledges that there are no “generally acceptable” load
equivalents for harmonic analysis and, in each case, the derivation of equivalent models from specified
P (active) and Q (reactive) power flows will need extra information on the actual composition of the load.
In line with the recommendations from CIGRE WG 36.05 (Section 3.4.1.1) and CIGRE WG CC02,
(Section 3.4.1.2), CIGRE JTF 36.05.02/14.03.03 proposed three types of representation depending on
the load composition; namely passive, motive and power electronic. In all cases, power factor correction
capacitance should be estimated as appropriately as possible and allocated at the corresponding
voltage level.
Method 1 (CIGRE JTF 36.05.02): Passive Load (Domestic)
This type of load can be represented by a series combination of a resistance R and
reactance X as shown in Figure 3-37. The parameters are calculated according to Equation
3.30.
69
Figure 3-37 Method 1 (CIGRE JTF 36.05.02/14.03.03) “Passive” Load Model (Domestic)
The equivalent resistance is estimated from the active power at fundamental frequency as
per Equation 3.31, where U is the nominal voltage of the network and P is the active power
of the load (at the fundamental frequency). The factor √ℎ captures the frequency
dependence of the resistive damping. The equivalent reactance is calculated using
Equation 3.32 and represents the relatively small motor content assumed in domestic load.
In some regions, particularly where residential air conditioning is prevalent, motors can
comprise a major if not dominant portion of domestic loads.
𝑍𝑟 (𝜔) = 𝑅(ℎ) + 𝑗𝑋(ℎ) = 𝑅𝑟 √ℎ + 𝑗𝑋𝑟 ℎ Equation 3.30

2
𝑈
𝑅𝑟 = Equation 3.31
𝑃
𝑈2
𝑋𝑟 = Equation 3.32
𝑄
Method 2 (CIGRE JTF 36.05.02): Motive Load

An approximate representation of induction motor load is proposed as a series combination
of a resistance, Rm, and reactance, Xm, as shown in Figure 3-38.
jXmh
Rmh
Figure 3-38 Method 2 (CIGRE JTF 36.05.02/14.03.03) “Motive” Load Model
𝑍𝑚 (𝜔) = 𝑅𝑚ℎ + 𝑗𝑋𝑚ℎ Equation 3.33

𝑋𝑚ℎ = ℎ 𝑋𝐵 Equation 3.34
𝑏
𝑅𝑚ℎ = 𝑅𝐵 (𝑎𝐾𝑎 + 𝐾 ) Equation 3.35
𝑆ℎ 𝑏
𝑆ℎ = (±ℎ − 1)/(±ℎ) Equation 3.36
RB is the total motor resistance with the rotor locked
70
XB is the total motor reactance with the rotor locked

Ka and Kb are the correction factors for the skin effect in the stator and rotor, respectively.
Method 3 (CIGRE JTF 36.05.02): Power Electronic Load

Representation of power electronic loads is complex as this type of load does not present
a constant R, L, C configuration. In addition, their non-linear characteristics cannot be
accurately captured in a harmonic equivalent model for frequency-domain analysis. In the
absence of detailed information, it is recommended to represent this type of load as an
open-circuit when calculating harmonic impedances. However, their harmonic injections
must be represented with an active source. See section 4.7 for detailed guidelines on
modelling variable speed drives.
Method 4 (CIGRE JTF 36.05.02): Measurement-based Load Model
In addition to the above component-based load representation, CIGRE JTF
36.05.02/14.03.03 presented a measurement-based load model. This representation is
normally more suitable for studies concerning the transmission network, where the
downstream distribution networks are represented as network equivalents and the only
information about the load is the consumption of active and reactive power. A parallel
combination of a resistance, R, and a reactance, X, is proposed as shown in Figure 3-39:
R jhX
Figure 3-39 Method 4 (CIGRE JTF 36.05.02/14.03.03) “Measurement-based” Load Model
𝑈2
𝑅= Equation 3.37
(0.1ℎ + 0.9)𝑃
𝑈2
𝑋= Equation 3.38
(0.1ℎ + 0.9)𝑄
Where P and Q are the fundamental frequency active and reactive power measured at
voltage level U.
3.4.1.4 IEEE Task Force

The IEEE Task Force on Harmonic Modelling and Simulation [55] reviewed the common practices and
analytical models for the representation of aggregate loads. The Task Force compiled descriptions of 6
models based on estimates of active and reactive power and information about composition and
characteristics of the load. These models are essentially the same as those presented in the preceding
sections (or present in the minor variations thereof). For completeness, these models are summarised
in Figure 3-40 [55]. In this figure, K represents the participation factor of motor load in the total demand.
Results of the sensitivity analysis performed to determine the impact of load modelling on the harmonic
propagation are also presented in [55]. The Task Force recommend a comprehensive modelling
approach that accounts for all major components of the loads including electronic loads and background
distortion.
71
Model Parameters
Model 1 (IEEE; “RL Series”). Series
jhX
𝑉2
𝑅=𝑃∙
𝑃2 + 𝑄2
𝑉2
𝑋=𝑄∙
R 𝑃2 + 𝑄2
Model 2 (IEEE; “RL//”). Parallel
𝑉2
𝑅=
𝑃
R jhX 𝑉2
𝑋=
𝑄
Model 3 (IEEE). Skin effect
𝑉2
𝑅(ℎ) =
𝑚(ℎ) ∙ 𝑃
𝑉2
R(h ) jhX ( h ) 𝑋(ℎ) =
𝑚(ℎ) ∙ 𝑄
𝑚(ℎ) = 0.1ℎ + 0.9
Model 4 (IEEE). Induction motors

𝑉2
𝑅2 =
Resistive Motive (1 − 𝐾) ∙ 𝑃
𝑉2
𝑋1 = 𝑋𝑀
R2 jhX 1 𝐾𝑚 ∙ 𝐾 ∙ 𝑃
𝐾𝑚 is the install factor (≈ 1.2). 𝑋𝑚 is the pu value of the
motor locked rotor reactance expressed on the motor
rating ≈ 0.15 − 0.25, and 𝐾 is the fraction of motor load.
72
Model 5 (IEEE). CIGRE/EDF
𝑉2
𝑅2 =
(1 − 𝐾) ∙ 𝑃
jhX 2 𝑋2 = 0.073 ∙ 𝑅2
𝑉2
jhX 1 𝑋1 =
𝐾 ∙ 𝑃 ∙ (6.7 ∙ tan 𝜙 − 0.74)
R2 𝑄
tan 𝜙 =
𝑃
Model 6 (IEEE; “2RL//”). Inclusion of

load transformer and motor damping
𝑋1 and 𝑅2 as in Model 4.
jhX 2 jhX 1
𝑋2 = 0.1 ∙ 𝑅2
𝑋1
𝑅1 =
𝐾3
R2 R1
𝐾3 is the effective quality factor of the motor circuit (≈ 8)
Figure 3-40 IEEE Task Force: Summary of Load Models [55]
3.4.1.5 Large Area Asynchronous Motor Load Model

As accurate data of distribution networks suitable for load composition and behaviour is not normally
available, a complex load model was proposed in [56] to represent large asynchronous load areas for
harmonic penetration studies. This load model was validated through field tests and was shown to be
critical for accurate prediction of some system configurations leading to resonance.
The proposed equivalent is shown in Figure 3-41 and the calculation of its parameters follow.
jXS jXa
RS Ra
Static Rotating
part part
Figure 3-41 Large Asynchronous Load Area Equivalent Model [56]

𝑈2
𝑅𝑠 = Equation 3.39
𝑃(1 − 𝑝)
𝑋𝑠 = 0.05 ℎ𝑅𝑠 Equation 3.40
𝑅𝑎 = 𝑅𝑚 [1 + 𝑘(ℎ 𝑓0 )0.5 ] Equation 3.41
73
𝑈2
𝑋𝑎 = 𝑋 ℎ [2(ℎ 𝑓0 )𝛽 ] Equation 3.42
𝑃 𝑝 𝑙𝑟
Where P is the active power absorbed by entire load area, p is the percentage of rotating loads, Xlr is
the locked rotor mean equivalent reactance, Rm is the equivalent series resistance, f0 is the fundamental
frequency, h is the harmonic order, while β and k are the parameters of the proposed model.
Typical parameters for use in this model are provided in [56].
A refinement to the model in [56] was presented in [57], also validated through experimental tests in
Italy. The equivalent circuit was identical to the one proposed in [56]. All equations for determination of
parameters were the same, except for the modification in the expression for Xs as shown below:
𝑋𝑠 = 𝛼 ℎ𝑅𝑠 Equation 3.43
The parameter α is an estimate of an average power factor of the static load (i.e. known Q/P). The
average values for k were found to be between 0.25 and 0.75. The field tests determined that the
parameter β is between -0.12 and -0.17. A value in the range of 0.03-0.04 pu (20% of Xlr) is
recommended for Rm.
3.4.1.6 Transmission Utility Composite Aggregate Harmonic Load Models

This subsection presents examples of aggregate models which use component- and measurement-
based approaches to establish the load composition for two UK transmission system operators (TSOs)
– (1) National Grid TSO in England and (2) Scottish Power TSO in Scotland.
Caution should be exercised regarding the generic applicability of these equivalent models on different
systems as the methods of connection of domestic customers and the topology of distribution networks
may vary significantly from country to country.
3.4.1.6.1 National Grid Equivalent Load Network Model

The model developed by National Grid is shown in Figure 3-42. In this model, BHV represents the total
cable capacitance at the HV supply bus and XT is the equivalent reactance representing all transformers
between the supply bus and LV connected customers. An equivalent LV bus is represented by a parallel
combination of resistive MW load (predominately heating and lighting), RLV, and LV capacitance, BLV,
to include power factor correction in the load and 400/415 V cable capacitance. If the loads have a
significant motor component, a series resistance and reactance can be added in parallel with RLV and
BLV. The motor should be represented at harmonic frequencies by its sub-transient reactance plus a
resistive element equal to its losses.
BHV
XT
0,415 kV
RLV BLV
Figure 3-42 Downstream Harmonic Component Load Model Representation at Supply Bus GSP (Typically
132kV) as used by National Grid TSO
The selection of the model parameters depends on the maximum permitted MW load seen at the supply
bus, the topology and the load composition. The load composition will depend on whether the load is
domestic, agricultural, commercial, commercial/domestic, industrial, lighting or traction, etc.
74
For National Grid, RLV is based on the proportion of load energy sales, where the various categories are
compiled according to domestic, farm, commercial, industrial and rail for the resistive and motor load
breakdowns with its assumed power factor correction (PFC). Table 3-9 gives an example of typical
proportions based on the maximum Average Cold Spell (ACS) demand (MW) to calculate RLV damping
(MW) and PFC (Mvar). It is important to note that these factors are pessimistic and are representative
of the typical load profile in the entire National Grid system.
Table 3-9 Typical values of RLV and PFC *courtesy National Grid
% of ACS Resistive MW PFC Mvar

100% 0.645*A 0.17*A
50% 0.25*A 0.125*A
30% 0.09*A 0.084*A
20% 0.06*A 0.06*A
The actual Mvar values for BHV and BLV are based on a nominal system feeding a 10 MW load from a
typical 132kV supply distribution system based on work carried out by the UK Electricity Council and
published in CIRED 1983. This has been used to derive the typical equivalent transformer reactance XT
and BLV as given in Table 3-10.
Table 3-10 Typical BLV (Power Factor Correction (PFC) + Cable Capacitance) and Equivalent XT
Transformer Reactance *courtesy National Grid
Approximate
Permitted Supply Bus Step-Down XT to 0.415kV BLV (Total 0.415kV
Max. MW Voltage (kV) Transformers (% on 100MVA) Capacitance) Mvar
Load
400 132 16 x 60 MVA 3.7 PFC + 26
350 132 14 x 60 MVA 4.1 PFC + 23
300 132 12 x 60 MVA 4.9 PFC + 20
250 132 10 x 60 MVA 5.7 PFC + 17
200 132 8 x 60 MVA 7.3 PFC + 13
150 132 6 x 60 MVA 9.5 PFC + 11
100 132 4 x 60 MVA 14.7 PFC + 7
50 132 2 x 60 MVA 26.3 PFC + 4.0
180 33 18 x 23 MVA 5.1 PFC + 3.6
150 33 15 x 23 MVA 6.1 PFC + 3.0
120 33 12 x 23 MVA 7.7 PFC + 2.4
90 33 9 x 23 MVA 10.2 PFC + 2.0
60 33 6 x 23 MVA 15.3 PFC + 1.0
60 11 3.7 PFC
50 11 4.4 PFC
40 11 5.5 PFC
30 11 7.3 PFC
20 11 60 x 500 KVA plus 11.0 PFC
200 x 50 KVA
The equivalent BHV is calculated based on the composition of cables in this typical distribution system
considering the cable capacitance at 33kV and 11kV. This is summarised in Table 3-11.
75
Table 3-11 Typical Cable Capacitance, BHV
Supply Bus Voltage (kV) BHV Mvar

132 Actual cable Mvar
33 2.8 Mvar per 60 MW
11 0.4 Mvar per 20 MW
3.4.1.6.2 Scottish Power Equivalent Load Network Model

The model shown in Figure 3-43 is used to represent the aggregated harmonic load model at GSPs in
the Scottish power system. In comparison to the National Grid model shown in Figure 3-42, it has an
additional motor component represented.
BHV
''
Z shc XT
0,415 kV
M
RLV BLV
Figure 3-43 Downstream Harmonic Component Load Model Representation at Supply Bus GSP as used
by Scottish Power TSO
The components for this model are as follows:

BHV represents the total capacitance at the GSP;
XT represents the equivalent reactance of the step-down transformers;
BLV is assumed to be scalable at 1% of actual P MW in Mvar;
RLV represents the resistive MW or heating component of the load which provides damping of the
resonances. Table 3-12 provides typical values for different load categories.
Table 3-12 Typical values of RLV for Different Load Categories
Heavy Light Commercial

Domestic
Industrial Industrial
RLV 0.9*A 0.3*A 0.5*A 0.7*A
To represent the network supplied from the GSP, Scottish Power has adopted a different approach than
that used by National Grid for deriving the equivalent BHV, BLV and XT and additional Zsc” values:
1. BHV is calculated by assuming a circuit capacitance of 0.1 µF/km at each GSP using an estimated
total length of circuits for the distribution system being fed from that GSP;
2. BLV is calculated by setting it to 1% of the actual P MW load in Mvar;
3. XT is constant (not scaled) and is assumed to be at 13.4% on 12 x n MVA, where n is the number
of primary transformers at each GSP;
4. Zsc” represents the non-static part of the load (motor) and is scalable at 1MVA per MVA fault infeed
to give a more pessimistic result for low demand conditions.
76
3.4.2 Examples of Load Modelling Approaches in Real Power Systems
3.4.2.1 Example 1: Load Behaviour and Modelling at Hydro Quebec (HQ)

A detailed study was conducted at HQ to investigate the effect of loads on the grid impedance seen
from a wind farm. The influence of selected load models from Figure 3-40 was also investigated. The
test case is shown in Figure 3-44.
Figure 3-44 Test Case
Three identical feeders were used, each having an average load composition. Only feeder 235 was
modelled in detail and its composition is given in Table 3-13. Each feeder has approximately 1.8 Mvar
of PFCs. The assumed load compositions are provided in
Table 3-14.
Table 3-13 Composition of 25 kV Load Feeder 235
Feeder 235
Residential 31.0 %
Commercial 3.5 %
Industrial 62.0 %
Others 3.5 %
Total 11.6 MVA
77
Table 3-14 Residential-Commercial Assumed Load Compositions

Peak Load Light Load
(February 7PM) (August 4AM)
Heating 50% 0%
Water heating 8% 10%
Electric stove 10% 0%
TV/PC 2% 5%
Lighting 10% (Incandescent: 7.2%; 35% (Incandescent: 25%;
Compact fluorescent: 2.8%) Compact fluorescent: 10%)
Single-phase motors 18% 50%
Dryers 2% 0%
The industrial loads were comprised of 65% small motors and 35% large motors (constant MVA).
Feeders with loads were modelled in detail and due to non-linear elements (e.g. power electronic
devices), the frequency responses were obtained by time-domain simulations carried out in batch mode.
78
3.4.2.1.1 Frequency Response of a 25kV Feeder

The frequency response of the unloaded feeder can be seen in Figure 3-45.
Feeder with Feeder with

light load; heavy load;
detailed detailed
model model
Unloaded feeder can be

conveniently represented
with a PI (see inset)
Figure 3-45 Modelling of 25kV Feeder with Various Load Levels
79
3.4.2.1.2 Model Comparison

Three load models (Method 1 (IEEE; “RL series”), Method 2 (IEEE; “RL//”) and Method 6 (IEEE; “2
RL//”)) from Figure 3-40 were compared against the detailed feeder model for the case of the feeder
with heavy load. The frequency scan of the feeder with heavy load is shown in Figure 3-46.
Feeder with heavy load;

detailed model
Figure 3-46 Model Comparison: Feeder with Heavy Load
From the results, Method 6 (IEEE; “2 RL//”) seems to be the most accurate representation when
compared to the detailed feeder model. Method 1 (IEEE; “RL series”) and Method 2 (IEEE; “RL//”) do
not fit the impedance angle below 260 Hz.
The three load models (Method 1 (IEEE; “RL series”), Method 2 (IEEE; “RL//”) and Method 6 (IEEE; “2
RL//”)) from Figure 3-40 were then compared against the detailed feeder model for the case of the feeder
with light load. The frequency scan of the feeder with light load is shown in Figure 3-47.
80
Feeder with light load;

detailed model
Figure 3-47 Model Comparison: Feeder with Light Load
From the results, Method 6 (IEEE; “2 RL//”) seems to be the most accurate representation when
compared to the detailed feeder model. Method 1 (IEEE; “RL series”) and Method 2 (IEEE; “RL//”) do
not fit the impedance angle below 250 Hz.
81
3.4.2.1.3 Frequency Response as seen from the Wind Farm (120kV)

The frequency response as seen from the 120kV wind farm (33km away) can be seen in Figure 3-48.
With feeders Feeders with light load; No

but no load detailed model feeders
Figure 3-48 Frequency Response as seen from the Wind Farm (120kV)
Key Points:
 There is an important damping effect at resonance even under light load;
 It is important to represent the load feeders to obtain an accurate resonance frequency.
82
3.4.2.1.4 Model Comparison

The same three load models (Method 1 (IEEE; “RL series”), Method 2 (IEEE; “RL//”) and Method 6
(IEEE; “2 RL//”)) from Figure 3-40 were compared against the detailed feeder model for the frequency
response as seen from the 120kV wind farm. The frequency scan results are shown in Figure 3-49.
Feeders with light load;

detailed model
Figure 3-49 Model Comparison as seen from the Wind Farm (120kV)
Key Points:
 With feeder representation, all models reproduce the correct resonance frequency.
 The damping effect of the load is not well represented by any of the 3 models.
83
3.4.2.1.5 Effect of Load Proximity to Point of Interconnection

The effect of load proximity to the point of interconnection (33km and 0km) was investigated and the
results shown in Figure 3-50.
Light load; detailed Light load; detailed

model at 33km model at 0km
Figure 3-50 Effect of Load Proximity to Point of Interconnection

Key Points:
 Load damping effect increases with load proximity to the point of interconnection;
 Method 6 (IEEE; “2 RL//”) produces less damping.
3.4.2.1.6 Results
This section presented an example in which Method 1 (IEEE; “RL series”), Method 2 (IEEE; “RL//”) and
Method 6 (IEEE; “2 RL//”) were compared. The key points observed from the study results are:
Key Points:
 Method 6 (IEEE; “2 RL//”), without being perfect, was found to most accurately model the load
behaviour with respect to the frequency.
 Detailed feeder and load modelling are the only way to precisely study the damping effects of the
load.
Consideration of the non-linearities introduced by components such as diodes in the power supplies of
televisions, computers and compact fluorescent lighting, may sometimes require that load modelling be
performed in the time-domain in batch mode.
84
3.4.2.2 Example 2: Load Behaviour and Modelling in Brazil

An example which illustrates the impact of load representation at different voltage levels and by
considering different load models is provided in Appendix E. It is based on a typical network in Brazil
but could be extended to other countries as well.
Key Points:
 The study example demonstrated that the load representation at 220 V with a detailed
representation of 69 kV and 13.8 kV network and through a system connection equivalent can
predict the system resonance.
 Load models at 230 kV and 69 kV (that did not represent the downstream impedance) do not capture
resonance.
 Results confirmed the importance of connecting the load model through an equivalent downstream
impedance.
3.4.3 Comparison of Load Models in a Test Grid

The aim of this section is to present and compare the distinct characteristics of different harmonic load
models (e.g. those recommended in IEEE and CIGRE publications) and those used in industry for
harmonic studies. The CIGRE 14-bus model (Figure 3-51) was used to connect the loads at different
voltage levels. This model was extended in order that the loads be incorporated at the MV level (33kV)
to properly study their impacts.
The selection of the linear load models and studies undertaken are summarised below:
1. Six load models comprising (IEEE, CIGRE) ) as provided in Figure 3-40 were modelled using the
extended CIGRE 14-bus model as shown in Figure 3-51 described in Section 3.4.1.
2. An aggregated load model used in a UK utility (shown in Figure 3-42) to represent a typical
distribution network was modelled in the extended CIGRE 14-bus model (Figure 3-51) described in
Section 3.4.2.
3. The extent of sensitivity of all these load models were assessed against different system demand
levels. This was studied considering three different system demand levels: winter, autumn and
summer demands to represent the seasonal variations of the year.
Figure 3-51 Extended CIGRE 14-Bus System for Comparison of Harmonic Load Models
85
This section presents the frequency impedance characteristics of the harmonic load models at the
busbars where the loads are connected (at the 33kV distribution level) eventually stepping up two
voltage levels through the grid transformers (110kV and 220kV) to the transmission system. A
combination of OHL and cable circuits were extended from the 110kV bus and modelled to represent a
simple connected 33kV distribution network as detailed in Table 3-15. Three loads at 33kV were
modelled and connected at “Bus3 33kV”, “Bus4 33kV” and “Bus5 33kV”, respectively. The profiles of
the loads representing winter, autumn and summer are provided in Table 3-16.
Table 3-15 Distribution-Level Circuits
Circuits Type Length (km) R, L and C Values

OH1 33kV Overhead line 10 0.21 Ω/km, 1.24 mH/km, 0.0089 µF/km
OH2 33kV Overhead line 5 0.21 Ω/km, 1.24 mH/km, 0.0089 µF/km
C1 33kV Cable 5 0.073 Ω/km, 0.3 mH/km, 0.26 µF/km
Table 3-16 Load Profiles Modelled
P (MW) Q (Mvar) % Motive Part Comments

90 10 10 Mostly resistive (winter)
60 40 40 Moderate inductive (autumn)
30 70 70 Highly inductive (summer)
The results at these busbars are described in 3.4.3.1 - 3.4.3.6 and a summary of results is provided in
3.4.3.7.
3.4.3.1 Comparison of Results Two Voltage Levels above Load Placement (at
“SB4 220kV”)
The results for the different load models show no observable difference for the harmonic frequencies
seen at 220kV for all demand levels (frequency scans of impedance are shown in Figure 3-52 to Figure
3-54). Only Model 1 using the series R and L was observed to show reduced damping when demand
levels decrease at the 11th harmonic order (Figure 3-54 shows a higher peak than Figure 3-53).
Figure 3-52 Impedance at SB4 220kV – Winter Load
86
Figure 3-53 Impedance at SB4 220kV – Autumn Load
Figure 3-54 Impedance at SB4 220kV – Summer Load
3.4.3.2 Comparison of Results One Voltage Level above Load Placement (at
“Bus1 110kV”)
At “Bus1 110kV”, the effects of the different load models at harmonic frequencies for all demand levels
(frequency scans of impedance are shown in Figure 3-55 to Figure 3-57) can be seen.
Model 1 using the series R and L showed significant differences as the demand levels reduced, with a
much higher parallel resonance around the 35th harmonic order for the autumn load, shifting to around
the 32nd for the summer load. This shift is expected due to the increased inductive component of the
load, and because this model is the only series-connected model (the others are parallel-connected).
This affected the resonance noticeably at the higher harmonic orders (above the 22nd in our test CIGRE
network model) when the demand levels were lower (Figure 3-56 and Figure 3-57).
During the winter load, as shown in Figure 3-55, the UK utility load model (NG Model) is largely in
agreement with Models 2 - 6 but has introduced an additional peak resonance near the 17th harmonic
order. This is due to the capacitance representing the cables and power factor correction in the
distribution network.
87
Figure 3-55 Impedance at Bus1 110kV – Winter Load
Figure 3-56 Impedance at Bus1 110kV – Autumn Load
Figure 3-57 Impedance at Bus1 110kV – Summer Load
88
3.4.3.3 Comparison of Results at “Bus2 33kV”

At “Bus2 33kV”, the results show that some of the load models differ for the harmonic frequencies at all
demand levels (frequency scans of impedance are shown in Figure 3-58 to Figure 3-60).
Models 2-6 show similar load impedance characteristics for all demand levels, however Model 1 resulted
in a significant difference as a series R and L model. Model 1 yields a significantly higher parallel
resonance around the 35th harmonic order for the autumn load, shifting to around the 32nd for the
summer load (as previously explained). This effect is seen at the 110kV bus.
The NG Model is largely in agreement with Models 2-6 up to the 12th harmonic order, after which three
parallel resonances are introduced (around the 17th, 19th and 22nd harmonic orders). The effect of this
load model resonance at the 17th harmonic order also affects the impedance as is evident in the results
for “Bus1 110kV”; therefore not only affecting the 33kV buses.
These are the impedance characteristics seen from the 33kV terminal of the 110kV/33kV distribution
transformer with combined effects of (i) all three loads; and (ii) all 33kV circuits connecting these loads.
89

At “Bus3 33kV”, the results show that some of the load models differ at the harmonic frequencies for all
demand levels (frequency scans of impedance are shown in Figure 3-61 to Figure 3-63).
Models 3-6 show similar load impedance characteristics for all demand levels except Model 1 and Model
2, which show significant differences at the 33kV buses to which they are directly connected. The
difference between Model 1 and Model 2 is the series and parallel topology with a constant P and Q
used for these two models. Model 1 causes a resonance around the 35th harmonic order for the autumn
load, shifting to around the 32nd harmonic order for the summer load (as previously observed for “Bus2
33kV”).
The NG Model is largely in agreement with Models 3-6, except between the 12th and 28th harmonic
orders, where three parallel resonances are introduced (around the 17th, 19th and 22nd harmonic
orders). This was also seen for “Bus2 33kV”. The resonance seen at “Bus3 33kV” coincides and is
directly influenced by this load.
90

At “Bus4 33kV”, the results show that some of the load models show differences at the harmonic
frequencies for all the demand levels (frequency scans of impedance are shown in Figure 3-64 to Figure
3-66).
Models 3-6 show similar load impedance characteristics for all demand levels, except Model 1 which
show significant differences at the 33kV buses to which it is directly connected. Like the other 33kV
buses, this load model causes resonance around the 35th harmonic order for the autumn load, shifting
to around the 32nd for the summer load.
It is seen that the NG Model is largely in agreement with Models 2-6, except between the 12th and 28th
harmonic orders where three parallel resonances are introduced (around the 17th, 19th and 22nd
harmonic orders). This was also seen for the other 33kV buses. The resonance seen at this “Bus4 33kV”
coincides with that seen at the other 33kV buses and is also directly influenced by this load.
91
92

At the “Bus5 33kV”, the results show similar behaviour to the other 33kV buses for the harmonic
frequencies, and the same explanation applies (frequency scans of impedance are shown in Figure 3-67
to Figure 3-69). A 33kV cable connection is used instead of an OHL. The peaks of the resonance are
directly influenced by the NG Model for all 33kV buses.
93
3.4.3.7 Results
From the results comparisons, it was seen that different load models used will have a significant
influence on harmonic characteristics depending on (i) where the models are connected (e.g. connected
at MV and influence at EHV) and (ii) which models are used. The key points observed from the study
results are:
Key Points:
 The CIGRE and IEEE load models need to include other distribution components in the transmission
network when used to represent the demands on a distribution system. If specific knowledge of load
composition at the connected node is known, the appropriate CIGRE and IEEE models can be used.
 Appropriate load models should be used as these can affect what is seen at EHV and HV levels
even when connected at MV (e.g. at 33kV in the 14-bus IEEE model). The differences between the
load models seen at EHV and HV buses diminish as the number of transformation levels increases.
 Different load models are required to represent the various categories of loads in the network
(distribution loads, industrial loads, power station loads etc.).
 The utility aggregated load model used for the UK represents the typical load composition of a
distribution load in an urban area. The various parameters for the load model are therefore not
representative for different regions and countries.
3.4.4 Summary
Providing the engineer with clear guidelines to develop equivalent harmonic load models is a non-trivial
task.
Considering load modelling from a system point of view, lumping load at a certain voltage level (probably
in the range of MV), the following should be considered:
i. According to the different operating conditions of the network (light/peak, summer/winter), the
equivalent lumped load could change not only in terms of rated power but also in terms of load
type (for example modelling operating conditions with different percentages of rotating loads
with respect to static loads);
ii. The equivalent lumped load, seen from a network bus, could consist of different contributions
to be used to represent the main characteristics of the distribution network. For example, a
lumped load could consist of motors, static loads and shunt capacitor banks of the distribution
network (or cables). These contributions can have different weightings according to the different
operating conditions of the network.
The equivalent load is a particular component of the network model, and it can assume different
operating states to model the different operating conditions of the distribution network. This factor can
94
result in many cases for harmonic studies (to be combined, for example, with different contingencies).
Guidelines for equivalent load representation should also consider consequences regarding the number
of cases that will then need to be studied. This number will be largely determined by a combination of
load conditions, contingencies and other possible factors.
Loads usually have an important damping effect near harmonic “emitting” installations such as SVCs,
HVDC and wind farms. Analyses presented in this section highlighted the difficulties in obtaining an
accurate assessment of the damping using various frequency-domain models. For a highly accurate
assessment of the damping effect of loads, detailed time-domain modelling may occasionally be
preferable. Neglecting loads may result in a pessimistic evaluation of harmonic distortion and
subsequent over-filtering.
The models described in this section are linear load models. A significant increase in power electronic
type loads is expected (in the form of embedded PVs, EVs and standalone PVs, wind farms, charging
sites for EVs at distribution levels, data centres, etc.), meaning that it will become increasingly important
to consider the non-linear effects of these types of loads in the models. This is an area where further
research work is required to develop enhanced load models that accurately capture the behaviour of
these new types of devices distributed throughout the power system.
Detailed modelling of loads is a very difficult task primarily because of the lack of information about the
load composition, type of feeders (overhead or underground) and locations of distributed power factor
correction capacitors. Correct consideration of the impacts of loads in harmonic studies is a large task
and requires time-consuming efforts in gathering information and performing component modelling.
Depending on the specific installation, it is up to the engineer to judge whether such efforts are
economically justified.
3.5 Synchronous Generators

3.5.1 Models
For generators subjected to a harmonic disturbance at frequency hω, harmonic currents are drawn at
hω and at (h +/-2ω), where h is the harmonic order and ω is the angular frequency. Therefore, generators
cannot be modelled with full accuracy by an impedance at a single frequency.
Frequency-domain models of synchronous machines suitable for harmonic studies have been well-
described in the literature [8], [58]. At harmonic frequencies, a synchronous generator can be
represented as a passive impedance by a reactance 𝑋𝑔𝑒𝑛ℎ and a resistance 𝑅𝑔𝑒𝑛ℎ . The representation
of the harmonic impedance of a generator is shown in Figure 3-70.
Figure 3-70 Synchronous Generator Harmonic Impedance
As the rotating magnetic field created by stator harmonics rotates at a speed significantly higher than
that of the rotor, the harmonic impedance of the generator approaches its negative sequence impedance
[8]. The IEEE Task Force on Harmonic Modelling and Simulation suggests that the synchronous
machine harmonic reactance be taken to be either the negative sequence impedance or the average of
95
the direct and quadrature sub-transient reactance. The generator reactance is commonly represented
by [8]:
(𝑋𝑑" + 𝑋𝑞" )
𝑋𝑔𝑒𝑛ℎ = ℎ ⋅ 𝑋2 = ℎ ⋅ Equation 3.44
2
Various expressions are available to describe 𝑅𝑔𝑒𝑛ℎ which is highly frequency-dependent. Figure 3-71
illustrates the variation of 𝑅𝑔𝑒𝑛ℎ and 𝑋𝑔𝑒𝑛ℎ as a function of harmonic order, derived by the manufacturer
of a 370 MVA salient pole generator owned by Hydro-Quebec (HQ). The sub-transient reactance of this
machine is 0.24 pu at 13.8 kV.
Figure 3-71 Sample 370 MVA Synchronous Generator Harmonic Equivalent Resistance, Rgenh, and
Reactance, Xgenh
The following subsections describe three common approaches used to represent the harmonic
impedance of synchronous generators:
3.5.1.1 Model 1: IEEE

The IEEE model [49] uses a frequency-domain equivalent passive shunt impedance, based on the
negative sequence impedance. This is illustrated in Figure 3-72.
Figure 3-72 Model 1: IEEE
𝐿1 = 𝑋2 /(2 ⋅ 𝜋 ⋅ 𝑓nom ) [𝐻] Equation 3.45

𝛼
𝑅ℎ = ℎ ⋅ 𝑅2 [0.5 ≤ 𝛼 ≤ 1.5] Equation 3.46
96
Where R2 and X2 are the negative sequence resistance and reactance, respectively, measured at power
frequency.
Measurement data shows that the resistance, R h , is highly frequency dependent. An α value of 0.5 is
theoretically correct, however a value of 1.0 was verified through limited tests [59].
3.5.1.2 Model 2: Electra 167

The Electra 167 model [2] uses a frequency-domain equivalent passive shunt impedance, derived only
from the sub-transient reactance.
Figure 3-73 Model 2: Electra 167
𝑅1 = 0.1 ⋅ 𝑋𝑑" [𝛺] Equation 3.47

𝐿1 = 𝑋𝑑" /(2 ⋅ 𝜋 ⋅ 𝑓nom ) [𝐻] Equation 3.48
𝑅ℎ = √ℎ ⋅ 𝑅1 [𝛺] Equation 3.49
where Xd” is the generator sub-transient reactance, measured at power frequency.
3.5.1.3 Model 3: HQ
The Hydro Quebec (HQ) model is an in-house model developed to fit manufacturer data shown in Figure
3-71. It models an equivalent passive shunt impedance and is applicable in both the frequency domain
and time domain.
Figure 3-74 Model 3: HQ
𝑋 = 𝑋2 [pu] Equation 3.50
97
𝑋
𝑅1 = 𝑍𝑏𝑎𝑠𝑒 ⋅ [𝛺] Equation 3.51
10
𝑋
𝐿1 = 𝑍𝑏𝑎𝑠𝑒 ⋅ [𝐻] Equation 3.52
2 ⋅ 𝜋 ⋅ 𝑓nom
𝑅2 = 𝑍𝑏𝑎𝑠𝑒 ⋅ 𝑋 ⋅ 100 [𝛺] Equation 3.53
20
𝐿2 = 𝑍𝑏𝑎𝑠𝑒 ⋅ 𝑋 ⋅ [𝐻] Equation 3.54
2 ⋅ 𝜋 ⋅ 𝑓nom
𝑅 = 𝑍𝑏𝑎𝑠𝑒 ⋅ 𝑋 ⋅ 500 [𝛺] Equation 3.55
3.5.2 Comparison of Synchronous Generator Models in a Transmission Grid

The effect of synchronous generator modelling (impedance and loss) may be insignificant in the
presence of other series impedances and losses (transformers, lines and loads, without accurate
modelling), and the location of the generator in relation to loads. Various case studies which illustrate
these modelling aspects are presented here.
The three synchronous generator models described in Section 3.5.1 were compared in a full system
model representing the transmission grid in Ireland. The base-case reflects a model without frequency
dependency in the resistive components of the generator and each of the subsequent cases
incorporates the frequency dependent characteristics described by Equation 3.46, Equation 3.49 and
Equation 3.55. For the IEEE model, three values of the alpha () exponent were tested.
Frequency scans were performed at 18 transmission nodes scattered throughout the system, giving a
representative overview of the performance of the different synchronous generator models at various
locations in the grid. Two characteristic cases are shown in Figure 3-75 . The graph on the left (Figure
3-75 (a)) represents a transmission node close to a large synchronous generator. The amplitude of the
harmonic resonances is affected by the selected model. The graph on the right (Figure 3-75 (b))
represents a transmission node electrically distant from any synchronous generator. In this case, it can
be observed that the selection of a synchronous generator model (from those studied) has no material
impact on the harmonic impedance at that point.
From the three models under assessment, the IEEE representation with an exponent of 1.5 provided
the highest damping whereas the Electra 167 representation provided the least damping. It was also
observed that the choice of model did not affect the location of the resonances.
Z(h) [ohm] Close To Generation Z(h) [ohm] Far From Generation

800 1200
1000
600
800
400 600
400
200
200
0 H 0 H
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
base case IEEE-ALPHA_0.5 IEEE-ALPHA_1 base case IEEE_ALPHA_0.5 IEEE_ALPHA_1
IEEE-ALPHA_1.5 ELECTRA_167 HQ IEEE_ALPHA_1.5 ELECTRA_167 HQ
Figure 3-75 Effect of Detailed Synchronous Generator Models on Harmonic Impedance in a Real
Transmission System (a) Close to Generation and (b) Far from Generation
To better compare the models, damping factors were calculated as the ratio between the harmonic
impedance obtained with each model to the harmonic impedance obtained in base case, with a constant
generator resistance. A summary of the damping ratio factors provided by each model is presented in
Figure 3-76 for each of the 18 transmission nodes. In this example, the Electra 167 model gave damping
ratio factors close to 1 at all transmission nodes, indicating that this representation is almost equivalent
to neglecting frequency dependency of resistive loses (i.e. base case). On the other hand, the IEEE
model with an exponent of 1.5 gave the lowest damping ratio factors, ranging from 0.58 to 0.94
depending on the transmission node. This indicates that, in the most extreme case (transmission node
#17 – conventional generation plant), the choice of model could result in a reduction of the harmonic
impedance by almost half. Those nodes electrically close to synchronous generation (e.g. nodes from
98
#12 to #17) are most affected by the choice of model, as expected. Therefore, when performing
harmonic analysis in areas close to synchronous generation, the use of frequency-dependent data from
manufacturers should help obtain the most accurate results.
1
0.95
(with respect to Base Case)
0.9
0.85
Damping Factor
0.8
0.75
0.7
0.65
0.6
0.55
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Transmission Node
IEEE_alpha_0.5 IEEE_alpha_1 IEEE_alpha_1.5 Electra_167 HQ
Figure 3-76 Maximum Damping Factor (with Respect to Base Case) Provided by each Synchronous
Generator Model
3.5.3 Summary
Three models were described in this section, and case studies presented highlighting various important
aspects of synchronous generator modelling. The value of the alpha exponent used in the IEEE model
should be selected appropriately. The accurate modelling of the (highly) frequency-dependent
resistance, R, depends on having accurate data for R and on the exponent used. Where possible it is
highly recommended in [59] to obtain this data either via the manufacturer or via measurement.
A comparison of the models via frequency scans using a model of the Irish transmission system suggest
that the selection of a synchronous generator model only has a noticeable effect at the parallel resonant
points and at nodes that are electrically close to synchronous generators. The IEEE model with an alpha
exponent of 1.5 provides the highest level of damping and the Electra 167 model provides the lowest
level of damping. It is expected that this model yields the most pessimistic results as it considers √ℎ.
Aspects such as model validation, sensitivity analysis and model simplification were illustrated in
Appendix F via measurement data and simulation results. The accuracy of the HQ model was confirmed
via measurements; the importance of modelling high frequency losses was shown, and the accuracy of
a simplified model demonstrated.
99
3.6 Shunt and Series Compensation

3.6.1 Shunt Capacitors
Shunt capacitors are represented as simple lumped components with a linear susceptance calculated
from Equation 3.56 where h is the harmonic order, Q is the Mvar output at fundamental frequency and
1 pu voltage and V is the system nominal voltage.
Resistive losses tend to be low and are typically neglected, however if information is available, they can
be represented as a lumped resistor connected in series with the capacitor. Any associated inrush
current-limiting reactors or discharge resistors, if applicable, also need to be represented in a similar
fashion.
1 ℎ∙𝑄
𝐵= = 2 Equation 3.56
𝑋𝐶 𝑉
In order to avoid low order harmonic resonances with the transmission network, it is common practice
in some countries to detune the capacitor banks (otherwise known as mechanically-switched capacitors,
MSC). This is achieved by installing a small reactor in series with the capacitor or by adding a damping
network (MSCDN) – see Figure 3-77. The capacitor bank effectively becomes a single-tuned filter (MSC)
or a C-Type filter (MSCDN). These devices need to be explicitly represented in the system model as
they have a significant impact on the harmonic performance of the network.
Reactor losses are determined by the quality factor (Q = X/R), which is normally defined in the customer
specification or in the manufacturer data sheet. It is advisable to incorporate these losses as a lumped
resistor in series with the reactor. As data on frequency dependency of the resistive losses is rarely
available (and these losses tend to be low), it is generally acceptable to assume a constant resistance
value.
For the purposes of harmonic modelling, each component is explicitly represented as a lumped element,
connected in series or parallel as per design scheme (Figure 3-77). The phase connection mode (wye
or delta) and treatment of the neutral point (grounded or ungrounded) should be accurately represented
as it affects the flow of zero-sequence harmonic currents.
Figure 3-77 Detuned Shunt Capacitors

As an example, Figure 3-78 illustrates the effect of the connection of shunt capacitors in the harmonic
impedance of a real system. The Irish transmission network model has been used in this example. The
frequency scan results show that the connection of a detuned capacitor bank (MSC) with or without
additional damping does not affect significantly the harmonic impedance of the network. The only visible
impact can be observed at the detuning frequency (h = 2.8 in this example) where the device behaviour
changes from capacitive to inductive, introducing a spike in the harmonic impedance (purple curve).
This effect can be minimised with the introduction of a damping resistor in parallel with the reactor (green
curve). Figure 3-78 also shows that the connection of a “pure” capacitor bank (brown curve) causes a
material change in the impedance of the network over a wide range of frequencies. Of particular interest
is the shift of parallel resonance in the low harmonic order range. This example demonstrates the
importance of accurate modelling of shunt capacitor banks.
100
500
Z [] Natural resonance in
network without
400 capacitor banks or with
detuned capacitor
Shifted resonance
banks
due to capacitor
300 bank without
detuning
Capacitor
200 bank
detuned
frequency
100
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Harmonic Order
Without Capacitor Banks With 30MVAr Capacitor Bank
With detuned 30MVAr Capacitor Bank With detuned 30MVAr Capacitor Bank - no damping R
Figure 3-78 Effect of connecting Shunt Capacitors in a Transmission Network

Key Points:
 Resistive losses should be included if data is available;
 Incorporation of reactor losses via a lumped resistor in series with the reactor is recommended;
 The accurate modelling of shunt capacitor banks is important in harmonic studies as they can
significantly affect the system harmonic impedance.
3.6.2 Shunt Reactors

Shunt reactors are represented in the same fashion as shunt capacitors – i.e. as lumped components.
The same considerations for resistive losses and component tolerances also apply here. Stray
capacitances can normally be neglected for the range of frequencies considered in typical harmonic
studies.
There are two types of reactors: air-core and oil-filled, with the main difference being the presence of a
magnetic core, with associated saturation effects and magnetic coupling between phases. However, for
the purposes of most steady-state harmonics studies, these effects can be neglected. In special cases
where more detailed modelling is required, the modelling guidelines provided for power transformers
can be applied to oil-filled reactors.
Typical applications of shunt reactors include:
 Connected to the LV delta windings of a transformer: for voltage control;
 Connected at station buses: for voltage control;
 Connected at line terminal(s): for voltage control;
 Connected at line terminal(s) with a neutral reactor: for voltage control and for single-pole trip
and reclose (neutral reactor may be required for secondary arc extinction considerations).
The effect of shunt inductive compensation on the harmonic performance of a transmission line has
been illustrated in [3]. It has been shown that, while the shunt compensation improves the voltage profile
at the fundamental frequency, it will also affect the harmonic profile. It is therefore important to represent
these shunt devices in the models used for harmonic propagation studies.
Key Points:
 Shunt reactors should be accurately represented in harmonic studies;
 Stray capacitances can usually be neglected.
101
3.6.3 Passive Harmonic Filters

Passive filters are the most commonly-used form of harmonic mitigation. A filter is essentially a
combination of inductive, capacitive and resistive components carefully selected and arranged to form
a resonant circuit at the tuned frequency (or frequencies) and/or to introduce resistive damping. Some
of the most commonly used filters are shown in Figure 3-77 (MSCDN or C-Type) and Figure 3-79. A
detailed description of each type and discussion on their performance, advantages and disadvantages
can be found in CIGRE TB139 [80].
Passive filters are modelled as series and/or parallel combinations of lumped individual components
arranged as per selected scheme. The modelling approach is the same as that for shunt capacitors and
shunt reactors – i.e. account for resistive losses (and frequency dependency if data is available),
component manufacturing tolerances and detuning due to fundamental frequency and/or ambient
temperature deviations (see example in section 4.3.1.3).
Single-tuned Triple-tuned Damped

bandpass Double-tuned bandpass highpass
filter bandpass filter filter filter
C1 C1 C1
C1
R1 L1 R1
L1 L1 R1 L1
R1 R2 L2 C2 R2 L2 C2
R3 L3 C3
Figure 3-79 Examples of Typical Filter Configurations
3.6.4 Series Capacitors

Series capacitors are represented as lumped parameters connected between two nodes. For modelling
purposes, they are considered as homogeneous sub-sections within a transmission line, and the overall
line model comprises the cascade connection of all discrete sub-sections, including the series
capacitors, as illustrated in section 3.1.3.
Manufacturing tolerances and capacitance variation with temperature need to be taken into
consideration in the model. Resistive losses can be represented by a series-connected resistor if
information is available.
Associated protection equipment such as surge arresters, spark gaps, by-pass circuit breaker, current
limiting reactor, damping resistor, etc. do not normally need to be captured in the model for steady-state
harmonic studies. The reader should be aware, however, that the bypass circuit is normally tuned to a
particular frequency in the range of around 10th to 20th harmonic. While bypass circuit operation is
transient in nature, harmonic resonance conditions could be problematic. In some cases, depending on
level of background harmonics in that frequency range, the tuning of the bypass circuit may need to be
investigated and its components represented in detail.
Key Points:
 Resistive losses and manufacturing tolerances should be included if data is available.
 Representation of the bypass circuit is not normally required, unless tuning of this circuit needs to
be investigated due to high background harmonic.
102
3.6.5 Series Reactors

Only air-core series reactors are typically used in transmission applications. These devices can be
represented in the same fashion as air-core shunt reactors described in section 3.6.2.
3.6.6 Frequency-Dependency of Passive Components

There is limited literature concerning the frequency dependence of passive components used in passive
filters. For example, with the large-scale adoption of VSC with active switches and PWM, the series
reactor used at the output of the VSC is subject to rectangular excitation due to the PWM with much
higher frequency than the fundamental frequency. The resulting losses in the filter reactor can be as
high as 1.5-2% depending on the adopted design [117]. Due to wide spectrum harmonics related to the
PWM, is not yet clear how these losses affect the frequency dependence of the reactors and its
equivalent damping. For series reactors driven by sinusoidal excitation (e.g. sine wave filters), the
equivalent model is similar to that of the power transformer; i.e. it has a relatively linear inductance with
current and frequency while the equivalent resistance is frequency dependent. Shunt reactors are like
series reactors, with the difference that the applied excitation voltage is much lower and includes
harmonic components.
The equivalent models of resistors and capacitors can be taken as ideal if the frequency of interest is
relatively low and they have a low temperature coefficient.
3.6.7 Summary
The modelling of shunt capacitors, shunt reactors, series capacitors and series reactors as simple
lumped components was discussed in this section. The effect of capacitor detuning and resistive losses
modelling on the harmonic impedance was demonstrated through an example using a shunt capacitor
in a real transmission system model. The effect at low-order harmonic frequencies was highlighted,
showing that modelling as accurately as possible is important. Passive harmonic filters are modelled as
combinations of lumped components, and typical filter configurations were illustrated.
3.7 Network Equivalent for Harmonic Studies

In order to calculate propagation of harmonic currents through the network accurately, it would be
necessary to have a good representation of all network elements (as described in this chapter) as well
as all harmonic distortion sources (as described in chapter 4). For many studies however, the extent of
the network model may be limited for a number of practical reasons. These could include model data
sharing between neighbouring network owners/operators, different specialist software packages,
division of work between teams or people, protection of user data confidentiality, simplicity of exchanged
data, etc. It might simply be a practical requirement to reduce the network model in order to speed up
studies.
In such cases, the rest of the network for which the system elements (lines, cables, transformers, loads,
etc.) are not represented in detail, can be represented by an equivalent network. Such an equivalent
represents the network “as seen” from the point(s) of connection, in terms of the background harmonic
voltage distortion and the network impedance as a function of frequency. By use of such an equivalent,
the assessments can be undertaken based on general rules of procedure.
The equivalents are generally provided as either Thévenin or Norton equivalent [61], which are
effectively identical in their performance and can be converted from one to another.
For harmonic studies in frequency domain, the Thévenin equivalent consists of impedance matrix and
series voltage sources whereas the Norton equivalent consists of an admittance matrix (network) and
shunt current sources. The equivalents are generally given for each frequency and a number of
operating conditions represented by discrete points on the R-X diagram or as an area in the R-X diagram
[80], normally defined as an envelope of all possible operating conditions.
The background harmonic distortion can be given as a fixed value based on measurements and/or
estimations or it can be given based on certain restrictions, such as compatibility levels, planning levels
or otherwise determined permissible incremental or aggregate values.
The reader is referred to chapter 5 for detailed discussions and recommendations on the extent of the
model (5.3.1), scenarios and contingencies to consider (5.3.2), development of harmonic impedance
envelopes (5.4) and measurement/estimation of background distortion (5.8).
103
104
4. Power electronic based network element models

4.1 Introduction
This chapter addresses the representation of non-linear devices connected to power systems acting as
sources of harmonic distortion. The mechanisms of harmonic distortion generation are explained and
recommendations on the modelling structure and input data requirements are provided for each device.
A generic structure for the harmonic models of these devices is presented first, followed by sub-sections
dealing with the most common sources of harmonic distortion based on power electronic devices.
4.1.1 Generic Norton/Thévenin Equivalent Harmonic Model Structure

The traditional approach to represent harmonic-generating devices in harmonic studies has been as
ideal (constant) current sources. This approach originates from times when penetration levels of power
electronic devices in the grid were very low and mostly dominated by load-commutated converters and
diode rectifiers, for which the constant current assumption can be a reasonable simplification. However,
recent proliferation of converter-based devices connecting at all voltage levels, being mostly thyristor or
IGBT based, drives the need for a more accurate representation taking into account not only the
harmonic emissions but also the harmonic impedance of the device, which can interact with the feeding
network causing harmful resonances.
A generic harmonic model structure represented as a Norton/Thévenin equivalent circuit is proposed
here for all type of power electronic based equipment – see Figure 4-1. The Norton/Thévenin equivalent
circuit is represented by means of an equivalent ideal current/voltage source (Table 4-1) and equivalent
impedance for each frequency of interest (Table 4-2). This model structure has already been proposed
to represent converter-based wind generation in IEC TR 61400-21-3 ED1 [62]. It is proposed here to
extend its application to all other harmonic generating devices based on power electronic converters.
This model structure will be applied to all devices described in this chapter.
Thévenin Equivalent Circuit Norton Equivalent Circuit
Thévenin Equivalent Norton Equivalent

Harmonic Impedance Harmonic
Impedance
Thévenin Equivalent Norton Equivalent

Harmonic Voltage Harmonic Current
Source Source
Figure 4-1 Norton/Thévenin Harmonic Model of Power Electronic Converter (source [62])
105
Table 4-1 Example Representation/Template of the Harmonic Voltage/Current Source
Harmonic Order Frequency Harmonic Voltage/Current V/I

Magnitude (RMS)
[-] [Hz] Phase []
[V/Amps]
2
3
4
…
Table 4-2 Example Representation/Template of the Harmonic Equivalent Impedance
Harmonic Order Frequency Harmonic Impedance Z

Resistance Reactance
[-] [Hz]
R [Ω] X [Ω]
2
3
4
…
4.1.2 Considerations on Norton/Thévenin Modelling Limitations
4.1.2.1 Cross-Sequence Coupling and Cross-Frequency Coupling

The underlying assumption of Thévenin or Norton equivalent modelling is a linear system. This
assumption may not always prove accurate for power converters. There are two types of coupling
present in power converters. The first is cross-sequence coupling which is the coupling between the
positive sequence and the negative sequence. It means that a positive sequence source not only
produces a positive sequence response, but also a response from the negative sequence. This also
applies to a negative sequence source. The impedance matrix must therefore include the sequence-
coupling impedance. The second type of coupling is cross-frequency coupling, which implies multiple
frequency responses by a single frequency source. These coupling phenomena are associated with the
design of converter controls including PLLs. To completely cover the abovementioned coupling issues,
the more sophisticated harmonic domain approach (see Section 2.5) would be required. However, from
the 2nd harmonic order and above, cross-sequence couplings and cross-frequency couplings are minor
and can be ignored in most system-wide harmonic studies. Should further information be available which
indicates strong couplings at harmonic orders of interest, then more detailed model and analysis
techniques would be required (note: these are outside the scope of this Technical Brochure).
4.1.2.2 Active Filtering

Active filtering in power-electronic converters is generally defined as any control functionality which
leads to intentional harmonic distortion level improvement at the point of interest; i.e. extra noise
rejection capability, damping, harmonic emission reduction from internal non-linear components or
controlled harmonic current injection. ‘Active’ in this context means a control circuit comprising active
components; e.g. a circuit that transfers energy from its power supply to the load.
Active filtering in power-electronic converters (such as those deployed with wind turbines and PV) will
influence their harmonic performance. Thus, the standard “static” Norton/Thévenin equivalent harmonic
model comprising harmonic source and harmonic equivalent impedance will need to adapt its
characteristic to accommodate the new “active” component. Depending on the applied mitigation
strategy, the changes can affect the harmonic source, the harmonic impedance or both. This could be
incorporated, for example, as dynamic look-up tables although it is anticipated to be a complex process
due to the variability of the “active” outputs. As this technology is still maturing, further work is required
to develop robust methods to adapt the frequency domain harmonic model to the selected strategy and
operating point.
106
4.2 Converter-Based Generation

Electrical power systems worldwide are undergoing a transition towards increased penetration of
emerging power electronic technologies. This affects the background harmonics content at all levels,
from generation to transmission and distribution. Renewable energy sources, such as wind, solar, wave
and tidal, are accelerating this trend due to the advantages of control flexibility offered by power
electronic converters. These kinds of renewable generation employ power converters for energy
conversion, either with a partial power converter (e.g. DFIGs) or a full power converter. Appropriate use
of power electronics can improve power quality, however power converters used in renewable
generation can also be a significant source of harmonics [63].
The study of harmonics in converter-based generation is more challenging than in conventional power
generation due to the complexity of the power converters. Power converters can produce substantial
harmonics over a wide frequency range due to their non-ideal operation, non-linearity of hardware and
the selection of control strategy.
Although power converters are a source of harmonic distortion in electrical power systems, they also
exhibit frequency-dependent impedances which can shape resonance frequencies in the feeding
network. Precise prediction of resonance points can help to reduce unnecessary investment in filters.
For this reason, it is of importance to model converter-based generation in a sufficiently-detailed manner
where both the harmonic contribution and internal impedance are included, to accurately capture their
net effect on the harmonic performance of the power system. The general requirements for an accurate
converter-based generator model include:
 Correct representation of the converter reaction to background harmonic voltages in the connected
grid;
 Correct representation of the influence of the connected grid impedance on the harmonic emissions
from the converter;
 The ability to be applied in conventional harmonic assessment studies.
The following subsections describe the most common types of converter-based generators and their
modelling on an individual device level. Aggregation of multiple components in a wind farm / PV farm is
covered in chapter 5, sections 5.6 and 5.7.
4.2.1 Converter-based Wind Generation

Converter-based wind turbine generators (WTGs) include type 3 and type 4 wind turbines. Both types
are widely installed in many countries. Because of their distinct use of power converter and control
strategies, there are differences that should be considered when developing a harmonic model.
4.2.1.1 Type 3 Wind Turbine Generation – Doubly-Fed Induction Generation

Doubly-fed induction generators (DFIGs) have been very popular for many years as they provide several
advantages such as [64], [65]:
 Higher yield compared to constant speed wind turbines;
 Reduced mechanical stresses;
 Cost effectiveness;
 Improved power quality (less flicker);
 High efficiency;
 Controllability of active and reactive power.
DFIGs are widely employed in MW-range wind turbines to facilitate variable speed operation. The DFIG
is essentially a wound rotor induction machine fed by both the stator and rotor. It is comprised of a back-
to-back voltage-source power converter: a rotor-side converter and a grid-side converter. The stator of
the DFIG is directly connected to the grid, and the two converters are included in the rotor circuit [64].
The grid-side converter regulates the DC bus voltage and can absorb and generate reactive power [66].
The rotor-side converter is responsible for controlling the active and reactive power output of the
generator. The DFIG is illustrated in Figure 4-2.
107
Figure 4-2 Doubly-Fed Induction Generator [65]
The presence of power converters based on forced-commutated power electronics causes the
production of harmonics and interharmonics. Harmonics are produced in DFIGs via three predominant
mechanisms [66]:
1) Grid-side converter: fast switching from the current-controlled VSC produces high-frequency
harmonics and interharmonics;
2) Rotor-side converter: low- and high-order rotor harmonic components propagate to the grid;
3) DFIG windings: high frequency harmonics in the generated voltage due to the non-linear effect in
the air-gap flux and slip.
4.2.1.1.1 Type 3 WTG Modelling

DFIGs are potential harmonic sources, whose harmonic current can vary widely, and should therefore
be modelled accurately for harmonic studies. Improper modelling (such a simple constant current
source) can lead to inaccurate harmonic distortion results and improper filter designs [67].
The traditional method used to represent harmonic-generating devices in harmonic studies has been as
ideal (constant) current sources. This approach originates from times when penetration levels of power
electronic devices in the grid were very low and mostly dominated by load-commutated converters and
diode rectifiers, for which the constant current assumption can be a reasonable simplification. However,
this assumption has since proven to be inaccurate for the modelling of DFIGs [67] and it can lead to
significant errors as the penetration level of Type 3 WTG in the grid increases.
The VSC power converters in DFIG may present very small harmonic impedance compared to a current
sourced converter. However, due to its frequency dependence with converter control, they cannot be
accurately represented by a constant current source. A Norton/Thévenin equivalent, as shown in Figure
4-1, is more appropriate, including an accurate representation of the frequency dependence of its
impedance ([67], [68]) which is determined by the converter’s control and its physical components. In
addition, the high-frequency harmonic filters typically used in DFIGs should also be considered as they
can influence resonances in the power system. These filters are often installed on the line side of the
VSC to shunt energy from the switching frequency. Even though the rating of each individual filter may
be very low, the cumulative effect of filters from all DFIG in a large wind farm can shift the natural
resonance of the system [67]. The harmonic sources in Thévenin or Norton equivalent modelling are
the aggregated sources of both stator and rotor, which shall cover all significant harmonic range,
including switching-related harmonics in a few kHz.
Some additional aspects related to modelling of DFIGs are briefly highlighted next:
 The harmonics produced by the rotor-side converter, as seen at the stator and the machine
terminals, will be shifted in frequency by the slip speed of the rotor. This produces, in general, non-
integer harmonics. Most modelling practices, however, ignore this effect due to its insignificance
level on the grid side.
 DFIG converters can be considered as two harmonic voltage sources, connected to the rotor and
stator, respectively [69]. The switching strategy of each converter determines the type of harmonics
produced, with the harmonics produced by the rotor converter being a multiple of slip frequency.
 The work in [69] has shown that the harmonic emissions of the DFIG are strongly affected by the
wind speed – i.e. at a lower wind speed, the wind turbine is operating at a lower mechanical torque,
thereby producing harmonics of different magnitude and frequency to those at a higher wind speed.
 The use of the standard wound-rotor induction machine model for harmonic analysis of DFIGs is
only valid if the stator and rotor are being fed by sources having the same frequency [70];
108
Key Points:
 An “ideal” current source representation is not adequate to capture the interactions between Type 3
WTGs and the harmonic network impedance.
 The use of a Norton/Thévenin equivalent model is recommended (Figure 4-1). This model must
capture the input impedance of the device comprising a “passive” part (determined by the grid-side
filter circuit elements) and an “active” part (determined by the internal control loops).
 Harmonic filters associated to the converters need to be captured in the model.
 The wind speed (and therefore torque) should be considered as it determines the harmonics
produced by the DFIG [69].
4.2.1.2 Type 4 Wind Turbine Generation – Full Power Converter Generation

Type 4 wind turbines are equipped with a full-scale power converter as seen in Figure 4-3. A full-scale
power converter is placed between a WTG and the power grid to transfer power from the generation
side to the grid with the required power quality. Typically, a back-to-back power converter with DC
capacitors in between is employed, which allows dynamic decoupling from the generation side to the
grid side. It offers full dynamic control capability on the grid side in both normal- and fault ride through
operation.
In an IGBT-based VSC, pulse-width modulation (PWM) is usually employed to regulate the voltage
output of the power converter. Relatively high switching frequencies (up to a few kHz), are often used
to achieve fast control dynamics. Sideband harmonics of the switching frequency are therefore a
consequence, located within the kHz range and centred around the switching frequency and multiples
of the switching frequency, which could easily be beyond the 50th order. Low-order characteristic and
non-characteristic harmonics are also often present with full-scale power converters, due to cross-
modulation between AC and DC, dead time of PWM, nonlinearity of IGBT, and imperfection of balancing
among converter modules.
Generally speaking, the level of low-order harmonics is insignificant. However, the amplification of
certain harmonics can be significant due to resonances within the local network. Passive filters are often
installed at the grid converter terminals (due to low cost and easy implementation) to suppress unwanted
harmonics. At the time of writing, increasing attention is being given to applications of active filtering to
cope with system dynamics and uncertainty of design, either by utilising the existing power converter
[71] or by adding a STATCOM [72].
The generation of interharmonics can be significant if the switching frequency of the power converter is
not properly selected to synchronise with the fundamental frequency. Sideband harmonics are usually
filtered internally by the WTG shunt filters to improve the power quality at the connection point.
Figure 4-3 Example Type 4 Wind Turbine Generator
4.2.1.2.1 Type 4 WTG Modelling

Similarly to type 3, the modelling of type 4 wind turbines for harmonic studies can be challenging
because of manufacturer-dependent power converter topologies and control strategies. Proprietary
protection also prevents detailed design information for harmonic modelling purposes. Certain level
simplifications and assumptions need to be made for feasibility of modelling, often conducted by TSOs,
consultants etc.
It is recommended to represent the type 4 WTG harmonic model as Norton/Thévenin equivalent circuit
at the frequencies of interest, in alignment with IEC TR 61400-21-3 [62]. This generic harmonic model
109
(Norton/Thévenin equivalent circuit) is shown in Figure 4-1. The Norton/Thévenin equivalent circuit is
comprised of an equivalent ideal current/voltage source and equivalent impedance for each
harmonic/frequency order of interest. This provides a common structure for various turbine/converter
vendors, regardless of their specific power converter designs. These two equivalent models are
theoretically identical, and interchangeable in terms of harmonic sources and harmonic impedances.
Regardless of the model used, the harmonic sources and harmonic impedances should be provided at
the frequencies of interest for the intended application.
The harmonic emission profile of a type 4 wind turbine is typically a function of operation points. To
simplify modelling and reduce the amount of input data required, the worst-case operating point at each
harmonic order is often used to yield a conservative result. However, the proposed model structure
allows for the implementation of look-up tables to represent the harmonic behaviour at different
operating points. This approach can be used for probabilistic analysis or sensitivity studies of various
operating points.
The equivalent harmonic impedance is the total converter output impedance which should include the
impact of closed-loop controls. Any other passive components, such as the smoothing reactor, turbine
filters and transformers need to be captured either as part of the equivalent harmonic impedance or
explicitly as an external component.
Unfortunately, it is not easy (especially at lower frequencies) to measure the frequency-dependent
impedance of passive components, which could introduce modelling errors, especially for harmonics
having a low magnitude. In addition, multiple parallel-connected converters are often used in
applications of MW power density, and these converters may introduce unbalance which can affect
internal harmonic current flow between converter modules. Even if coupled sharing reactors are
designed to limit the current imbalance, some asymmetry in harmonic generation between converter
modules can be seen. The manufacturing tolerances of electrical components such as reactors,
capacitors and resistors, must be considered in the uncertainty/tolerance of the modelling.
4.2.1.2.1.1 Harmonic Sources

A voltage or current source can be used to represent the harmonic emission profile of a power converter.
Converter harmonics are commonly generated by hardware nonlinearity and imperfection of the
switching mechanism. Sideband harmonics of the switching frequency can be calculated based on an
analytical equation and considering the specific PWM strategy being used. However, for kHz-range
power converters, there is no analytical method commonly available for the derivation of low-order
harmonics caused by dead-time and non-linearity of switching devices. Detailed time-domain modelling
of the converters and/or measurements can be used by the vendors to provide the required harmonic
source characterisation to the user (system operator, wind farm developer, consultant, etc) within the
specified frequency range.
Another factor influencing the harmonic profile of the WTG is the operation point. If the worst-case
harmonic is insufficient for detailed harmonic study or filter design, operation-dependent harmonic
sources might be needed to provide for further information.
Phase angles of harmonics can be of interest when performing harmonic penetration studies with
external background harmonics sources present. The stochastic nature of relative phase angles among
WTGs is useful for harmonic summation at the PCC. If it is not obtainable, a general conservative study
is often applied by synchronising phase angles of like-frequencies at multiple WTGs [146]. However,
this could result in overly-conservative results due to the random nature of even-order harmonics and
multiples of triplen harmonics. It is therefore recommended to acquire more information on stochastic
phase angle behaviour at harmonic orders of interest. At the time of writing, at least one European TSO
is already requesting the prevailing phase angle as part of the modelling requirements defined in their
Connection Code3.
3
TenneT has included the requirement to provide prevailing phase angle information in the German
implementation of the European HVDC Connection Code (TAR 4131).
110
4.2.1.2.1.2 Harmonic Impedance

The Thévenin or Norton equivalent impedance can capture either (i) only the internal power converter
impedance; or (ii) additionally include passive components such as the main reactor impedance, passive
filters and turbine transformers. All other electrical components need to be modelled separately.
Because of widely applied controls of power converters, it is important to note that the impact of closed-
loop controls on the overall harmonic impedance should be considered. This is of relevance up to a few
hundred Hz which is typically the bandwidth of current controls. Sometimes, negative resistances can
be present at certain frequencies because of impact of power converter control, but this is not in any
way indicating any problems of WT operation. For frequencies above this, the harmonic impedance is
typically dominated by any passive components, mainly the smoothing reactor and passive filters. An
example comparison can be seen in Figure 4-4.
Figure 4-4 Example Comparison of WT Harmonic Impedance and Smoothing Reactor of Converter
4.2.1.3 Comparative Study of WTG Modelling Approaches

When modelling WTGs for harmonic studies, it is important that the behaviour of all components is
accurately captured in the model. This is illustrated next with an example of a real wind farm using type
4 WTGs.
In this example, three modelling approaches for WTGs are compared:
A. Complete harmonic model (Figure 4-5 A): The WTG is represented as a full Norton/Thévenin
equivalent. Zh represents the converter harmonic impedance, including passive filters. This is the
most accurate representation of the WTG.
B. Ideal current source (Figure 4-5 B): The harmonic impedance is neglected, and the WTG is
represented only as a current source. This means that the interactions between the WTG impedance
(consisting of the converter and the filter) and the grid cannot be captured with this model.
C. Current source with passive filters (Figure 4-5 C): The WTG is again represented as a
Norton/Thévenin equivalent, but only the filter is considered (i.e. the converter harmonic impedance
is neglected). This means that the interactions between the converter internal impedance and the
grid cannot be captured with this model.
111
Ih Zh Filter Ih Ih Filter
A) Complete harmonic model B) Ideal current source C) Current source with passive filters
Figure 4-5 WTG Modelling Approaches
Figure 4-6 shows a comparison of the three WTG modelling approaches in a frequency scan of the
entire wind farm as seen from the HV side of the wind farm transformer. The internal collector network,
WTG transformers and WTGs are individually represented. It should be noted that in this example the
wind farm is not connected to an external grid in order to highlight the differences observed at the
Connection Point without any masking effects from the external grid harmonic impedance. Approach A
(complete harmonic model) is the most accurate and is therefore defined as the reference case. Figure
4-6 shows that the simulation results produced with the other two modelling approaches fail to reflect
the correct interactions between the WTGs and the system under study (wind farm internal collector
network). In this example, the effect of the converter internal impedance is dominant in the low frequency
range, therefore modelling approaches (B) and (C) are not adequate in that range. It can be concluded
that approaches B (ideal constant current source) and C (constant current source with passive filters)
can yield incorrect results when studying the impact (and interactions) of this wind farm on a power
system.
Key Points:
 Converter harmonic impedance and filters should be included when modelling type 3 and type 4
WTGs. Therefore, the use of Thévenin/Norton equivalents are recommended instead of the
constant current source approach.
 The user should request detailed WTG models capturing these elements from vendors to improve
accuracy in harmonic studies involving wind farms.
112
10000 (A) Complete harmonic model

(B) Ideal current source
(C) Current source with passive filters
1000
Impedance [Ohm]
100
10
1
0 500 1000 1500 2000 2500 3000
Frequency [Hz]
150 (A) Complete harmonic model

(B) Ideal current source
(C) Current source with passive filters
100
Angle [deg]
50
0
0 500 1000 1500 2000 2500 3000
-50
-100
Frequency [Hz]
Figure 4-6 Frequency Scan of a Wind Farm with different WTG Harmonic Models
4.2.2 Converter-based PV Generation

As the number of PV systems increases there is a need to evaluate their impact on the grid, both
considering aggregation of several smaller PV installations in distribution and low voltage networks, as
well as large PV parks connected to higher voltage levels. Most of the literature on the subject focuses
on the former [73]-[78]. To perform harmonic studies, adequate models of grid-connected PV systems
are required. Figure 4-7 shows an example of a typical PV configuration.
Figure 4-7 Typical PV System Overview
113
4.2.2.1 Modelling PV inverters for harmonic studies

Harmonic models for PV inverters require a lot of information from the manufactures, which is usually
kept confidential. Due to this lack of knowledge, suitable, manufacturer-specific harmonic models
capable to be used for studies of harmonic emission, harmonic instabilities or harmonic resonance are
not widely available. This includes also aggregated harmonic models, e.g. for representing a PV plant
consisting of multiple PV-inverters. Measurement based models seem to be a possible approach to
improve the model accuracy compared to the simple models based on constant current sources,
however the superposition of the different impact factors, like supply voltage distortion, network
impedance, magnitude of supply voltage and output power of the PV-inverter is still not validated [79].
PV systems are often modelled as an ideal constant current source for harmonic studies, however this
approach does not capture the interaction between the output filters and the grid impedance or the
influence of background distortion on the current emissions [75], [78]. These factors can have a
dominant effect in certain frequency ranges and must be included when modelling PV for harmonic
studies (see Section 4.2.1.3 for an illustrative example). The use of constant current source models
should be discouraged.
For realistic harmonic studies it is recommended to model PV inverters as a Norton or Thévenin
equivalent capturing the dependency of harmonic currents of PV inverters on supply voltage distortion
and network impedance as well as the input impedance characteristic, which can cause (or alter)
resonances in the networks [79]. This modelling approach is the same as for type 4 WTGs ([75], [78]),
as shown in Figure 4-1.
The values of the harmonic impedances can be obtained in a number of ways, e.g. taken from
manufacturer data or via the use of measurements [73], [75], [78]. In [77], measurements of voltage and
currents are used to iteratively calculate the impedance value for each frequency. In [78], an alternative
approach is suggested where the output impedance is represented by physical elements (capacitor,
inductor, resistor etc.). Measurements are used to determine the parameter values and topology. It
should be noted that although measurements are critical to evaluate existing and future models,
measured harmonics will in turn be impacted by the grid characteristics.
As with type 4 WTGs, the control strategy and inverter topology will also have an impact on the resulting
harmonic emission and impedance, and therefore a range of different operating points can be
considered when applying a harmonic model for certain type of studies such as statistical assessments
or sensitivity investigations, as opposed to a single value worst case approach with the highest emission
and impedance values. It should be noted that the worst-case distortion for different harmonic orders
may appear at different operating points; therefore, the worst-case operating point at each harmonic
order will lead to conservative results.
In case of centrally located PV inverters (e.g. in large PV plants), particularly in case of similar inverter
models, CIGRE TB 672 [79] recommends aggregating harmonic currents by adding them arithmetically,
independent of the harmonic order.
4.2.3 Other Types of Converter-based Generation

Other types of converter-based generation include e.g. wave or tidal power. There is not much written
concerning the harmonic performance of such systems at the time of writing this TB, but in general the
same modelling requirements as are described for type 4 WTG or PV should apply here as well.
4.2.4 Summary
Power converters are a source of harmonic distortion. The converter also acts as frequency-dependent
impedance which can impact the resonance frequencies seen in the network. To be able to fully assess
the impact of converter-based generation on the harmonic performance of the grid it is essential that the
models capture both the harmonic emission and the impedance characteristics of the power converter.
It is recommended to model converter-based generation as a Norton or Thévenin equivalent circuit at
the frequencies of interest (typically 2 nd to 50th harmonic order), in accordance with IEC 61400-21-3 [62].
The equivalent circuit consists of an equivalent voltage/current source and a harmonic impedance. The
harmonic impedance must capture the power converter impedance characteristics; it may also include
passive components such as the main reactor impedance, passive filters and turbine transformers,
which will otherwise have to be modelled separately.
114
The emission and impedance characteristics depend on the operating point of the power converter,
where the worst-case operating point may differ for different harmonic orders. However, for simplicity,
the worst-case for each harmonic order is often used to yield conservative results.
The above detailed harmonic model needs to be provided by the manufacturer (or designated party) to
the user (system operator, wind farm developer, consultant, etc) to facilitate harmonic studies associated
with the connection of the converter-based generation devices to the grid. Some system operators are
already including the provision of these models (and associated documentation) as a pre-connection
requirement within the connection offer process.
The reader is reminded that the above considerations apply to the modelling of converter-based
generation on an individual device level. Aggregation of multiple components in a wind farm / PV farm
is covered in chapter 5, sections 5.6 and 5.7.
4.3 HVDC Converters

An HVDC converter station can be one of the most complex and difficult elements to model for harmonic
studies, for reasons which will become clear below.
Harmonic modelling for HVDC can be envisaged in two main categories:
1) Highly detailed modelling for the design studies of the HVDC station, including filter design and
calculation of harmonic performance and rating. Such studies are normally undertaken by the
HVDC suppliers using specialized software. The external AC network, as seen from the connection
point, is generally represented as a Z-plane equivalent envelope in such studies (see section 5.4).
2) Inclusion of an HVDC station in a broader network harmonic study, typically undertaken by a utility
or consultant. In this case some level of simplified representation may be desirable, depending on
the objectives of the study. It is this category which is the focus of the material presented in this
Technical Brochure.
The following sections first give a brief description of the main elements and behaviour of an HVDC
converter station, since they are relevant to understanding of its harmonic characteristics. Subsequently,
the important aspects to be considered in modelling are highlighted.
4.3.1 HVDC - LCC

The harmonic characteristics, behaviour and modelling of HVDC line-commutated converters (LCC) are
all very well-known and documented. CIGRE Technical Brochures 139 [80] and 553 [23] are devoted to
the subject, and these have been converted to IEC TR 61000 Parts 1-4. Numerous textbooks and
papers also cover the subject. To go into detail on any aspect of the harmonic modelling would take
excessive space in the present Technical Brochure. Therefore, the discussion below simply aims to
describe the concepts in a general way, suitable for a non-specialist, with the intention of highlighting
the key areas which can then be followed up in detail elsewhere.
The core technology of HVDC LCC has become fairly standardised. Conversion is by means of thyristor
valves in a three-phase Graetz bridge configuration. Twelve-pulse operation is achieved by connecting
two six-pulse converters through Y-Y and Y-D connected transformers to the AC supply network. These
converters are series connected on the DC side to achieve the desired high direct voltage. Control of
the power transmission is achieved by controlling the delay angle of the thyristor firing. The voltage is
normally controlled at the receiving (inverter) station and current at the sending (rectifier) station.
4.3.1.1 Reactive Power

The current through the HVDC LCC converter lags the supply voltage due to the thyristor firing angle
delay and the overlap angle as the direct current commutates from one phase to the next. This phase
lag results in a high consumption of reactive power by the converter – typically in the order of 50-60%
of the active power when operating at nominal design conditions.
As it is not feasible to supply such amounts of reactive power from the AC system, local compensation
must be provided for most of this requirement, in the form of AC filters, capacitor banks, STATCOMs or
synchronous condensers. Because the reactive power demand of the converter varies with load level,
the individual switchable elements (AC filters or capacitors) must be limited in size such that they can
be switched sequentially to follow the reactive power demands of the converter while restricting the
115
reactive interchange with the AC network within specified limit bands (see Figure 4-8). Hysteresis is
normally introduced in the switching sequence to avoid excessive switching during minor variations in
converter power.
0,8
Q Converter
0,6
Q Filters
Reactive Power (pu)
0,4 (ramp down)

Q Filters
0,2 (ramp up)
Net Q (ramp
0 down)
Net Q (ramp
-0,2 up)
AC Q Limits
-0,4
0,00 0,20 0,40 0,60 0,80 1,00 1,20
Active Power (pu)
Figure 4-8 AC Bank Switching Steps following HVDC LCC Converter Q Curve
4.3.1.2 Harmonic Generation

The AC current through an HVDC LCC is a 12-step approximation to a sinusoid, the steps being more
or less smoothed depending on the commutation period between phases. This current is therefore rich
in harmonic content. Under idealized conditions it consists of “characteristic” harmonics of order 11, 13,
23, 25 …that is 12n1, with magnitudes diminishing with increasing frequency.
Harmonics of order 12n-1 are of negative sequence and those of order 12n+1 are of positive sequence.
No zero-sequence harmonics are produced by a HVDC LCC.
Non-idealities in either the converter (small unbalances between transformer phase reactances, slight
differences in valve firing instants, ripple in the direct current) or in the supply voltage (negative
sequence voltage, pre-existing harmonics) will result in the generation of “non-characteristic” harmonics
of all orders. These non-characteristic harmonics are generally of relatively low magnitudes compared
to characteristic harmonics [81].
Reference [82] discusses in detail the non-characteristic harmonics arising from different asymmetries
found in practical converters. The most significant non-characteristic harmonic is usually the 3rd, which
is caused mainly by unbalance (negative sequence) in the AC network fundamental frequency voltage
and appears as positive phase sequence.
The magnitude of the generated characteristic harmonics varies with transmitted power, increasing non-
linearly. The higher order harmonics increase with a superimposed cyclical variation, and so the
percentage content of each harmonic in the total current is different at each operating point of the
converter. The magnitude and variation with load is dependent on the factors which affect the overlap
angle, namely the commutating reactance (which is mainly the transformer leakage reactance) and the
control angle. (For HVDC LCC the control angle is normally restrained in steady-state to a narrow range,
typically around 12.5°-17.5° for rectifiers and 16°-18° for inverters, although these ranges may vary for
specific applications or different control system designs). Figure 4-9 shows the variation of the
characteristic harmonics with DC power transfer for illustrative circuit data (U ac= 1 pu, reactance= 0.145
pu, control angle= 17°).
116
0,05
AC Harmonic Current (pu*)

0,04
I11
0,03
I13
I23
0,02
I25
0,01
I35
I37
0
0,00 0,20 0,40 0,60 0,80 1,00
DC Power (pu)
* pu of fundamental current for nominal DC power
Figure 4-9 Illustrative Variation of Characteristic Harmonics Magnitude vs Load (HVDC LCC)
Key Point:
 For initial basic analysis, the converter is often regarded as a harmonic current source, that is, the
harmonic current emission does not vary with the AC side impedance. While this simplification gives
approximately correct results at higher orders, it can be unsatisfactory for low orders, particularly for
the 3rd harmonic. Therefore, a more complex analysis is necessary for such low order harmonics,
as discussed further below under “cross-modulation”.
4.3.1.3 Harmonic Filters

The converter harmonic current is of such magnitude that if allowed to flow into the AC network it would
result in unacceptable distortion levels. Filtering is therefore required and is normally achieved by
adapting some or all of the required shunt capacitive reactive compensation to become tuned harmonic
filter banks. For guidelines on modelling these harmonic filters the reader is referred to section 3.6.3.
The filter units are usually arranged hierarchically as branches, sub-banks and banks, as shown in
Figure 4-10. The sub-banks are switched such that both adequate reactive compensation and adequate
filtering is always connected at each point or range of the HVDC converter operation.
The individual filter branches may be of different types and tuning. The main characteristic harmonics
for a 12-pulse converter are dealt with by 11th and 13th harmonic filtering, which is often in the form of
double-tuned 11th/13th filters, or a high-pass broad-band filter tuned to the 12th harmonic to cover both.
Higher order harmonics are often mitigated by a high-pass 24th filter, or by different double-tuned
branches. Sometimes, a 3rd harmonic filter is also required. In a very few HVDC projects, additional
series filters between the shunt filters and the Point of Connection have also been used to impede the
flow of a range of harmonic currents into the AC system.
Tuned shunt filters present a low impedance at and around their tuned frequency, but parallel filters
tuned to different frequencies will create an antiresonance at an intermediate frequency, with possibly
very high impedance (see Figure 4-11). The resulting impedance profile across the complete harmonic
spectrum of interest must therefore be considered, with damping selected such as to optimize the overall
harmonic performance.
The filters in an HVDC LCC station are usually designed considering the worst-case performance, which
is normally when they are at their maximum point of detuning. Component manufacturing tolerances,
partial capacitor failures, and capacitor variation with temperature are causes of filter detuning, which
117
must be considered together with the effect of network frequency variations for an accurate assessment
of the HVDC LCC installation (see example in next page). Variation of resistance with temperature will
affect the filter damping.
It should be noted that the above considerations are normally only captured in the detailed design and
rating studies performed by the manufacturer or specialist designer of the HVDC LCC installation.
System-wide harmonic studies are not normally concerned with these minor parameter variations and
filters are typically represented with their nominal values. However, the user (system operator or
consultant) may deem prudent to include the filter detuning aspects when performing studies in an area
close to a HVDC LCC installation, especially if resonances are expected near the filter tuned
frequencies.
SUB-BANK SUB-BANK
BRANCH BRANCH
BANK SUB-BANK
Figure 4-10 Filter Bank, Sub-Bank and Branch Definition
Key Point:
 Reactive Power compensation devices (especially switchable filters and/or capacitors) need to be
captured in the HVDC LCC converter model as they will interact with the harmonic impedance of
the grid. A range of operating points, and associated status of filters/cap banks, may need to be
analysed to ensure that they do not introduce excessive resonances or harmonic distortion in the
grid.
118
Example: AC filters impedance and detuning

An example 1000 MW monopolar HVDC LCC converter station requires a doubled-tuned filter for the
11th and 24th and a triple-tuned filter for the 13th, 36th, and 48th in order to operate at minimum DC
power. Figure 4-11 shows the impedance obtained with these filters connected to the system.
10000
1000
|Z| Ω)
100
DT (11th, 24th) + TT
10 (13th, 36th, 48th)
1
0 10 20 30 40 50
Harmonic Order
Figure 4-11 Harmonic Filter Impedance - Minimum DC Power Configuration
For this particular system, the connection of an additional triple-tuned filter for the 3rd, 5th and 11th is
required in order to meet harmonic performance at maximum DC transfer. Figure 4-12 shows the
impedance of this filter.
10000
1000
|Z| Ω)
100 TT Filter (3rd, 5th,

11th)
10
1
0 10 20 30 40 50
Harmonic Order
Figure 4-12 Harmonic Filter Impedance - Triple-Tuned Filter
Figure 4-13 shows the resulting impedance when having all the filters required for maximum DC
transfer connected to the system. This plot also shows the impedance change for +3% and -1.5%
filter detuning, which were the maximum values taken into consideration to study the harmonic
performance of this sample HVDC system during the design and rating phases.
119
Figure 4-13 Harmonic Filter Impedance - Maximum DC Power Configuration
4.3.1.4 Cross-Modulation
The assumption that an HVDC LCC converter may be treated as a pure harmonic current source is a
convenient one for practical analysis and is frequently used. For practical design purposes, the errors
introduced by this assumption may be typically around 10% at the 11 th and 13th harmonics and at higher
harmonics much lower still.
The treatment of the converter as a pure current source rests on the assumption that the direct current
is perfectly smooth. In practice however, it always has some “ripple”, that is, a harmonic content. The
mechanism by which the AC and DC side harmonics are related is known as “cross-modulation” and is
discussed in CIGRE Technical Brochures 143 [83] and 553 [23] and other numerous publications.
Analysis of cross-modulation effects is a complex matter, as there exists a circular interaction between
the harmonic voltages, currents and impedances on the AC and DC sides of the converter. In addition,
the transfer function of the converter varies with the operating point and load level. There are methods
in the frequency domain which provide approximate results, and the harmonic domain approach (section
2.5) aims to achieve a more accurate solution. However, to fully account for all the factors involved, and
to take into account the effect of the HVDC LCC converter control, a time domain solution (section 2.3)
using a full representation of the converter and its control system is often the most appropriate approach.
Ignoring cross-modulation effects will have the greatest impact at the low order non-characteristic
harmonic interaction resulting from AC system voltage unbalance, that is, the presence of a small
negative sequence component in the supply voltage. This results in generation of DC side 2 nd harmonic
voltage and consequently current, which in turn is cross-modulated back to the AC side as 3rd harmonic
current of positive sequence (and negative sequence fundamental). The magnitude of this 3 rd harmonic
current can be highly significant, with major implications for whether specific 3 rd harmonic filtering is
required.
When considering cross-modulation, it is no longer valid to assume that the highest voltage distortion
will occur at the highest value of total AC side harmonic impedance (filter plus network). As the cross-
modulation analysis has to consider the complete system of AC and DC side impedances, it is not
possible to make a general statement about what AC side impedance will produce the highest distortion.
This is particularly important for the 3rd harmonic, where the need for filtering may depend on such
assumptions.
120
4.3.1.5 Converter Harmonic Impedance

An HVDC LCC presents a harmonic impedance which is a function of its physical characteristics
(converter transformer reactance, valves, smoothing reactors, leakage path to ground, etc.), DC side
harmonic impedance, operating point, and control system action. This harmonic impedance will be seen,
together with the filter impedance, by pre-existing harmonic voltages on the AC supply network, and
these together will determine any possible magnification of such harmonics. In practice, the harmonic
impedance of the HVDC LCC installation is normally dominated by the harmonic filters at most harmonic
orders.
4.3.1.6 Modelling Recommendations

Drawing together the implications of the above paragraphs, some important principles for harmonic
modelling may be stated:
(1) Depending on the study objective, it may be necessary to model the HVDC LCC converter station
as a complete entity including the converters, the associated harmonic filters and other shunt
compensation.
(2) The station will act as both a source of harmonic currents and as a harmonic impedance, with the
filters dominating the harmonic impedance at most harmonic orders. The harmonic impedance of
the converter itself may however be significant, particularly at low harmonic orders.
(3) The harmonic current from the converter itself may be calculated quite precisely at any given
operation point, given full knowledge of all relevant parameters. Alternatively, it may be calculated
as a worst-possible value over a range, considering worst-case values of certain parameters.
(4) A large proportion of the converter harmonic current will be absorbed by the filters, rather than
passing into the AC network. However, the degree of this mitigation will depend on both the
detuning of the filters and the actual harmonic impedance of the AC network at that time. Except
for a very special analysis of some specific condition, it is common to make worst-case
assumptions regarding filter detuning and AC network impedance values.
(5) The different possible configurations of the HVDC station pose a complicating factor. A typical
bipolar HVDC LCC transmission will consist of two poles, identical in terms of AC side harmonic
behaviour. Either one may be in operation singly (monopolar operation) or both together in bipolar
operation. The harmonic generation will clearly be different in the two cases, as will the
configurations of AC filters required to be connected.
(6) As the DC power is varied through the operating range, the generation of harmonic currents will
vary, and the configuration of types and sizes of AC filters connected will change.
(7) Because the harmonic generation of a converter is disproportionally high in relation to
fundamental frequency at low DC power, and because reactive balance considerations may limit
the filters which can be connected at such DC power levels, the worst case for harmonic distortion
performance may in fact occur at low DC power transmission levels rather than at maximum DC
power.
(8) Considering the above factors, it becomes clear that both the harmonic current flow from the
HVDC station into the AC system, and the harmonic impedance of the station as seen externally,
will vary over very wide ranges in the course of operation. It is essential that this is fully understood
when modelling, and all such variations considered in a suitable way.
The magnitude and variation with load of converter harmonic generation, and the detailed design of the
filters, are matters which cannot be predicted at the planning stage of an HVDC LCC project. They will
only become clear as part of the integrated design of the station, and will depend on such factors as the
choice of converter transformer reactance, the permitted harmonic distortion levels, and the AC network
harmonic impedance. The design will also depend on economic factors, physical layout constraints and
outage requirements. All this implies that it is especially difficult to incorporate a future HVDC LCC
station in a network harmonic model at the planning stage. Very general assumptions regarding
harmonic generation and filtering have to be made.
The following two sections discuss modelling aspects from the point of view of an Owner, TSO,
consultant or other third-party who may not have access to all the detailed design information available
to HVDC suppliers. These users may need to model an HVDC LCC system for a harmonic study or may
simply wish to acquire a basic understanding of detailed design studies performed by technology
providers.
121
4.3.1.6.1 Modelling HVDC LCC as a new Connection for Compliance Assessment

The assessment of compliance with emission limits normally has two components: (i) assessment of
the new harmonic distortion caused solely by the current/voltage harmonics generated by the new
installation and (ii) assessment of amplification of pre-existing background distortion caused by the
interaction between the harmonic impedance of the new installation and the impedance of the grid. Both
components can be analysed independently based on superposition. Modelling recommendations for
each type of assessment are described next.
 New distortion caused by harmonic currents generated by the HVDC LCC
Figure 4-14 shows a typical simplified active model of a HVDC LCC installation as a current source.
This representation is normally adopted to assess the harmonic distortion caused at the point of common
coupling (PCC) by the converter emissions.
The converter can be represented as an ideal current source. The filters should be represented in detail
(as described in section 3.6.3). The impedances of the filters and the AC network are selected from
within their feasible ranges to give the highest resulting impedance and consequently the highest
harmonic distortion at the PCC. In a frequency domain study, this model is applied at each separate
harmonic frequency.
As noted in 4.3.1.2 different harmonics will be inherently of positive or negative sequence, but none of
zero sequence. Normally it is valid to assume that the positive and negative sequence impedances of
the circuits used in this and the subsequent sub-section are equal. A single-phase equivalent model is
normally used.
PCC
I LCC_n Z AC Z AC
Filters System
Figure 4-14 Simplified Equivalent Circuit for Harmonic Emission Assessment at PCC (HVDC LCC)
Normally it is necessary to study different HVDC LCC configurations and output levels. For each of these
possible scenarios, the following considerations must be made:
 Converter Harmonic Source (ILCC_n): Typically, the complete HVDC LCC converter, including
converter transformers, is treated as an ideal current source. The level of detail that can be attained
when constructing a model is limited by the data provided by the HVDC LCC supplier. In general,
the technology providers are required to make available, as a minimum, the highest levels for each
individual harmonic over the full range of operation – known as a “non-consistent” set of harmonics.
This may be a conservative approach since the worst levels of each harmonic do not always occur
at the same operating level or configuration, so the total will give an unrealistically high THD.
Additionally, the value of all individual harmonics at a defined set of operating conditions may be
provided (e.g. at 10% steps of transmitted power.) If such more detailed data is available, it must
be decided whether it is required to model the specific harmonic levels for each desired scenario or
if the assumption of coincident worst levels is sufficient. For example, each harmonic source can be
modelled with a look-up table to produce the right output based on the DC transfer level. If no data
is available, typical idealized characteristic harmonics can be calculated for the desired operating
122
conditions using well-known equations for the converter, as provided in many references (e.g. Part
1, Annex B of [84]).
 AC Filter Impedance (ZAC Filters): This impedance must be adjusted depending on the number of
filters in service for a given HVDC LCC configuration and DC transfer level. The different impedance
values are usually available in the Harmonic Performance Study report provided by the supplier.
The calculations carried out to determine these values typically already account for filter detuning
and frequency variations, but it must be verified if the worst-case conditions assumed in the
supplier’s study are still valid. Alternatively, the values of all filter components and the detuning
coefficients may be provided, which would enable an accurate model to be created.
The external system impedance (ZAC System) is normally provided by the system operator in the form of
harmonic impedance envelopes (see Section 5.4).
 Amplification of background harmonic distortion caused by the HVDC LCC

Figure 4-15 shows a simplified passive model suitable to analyse the voltage distortion at the PCC due
to existing AC system harmonics, which may be amplified or damped by the HVDC LCC converter
station. On the AC side a voltage source is used to represent the pre-existing harmonics to be studied.
PCC
Z Z
Transformer AC System
Z Ufn
LCC Z AC
Conv. Filters
Figure 4-15 Simplified Circuit to Assess Magnification (or Damping) of Background Voltage Distortion at
PCC (HVDC LCC)
For most studies, it may suffice to model the AC filter impedance (ZAC Filters) under the assumption that
the impedances of the converter and the transformer are much higher than those of the filters and can
be assumed to be an open–circuit in the model. Typically, this simplified approach might be used to
calculate the harmonic voltage distortion at the PCC due to background harmonics, and the component
rating stresses due to background harmonics.
This simplified approach, ignoring the converter harmonic impedance, would provide a general
assessment of the converter station as a harmonic load, but the results may not be very accurate when
considering low order harmonics. More precise calculations may be desirable, not only for normal design
or compliance studies but also for analysis of particular behaviour in the low-order harmonic range, and
can be achieved through detailed consideration of the internal impedance of the converter, which
consists of two elements:
 Converter Transformer (ZTransformer): Information on the transformer impedance should be available
in the design documentation. Tap-changer range should be included. See section 3.3 for guidelines
on transformer modelling.
 Converter impedance (ZLCC Conv): This includes the impedances of the DC line, DC filters, converters
valves, smoothing reactor, etc., transferred to the AC side by the cross-modulation function of the
conversion process. For cases in which the DC side can be considered decoupled, it may be
sufficient to model the time average commutating impedance per bridge considering overlap. If the
123
converter impedances for different conditions are not provided by the supplier, it is possible to
calculate the required impedance knowing certain parameters such as the commutating reactance
per bridge and the overlap angle [85], [86].
harmonic impedance envelopes (see Section 5.4). The pre-existing background distortion (Ufn) is also
provided by the system operator and it is normally obtained from measurement campaigns or
simulations (see Section 5.8).
4.3.1.6.2 Modelling HVDC LCC in a System-Wide Harmonic Study

Figure 4-16 shows a complete model of a HVDC LCC installation which can be included in a larger
system model suitable for general harmonic studies.
The representation of the individual components, ILCC_n, ZLCC Conv, ZTransformer and ZAC Filters is as described
in the previous sub-section.
For system-wide studies, the current source and status of the filters (and any other shunt compensation
devices) should match the operating point in order to have a realistic representation of the emissions
and impedance interactions with the AC system. Depending on the scope of the study a range of
operating conditions will need to be assessed and the reader is reminded that the worst case may occur
at low DC power transmission levels rather than at maximum DC power.
PCC
Z
Transformer
Z
I LCC_n LCC Z AC
Conv. Filters
Figure 4-16 HVDC LCC Model for System-Wide Studies
4.3.2 HVDC – VSC

The subject of HVDC VSC harmonics is the task of a parallel CIGRE Working Group B4.67 and the
resulting TB754 [5] provides a comprehensive treatment of the topic, including modelling aspects. The
information below is therefore restricted to serve as an introduction and a pointer to specialized
references.
VSCs use electronic switches, typically insulated-gate bipolar transistors (IGBT), which can be turned
off independently of the polarity of the applied voltage. Early designs were based on a two-level
converter capable of producing a square wave, but switching techniques such as Pulse Width
Modulation were used in order to improve the harmonic distortion of the converter. This however resulted
in high losses since multiple switching events were required within one cycle.
Modular Multi-Level Converters (MMC) are becoming the most common topology of VSC with different
proprietary solutions available depending on the supplier. The basic configuration consists of several
VSC modules, which are formed by IGBTs and their antiparallel diodes connected in series across a
storage capacitor. The converter is then connected to the AC system through a phase reactor and
usually a transformer. Each VSC module is individually controlled to produce a very close approximation
to a pure sinusoidal voltage at the AC terminals with the desired phase angle in relation to the network
124
voltage. This configuration permits using lower switching frequencies compared to PWM, which also
leads to lower losses.
Figure 4-17 shows the typical configuration of an MMC VSC converter, as well as the two main types of
modules used. The half-bridge configuration has the lowest number of components (which also means
lower cost and losses), but its output voltage only has one possible polarity. The full-bridge configuration
allows an output voltage with either polarity, which makes it possible to block the current from DC faults
and provides more flexibility at the expense of using more switching devices.
Figure 4-17 a) MMC VSC Configuration, b) Half-Bridge Module (c) Full-Bridge Module
4.3.2.1 Harmonic Generation

The AC side harmonic generation of a VSC depends greatly on the valve configuration and the switching
techniques used, which vary among suppliers and with successive generations of designs. In general
terms, HVDC VSC generated harmonics can be characterized as follows.
 The converter appears essentially as a harmonic voltage source, rather than as a current source.
 The fast switching techniques used combined with multi-level technology extend the range of
significant harmonic generation in the high frequency range, above the limit of the 50 th harmonic
typically considered for HVDC LCC. Harmonics are dependent on the modulation index but
independent of load current.
 The resulting harmonics are of much lower magnitude than for HVDC LCC, such that AC filters
are either very small and permanently connected, or not required at all.
 Interharmonic frequencies can be of comparable magnitude to integer harmonics due to switching
strategies not always being periodic within one cycle of the fundamental frequency.
 If necessary, partial suppression of lower order harmonic generation can be achieved by
introducing such functionality in the converter switching control.
125
4.3.2.2 Cross-Modulation
Harmonic interaction between the AC and DC sides of an HVDC VSC occurs but is generally less
significant and more complex than for HVDC LCC. The existence of capacitive energy storage within
the converter tends to partially decouple the two sides of the converter in terms of harmonic interaction.
4.3.2.3 Converter Harmonic Impedance

The total harmonic impedance of the HVDC VSC converter consists of the series combination of the
passive components of the converter (phase reactor and transformer) and the active internal impedance,
which is dependent on the control characteristics. The active internal impedance may vary between
inductive and capacitive appearance with frequency, and may possibly display a negative resistive
characteristic at some frequencies. In the absence of AC filters, the overall converter impedance
determines whether amplification of pre-existing network harmonics will occur.
In order to permit modelling of the active internal converter harmonic impedance, the HVDC VSC
supplier must provide the calculated impedance characteristic or a digital model of the converter which
is adequate to represent its harmonic impedance.
4.3.2.4 Modelling Recommendations

Summarizing the above, it is clear that in many ways, modelling a HVDC VSC station in the context of
a network study should be much simpler than with HVDC LCC. Assuming that any filter is small and
broad-band, then the issues identified above for an LCC station are not applicable (filter switching, filter
detuning, separate calculations for different power levels and filter configurations, etc.).
As the harmonic generation is typically very low, it is normally sufficient to calculate a non-consistent set
of maximum harmonic voltages over the complete range of operation of the converter, and use this as
the modelled harmonic generation, represented as a voltage source behind the converter internal
harmonic impedance.
The correct modelling of the converter harmonic impedance is therefore highly important, both in
modelling the harmonic generation and also in representing the converter as a harmonic impedance as
seen from the network. The modelling must include the passive components (power transformer and
any filtering) as well as the internal impedance of the converter as discussed above.
The next two sections provide some guidelines to model HVDC VSC systems in a harmonic study for
users who may lack detailed HVDC VSC modelling parameters (e.g. TSO, utility, consultant or other
third-party).
4.3.2.4.1 Modelling HVDC VSC as a new Connection for Compliance Assessment

The assessment of compliance with emission limits normally has two components: (i) assessment of
the new harmonic distortion caused solely by the current/voltage harmonics generated by the new
installation and (ii) assessment of amplification of pre-existing background distortion caused by the
interaction between the harmonic impedance of the new installation and the impedance of the grid. Both
 New distortion caused by harmonics generated by the HVDC VSC
Figure 4-18 shows the typically used equivalent circuit to carry out an assessment in the frequency
domain of the harmonic distortion at PCC caused by the converter emissions. Note that the AC filters
can be located either on the low voltage side of the transformer or at the point of common coupling, with
newer installations favouring the latter in order to improve filtering effectiveness at the PCC.
126
PCC
Z
Z
converter
Transformer
reactor
Z VSC
Conv.
Z AC Z AC
Filters System
U VSC_n
PCC
Z
Z
converter
Transformer
reactor
Z VSC
Conv.
Z AC Z AC
Filters System
U VSC_n
Figure 4-18 Simplified Equivalent Circuit for Harmonic Emission Assessment at PCC (HVDC VSC)
The main aspects to be considered for each element of the circuit are discussed below.
 Voltage harmonics source (UVSC_n): HVDC VSC-generated harmonics are in general of very low
magnitudes. With VSC technology we do not have the option to calculate characteristic harmonics
using well-known equations. The representation of the voltage source will solely depend on the
available data provided by the supplier, since the generated harmonics are unique to their
proprietary solution. The supplier can provide the harmonic voltages produced for each operating
point or, at the least, the maximum individual harmonic voltages considering a range of operating
points. Once again, it is important to consider that the maximum distortion for each frequency does
not occur at the same VSC operating point, so modelling the voltage source based on these values
will lead to conservative results and will not match measurements.
 Converter impedance (ZVSC Conv): This impedance represents the VSC module elements, such as
the valves and the capacitors and the influence of the control system. In order to model this
impedance, the supplier must provide values for different operating points.
 Converter reactor impedance (Zconverter reactor): Values for the reactor impedance are readily available
in the design documentation.
 AC filters (ZAC Filters): AC filters are not required in many HVDC VSC installations using the latest
technologies. Where a filter is used, it normally consists of a single high-pass branch which remains
connected at all power transfer levels. Its impedance data may be readily obtained. Section 3.6.3
provides guidelines for the modelling of harmonic filters.
127
 Transformer impedance (ZTransformer): It may suffice to use the leakage reactance of the converter
transformer considering its tap-changer range. However, depending on the study objectives it may
also be necessary to include imbalances between phases and converter bridges. At high
frequencies (above 50th harmonic order), the transformer stray capacitances may become
significant and should be modelled. See section 3.3 for detailed guidelines on transformer modelling.
harmonic impedance envelopes (see Section 5.4). As mentioned above, HVDC VSC systems may
generate harmonics of very high order. If the effects of these harmonic flowing into the AC network are
of concern, then the AC system harmonic impedance beyond the 50 th harmonic must be taken into
consideration.
 Amplification of background harmonic distortion caused by the HVDC VSC
Figure 4-19 shows an equivalent passive circuit diagram suitable to assess the modification of
background voltage distortion at PCC caused by the HVDC VSC harmonic impedance.
In this case, the source Ufn is used to represent the pre-existing AC harmonics to be studied. This is
provided by the system operator and it is normally obtained from measurement campaigns or
simulations (see Section 5.8).
The modelling aspects of the other circuit elements have been covered in the previous subsection.
PCC
Z
Z Z
converter
reactor
Z VSC Z AC Ufn
Conv. Filters
PCC
Z
Z Z
converter
reactor
Z VSC Z AC Ufn
Conv. Filters
PCC (VSC HVDC)
128
4.3.2.4.2 Modelling HVDC VSC in a System-Wide Harmonic Study

Figure 4-20 shows a complete model of a HVDC VSC installation which can be included in a larger
system model suitable for general harmonic studies.
The representation of the individual components, UVSC_n, ZVSC Conv, Zconverter reactor, ZTransformer and ZAC Filters
is as described in the previous sub-section.
PCC
Z
Z
converter
Transformer
reactor
Z VSC
Conv.
Z AC
Filters
U VSC_n
PCC
Z
Z
converter
Transformer
reactor
Z VSC
Conv.
Z AC
Filters
U VSC_n
Figure 4-20 HVDC VSC Model for System-Wide Studies
129
4.3.3 Summary
Figure 4-21 summarises the basic guidelines to follow when modelling HVDC systems for harmonic studies.
HVDC Modelling
Guidelines for Harmonic
Studies
Other study Other study Other study

HVDC System to be Assessing existing objectives that objectives that objectives that
added to the HVDC generated require modelling of require accurate require modelling of
network harmonic levels HVDC as harmonic assessment of AC/ HVDC as harmonic
source DC interactions impedance
Low order
harmonics and
interharmonics are
of concern?
Design Stage: Utility/Contractor

Feasibility Study: to take LCC Simplified VSC Simplified LCC Simplified VSC Simplified
Detailed study Model HVDC
Detailed measurements Modelling: Modelling: Modelling: filter Modelling: valve +
shall be Measurements Converter in
Harmonic Study over sufficiently Current Source + Voltage Source + impedance is phase reactor
undertaken by not possible/ detail (see TB
unlikely to be long period (TB Impedance (see S. Impedance (see S. dominant (see S. impedances are
technology accurate 553, JWG B4.67)
needed (high- 553, IEC 61000-4- 4.2.1.6.1) 4.2.2.4.1) 4.2.1.6.2) dominant (see S.
provider 4.2.2.4.2)
level estimations 7)
can be made)
Use Use
Utility to provide Option to Use digital
manufacturer manufacturer
harmonic limits calculate LCC model capable
data to model data to model
and network characteristic of representing
generated converter
impedance data harmonics system controls
harmonics impedance
Model imbalance:
AC phases, More accuracy
transformer and needed
bridge impedance
Figure 4-21 HVDC Modelling Guidelines for Harmonic Studies
130
4.4 FACTS Devices

Flexible alternating current transmission systems (FACTS) are a family of equipment consisting of a
combination of power electronics and static components. They are designed to control one or more
electrical parameters of a transmission system to increase its power transfer capability and improve
voltage control and system stability.
FACTS devices can be split into two groups, based on the type of switching technology they use:
Thyristor-based line-commutated FACTS devices, which include:
 Static Var Compensator (SVC);
 Thyristor-Controlled Series Compensator (TCSC).
FACTS devices based on self-commutated converters (typically Insulated-Gate Bipolar Transistors -

IGBTs), which include:
 Static Synchronous Compensator (STATCOM);
 Solid State Series Compensator (SSSC);
 Unified Power Flow Controller (UPFC);
 Interline Power Flow Controller (IPFC).
FACTS devices generate harmonics depending on their configuration, control strategy and switching
scheme. Therefore, certain device parameters must be taken into consideration and adequately
modelled for harmonic studies.
A comparison of the harmonic performance of the most common LCC and VSC FACTS devices (i.e.
SVC and STATCOM) is given below as a brief overview of the main aspects to consider with each
technology:
Table 4-3 Comparison of Harmonic Performance in Typical FACTS Devices
STATCOM SVC
 Harmonic emission related to VSC  Harmonic emission related to thyristor-
topology. based converter topology.
 Harmonics related with frequency  Harmonic emission is dependent on
converter modulation. operating point.
 Converter control determines also the  Significant power system characteristic
harmonic profile, i.e. harmonic harmonics due to non-linear nature of
contribution / attenuation and impedance semiconductor thyristors.
shaping.  Possible stability issues for frequencies
 Power system characteristic harmonics below the fundamental component.
due to nonlinear components.  Possible resonances can be seen
 Possible stability problems within between TSCs, TCRs and other wind
harmonic range as well as frequencies power plant components.
below the fundamental component.
Table 4-4 Typical Harmonic Mitigation Solutions in FACTS Devices
STATCOM SVC
 Optimized modulation strategy.  Harmonic filters tuned to low-order power
 Small harmonic filters tuned to system characteristic harmonics.
frequencies related with switching  Possible extra need of harmonic filters or
operation. detuning of TSCs.
 Possible active filtering in terms of  12-pulse (or higher pulse) configurations
compensation and damping. possible to limit filter size.
 Can contribute to active damping of
harmonics generated by other
components
131
The following subsections provide harmonic modelling guidelines for selected FACTS devices. Due to
their popularity, SVCs and STATCOMs are discussed in detail. The other FACTS devices are not as
widely used in power systems, but they are also briefly covered, and can be modelled using similar
approaches.
4.4.1 Line-Commutated FACTS Devices
4.4.1.1 Static Var Compensator (SVC)

SVCs typically include a combination of some or all of the following elements connected as indicated in
the Figure 4-22:
 thyristor-controlled reactors (TCR);
 thyristor-switched capacitors/reactors (TSC / TSR);
 mechanically-switched capacitors/reactors (MSC / MSR);
 harmonic filters.
A TCR consists of a reactor in series with a switch often made of antiparallel thyristors. The reactor is
typically split into two parts, with the valves placed in between to protect them from short-circuit currents.
The TCR can be switched to smoothly vary the susceptance seen by the power system, effectively
controlling the reactive power injected to the network to maintain the required voltage set point. This is
achieved by delaying the firing angle of the thyristors to reduce the inductive current drawn by the
reactor.
The TSC/TSR is made of groups of capacitor/reactor banks that can be switched in and out of the
system, typically using a bidirectional thyristor switch. They usually include a current limiting reactor to
damp inrush current, reduce switching transients and prevent resonances with the AC system. SVCs
only having a TSC/TSR can simply provide stepwise control of the reactive power. However, the TSC
can be used to increase the capacitive range of the SVC, thereby providing a faster response compared
to mechanically-switched capacitors.
For SVCs having both a TCR and a TSC, the control of these two devices can be coordinated to provide
linear control of the reactive power support. In addition, the SVC has the capability to control
mechanically-switched elements to meet the network’s steady-state needs while saving the use of its
thyristor-based elements for a fast dynamic response. SVCs also usually require the use of permanently
connected filters to meet harmonic performance requirements.
Figure 4-22 shows a typical configuration of an SVC. The TSC and TCR are usually delta-connected to
trap the triplen harmonics in the delta. The filters and fixed shunts are generally connected in an
ungrounded wye arrangement to block zero-sequence and third harmonic currents from entering these
components. The SVC transformer is typically delta connected on the secondary side to prevent the
flow of zero sequence components into the grid.
TCR Harmonic TSC

Filter
.
Figure 4-22 SVC Typical Configuration
132
The following sections focus on the harmonic generation mechanism and characteristic harmonics of
an SVC as well as their modelling aspects. For a more in-depth overview of SVC types, characteristics
and overall performance capabilities, the reader is encouraged to consult CIGRE Technical Brochure
25 [87].
4.4.1.1.1 Harmonic Generation

TSCs are generally controlled in a way that does not generate harmonics. The thyristors will switch off
the capacitor when the current crosses zero, which corresponds to the capacitor’s peak voltage. Unless
the capacitor remains disconnected from the system for a very long time (or it is discharged by other
means), it will be partially charged when it is switched back in. To avoid undesired oscillatory currents,
the controls switch in the capacitor when its terminal voltage matches the system voltage. This switching
strategy does not produce harmonics, since once the capacitor is switched in, current flows for the
remainder of the cycle.
On the other hand, the operation of TCRs produces harmonic currents that vary depending on the
operating point. There are several references that provide detailed analyses of the characteristic
harmonics produced by the TCR [3], [87]. The TCR firing angle can range between 90º and 180º. At
90º, the thyristors conduct during the full cycle resulting in a sinusoidal current. As the firing angle is
increased, the current is reduced and flows in discontinuous pulses, which results in the generation of
harmonics.
For each phase, the thyristors are ideally fired at the same point of their respective half-cycle, resulting
in a symmetrical current waveform and thereby eliminating even harmonics. Furthermore, since TCR’s
are generally delta-connected, the triplen harmonics remain in the delta and do not flow into the AC
system (assuming a perfectly balanced system). The characteristic harmonics of a TCR are therefore
of order 6n±1, with the ones of order 6n+1 being of positive sequence and those of order 6n-1 being of
negative sequence. Figure 4-23 shows an example of the typical harmonic currents generated by the
TCR for different firing angles. It can be observed that maximum amplitude of harmonic current does
not occur at the same firing angle for all harmonic orders.
0.06
I5
0.05
I7
Harmonic Current (p.u.*)
0.04 I11
I13
0.03
I17
0.02
I19
0.01
0
90 110 130 150 170
Delay Angle (deg)
* pu of nominal (fundamental frequency) current
Figure 4-23 Typical Harmonic Currents Generated by a TCR [87]
Since power system conditions are never ideal, there may be non-characteristic harmonics produced
by the TCR depending on the following conditions:
 Even harmonics can be created by the asymmetrical firing of the antiparallel thyristors. These
harmonics are generally of much smaller magnitude than the odd characteristic harmonics, but in
some cases it may necessary to include the 2nd harmonic in the performance calculations.
133
 Triplen harmonics could flow into the network due to system unbalances, such as firing angle or
inductance differences between phases. The negative sequence component in the system
voltage will also result in triplen harmonics escaping the delta, with the 3rd harmonic having the
most significant impact. The magnitude of these harmonics depends on the severity of the
unbalance, so it is recommended to assess these harmonics taking into consideration the specific
network conditions and the TCR equipment tolerances.
4.4.1.1.2 Modelling Recommendations
4.4.1.1.2.1 Modelling an SVC as a new Connection for Performance Assessment

If needed, the harmonic performance assessment of an SVC can be divided in two components: (i)
assessment of the new harmonic distortion caused solely by the current harmonics generated by the
TCR and (ii) assessment of amplification of pre-existing background distortion caused by the interaction
between the harmonic impedance of the new installation and the impedance of the grid. Both
 New distortion caused by harmonic currents generated by the SVC
Figure 4-24 shows an equivalent circuit that can be used to assess the harmonic distortion produced by
the SVC emissions at its point of common coupling.
The TCR is represented as an ideal current source (ITCR_n) and the other SVC components are
represented by their harmonic impedance. This model must be applied for each separate harmonic
frequency, and can be used for harmonic performance calculations or for rating purposes. The
harmonics assessment must be carried out for all possible SVC configurations (TSC on and off),
different firing angles and network impedance scenarios. Look-up tables can be used to capture the
dependency of the harmonic current source with the firing angle.
PCC
Z(h)
Transformer
Z(h) AC Z(h) AC Z(h)

ITCR_n Z(h) Filters Filters AC
TSC (LV (HV System
Side) Side)
Figure 4-24 Simplified circuit for Harmonic Emission Assessment at PCC (SVC)
As discussed above, SVC configurations vary so not all the components shown in the generic equivalent
circuit (Figure 4-24) may be present in a given project. For example, an SVC may not include a TSC
and may only have filters on the secondary side of the transformer.
The maximum and minimum harmonic impedance values for each of the SVC branches are generally
provided in the Main Component Design report, which is part of the standard documents provided by
the manufacturer as part of the engineering design and review process. These values typically include
the worst-case combination of component tolerances and frequency variations.
The TCR-generated harmonic currents are not always provided by the supplier for different operating
points, but the Harmonic Performance Study should at a minimum provide the highest level observed
for each individual harmonic over the whole range of operation, typically at least up to the 25 th harmonic
134
order. These maximum levels do not occur at the same operating point or SVC configuration, so the use
of this non-consistent set of harmonics will lead to a conservative THD assessment. Table 4-1 shows a
template that can be used to gather the required data to model the TCR generated harmonics.
Alternative formats, such as definition of current relative to the fundamental component of the AC line
side, are also possible. If no data is available, the ideal characteristic harmonics can be calculated for
the desired scenario using the equations for the TCR available in many references [3], [87].
The external system impedance (Z(h)AC System) is normally provided by the system operator in the form
of harmonic impedance envelopes (see Section 5.4).
 Amplification of background harmonic distortion caused by the SVC
Figure 4-25 presents a simplified passive model that can be used to analyse the modification of the
background voltage distortion at the PCC caused by the SVC harmonic impedance interacting with the
power system impedance. A voltage source (Ufn) is used to represent the existing background
harmonics on the power system.
PCC
Z(h) Z(h)
Z(h) AC Z(h) AC
Z(h) Z(h) Filters Filters Ufn
TCR TSC (LV (HV
Side) Side)
PCC (SVC)
As discussed in the previous subsection, the Main Component Design report should at least provide a
maximum and minimum value for the impedances of each SVC component. However, depending on
the objective of the study, it may be necessary to assess the impedances over the whole SVC operating
range; and this information may not be included in the report.
The TCR impedance varies depending on the firing angle and these values can be manually calculated
using the TCR equations [88], or they can be obtained using a software model of the SVC that includes
a representation of the controls.
The TSC may form a resonance with the system impedance that can cause large harmonic currents in
the SVC components and voltage distortions in the network. In the SVC design process, these
resonances must be taken into consideration for all harmonic orders of interest to ensure that the
equipment is rated to withstand the resulting currents. In some cases, it is necessary to include a filter
to damp these resonances.
4.4.1.1.2.2 Modelling an SVC in a System-Wide Study

Figure 4-26 shows a complete model of a SVC installation which can be included in a larger system
model suitable for general harmonic studies.
The representation of the individual components, ITCR_n, Z(h)TSC, Z(h)Transformer and Z(h)AC Filters is as
described in the previous sub-sections.
135
PCC
Z(h)
Transformer
Z(h) AC Z(h) AC
ITCR_n Z(h) Filters Filters
TSC (LV (HV
Side) Side)
Figure 4-26 SVC Model for System-Wide Studies
4.4.1.2 Thyristor-Controlled Series Compensator (TCSC)

A TCSC is a series-connected, controlled, typically capacitive reactance that can provide continuous
control of power flow. The variable series reactance is provided by a parallel combination of a fixed
capacitor and thyristor-controlled reactor (TCR), as illustrated in Figure 4-27.
Figure 4-27 Principal Diagram of TCSC
A practical arrangement of the TCSC may consist of several units connected in series and includes
other elements, protection and control.
There are three modes of TCSC operation: (i) bypassed-thyristor mode when the thyristors fully conduct;
(ii) blocked-thyristor mode where the thyristor valves are blocked (do not conduct at all); and (iii) Vernier
mode which allows the TCSC to behave either as a continuously-controllable capacitive reactance or
as a continuously-controllable inductive reactance.
In capacitive Vernier control mode, the thyristors are fired when the capacitor voltage and capacitor
current have opposite polarity which causes a TCR current to have a direction opposite to the capacitor
current resulting in an increase of the equivalent capacitive reactance.
In inductive Vernier mode, the direction of the current circulating between the reactor and capacitor is
reversed and the controller presents a net inductive impedance.
The resulting equivalent reactance of the TCSC is given in Figure 4-28.
136
Figure 4-28 TCSC Impedance as a Function of Firing Angle (source [89])
TCSC operation with partial conduction will cause some harmonics to be injected into the power system.
The TCSC will generate all odd harmonics due to the operation of the TCR, which are well known. Most
of the harmonic currents will flow through the parallel capacitor bank and the effect on the power system
will be greatly attenuated – see Figure 4-29.
Harmonics are not expected to be an issue with TCSC applications, especially if several modules are
connected in series. Typically, only the lower-order harmonics (the 3rd, 5th, and 9th) dominate and need
to be considered.
Because of parallel combination of the TCR and a capacitor bank, the harmonic generation from a TCSC
appears as a low impedance voltage source in series with the line which drives the harmonic current
through the system. As TCSCs are likely to be applied to long lines, the impedance of the line results in
relatively small harmonic current.
Figure 4-29 Flow of Harmonic Currents in a TCSC
The exception to the situation described above is when there is no capacitor and the installation is
reduced to a variable thyristor-controlled reactor. In this case, there will be significant harmonics injected
137
into the power system. Based on existing TCSC installations, this type of installation is more theoretical
than practical.
As the TCSC would be connected to two electrically separate busbars of the power system, both
busbars will be affected by the harmonic injection of the device. To correctly model and calculate the
harmonic propagation and distortion in such a case, it would be necessary to have a detailed model of
the network or alternatively, a model of the network that will represent impedance at both busbars where
TCSC is connected as well as the mutual impedance between the busbars. Such an equivalent model
must represent all feasible outage conditions on the system.
4.4.1.3 Summary of Modelling Considerations for Line-Commutated FACTS

Devices
Table 4-5 Modelling Considerations for Line-Commutated FACTS Devices
SVC TCSC
 Harmonic emission related to thyristor-  Harmonic emission related to TCR
based converter topology. switching.
 Harmonic emission is dependent on  Harmonic emission is dependent on
operating point. operating point.
 Significant power system characteristic  Significant power system characteristic
harmonics due to non-linear nature of harmonics due to non-linear nature of
semiconductor thyristors. semiconductor thyristors.
 TCR emissions can be represented as an  TCR emissions can be represented as an
ideal current source. ideal current source.
 Harmonic impedance is determined by  Harmonic impedance is determined by
filters, TCR (as a function of the firing TCR (as a function of the firing angle)
angle), TSC and transformer impedance. and capacitors.
 Possible resonances between TSCs,  Typically, only the lower-order harmonics
TCRs and other power system (the 3rd, 5th, and 9th) dominate and need
components can drive the need for to be considered.
additional filters.
4.4.2 Self-Commutated FACTS Devices
4.4.2.1 Static Synchronous Compensator (STATCOM)

As with HVDC voltage source converters (VSCs), the modular multi-level (MMC) topology is becoming
the most widely used for STATCOMs.
For the STATCOM with a modular multi-level topology (Figure 4-30), each module with power electronic
devices has a switching frequency that is several times that of the power system fundamental frequency,
resulting in low power losses. The STATCOM has several modules connected in series and its output
voltage is the summation of each module output voltage. The amplitude of the STATCOM output
harmonic voltage depends on the switching frequency, the number of levels and the sampling cycle.
STATCOMS have lower harmonic emissions than SVCs and, in many cases, do not require harmonic
filters.
138
VSC
VS ule
module
C
o d
m
VS ule
C
d
VSC
mo
module
VSC
mo SC
le
module
du
V
mo SC
le
du
V
Figure 4-30 Simplified Three-Phase Diagram of the Modular Multi-Level STATCOM
The following sections focus on the harmonic generation mechanism and characteristic harmonics of a
STATCOM as well as their modelling aspects. For a more in-depth overview of STATCOM
characteristics and overall performance, the reader is encouraged to consult CIGRE Technical
Brochures 144 [90] and 663 [91].
4.4.2.1.1 Harmonic Generation

As previously discussed for HVDC VSC converters (section 4.3.2), the STATCOM behaves as a
harmonic voltage source, rather than as a current source.
For both the two-level and modular multi-level topology of STATCOM applications, the harmonic voltage
is generated not only due to the hardware (i.e. the non-linear behaviour of semiconductor, dead-time
topology and modulation) but also by the control techniques of the software (e.g. the power electronics
control). For the hardware component, the non-linear behaviour of the semiconductor will typically
generate harmonics well exceeding 2.5 kHz whilst the dead-time/blanking-time necessary for the correct
operation of half-/full-bridge mainly contributes to the generation of low-order harmonics (e.g. 5th and 7th
harmonics).
The harmonics generated by the switching modulation are largely linked to its power circuit topology. In
the case of a two-level VSC, switching harmonics are found at the sidebands of the integer multiples of
the switching frequency. In the case of a modular multi-level VSC, the switching harmonics are found at
the sidebands of integer multiples of the equivalent switching frequency, which is determined by the
number of interleaved cells/levels.
For the software component, power electronics control boards sense the grid voltage and regulate the
output current in a closed-loop manner. Figure 4-31 provides an overview of generic stationary frame
STATCOM control, where the grid voltage, 𝑒, is sensed and the phase angle, 𝜃, is determined by the
phase-locked loop (PLL). A cascaded control is applied, where the inner current control loop regulates
the inverter side current, 𝑖1 , to its setpoint, 𝑖𝑑∗ & 𝑖𝑞∗ , whilst the outer loop regulates the DC bus voltage,
𝑣𝑑𝑐 , to a constant and generates the q-axis current setpoint, 𝑖𝑞∗ , as per the reactive power setpoint, 𝑄1∗ .
In the case of harmonic distortion on the background voltage, 𝑒, the combined effect of the PLL, outer
loop, and inner loop will cause distortion on the voltage reference input to the PWM modulator, thus
influencing the STATCOM output voltage harmonics at the terminal, 𝑣𝑖𝑛𝑣 . In Figure 4-28, 𝑓1 and 𝑓𝑠
represent the fundamental frequency and digital sampling frequency of power electronics, respectively.
139
Figure 4-31 STATCOM Control and Oscillation Phenomena Impact Overview
4.4.2.1.2 Harmonic Impedance

The following considerations regarding the harmonic impedance of a STATCOM should be noted:
 The harmonic impedance is operating-point dependent (e.g. STATCOM maximum capacitive or
inductive output) for the low-frequency range up to the bandwidth of the PLL (i.e. typically up to
the 5th harmonic order). Above that frequency, in the range where the impact of the inner current
controller is dominant, the converter input impedance can be considered as setpoint independent.
Above the active bandwidth of the inner current controller of the VSC, the passive component will
dominate the frequency behaviour of the input impedance of the VSC and generally it contributes
additional resistive damping to the system as a result of the skin effect.
 The input impedance model of the STATCOM may need to consider the PLL impedance shaping
effect when the Short-Circuit Ratio (SCR) is near or below 3. In this case, detailed information
needs to be provided by the vendor to represent cross-sequence and cross-frequency coupling.
These effects are difficult to capture in standard frequency domain analysis. If it becomes a
concern, more specialised tools and detailed converter models are required, but these are
normally only within the control of the STATCOM vendor
4.4.2.1.3 Modelling Recommendations

Some guidelines are provided below to model a STATCOM in a typical system-wide harmonic study (in
frequency domain) for users who may lack detailed STATCOM modelling parameters or specialised
tools (e.g. TSO, utility, consultant or other third-party). The equivalent circuit is the same as for HVDC
VSC, as shown in Figure 4-20.
 Harmonic Source
The representation of the harmonics generated by a STATCOM is the same as for the HVDC VSC – i.e.
as an active harmonic voltage source.
In general, the harmonics generated by the STATCOM are of very low magnitude, operating point
dependent and unique to the propieraty solution adopted by the vendor. The supplier can provide the
140
harmonic voltages produced for each operating point or, at the least, the maximum individual harmonic
voltages considering a range of operating points. Table 4-1 provides a template for gathering the
required information.
As the harmonic generation is typically very low, it is normally sufficient to calculate a non-consistent set
of maximum harmonic voltages over the complete range of operation of the converter, and use this as
the modelled harmonic generation, represented as a voltage source behind the converter internal
harmonic impedance.
 Harmonic Impedance
With reference to Figure 4-20, the harmonic impedance model must include not only the passive
components of the STATCOM (e.g. harmonic filter, series reactor, step-up transformer) but also the
VSC internal impedance defined by the operational mode, the controller transfer functions and its
corresponding parameters. The passive components normally dominate the equivalent input impedance
within the medium to high frequency range. However, for low frequencies within the bandwidth of the
converter inner current control loop, the influence converter impedance may dominate. Therefore, the
harmonic impedance is operating point dependent in the low frequency range (typically up to the 5 th
harmonic order).
In order to assess the interactions between the STATCOM harmonic impedance and the power system,
the vendor needs to supply the harmonic impedance for a range of operating points.
4.4.2.2 Solid State Series Compensator (SSSC)

Like the TCSC, the SSSC is a series-controlled reactance that can provide continuous control of power
flow. The principal difference between the TCSC and SSSC is the type of converter used. The SSSC
uses a self-commutated converter connected into the regulated circuit via a transformer, as illustrated
in Figure 4-32. The variable power flow control is achieved by varying the apparent series voltage
injected into the circuit by the converter via the series connected transformer.
Figure 4-32 Principal Diagram of SSSC

Harmonics generation is dependent on the type and topology of the converter, which is already
described in detail in previous sections. The main characteristic of these devices is that the harmonic
generation from a SSSC appears as a voltage source in series with the line (Figure 4-32).
4.4.2.3 Unified Power Flow Controller (UPFC)

A unified power flow controller (UPFC) combines a shunt and a series-connected controller in one device
and as a result can control active and reactive power flows on a transmission line. The principal
schematic of the device is illustrated in Figure 4-33.
141
Figure 4-33 Principal Diagram of UPFC
As the device utilises two converters, both need to be accounted for in harmonic studies. Harmonics
generation is dependent on the type and topology of the converters, as described in previous sections.
4.4.2.4 Interline Power Flow Controller (IPFC)

An Interline Power Flow Controller (IPFC) is a device similar to UPFC but differs in that a UPFC connects
to one line, whereas an IPFC consists of a set of converters that are connected in series with different
transmission lines. In addition to these series converters, an IPFC may include a shunt converter. As
with the UPFC, the converters are connected through a common DC link.
The representation in harmonic studies needs to include all converters with harmonics generation
dependent on the type and topology of the converters, as described in previous sections.
4.4.2.5 Summary of Modelling Considerations for Self-Commutated FACTS

Devices
Table 4-6 Modelling Considerations for Self-Commutated FACTS Devices
STATCOM SSSC
 Represented as a harmonic voltage  Harmonic voltage source connected in
source behind the converter internal series with the line impedance.
harmonic impedance.
 Harmonics related with frequency UPFC
converter modulation.  Harmonics can be generated by either
 Converter control determines also the the series or the shunt connected
harmonic profile, i.e. harmonic converter (or both).
contribution / attenuation and impedance  For simplicity, harmonic generation can
shaping. be represented by a single harmonic
 Harmonic impedance is operating point voltage source combining the effect of
dependent and needs to be supplied by both converters.
the vendor for a range of conditions.
 Harmonic generation is typically low. IPFC
Assessment of a non-consistent set of
maximum harmonic voltages (to be  Harmonic voltage sources connected in
provided by the vendor) is normally series with each of the lines’ impedance.
sufficient.
142
4.4.3 Active Harmonic Filtering

The latest VSC generations (especially modular multilevel cascaded converters, MMCCs) have
significantly reduced their harmonics emissions. However, the pre-existing harmonic pollution in the
networks has increased due to the popularity of non-linear components as well as poorly damped
resonances within the harmonic range. Therefore, there is great interest in the development of active
filtering functionalities as an option for harmonic distortion reduction in the grid.
Grid-connected voltage source devices (e.g. WTs, STATCOMs) can operate as active power filters
(APFs). To provide filtering action at frequencies above the fundamental, converters typically employ an
effective switching frequency of several kHz or more to assure appropriate control bandwidth. Only a
very limited set of high power semiconductor devices can operate at these frequencies, with the
insulated-gate bipolar transistor (IGBT) being a device commercially available from a wide range of
manufacturers. Intrinsic characteristics of the IGBT make these devices better suited for voltage source
converter (VSC) applications.
VSCs themselves are available in various configurations and include: single-phase, 3-wire 3-phase, 4-
wire 3-leg 3-phase or 4-wire 4-leg 3-phase. Decision on the specific configuration is based on
application, cost, complexity, and modularity of design. Of these four, it is the 3-phase, 3-wire converters
that are typically the most cost effective, but with the caveat that they are unable to provide
compensation of zero-sequence current components. Depending on the application, either a fixed
switching frequency scheme, such as sinusoidal PWM and space vector modulation, or a variable
switching frequency hysteresis control scheme may be selected.
Active filtering techniques to reduce harmonic distortion are generally used in combination with the main
control functions of other devices (e.g. MMCCs in STATCOMs or HVDC) without the use of additional
equipment. Furthermore, the device may be designed to operate as a virtual resistance in a certain
harmonic frequency range to damp any resonance between the network and the device itself.
The term APF in this context means any VSC whose controller is designed to reduce harmonic (voltage
or current) distortion at the point of interest which can be either at the converter terminals or at a remote
node within the power system.
The most commonly seen APF control schemes can be categorized in the following ways:
 Reducing the emission levels of harmonic sources by dedicated control algorithms and control
designs.
 Additional digital filters in the main control chain to improve the harmonic rejection capability. This
will affect the overall converter control frequency response (or equivalent impedance).
 Dedicated control loop to shape the virtual converter impedance according to the specified
performance / design criteria. As above, it will introduce changes in the converter equivalent
impedance.
 Separate harmonic (voltage or / and current) controller / compensator to fulfil a certain control
operation, i.e. follow a specific setpoint.
Thus, the grid converter can improve the harmonic distortion either by active harmonic current
contribution (e.g. harmonic current compensation) or by impedance shaping (e.g. resonance damping).
Whether the active harmonic filter is series- (e.g. WT, HVDC) or shunt- (e.g. STATCOM) connected, it
can be represented for harmonic studies as an equivalent Thévenin or Norton circuit where the voltage
or current source (respectively) will represent APF harmonic contribution and the associated impedance
will reflect the converter control scheme.
Typically, APFs can be modelled in the harmonic or frequency domain to perform standard harmonic
propagation studies. However, some harmonic controllers are difficult to linearize, due to their highly
non-linear behaviour (e.g. limiters, instability detection, auto tuning). For such systems, detailed
modelling using time-domain or real-time simulations are more adequate. APF’s can fulfil a given control
function by minimizing the error between measurements and a harmonic reference. In this case, they
act as a controlled source for that specific frequency, e.g. harmonic current compensation with defined
reference current can approximately be considered at that harmonic frequency as a current source with
current magnitude equal to the setpoint / reference value.
It should be noted that active filtering is an area of ongoing work and development. For further
information the reader is referred to [92] to [98].
143
4.5 Traction Systems

4.5.1 AC Electrified Railways
For the purposes of harmonic distortion studies, the railway electrification system can be said to consist
of three principal parts, the transmission/distribution network, the traction power system and the
interface between these two systems, which may consist of traction transformers, compensation
equipment, filters, etc (see Figure 4-34). Representation of these individual components is described in
chapter 3.
Figure 4-34 Principal Arrangement of Different Sub-Systems under Assessment
One of the main challenges when assessing the harmonic distortion of electric railways is the distribution
of the harmonic generation along the railway route in both time and space. Electric railways can be very
long and as such supplied from the transmission/distribution network at many points which to some
degree have to be assessed as one system, not as individual independent connections. One way of
achieving this is to represent the traction power system by a number of equivalents which together
represent the entire railway effect on the transmission/distribution system [99].
The railway system representation must be adequate and it has to include suitable representation of all
trains operating on the system to a given schedule or timetable. Detailed representation of the trains in
harmonic studies can be very complex and it is common to employ some method of a simplified, yet
adequately accurate representation. Such simplification can utilise network equivalents of each feeder
station or traction supply point which accounts for the effects of the ac traction system impedance on
the harmonic propagation through the system and the aggregated traction load.
4.5.1.1 Traction Load Equivalent

Each train acts as a harmonic current source with the harmonic currents propagating through the
catenary system, traction transformer and into the EHV/HV system, as illustrated on Figure 4-34. The
harmonic current distribution will depend on the catenary system parameters, train parameters and train
location.
The network equivalent must adequately represent the system at each traction power supply point in
order to account for all parameters. The equivalent must account for several trains (potentially of
different types) running to a given timetable. Different equivalents may be required for different railway
sections, modes of train operation, timetables, train types, etc. Some averaging methods may be
employed to simplify the process.
A sufficiently accurate account must be taken of all trains moving along the railway route in line with the
timetable which will accelerate, coast, brake and stop at stations in line with the timetable and required
driving strategy, such as the rate of acceleration and braking. It is quite possible that these different
modes of train operation generate different harmonics spectrum which has to be suitably represented.
Each train travelling through along the railway route will encounter different system impedance
depending not only on the frequency but also on the train location. It is even possible to reach a state
of resonance causing harmonic current amplification. To account for this, it is necessary to adequately
simulate the train movement and undertake harmonic current propagation calculations.
144
Finally, the harmonic currents generated by all trains operating on the railway are aggregated using a
suitable aggregation rule. Power electronic converters typically used in trains may require different
summation law to that generally recommended [100]. A classification of the train generated harmonics
the point of view of synchronization and frequency components distribution can be grouped as follows
[101]:
 Unsynchronised Emission Sources with Fixed Frequency
 Synchronised Emission Sources with Fixed Frequency
 Uncoupled Emission Sources with Variable Frequency
 Coupled Emission Sources with Variable Frequency
 Modulation Products Between Fixed and Variable Frequency Sources
4.5.1.2 Railways with Low Frequency AC Supply

Several countries have train systems with low frequency AC supply. Thus, the traction system is not
only connected by a transformer since frequency conversion is also required. Rotating converters were
widely used in the past for converting the frequency between the power supply and the traction system,
while power electronic converters are the preferred option nowadays. Modelling those components can
be done as described in section 3.5 (Synchronous Generators) for rotating converters or 4.2 (Converter-
Based Generation) for power electronic converters.
4.5.2 DC Electrified Railways

DC electrified railways generally utilise rectifiers to convert the AC supply derived from the HV network
to DC suitable for traction. It is therefore to be expected that the harmonic distortion on the HV side and
so on the HV network will be the result of rectifier operation, rather than of direct train operation.
To assess the effect of harmonics generated by the rectifiers on the overall system and the public
network, each rectifier is typically represented by a harmonic current source based on the level of the
actual traction load. This would normally be done for each step of the traction simulation which includes
simulation of train movement to a particular timetable. The behaviour of traction rectifiers in respect of
harmonic current generation is no different to the behaviour of LCC HVDC links although the traction
rectifiers are typically uncontrolled diode rectifiers.
Some railways systems utilize LV DC traction. Due to the low direct voltage and high currents, the track
sections served by each rectifier are quite short and many individual rectifiers will be distributed
throughout the traction system. A number of such rectifiers will be supplied at medium voltage AC from
a single transformer supply point on the HV network. Each individual rectifier will generate harmonic
currents whose magnitude will depend on its instantaneous load, which will vary as trains pass through
the line section served. The harmonic currents from each rectifier will combine at the supply point.
Typically there will be a wide variation in instantaneous magnitude and phase of the harmonic currents
from each rectifier and this diversity will determine the magnitude of the total harmonic current seen at
the supply point. The definition of suitable diversity factors for different harmonic ranges needs to be
considered carefully, depending on the particular nature of the system, and if possible verified with
empirical observations.
To reduce the calculation effort, harmonic generation of each rectifier may be determined from DC loads
averaged over a period of time for each rectifier unit. The averaging period will depend on the traffic
patterns and distance between DC traction feeder stations. The diversity of load coincidence can be
considered by using aggregation laws like those presented earlier in this section.
A harmonic study of the DC traction supply system must therefore consider:
 The train load and operating pattern on the line section served by each rectifier
 The harmonic generation from each individual rectifier on the MV AC system
 Any harmonic performance limits applicable on the MV AC supply to the rectifiers (note – as this
may be an internal, dedicated system, harmonic levels higher than usual may be tolerable)
 The summation of the harmonic generation of all rectifiers connected at MV to the same HV AC
supply point, considering diversity factors
145
Harmonic performance limits applicable to the HV AC side of the supply point, which will generally be
part of the public network and be required to conform to applicable harmonic standards.
4.6 Electric Arc Furnaces

4.6.1 General Information
Almost always connected to the transmission grid, Electric Arc Furnaces (EAF) usually have large power
requirements compared to the short circuit capacity at their connection point. They are known to be
some of the most disturbing applications from the perspective of power quality. Their non-linear
operation produces an input current that is very rich in harmonics. There are several distinct process
states during the cycle of operation of the EAF: initial arc striking, start of melting, melting, refining, etc.
The start of melting is a chaotic state and generally results in the worst case for harmonic generation
while the melting process is more stable and having generally lower levels of harmonic current injections,
but this state persists for longer periods of time than for initial melting. The harmonic behaviour strongly
depends on the process state, which changes the operating point of the furnace. Each state can be
characterized by a harmonic current spectrum. Thus, decisions will need to be made as to which process
states of the furnace cycle should be considered; perhaps a single “worst case” is sufficient for some
needs.
There are two kinds of EAF: AC arc furnaces, which represent the majority, and DC furnaces. Both kinds
produce odd and even harmonics and a continuous spectrum in the harmonic range. Submerged Arc
Furnaces (SAF) are a widespread sub-type of EAF. Their electrical characteristics and controls are quite
different than classical furnaces, which results in their distinct harmonic behaviour. It is thus a complex
problem to construct a generic frequency domain model to represent the EAF.
4.6.2 Harmonic Behaviour of Electric Arc Furnaces
4.6.2.1 Classical AC Electric Arc Furnaces:

A classical AC EAF has three graphite electrodes, electrode arms, a high current system with currents
up to 80 kA and a three-phase furnace transformer. In order to enable flexible power input at optimized
operating points, they have around 30 secondary voltage taps and an on load tap changer. A furnace
control system moves the electrode arms so that the power input is permanently adjusted according to
chosen operating points. These points are optimized using a furnace circle diagram, which shows circles
of constant secondary voltage [102], [103].
146
Figure 4-35 V-I-characteristic of AC-Arcs (1 to 6 for Low Currents and 7 to 8 for High Currents) [102] and
Example of a Measured Harmonic Spectrum of the Primary Currents [103]
The arc resistance is non-linear. Figure 4-35 shows the measured current harmonics representative of
a large AC EAF. The result is a harmonic distortion of the currents on the primary side of the furnace
transformer and consequently a corresponding voltage distortion that depends on the network
impedance. AC furnaces are important sources of odd and even harmonics, mostly below order 11 th
(Figure 4-35), but generate also a continuous spectrum in the harmonic range.
4.6.2.2 DC Electric Arc Furnaces:

A DC EAF usually has one or two graphite top electrode(s) (Cathode). The Anode is integrated in the
vessel as a bottom electrode. On the high current side, the converter technology used to supply a DC
arc furnace is usually a 12-pulse thyristor rectifier which is similar to classical LCC HVDC. The main
difference with HVDC is the load on the DC side of the converter, which can be considered to be constant
for an HVDC converter (at least in the short term), but is continually changing for a DC furnace due to
fast control of resistance. These load variations affect the input AC current spectrum as well as the
passive behaviour of the furnace, as seen from the grid. Two types of control are used for DC furnaces:
voltage control, which adjusts the electrode position so that the arc has the desired voltage (slow control)
and current control, which adjusts the firing angles of the thyristors to keep the electrode current constant
(fast control).
Figure 4-36 Typical Electrical Diagram of a DC Arc Furnace with Classic 12-Pulse Operation and
Harmonic Spectrum (Measured) of the Primary Currents (Maximum and Average Values)
Figure 4-36 shows the measured current harmonics representative of a large DC EAF. All even and odd
harmonics are present. Orders 11th and 13th are dominant since the converter is in 12-pulse operation.
They are, however, significantly smaller than the theoretical calculated harmonics and they become
even smaller if the furnace operation is more unstable. This is a well-known reaction of converters under
such conditions called the “principle of the air mattress” [103].
147
4.6.2.3 Submerged Arc Furnaces

Submerged Arc Furnaces (SAF) are a particular type of EAF which are usually very large and principally
used in the industry of ferroalloys and pig iron but also for some special applications such as waste
recycling. SAF can be AC furnaces with 3 or 6 electrodes or DC furnaces with 1 electrode.
Power and impedance controls are used for AC submerged arc furnaces, by the means of anti-parallel
thyristors (Figure 4-37). Power and resistance controls are used for DC SAF, usually using a full 12-
pulse thyristor rectifier.
An SVC or STATCOM is usually added on the MV side in order to compensate reactive power and
reduce voltage flicker.
Figure 4-37 Typical electrical diagram of a 3-electrode AC submerged arc furnace
4.6.3 Modelling of Arc Furnaces for Harmonic Studies

As discussed in other sections of this technical brochure, the effects resulting from the operation of a
relatively large source of harmonic currents, such as the EAF, together with the normal system
background harmonics will combine to form a total harmonic distortion picture. It is emphasised that the
harmonic performance of an EAF is determined by the aforementioned type and design of the
installation, its operating process state and the characteristics of the external grid. Therefore, it is
imperative that adequate models are made available to the utility or the TSO to accurately represent
their effects in the power system. It is equally important that the utility or TSO specify which furnace
processes are of most interest to be captured in the models. For existing EAFs, the models would have
typically been provided prior to first energization and validated throughout commissioning tests and on-
going monitoring. However, the main challenge arises when assessing the impact of potential new EAF
installations at the very early stage of the project; e.g. prior to decision on AC or DC technology. Close
co-operation between network owner/operator and EAF developer is necessary throughout the design
of the installation to define appropriate harmonic performance requirements and to develop accurate
models. Typically, the utility or TSO will provide a record of the background distortion measured (or
estimated) over a representative period of time in order to facilitate the design from a harmonic rating
and performance perspective. There is no reason to believe that models for one EAF installation would
be applicable to another furnace.
Additional equipment, such as harmonic filters, SVCs or STATCOMs, are frequently installed at EAF
plants to keep power quality (normally harmonics, flicker and unbalance) within specified limits [104]-
[106]. It is important to represent these components in the models as described in other sections of this
document.
The following subsections discuss the main features that should be captured in the EAF models.
Key Point:
 It is assumed that any discussion on measurements of harmonic currents at the arc furnace include
the effects of power factor correction capacitor banks, harmonic filter banks and any dynamic
voltage regulation/flicker reduction high power electronic equipment that will normally be associated
with the arc furnace. If the furnace model that is provided to the utility or the TSO does not include
these, they are discussed separately, elsewhere in this technical brochure.
148
4.6.3.1 Modelling of AC Electric Arc Furnaces

 Harmonic Sources
Traditional representation of the EAF as an ideal harmonic current source behind its transformer is a
gross oversimplification and it should be avoided. A preferred model of the AC EAF, shown in Figure
4-38, is one in which the furnace is represented by its transformer with a Thévenin (or Norton) equivalent
having a harmonic voltage (or current) source at each frequency along with its internal harmonic
impedance.
In some cases, multiple models may be required in order to capture different process states. This can
be implemented by means of look-up tables. It should be noted that the harmonic components generated
by AC EAF are not, in general, independent from one another and this should also be reflected in the
harmonic source model.
Figure 4-38 Thévenin Equivalent Model (Dual of the Norton Equivalent)
When frequency domain models are not sufficient to capture the complex behaviour of the EAF, time
domain simulations with detailed models of the installation, might be needed. For the case of submerged
arc furnaces, it means a detailed representation of the converter components (thyristors) and their
control laws. These time-domain models can also be used to develop or validate simplified models for
frequency-domain analysis according to the structure shown in Figure 4-38.
The reader is warned that performing measurements of voltages and currents at the EAF facilities, as
an alternative to the model development, is probably impractical if not impossible in many cases. There
are multiple difficulties when considering direct measurements: (1) the great problem of calculating the
influence of normal background harmonics on the measurements, (2) risk of an unintended trip-out of
part or all of the EAF installation when external personnel (unfamiliar with the facilities) are on site and
cause something to go wrong, and (3) arranging for a suitable time with the furnace operator/owner to
perform the measurements (which might only be possible after annual maintenance and startup).
 Harmonic Impedance
For the Thévenin (or Norton) equivalent model of the EAF (Figure 4-38), one equivalent impedance per
harmonic frequency for each stage of the process (or each relevant operating point of the furnace) is
needed. This impedance represents the passive behaviour of the furnace components (including any
power factor correction device, harmonic filters, etc.) and accounts for the influence of the furnace on
the network harmonic impedance. In addition to all passive assets, the influence of the controller (if
applicable) also has to be covered in the harmonic impedance.
When this complex behaviour cannot be captured with frequency-domain models, time domain
simulations may be necessary. Once again, the detailed time-domain models must reflect all the process
states that need to be included in the studies. For the case of the AC submerged arc furnace, the
harmonic impedance means a detailed representation of the converter components (thyristors) and their
controls is required.
4.6.3.2 Modelling of DC Arc Furnaces

The converter technology used to supply a DC arc furnace is usually a 12-pulse thyristor rectifier which
is similar to a classical LCC HVDC. The worst-case (in terms of absolute Amps) input current harmonic
spectrum of this kind of converter can be approximated as that corresponding to an ideal square wave
or stepped square wave but, in reality, it is more complex. The spectrum is affected by the firing angle
control and the switching strategy of the thyristors and smoothing conditions (AC short circuit level and
rating of the DC smoothing reactor). The main difference with LCC HVDC, however, is the load on the
DC side of the converter which can be considered as constant for an HVDC converter (at least in the
short term), but is continuously changing for a DC furnace due to the fast control of resistance. These
149
load variations affect the input AC current spectrum as well as the passive behaviour of the furnace, as
seen from the grid.
Similar to the AC EAF type, a Thévenin or Norton representation (Figure 4-38) capturing active emission
sources and passive harmonic impedance components is recommended.
 Harmonic Sources
The generation of harmonic currents by DC furnaces strongly depends on the rectifier type. Assuming
ideal commutation and smoothing conditions, the harmonics are generated at orders h=n*p±1 (p is the
pulse number and n is an integer) and can be estimated using the well-known equation based on the
ideal square wave or stepped square wave input current:
𝐼
𝐼ℎ = 1. Equation 4.1
ℎ
If a rectifier without special features such as free-wheeling diodes or shift control is used, then Equation
4.1 can be used to estimate the harmonic output of a DC arc furnace. The purpose of the special features
and controls is to minimize flicker, not harmonics. In real operation, commutation and smoothing are not
ideal and the firing angles are continually changing. This leads to a spectrum of harmonics that is time-
varying, does not follow the ideal results of Equation 4.1, and cannot be calculated easily. Due to this
complexity, the model for the DC arc furnace as a source of harmonic currents must be furnished by the
specialist supplier of the power quality control equipment, such as STATCOM, SVC and any associated
filtering.
In order to assess the harmonic impact of a DC arc furnace when planning new plants or evaluating the
performance of compensating devices such as passive or active filters, time domain models may be
more suitable. For the modelling of the 12-pulse converter, the reader should refer to LCC HVDC
modelling in Section 4.3.1. However, the representation of the DC arc dynamics is quite challenging. A
deterministic approach is usually insufficient [107]. However, chaos-based methods using Chua or
Lorentz equations [108] can also be used.
 Harmonic Impedances
It is a very complex problem to build a frequency domain model to represent a DC furnace or a DC
submerged arc furnace. Such models must be supplied to the utility owner or operator and, in reality,
will normally originate with the specialist contracted by the furnace manufacturer to specify the power
quality control equipment. Similar to the AC EAF case, the harmonic impedance model must capture
the passive behaviour of the furnace components (including any power factor correction device,
harmonic filters, etc.) and account for the influence of the furnace on the network harmonic impedance.
In addition to all passive assets, the influence of the controller also has to be captured in the harmonic
impedance.
Key Points
 The electric arc furnace, in whichever form, is a large and often dominating source of harmonics
which deserves careful modelling for harmonic studies. It is essential that accurate models are
made available to the utility or TSO to accurately represent their effects in the power system.
 It is equally important that the utility or TSO specify which furnace processes are of most interest to
be captured in the models. This implies close cooperation between the utility or the TSO and the
customer/furnace owner, especially for future new installations.
 Measurements of voltages and currents at the arc furnace has severe drawbacks, not limited to the
difficulty to calculate the influence of the background harmonics.
 A Thévenin or Norton representation of the EAF capturing both, active emission sources and
passive harmonic impedance components, is recommended.
4.7 Variable Speed Drives

4.7.1 General Information
Variable Frequency Drives (VFD) are significant components of most commercial or industrial loads
[109]. Their basic function is to synthesize the voltages and frequency applied to a motor to achieve
control of motor speed and/or torque. VFDs are based on power semiconductors and can be significant
sources of harmonic currents. The waveform of the input currents to the drive depends on the converter
transformer winding configurations used, as well as the topology, the technology and the controls
150
implemented in the drive. For purposes of modelling for harmonic studies, it is thus important to
recognize the differences in the spectrum and magnitudes of the harmonic currents that they generate.
Five types of drives are discussed in this section. Their technologies are briefly described below and
their harmonic modelling is discussed.
4.7.2 Type 1: Classical VFD with Diode Rectifier

A voltage source PWM inverter supplied from an uncontrolled (diode) front-end 6-pulse rectifier is one
of the most common configurations used in variable frequency AC drives (Figure 4-39). This drive is
found in LV and low power commercial and light industrial applications, from fractional kW to a few
hundred kW (driving fans, pumps, conveyors, compressor motors, etc.). The power semiconductor used
for PWM switching is the IGBT, which can be switched on and off at high frequency (several kHz). There
is a relatively large DC link capacitor to stabilize the DC voltage, which is controlled to be constant.
Rectifiers in larger drives could be 12-pulse, 18-pulse, or 24-pulse to achieve some harmonic
cancellation and to reduce filtering requirements. Type 1 drives cannot be used to actively slow down
the motor. The diodes prevent the flow of energy from the motor’s kinetic energy of rotation back into
the AC system.
Figure 4-39 Classical 6-Pulse Diode Bridge Rectifier and 6-Pulse PWM Inverter
 Modelling for Harmonic Studies:

For harmonic penetration studies in the frequency domain, the common model for diode front end
Voltage Source Inverter (VSI) drives is the ideal harmonic current source model, at the characteristic
(and non-characteristic, if applicable) harmonic frequencies (Figure 4-40). In addition to this model, any
power factor correction and harmonic filter capacitor banks for the drive must also be represented
separately.
Figure 4-40 Multi-Pulse VSI Model for Harmonic Studies
 Adjustment of model parameters:

If the DC output current is assumed to be smooth and the commutation effects are neglected - assuming
the AC input inductance to be zero and the DC output inductance to be infinite - then the input current
consists of a square wave (or a stepped square wave, depending upon converter transformer winding
configurations). Harmonic currents are then given by well-known Equation 4.2 and Equation 4.3 as
follows (h is the harmonic order, k is an integer, and p the pulse number) [109]-[112]:
Ih /I1 = 1/h Equation 4.2
ℎ = 𝑘 · 𝑝 ± 1, 𝑘 = 1, 2, 3, … Equation 4.3
151
The potential worst case in terms of Amps (highest magnitude of any harmonic component) is expected
to correspond to the maximum load.
In reality, the AC input inductance, due to impedance of the feeding network and supply transformers
as well as the configuration of the converter transformer winding and load current through the drive will
have a significant effect on the AC input current waveform [109]-[111]. The percent distortion currents
for an actual diode rectifier can be significantly higher than the theoretical values [109]-[112]. Figure
4-41 shows examples of AC input current spectrum for different 6-pulse VFDs at different load
conditions, with AC source inductance accounted for [113].
Figure 4-41 Examples of Input Current Spectrums for 6-Pulse VSI-PWM Drives ([114], [115], [116])
For more realistic adjustments of harmonic current magnitudes, a look-up table based on time domain
simulation results can be used [111]. Different load conditions can then be considered, as well as
different AC grid inductance or DC link inductance values. However, this approach does not consider
the rectifier operational mode (continuous or discontinuous conduction mode), on which harmonic
current emissions strongly depend. To do so, an analytical model can be used such as the one proposed
in [111].
The effects of unbalance in the supply voltage could also be considered [117], [118]. Indeed, unbalance
in the AC input voltages to the diode rectifier will result in non-characteristic triplen harmonics (3rd, 9th,
etc.) appearing in the input currents [117], [119]. The magnitudes of these harmonics will increase with
increase in degree of unbalance and there will be a corresponding decrease in the magnitudes of the
characteristic harmonics [117].
 Alternative Models:
For a more accurate approach than the ideal harmonic current source, a frequency-domain harmonic
matrix model can be used. This method allows consideration of the interaction between the grid and the
VFD. An analytical model is presented in [120] and is based on an accurate determination of operational
mode. This approach, for which voltage distortion at the connection point is considered, is recommended
for harmonic penetration studies.
As a last step in the accuracy of the model, time domain simulations with circuit-based representation
can be used. With this approach, the influence of unbalance and voltage distortion is included in
calculations. However, as for any simulation of this kind, it requires much more parameters and longer
calculation time.
4.7.3 Type 2: VFD with PWM Front End

For larger motors having ratings of up to several MW, the use of Type 1 drive would result in poor power
factor and the production of substantial harmonic currents that would manifest as significant problems
requiring mitigation. An effective and often cost-effective alternative to diode rectification is the switch-
mode PWM voltage source rectifier, which has the same topology and exploits the same power
semiconductor (IGBT) as the PWM voltage source inverter. Figure 4-42 shows the simplified electrical
single-line diagram for a variable frequency drive based on voltage source converters with full PWM
rectifier and inverter and “Active Front End” (AFE). The AFE, employing special controls to the PWM,
offers the features of:
 reduced low-order input current harmonics for smaller filter requirements,
 near unity power factor operation is possible,
152
 constant DC bus voltage necessary for voltage source converter operation [121]-[123],
 can be used to actively slow down the motor through the capability of dynamic braking by feeding
energy back into the AC system from the motor through the front end of the drive.
With some modifications, the drive with AFE has the capability of bidirectional power flow [124]
Figure 4-42 Full PWM Rectifier and Inverter

The topology of the Type 2 VFD is similar to the Type 4 wind turbine generator discussed in 4.2.1.2.1
which leads to some degree of similarity in their harmonic behaviour. The PWM rectifier with AFE can
be considered as harmonic sources yet, at the same time, as a mechanism for harmonic mitigation.
Under ideal balanced conditions the AFE produces nearly sinusoidal input currents. Total harmonic
distortion in the ac input currents is typically below 5%. However, the harmonic emissions from the
rectifier into the external system are sensitive to background harmonics due to interaction with the PWM
controls. For this reason, in the frequency domain, a representation as an ideal harmonic current source
is usually insufficient for this type of drives. It is hereby recommended to use a model structure
represented as Norton/Thévenin equivalent circuits at each harmonic frequency like the Type 4 wind
turbine generator, shown in Figure 4-1.
It should be noted that the presence of the rectifier’s PWM carrier frequency (typically from 3 kHz to 15
kHz) and its sidebands will be seen in the input current spectrum [123] but are usually above the
frequencies of interest for this technical brochure. However, in some “low end” drives, to reduce
switching losses, the carrier frequency can be lower, at around 31st harmonic. In that case, the harmonic
sidebands cannot be neglected.
Also, unbalance and distortion in the input voltages cause the appearance of odd order harmonics in
the input currents to the drive which can be mitigated by active PWM control schemes in the AFE, such
as the generation of negative sequence current to counteract the voltage unbalance and Selective
Harmonic Elimination (SHE-PWM) [125], [126] for the generation of certain harmonic currents to cancel
or minimize some background low-order harmonic voltage distortions [127]-[130].
 Adjustment of Model Parameters:
In the recommended equivalent Norton/Thévenin model, both the source and the impedance at each
harmonic frequency are dependent on the rectifier control strategies and control interaction to
background harmonics at each harmonic frequency. Two methods can be used to identify these
parameters:
 time-domain simulations in open or closed loop,
 measurements under controlled conditions.
Additional information on modelling the source as well as the operational impedance of the PWM rectifier
at the relevant harmonic frequencies can be found in 4.2.1.2.1.
If passive filters are used to mitigate the effects of the sidebands of the PWM carrier frequency, they
must be included with the Norton/Thévenin equivalent circuits. Below the frequencies of the sidebands,
the drive should not produce emissions that would require filtering. PFC capacitor banks also should not
be normally required if an AFE is used.
153
4.7.4 Type 3: PWM Current-Source Inverter Drives

Current-source inverter (CSI) technology used with Type 3 VFDs is well suited to medium voltage high-
power motor drives and can be considered as the dual of the Voltage Source Inverter (VSI). As with
line-commutated HVDC, the CSI uses a DC link inductor for energy storage rather than a DC capacitor.
The “PWM CSI” has been in use since the early 1970s. It is based on a PWM rectifier and inverter using
GTO or IGCT power semiconductors (Figure 4-43) which are capable of being switched off. The
switching frequency is normally below 500 Hz [131] to limit switching losses (which should be compared
to IGBTs that can be switched at several kHz). The rectifier may use Selective Harmonic Elimination
(SHE-PWM) modulation technique to simultaneously eliminate several low-order characteristic
harmonic emissions in the AC input currents [131]. Filter capacitors will attenuate the remaining
harmonics. For example, with a rectifier PWM switching frequency of 420 Hz, the 5thand 7th harmonic
components can be cancelled.
Figure 4-43 CSI Variable Speed Drive with PWM CSR and CSI

The PWM current-source inverter drive, like the PWM based VSC drive, can be regarded as both a
source of harmonic emissions as well as a mechanism for harmonic mitigation. Due to the relatively low
PWM switching frequency, this type of drive will be a source of harmonic emissions at the carrier
frequency and will have upper and lower sidebands within the frequency range of interest. In addition,
other harmonic components will appear in the input currents due to interaction between the converter
controls and the background harmonics. On the other hand, some lower order characteristic harmonics
could be missing from the emissions spectrum due to harmonic cancellation if the SHE-PWM control
strategy is implemented.
The steady state operation of a PWM CSI drive can be characterized by a Norton/Thévenin equivalent
representation (see Figure 4-1 ) based on equivalent ideal current/voltage sources and equivalent
operational impedance at each harmonic frequency of interest. The equivalent sources and impedances
will be dependent on the converter control frequency response (due to the interaction between the PWM
controls and the background harmonics) at a particular frequency. The equivalent voltage sources in
the model represent the disturbance caused by (non-ideal) PWM switching.
Any harmonic filter/PFC capacitor banks associated with the drive must be explicitly modelled or
included in the equivalent impedance.
Ideally, the equivalent sources and the operational impedance at each harmonic frequency of interest
should be available from the drive manufacturer over a range of loading levels for the particular PWM-
CSI drive. If not, then the harmonic behaviour of the CSI PWM drive can be investigated based on
simulations with a manufacturer’s model (if it is available) or else by direct in situ measurements,
including accounting for the natural background harmonics. The time domain steady-state response
should be reflected in the frequency/harmonic domain. If not available, measurements could be made
at the input to the actual drive. Measurements should consider the drive both in and out of service so
that the natural background harmonics can be estimated and taken into account.
4.7.5 Type 4: Load-Commutated Inverter Drives

The Load-Commutated Inverter (LCI) drive is also a type of current source inverter (CSI). The front end
of the drive is an SCR (Silicon Controlled Rectifier) bridge rectifier and the back end of the drive is based
on the SCR load-commutated inverter, one of the earliest inverters developed for application in variable
speed drives [125], [131]. The basic topologies of the LCI are indicated in Figure 4-44 and Figure 4-45.
The overall LCI drive topology resembles a miniature version of a classical HVDC system using natural
154
commutation. Relatively large DC smoothing reactors are used in the DC link. The rectifier (6-pulse,
12-pulse or 24-pulse) controls the DC current while the inverter controls the speed of the motor [125],
[131], [132]. LCI drives are used for large motors, typically over 15 MW (e.g. for the compression of
natural gas for storage and shipping), and LCI converter blocks can even be paralleled to deliver up to
about 120 MW. If the driven motor is asynchronous, capacitors are required in the inverter circuit to
assist commutation (Figure 4-44). If the motor is synchronous (Figure 4-45), then the excitation is
adjusted to provide slightly leading power factor and the inverter phase currents are commutated using
the back EMF of the motor. Power factor at the input to an LCI drive, before application of harmonic
filters and/or power factor correction capacitor banks, can range from 0.5 to 0.92 for a variable torque
load, thereby requiring a significant amount of correction in terms of capacitive Mvar. Additionally, the
LCI has large input current harmonics, with I-THD as high as 12%, requiring substantial filtering.
Utility
Grid A
Ia
B
Ib
IM
C Ic
Induction
Motor
Harmonic
Filters and PFC
Figure 4-44 Topology of a 6-Pulse LCI Drive for an Induction Motor
Figure 4-45 Topology of a 6-Pulse LCI Drive for a Synchronous Motor

LCI drives are used in applications requiring a large power transfer from the grid to the driven motor.
Also considering that the front end of the drive is basically an SCR multi-pulse rectifier, it can be
expected to have a harmonic current spectrum similar to a classical line-commutated HVDC system of
the same pulse order (see section 4.3.1).
The LCI drive is expected to have a model structure represented as Norton/Thevénin equivalent circuits
(see Figure 4-1) at each harmonic frequency similar to line-commutated HVDC having the same pulse
number. Any power factor correction and harmonic filter capacitor banks associated with the LCI drive
must be modelled or included in the equivalent harmonic impedance.
The drive manufacturer should be able to provide the operational impedances of the drive and the
injected current at each of the harmonic frequencies of interest at full load and at various firing angles
of the SCRs to account for a range of realistic operating conditions. If this approach is not possible, then
in situ measurements are recommended, preferably with and without the LCI operating and at various
loading levels up to the full load so that the natural background harmonics can be estimated and taken
155
into account. The reader is referred to 4.3.1 on LCC HVDC for some additional information on modelling
of current-source converters.
It is interesting to note that LCI drives do not have the inherent capability to mitigate any of their
harmonics, as with some PWM-based drives which use SHE. As an example, Figure 4-46 shows the
spectrum of harmonic emissions from a particular 12-pulse LCI drive for a 14,200 hp (about 13 MW)
synchronous motor for two different loading levels [132].
Figure 4-46 Input Current Harmonics for a 12-Pulse LCI Drive for a 12.47kV 14,200 hp Synchronous Motor
4.7.6 Type 5: Phase-Controlled Cycloconverters

The topology of the phase-controlled cycloconverter (CCV) is quite different than the topology of almost
all other variable frequency drives because there is no DC link. It is a high-power semiconductor (SCR)
device which directly converts an AC voltage waveform to another AC waveform having a lower
frequency. Power flow can be bidirectional, de facto four-quadrant operation. CCVs are applied with
very large synchronous motors, where control over a range of low speeds is required (e.g. cement and
ball mill drives, mine winders, and rolling mill drives or drives for hydro-electric pumped storage). This
section does not consider the matrix converter or single-phase-to-three-phase cycloconverters. These
are discussed in [125], [133].
The two basic topologies of the classical three-phase-to-three-phase cycloconverter are the half-wave
(3-pulse) configuration (18-thyristor circuit) and the more common 6-pulse bridge configuration
(36-thyristor circuit) [125]. Each phase to the motor is supplied from two fully controlled back-to-back
connected anti-parallel SCR bridges. One bridge is the positive converter and the other is the negative
one. The firing angle of each SCR phase group is modulated sinusoidally with a 2π/3 phase angle shift
to synthesize a three-phase balanced set of voltages at the motor terminals. There are also two
operating modes: blocking mode and circulating current mode, with the former being much more
commonly used. Blocking mode cycloconverters control motor speed over a range that is generally
below 20Hz and the circulating current types below 40Hz. Figure 4-47 shows a 6-pulse blocking mode
type cycloconverter with 36 thyristors. For the blocking mode CCV, while one bridge is on, the
antiparallel one is off, hence the term “blocking mode”. The circulating current type CCV maintains
simultaneous conduction of both the positive and negative bridges by placing large reactors between
the bridges. This prevents output current from becoming discontinuous as a result of switching. The
operating principles of cycloconverters are described in detail in chapter 4 of [125].
156
Figure 4-47 Topology of a 3-Phase to 3-Phase 6-Pulse Blocking Type Cycloconverter

The CCV is expected to have a model structure represented as a Norton/Thévenin equivalent circuit
(Figure 4-1) at each harmonic frequency of interest. For the particular cycloconverter of interest, being
a special-purpose application-dependent power converter and a significant source of a wide range of
harmonics, the specialist manufacturer should be able to supply the spectrum and the operational
impedance at each harmonic. If the cycloconverter uses dedicated harmonic filtering/PFC capacitor
banks, the design details of these should also be provided by the specialist manufacturer and captured
in the model.
Cycloconverters have a complex current spectrum [133] compared to other types of VFDs. The input
currents of the cycloconverter are expected to have complex harmonic patterns. The input current
harmonics are at frequencies:
(n·p ± 1)·fi ± m·f0 Equation 4.4
where fi is the input frequency to the rectifier and f 0 is the drive output frequency. “m” is constrained by
the following requirement:
(n·p ± 1) ± m = odd integer Equation 4.5
where p = pulse number, m = integer, and n = 0, 1, 2…. Due to the complexity of the input spectrum for
a cycloconverter, it is suggested the reader use reference [125] or [133] for a more detailed treatment
of harmonic emission.
As an example, Figure 4-48 shows the cycloconverter harmonic current spectrum for an 18 MW motor
operating at full speed (11.6 RPM), taken from Table 1 of Reference [134] and based on manufacturer’s
data. It is interesting to compare Table 2 (operation at 8.5 RPM) of [134] to Table 1, showing some
shifting in the spectrum of the interharmonics. The significant feature of the emissions from
cycloconverters compared to other VFDs is the relatively significant magnitudes at some of the higher
harmonics.
157
Figure 4-48 Harmonic Current Spectrum of an 18 MW Cycloconverter
4.8 Summary
This chapter has addressed the modelling of non-linear devices as source of harmonic distortion in the
power system. The most common devices are discussed, namely converter-based generation, HVDC
converters, FACTS devices, traction systems, electric arc furnaces and variable speed drives.
For each of these devices, the chapter explains the mechanisms for generation of harmonics and
provides recommendations for adequate modelling in frequency domain for harmonic studies. When
possible, limitations of the models are highlighted and recommendations are provided for more
sophisticated methods of analysis.
A common feature of most non-linear devices is the complexity of operation and the dependency of
many factors (e.g. operating mode, control loops, converter topology, external grid, etc.) for their
harmonic performance. As such, it is not possible to develop generic models for these devices. Instead,
a common model structure is proposed in this chapter based on a Norton/Thévenin equivalent circuit
(Figure 4-1). This recommendation aligns with the recent IEC TR 61400-21-3 ED1 [62], applicable to
the modelling of Wind Turbines (WT). The same principles can be extended to model all other non-linear
devices, especially those based on power electronic converters. In addition to the obvious advantages
of standardisation, the proposed structure offers the benefit of providing a comprehensive
characterization of the non-linear device without the need to disclose proprietary design features.
158
5. General considerations for harmonic studies

5.1 Introduction
As the level of harmonic generating devices connecting to the systems increases, the need for
performing detailed assessment of their impact on the grid (in terms of adverse impact on network
equipment and on other customers) is becoming more critical. Accurate modelling of each network
component and non-linear device is one key aspect of the analysis, but not the only one. Practical
considerations regarding the assumptions, inputs, interpretation of results, data exchange, etc are
complementary to the modelling aspects. This chapter provides guidelines and discussions on key
aspects to consider when performing harmonics studies related to connection of new non-linear
installations. These discussions aim to address both perspectives: system owner/operator and new
connectee with the overall objective of minimising risk of equipment failure due to excessive harmonic
distortion as well as unnecessary (preventive) investment in mitigation. While finding the right balance
is never an easy task, it is hoped that the considerations presented in this chapter will aid the engineer
in making informed decisions.
5.2 Types of Harmonic Studies

The proliferation of Power Electronic (PE) device connections increases the need and importance of
performing detailed harmonic assessments in power systems. However, the drivers and objectives of
those assessments can vary and so the modelling requirements. Here are some of the most common
type of harmonic studies:
1. Network assessment prior to the connection of new non-linear customer facility.
This study is normally performed by the System Operator and the objective is to assess the pre-
connection background distortion and to allocate emission limits to the new connectee. The system
harmonic impedance (envelopes) at the PoC is also calculated as part of this study.
2. Assessment of compliance with the emission limits issued by the System Operator.
This study is normally performed by the Customer as a pre-connection requirement for non-linear
installations. The new facility is represented in detail and the external system is represented by the
envelopes provided by the System Operator. The objective of the study is to demonstrate compliance
with the allocated emission limits for all operating points encompassed by the envelopes. Investigation
of mitigation options is also part of the study in case of non-compliance.
3. Harmonic assessment prior to a network expansion.

This study is normally performed by the System Operator and the objective is to assess the effect of the
connection of new circuits (typically large cable circuits), new capacitor banks or network re-
configurations on the system harmonic impedance. It is intended to identify harmonic resonances or
amplification that could result in excessive harmonic distortion at locations where customers are
currently connected or planned to connect. Investigation of mitigation options is also part of the study.
4. Trend investigation.
This study is intended to investigate trends in measurements from a network of monitoring devices and
to identify areas of the grid or operating conditions with increasing levels of harmonic distortion. The
objective is to understand the drivers behind the trends and to proactively implement actions that will
prevent breaches of statutory limits.
5. Forensic investigation following an incident.

This study is normally performed following a complaint or an incident such as equipment failure and the
objective is to accurately reproduce the event and to investigate the root-cause, identify responsibilities
and assess mitigation options to avoid reoccurrence.
159
6. Harmonic filter design.

This study is normally performed by manufacturers or filter specialists to meet the specified performance
requirements. The system harmonic impedance (envelopes) and emission limits at the connection point
are inputs to this study.
7. Functional specifications for HVDC or FACTS.

The objective is to define the harmonic performance requirements for new HVDC or FACTS
connections. The system harmonic impedance (envelopes) and emission limits at the PoC are
calculated as part of this study.
5.3 Considerations for Power System Representation

5.3.1 Extent of the Network Model
Typically, all utilities have a full steady-state model of their network that can replicate ordinary power
flows and be able to produce short circuit calculations. In addition, most utilities have a dynamic model
of their system adequate for stability studies. However, not many have a frequency dependent network
model suitable for the purposes of harmonic distortion analysis.
When a full network model is not available, one of the biggest challenges for performing a harmonic
study involve determining (a) how much of the network needs to be represented in detail and (b) how to
derive accurate equivalents of the rest of the network. This is typically a matter of finding a balance
between the required accuracy on one side and the availability of data and computational burden on the
other.
It is difficult to provide general recommendations in terms of minimum number of nodes or distance to
be accurately represented as it is highly dependent on the specific characteristics of the network and
the range of harmonic frequencies of interest. A common approach is to perform a sensitivity study
progressively extending the network model until the results from two consecutive iterations converge
[135]. Furthermore, these additional guidelines can also be considered [27]:
 Network to be represented in detail up to five buses and two transformations back from the point of
interest.
 Buses with significant levels of capacitive compensation should be represented.
 A logical place to truncate or equivalence the model is in the vicinity of large conventional generation
plants, which serve as sinks due to the relatively low sub-transient impedance.
The extent of the network represented in detail must be sufficient to incorporate all contingencies that
need to be studied. The reader is directed to chapters 3 and 4 of this TB for guidelines on accurate
representation of individual network components.
Accurate representation of the external network with a network equivalent can be a challenging task.
Various approaches for accurate representation of network equivalents are discussed in [135]. As a rule
of thumb, the use of Power Frequency Network Equivalents (short circuit impedance or other variations)
is only adequate when the external network does not exhibit any resonances within the frequency range
of interest (see section 5.5) or when the external network is electrically so far from the area of study that
does not have significant impact on the harmonic impedance at the buses of interest. In other cases, a
frequency dependent network equivalent (FDNE) is required.
If the frequency response of the network is known (from measurements or from a complete network
model used for a frequency scan), a network equivalent can be built that reproduces its frequency
response within the frequency range of interest.
The use of the FDNE instead of the full network model saves computation time and facilitates the
exchange of data between different companies as the network topology and its parameters are hidden
[136]. It should be noted that multiple FDNE may be required in order to capture a range of scenarios
and operating conditions. This is normally implemented in the form of harmonic impedance envelopes
capturing a wide range of credible operating conditions, as described in section 5.4.
Most commercial harmonic analysis tools based on frequency domain techniques include functionality
to represent the frequency variation of external grids as look-up tables or matrices. Alternatively, if EMT
160
tools are selected for the study, the frequency domain response of the external network can be fitted
with a rational function-based model which could be obtained using a technique such as vector fitting
[137] with passivity enforcement [138].
To get accurate results for the frequency response of the system, it is preferable to calculate the FDNE
from a detailed network model. However, if the detailed network model is not available in a simulation
tool, the network can be modelled in detail up to a certain distance from the point of interest and with
power frequency network equivalent, like Thévenin voltage source, beyond this distance. To determine
the minimum critical distance, CIGRE working group C4.307 [135] has suggested to “progressively
increase the distance (horizontal and vertical) of the detailed modelling until the results do not change
in a significant manner”. In other words, the detailed network is progressively extended horizontally,
starting from the highest voltage level, until the variation of the network impedance is not significant
anymore in the frequency range of interest. The network is then extended vertically to lower voltage
levels until the variation of the network impedance is not significant anymore. When extending the
network, priority should be given to the electrical nodes containing the highest capacitive components.
This method is extensively used by the French [136] and Spanish [139] transmission system operators.
Many harmonic studies have been recently conducted on the French HV grids. As a result, the detailed
network model has been progressively extended and it now includes the complete 400 kV and 225 kV
grids in an EMT tool. As an illustration, the complete 400 kV detailed network is displayed in Figure 5-1.
Other TSOs, such as those in Denmark, Ireland and the UK, have reported the use of full system models
in frequency domain for their harmonic studies.
One drawback of such large and complex network models is the continuous requirement to keep large
volumes of data up-to-date and to maintain coherency between the studies in different simulation
platforms for the same project. However, the quick accessibility to all parts of the model and increased
confidence on simulation results normally outweighs the data management overhead.
Figure 5-1 Complete 400 kV Detailed French Network
5.3.2 Power System Configuration

In order to provide practical guidelines concerning the power system configurations to be considered in
a harmonic study it is important to take into account the main purpose of the application, e.g. if the
harmonic analysis is for a future connection (planning study) or for an existing connection (operational
study).
In the case of a planning study, where a new connection to the grid is being investigated, it is suggested
to consider at least the year of connection as well as that corresponding to some years later or the last
year of the planning horizon. Some cases, such as HVDC projects, may require longer time horizons,
i.e. between 25 and 40 years ahead, to ensure robust designs of harmonic filters [23]. In this analysis,
161
the engineer should be cognisant of the increasing levels of uncertainty associated with the later years.
Concerning the evolving system configuration during this period, it is recommended to capture
significant development phases that could impact harmonic impedance, especially near the connection
point.
In the case of an operational study, where the installation and the system configuration are well defined,
it is important to replicate as much as possible the conditions that can be the cause of trouble or
investigation. In addition, the expected system configuration several years ahead may be considered to
investigate the risk of reoccurrence.
5.3.2.1 Demand Level

For each system configuration, it is common practice to assess at least two or three system demand
levels: minimum, intermediate and maximum. The objective is to capture the changes in system
resonances and damping associated with the various levels and composition of the load as well as the
status of reactive compensation equipment. These seasonal variations are illustrated in Figure 5-2 which
shows harmonic voltage measurements (95th percentile) in a transmission station in the UK over a period
of four consecutive weeks during summer (low demand), spring (medium demand) and winter (high
demand). This figure shows that the maximum recorded values do not correlate with the level of system
demand for all harmonic orders, suggesting seasonal shifts in harmonic resonances. While these
measurements cannot be generally extrapolated to other systems, they support the need to consider
various levels of demand in harmonic studies.
As a general rule, the following should be considered when feasible [23]:
 System load / generation variation for a maximum demand scenario.
 System load / generation variation for a minimum demand scenario.
 System load / generation variation for intermediate demand scenario(s).
 Different AC system generation connection conditions, e.g. differing mixes and locations of hydro,
nuclear, thermal, wind, solar, other HVDC links etc. Where nearby generation exists, it would be
interesting to consider not only the above, but also scenarios representing the lowest practical levels
of such generation on-line, to capture the weakest credible system conditions.
 Status of reactive compensation equipment, both dynamic and mechanically switched, required for
each level of demand.
0,70%
Vy ph spring 2014
% Harmonic voltage of fundamental
0,60%
Vy ph summer
0,50% 2014
Vy ph winter 2015
0,40%
0,30%
0,20%
0,10%
0,00%
5
6
2
3
4
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Harmonic order
Figure 5-2 Harmonic Voltage Distortion Measurements (95th Percentile) in a Transmission Station in the
UK for Three Levels of System Demand: Summer, Spring and Winter
162
5.3.2.2 Network Contingencies

The number and type of network contingencies set that should be considered in a harmonic study is
strongly related to the security criteria in which the utility or system operator plans and operates the
network. In general, the goal is to capture the largest feasible number of outages by running automation
scripts. However, it is very important to ensure that they represent credible operating conditions by
performing fundamental frequency load flow checks against planning and operational security
standards.
A further consideration relates to the scope and intended outcome of the study. For example, the
definition of “performance” and “rating” requirements for harmonic filters, HVDC, FACTS, etc are
normally based on a different set of contingencies, with the rating requirements defined by conditions
beyond those planned for under the network’s security standard [23]. Also, depending on the utility or
system operator criteria, a cost of mitigation vs risk and consequences of occurrence analysis could be
carried out to deal with extreme or unusual scenarios when the acceptance criteria is not met. In terms
of type of outages, in general, all possible types should be considered, e.g. planned or forced outages
of lines/cables, transformers, generator units, converter bridges, reactive compensation devices; etc.
The most common types of contingencies considered in planning studies are listed below:
 The well-known N-1;
 N-1-1, where the system is submitted to a forced outage of one component, being previously
operating without any other component due to a planned outage;
 N-2, where the system is submitted to a forced outage of two components simultaneously. The trip
of two independent circuits is a typical example of this group;
 Any combination of planned outage considered feasible under normal conditions;
 If there are converter bridges near the point of interest, the outage of part of the bridges (for example
one 6-pulse bridge outage in a 12-pulse bridge converter) is an interesting case due to the loss of
the harmonic compensation effect that occurs in the complete configuration. The necessity to
consider this type of outages depends on the specific design of the converter bridge, in terms of
possible operation modes with some parts out of service.
An example of harmonic impedance variation due to network contingencies is illustrated in Figure 5-3.
This graph shows frequency scans at a transmission node using a full detailed system model of the Irish
system. A single circuit outage introduces a large parallel resonance at approximately 800 Hz and a
double-circuit outage shifts this resonance towards 750 Hz.
1000
Intact (N) Condition
900
Single contingency (N-1) condition
800
700 Double contingency (N-2) condition

Impedance [Ω]
600
500
400
300
200
100
0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Frequency [Hz]
Figure 5-3 Harmonic Impedance in a Transmission Station in Ireland under Intact (N), Single Contingency
(N-1) and Double Contingency (N-2) Conditions
Depending on the complexity of each network and the range of harmonic frequencies of interest, it is
not straightforward to provide general guidance on the number and type of contingencies to be studied
for each system configuration or demand level. This choice should be made through engineering
judgement, based on knowledge and understanding of each network. To reduce number of cases, it is
163
generally suggested to restrict the contingencies up to the third node away from the point of interest
unless there is evidence that other contingencies beyond that boundary could result in more onerous
conditions within the range of harmonic frequencies of interest. The need to check the operational
feasibility of each contingency is emphasised.
5.3.2.3 Practical Recommendations

The combination of the above-mentioned factors (system configurations, demand levels and network
contingencies) can give rise to a very high number of cases to be investigated. In addition, the following
points should be emphasized:
 The demand level and network contingencies also have an influence on the status of the system
reactive compensation, both dynamic and fixed (e.g. mechanically switched capacitors and
reactors) types. In this respect, all possible realistic combinations of shunt reactive compensation
at or close to the connection point should be considered because, where more than one such
device is connected, these are likely to interact, thereby forming differing resonance conditions.
 The modelling of SVC and HVDC installations that are electrically close to the area of interest
should deserve special attention. In this case, they should be modelled explicitly at least for what
concerns their associated filters and shunt capacitors banks (or TCR banks for the SVCs), rather
than being included as a lumped element within the network. Many combinations of this
associated equipment should be established, considering the variation in the number and types
of banks, which may be connected to varying load. Furthermore, the effects of their detuning (due
to changes in system frequency, ambient temperature, capacitor element failures, etc.) should be
considered. These installations can have different harmonic behaviour according to the different
configuration of the filters and/or shunt capacitor banks (or TCR banks for the SVCs). In addition,
the different operating modes of an HVDC installation should be considered: different active
power flow level and different control modes (active power or DC voltage). These changes can
indeed affect the harmonic behaviour of the HVDC converter.
 Given that each of the above two factors usually give rise to many different operating states, this
can require for a high number of cases to be considered, to search critical resonance conditions.
It is therefore worthy to be noted that attention must be given to considering only realistic conditions
(practically feasible combination of system configuration / demand level / contingencies / capacitor
banks / SVCs and HVDC), in order to limit the computational burden and, more importantly, to avoid
overly-pessimistic results leading to investment in mitigation or over-design of harmonic filters.
Any network condition that can be deemed unrealistic, particularly in terms of generation and load
scenarios (i.e. those conditions that imply impossible operating scenarios or which might fail to provide
a convergent fundamental frequency load flow), should not be included in the harmonic study. If the
software does not allow for load flow calculations, it should at least be verified that the fundamental
frequency short-circuit impedance, calculated for the same cases as the harmonic impedance, is within
the anticipated range.
5.4 Harmonic Impedance Loci and Envelopes

Once the network harmonic model has been constructed from its constituent components as described
throughout this Technical Brochure and the harmonic impedance of the network calculated, as seen
from a certain node (for example, the point of connection of new plant), the final step in the system
studies is to present the calculated impedance in a form which is of use to third parties in the power
quality analysis process. Detailed discussion of this topic may be found in (TB139 [80], TB553 [23],
IEC62001 [84]) but for completeness and convenience a condensed description is included below.
The harmonic network impedance of the network is continuously changing due to different operating
configurations, connection of loads and generators, outages of major components, etc. Therefore it is
important that the information provided by system operators to third parties captures these variations
and accounts for future network expansions, as discussed in previous sections. The calculated harmonic
impedance for each scenario and operating condition can be presented in tabular form, but it may
consist of a large amount of data. For the purposes of data exchange, it is more practical to present the
information as envelopes in an R-X plane encompassing all possible operating conditions. The harmonic
impedance data can then be displayed graphically in a single envelope including all frequencies of
interest or as a family of envelopes each comprising one or more harmonic frequencies. With this
approach, it is sufficient to provide a definition of the envelope boundaries for each harmonic order (or
group of frequencies), which simplifies the process for exchange of data and subsequent analysis.
164
The concept of harmonic impedance loci and envelopes is illustrated with an example in Figure 5-4,
which relates to the impedance at a 110 kV substation in the Irish transmission system. Figure 5-4 (a)
shows the well-known frequency scan representation in which the magnitude of impedance is plotted
against frequency for a defined network operating condition. Peaks in frequency scan relate to parallel
resonances (e.g. 635 Hz) while valleys relate to series resonances (e.g. 698 Hz). Figure 5-4 (b) shows
the same harmonic impedance data but plotted in an R-X plane. This representation reveals the path
traced in the complex plane by a single network operating state with increasing frequency, also known
as impedance locus. It can be seen that the harmonic impedance exhibits a characteristic spiral-like
path with multiple loops.
Figure 5-4 (b) shows that the harmonic impedance at this transmission node for the assumed network
condition presents inductive behaviour for most of the frequency spectrum. However, it changes to
capacitive and then back to inductive at the major resonant frequencies (highlighted in the figure), which
is reflected as a major loop in the R-X plane. Major resonances can be identified in the R-X plane as
those points where the impedance path crosses the horizontal axis (i.e. where the impedance is purely
resistive). The impedance locus also shows a number of minor loops in which the sign of the impedance
does not change. It is worth mentioning that the characteristic behaviour around major resonances is
typically defined by large changes in impedance associated with small changes in frequency, as can be
seen in Figure 5-4 (b) in the frequency range from 625 Hz to 640 Hz.
Figure 5-4 (c) displays a group of frequency scans corresponding to different scenarios and operating
conditions selected as per guidelines described in section 5.3.2. It can be seen that the network
impedance exhibits a number of series and parallel resonances whose amplitude and location can vary
significantly with different operating conditions. Figure 5-4 (d) displays the same harmonic impedance
data in the R-X plane, comprising the loci for all operating conditions. The red dashed line is the
harmonic impedance envelope. The network harmonic impedance can present any value on the
boundary or within this envelope.
The harmonic impedance data can be subdivided into smaller envelopes covering narrower frequency
ranges as shown in Figure 5-4 (e) for 5th ≤ h ≤ 7th and Figure 5-4 (f) for 26th ≤ h ≤ 35th. Figure 5-4 (g)
compares the reduced frequency range envelopes with the larger single frequency range. This
subdivision allows a more realistic power quality assessment, especially in the low order harmonic
range.
It should be noted that this section concentrates on harmonic impedance envelopes as the most
commonly used approach in industry to define system impedance for compliance with harmonic
emission limits and/or for filter design. However, some utilities may adopt other approaches, such as
the provision of families of trajectories for defined frequency ranges or the provision of frequency scans
covering the frequency range of interest for all scenarios. Independently of the format selected to present
the data, the practical considerations described in section 5.4.2 still apply.
165
Z [ohm] X [ohm]
400 350
350 300
Parallel 635 Hz
300 Resonance 250
625 Hz
200
250
150 630 Hz
200
100
150 R [ohm]
50
100 0
0 50 100 150 200 250 300 350
50 -50
Series
Resonance 698 Hz -100 635 Hz
0
640 Hz
0 200 400 600 800 1000 1200 1400 1600 1800 2000
-150
Frequency [Hz]
(a) Impedance vs frequency for one operating condition (b) Harmonic impedance locus from 2nd to 40th harmonic orders
Z [ohm]
X [ohm]
1000
800
900
800 600
700
400
600
500
200
400
R [ohm]
300 0
0 100 200 300 400 500 600 700 800 900
200
-200
100
0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 -400
Frequency [Hz]
(c) Impedance vs frequency for multiple operating conditions (d) Harmonic impedance loci from 2nd to 40th harmonic orders
X [ohm] X [ohm]
250 600
500
200
400
150 300
200
100
100
R [ohm]
0
50 0 100 200 300 400 500 600 700 800 900
-100
R [ohm]
0 -200
0 50 100 150 200 250
(e) Harmonic impedance loci from 5th to 7th harmonic orders (f) Harmonic impedance loci from 26th to 35th harmonic orders
X [ohm]
800
2 ≤ h ≤ 40
600
26 ≤ h ≤ 35
400
5≤h≤7
200
R [ohm]
0
0 100 200 300 400 500 600 700 800 900
-200
-400
(g) Harmonic impedance envelopes
Figure 5-4 Example of Harmonic Impedance in a Transmission Network: Locus, Loci and Envelope (5 Hz
Calculation Resolution)
166
5.4.1 Types of Harmonic Impedance Envelope

Various types and shapes of impedance envelopes can be found in literature and are in practical use.
The selection of a particular envelope type depends on a number of factors such as the amount and
quality of network data available, the network characteristics and frequency behaviour, the intended
application, etc. A brief description of the most common types follows.
 Circles
This is the simplest envelope shape, as illustrated in Figure 5-5. The envelope is defined by a circle
centred in the horizontal axis and passing though the origin. Minimum impedance and angle boundaries,
applicable to the frequency range captured by the envelope, are added. The circle diameter corresponds
to the maximum harmonic impedance (Zh(max)), also for the frequency range captured by the envelope.
Circles can reproduce very closely the R-X scatter plots around major resonance conditions. However,
in other conditions they can result in the inclusion of a large non-applicable area, particularly in the
capacitive region.
X [ohm]
300
200
Zh(max)/2
100
h(max)
0 R [ohm]
0 100 200 300 400 500 600
h(min)
-100
Rh(min)
-200
-300
Figure 5-5 Generic Circle Envelope
 Sectors
This is a variation of the circle envelope that restricts the area by setting upper and lower limits to the
magnitude and angle of the harmonic impedance as shown in Figure 5-6. Each sector is defined by the
following four parameters: Zh(max), Zh(min),  h(max),  h(min), applicable to the frequency range captured by
the envelope. Similar to the circle representation, sectors are very quick and easy to calculate.
X [ohm]
300
250
200
Zh(max)
150
100 h(max)
50 Zh(min) h(min)
R [ohm]
0
0 50 100 150 200 250 300
Figure 5-6 Generic Sector Envelope
167
 Polygons
Discrete polygons with variable number of vertices can provide a close delimitation of the impedance at
specific harmonic orders without including excessive empty areas with no calculated impedance points.
A basic definition of a polygon envelope is shown in Figure 5-7, based on maximum and minimum values
of resistance, reactance and impedance angles within the frequency range captured by the envelope.
More complex and irregular polygon shapes can also be defined if needed to provide a better fit to the
calculated impedance values (see Figure 5-4 and Figure 5-8), for example by means of convex-hull.
Polygons are typically defined by the co-ordinates of the vertices, provided in tabular format. In general,
this type of envelope requires the most effort to produce, however it also provides the best definition for
the harmonic behaviour of the system impedance.
Figure 5-7 Generic Basic Polygon Envelope
5.4.2 Practical Considerations for Creating Harmonic Impedance Envelopes

As demonstrated in previous chapters, the impedance of a meshed power system can be very sensitive
to frequency. Component tolerances, circuit outages, network re-configurations, nominal frequency
variations, etc. can lead to some shifts in resonant frequencies. If resonances are poorly damped, e. g.
if several converters are connected by cable systems, the exact location of the resonance can have
significant impact on performance. As an example, the change in system impedance due to the outage
of a string of wind turbine generators in a large wind farm may cause the difference between a grid that
can be normally operated or the presence of harmonic problems. Therefore it is important to capture all
relevant operational configurations.
A suitable frequency step resolution needs to be selected in order to capture all resonances in the
frequency scan calculations. If too large frequency steps are adopted, inter-harmonic resonances will
be missed. These resonances could easily shift to the nearest integer harmonic order and produce
excessive harmonic distortion in the system. In general, a frequency step no larger than (fn/10) should
be adopted, although smaller intervals may be required in different systems depending on their
frequency response4. Some commercial tools include variable frequency step to optimise the
computation time. A balance must be sought between accuracy and computational burden.
In reality, each phase can present different harmonic impedances due to system asymmetries (see
section 2.2). Therefore, it is advisable to capture the harmonic impedance of all three phases in the
envelopes. This is important even if phase transposition is implemented, as it may not be effective at
high frequencies [29]. The three phase impedances can be combined into a single envelope for
convenience.
The use of a single envelope encompassing impedance for all harmonic orders should be strictly
avoided. Instead the harmonic spectrum should be divided into individual harmonic orders or into
4
A frequency step of (fn/100) is reported in [140] as typically used by a European TSO.
168
frequency bands, with some overlapping as discussed next. It should be noted that the
recommendations below are of general nature and mostly applicable to typical power systems
dominated by overhead lines. Systems with large share of cable circuits will typically exhibit resonances
at low harmonic orders and the engineer will need to take account of that characteristic behaviour when
defining the most appropriate frequency width and overlap of harmonic envelopes.
The choice of frequency bands is very much related to the frequency behaviour of the network. In
general, since low harmonic orders do not tend to exhibit large resonances, the impedance can be
adequately defined with individual harmonic order envelopes or with narrow frequency bands (e.g. two
to four harmonic orders); see Figure 5-4 (e). The use of envelopes covering a wide frequency range is
not advised for low harmonic orders unless resonances are observed in that frequency range as it will
typically result in unrealistic large impedance areas and, hence, too onerous conditions for assessment
(see Figure 5-4 (g)).
At higher frequencies, numerous resonances tend to appear at multiple frequencies and it is not possible
to capture this complex behaviour with narrow frequency band envelopes. In order to confidently capture
possible resonance shifts and/or modelling uncertainties it is necessary to widen the size of the
frequency bands. As far as practicable, the boundaries of the bands should avoid any resonant
frequency; see Figure 5-4 (f) and (g).
The introduction of some overlapping between subsequent envelopes is advised as a safety margin to
account for uncertainties such as small variations of system frequency, equipment tolerances,
parameter variation with temperature, modelling errors, etc. At low frequencies the uncertainties tend to
be low, therefore, widening the size of the envelope by ±½ harmonic order is normally adequate if there
are no significant resonances. For example, an envelope covering the 4th and 5th harmonic orders could
include data from the 3.5th to the 5.5th. At higher frequencies, the modelling uncertainty tends to increase
and accuracy tends to decrease, hence higher overlapping between envelopes is justified. As a general
guideline, a ±10% tolerance of the harmonic orders covered in the envelope can be used (although
larger or smaller tolerance may be required in different systems depending on their frequency response
and the modelling accuracy). This means that an envelope covering the harmonic range from 21 st to
34th should be extended to cover from 18.9th to 37.4th. As there will be an intrinsic overlap with the bands
for the higher and lower harmonic groups, the analysis of frequencies within the overlap range needs to
consider both envelopes (or a composite of the two).
The shape of the envelopes should be selected to minimise “empty” areas without realistic impedance
points. Polygons are very widely used as they can provide a very close fit to a set of calculated
impedance points. Typically, polygons can be defined by six to ten vertices, but higher number may be
required in cases. The larger the number of vertices, the best fit is achieved. The use of convex-hull
polygons provides the closest possible fit (i.e. minimum “empty” areas) at the expense of large number
of vertices, which can increase the computation burden for the assessing party.
An illustrative example comparing various envelope shapes is shown below. Figure 5-8 (a) compares
the fit of the generic envelopes described in section 5.4.1 for a set of calculated harmonic impedance
values in a defined frequency range (from 5th to 7th harmonic orders). In this example, it is clear that the
full-circle envelope produces an area that is too large and does not reflect the behaviour of the system
impedance in this frequency range (e.g. the circle suggests a large area of capacitive behaviour that is
not present in the calculated impedance data set). The size of “unrealistic” impedance area reduces
sequentially with the use of sector, basic polygon and convex-hull fit polygon.
The possibility of adapting polygons to reduce empty areas is shown in Figure 5-8 (b). In this example,
the basic polygon has been enhanced by cutting the empty top left corner, thus introducing an additional
vertex. The top-left area has been selected because it is normally one of the most critical regions for
compliance due to the low resistance values. The same approach can be adopted in all other corners if
necessary. In this example, the resulting number of vertices is 6, 7 and 27 for basic polygon, enhanced
polygon and convex-hull, respectively. The user needs to balance the requirements for accuracy (i.e.
minimum empty areas) and computational burden (i.e. number of vertices).
169
X [ohm] X [ohm]
300 250
200
200
X [ohm]
100 120
150 100
R [ohm]
0 80
0 100 200 300 400 500 600
100 60
-100
40
50 20
-200 R [ohm]
0
0 50 100
R [ohm]
-300 0
0 50 100 150 200 250
(b) Harmonic impedance envelopes: Basic Polygon
(a) Harmonic impedance envelopes: Circle,
(continuous line), Enhanced Polygon (dashed line),
Sector, Basic Polygon, Convex-Hull Polygon
Convex-Hull Polygon
Figure 5-8 Example of Harmonic Impedance Envelopes
For the purpose of harmonic filter design and definition of resonance between de-tuned filters and the
network impedance, the critical points on the harmonic impedance envelope always lie on the left-hand
boundaries of the envelope (TB553 – appendix 2.1 [23]). Thus, most care must be taken in the definition
of this boundary – especially with regard to the maximum angle and the maximum and minimum values
of impedance.
In some instances, different configurations of the power system may create clouds of impedance points
which lie in distinctly different areas of the R-X plane (for example, in the event of a major line outage).
In such cases it may be expedient to create two envelopes, each containing a distinct area of the R-X
plane, as being the best way to minimize the inclusion of unrealistic harmonic impedances. See example
in Figure 5-9.
Figure 5-9 Example of Two Impedance Envelopes for a Particular Band
170
Future-proofing
While it is possible to derive harmonic impedance envelopes to quite accurately represent the existing
or future planned states of the power network, treatment of developments in the more distant future is
more difficult.
Electrical power systems are highly likely to change significantly in nature over the lifetime of any new
plant being installed now. Renewable energy sources, distributed generation, electric vehicle charging,
and the introduction of very large numbers of associated power electronic converters may have a
significant and largely unpredictable impact on the network harmonic impedance.
Future changes in the network should not necessarily imply larger impedance envelopes – as networks
increase in capacity the impedances may well decrease and the harmonic impedance envelopes shrink
in some cases (note that the use of insulated cables can have the opposite effect by means of
introduction of low order harmonic resonances). This has often been the case with what are now
regarded as conventional power networks. But with new technological paradigms entering the industry,
future predications become more uncertain.
The expected life of the newly-connected plant and associated filters may be in the range of 30-40 years.
It is reasonable to expect that harmonic performance of the plant should continue to be acceptable
during much or all of that period. But if the network harmonic impedance changes substantially in the
future, such that it transgresses the impedance envelopes used for the design, then the harmonic
performance may exceed the agreed limits – through no fault of the plant owner. Any consequential
measures required for mitigation could be problematical – not only in terms of contractual/legal
discussions regarding responsibility for the cost, but also in terms of physical space for additional filtering
and the impact on the plant availability during modification works.
Consequently, there is an argument for including additional margins in the network impedance
envelopes to be used for design purposes, to create a degree of “future-proofing”. Such additional
margins, implying expansion of the areas covered by envelopes in the Z-plane, may well result in
additional filtering capacity being required and a higher degree of damping in the filter design, with
concomitant higher losses. Both imply an increased cost for the connection.
Whether or not to include margins for future-proofing in the definition of harmonic envelopes is not a
question on which there can be any general recommendation. Decisions must be made on a case-by-
case basis, depending on the contractual relationship of the parties involved and the assessed risk and
cost of any future modifications.
5.5 Use of Power Frequency Short-Circuit Thévenin Equivalent

The provision of system (harmonic) impedance data to third parties based on maximum and minimum
fault levels at the connection point has been common practice in the past, mostly driven by the lack of
detailed harmonic models by system owners/operators. A discussion about the adequacy of this
information for the purposes of harmonic studies follows.
Unlike some distribution systems, the harmonic impedance of transmission networks normally exhibits
a complex behaviour with sudden changes from inductive to capacitive mode, resulting in parallel and
series resonances where the network behaves in resistive mode. This complex behaviour is the result
of interactions between multiple (distributed or lumped) inductive and capacitive components in the
system at different frequencies and it is constantly changing according to the levels of demand, network
topology configuration (network sectionalising/re-configuration, circuit outages, etc.), generation
dispatch, etc.
A linear impedance representation derived from a power frequency short-circuit impedance cannot
reproduce this complex behaviour and, therefore, would only be adequate for frequencies below the first
resonant point, where the network still behaves in inductive mode.
A detailed investigation into the representation of the external grid impedance in harmonic studies for
offshore wind farms is reported in [141]. This work compares the effect of representing the external grid
by a simplified impedance model, based on short-circuit power and X/R ratio, versus envelope diagrams
given for a range of harmonic orders. Results of harmonic penetration studies performed for a generic
off-shore wind farm reveal that the simplified impedance model introduces significant errors in areas
where resonances occur and, therefore, is not suitable for the purposes of assessing compliance with
emission limits or design equipment for mitigation.
By means of illustration, Figure 5-10 shows the network impedance calculated at a network node in an
off-shore grid in Germany. The graph includes a frequency scan derived from a detailed full network
171
model and also a linear representation derived from the short-circuit impedance. The results obtained
with both approaches deviate before the main parallel resonance (at approximately 450 Hz) and do not
converge again. The limitations of this approach are obvious and can lead to over- or under-estimation
of the actual network harmonic impedance depending on the system characteristics and frequency
range.
500
Impedance in Ohm
400
Short-Circuit Equivalent
300
200
Detailed Frequency Dependent
100 Model
0
0 500 1000 1500 2000 2500
Frequency in Hz
100
75
50
Angle in degree
Short-Circuit Equivalent
25
0 Detailed Frequency Dependent

0 500 1000 1500 2000 2500 Model
-25
-50
-75
-100
Frequency in Hz
Figure 5-10 Comparison of Harmonic Impedance Derived from a Power Frequency Short-Circuit
Equivalent and from a Detailed Frequency-Dependent Model
It can be concluded that linear short-circuit equivalents are only suitable for frequencies well below the
first network resonance. In most cases, this frequency can be estimated using Equation 5.1 where f1 is
the fundamental frequency, SSC is the short-circuit power at the point of interest and Q is the reactive
power of the capacitive components directly (or close-by) connected to the point of interest. A more
detailed representation of the network is required if the scope of the harmonic analysis includes
frequencies close to (or above) f1.
𝑆𝑆𝐶
𝑓𝑟 = 𝑓1 √ Equation 5.1
𝑄
Reference [8] includes a discussion on this topic concluding that that the use of a simple inductance
based on the fault level contribution gives unsatisfactory results when applied to transmission systems.
A simple equivalent based on a “T circuit” is shown to give acceptable results reproducing the harmonic
impedance of a generic transmission system with a dominant resonance. While this approach is an
improvement from the basic simple inductance representation, caution is advised when applying it to
systems with more complex behaviour showing multiple dominant resonances (e.g. Figure 3-30, Figure
3-75, Figure 5-3 and Figure 5-10). The use of FDNE is recommended in those situations (see section
5.3.1).
Key Point:
 The practice of using short-circuit impedance to derive harmonic impedance at the PoC must be
discouraged for applications in meshed power systems.
172
5.6 Representation of Customer installation

When a customer is developing a new installation, harmonic studies need to be carried out in order to
demonstrate compliance with harmonic requirements in the applicable Grid Code or specified emission
limits prior to the commissioning, and, more importantly, to ensure compliance with these same limits
once the installation is in operation. The peculiarity of these studies relies on the scarcity of real data
available and the risk of oversizing mitigation solutions when a non-compliance appears during the
design stage.
The customer installation consists of active sources of harmonic distortion and passive elements. The
distortion sources are mainly in the generating units (e.g. PE converters) plus any other active
equipment, such as the FACTS devices required for continuous regulation of reactive power. Modelling
the harmonic sources with an acceptable level of accuracy has a high impact on the study results. If test
data is available, this should naturally be considered for the study; if this is not the case, simulation
models from the manufacturers should suffice. The modelling guidelines described in chapter 4 can be
applied here.
For wind generation, IEC 61400-21 establishes the harmonic information to be issued by a wind turbine
manufacturer as only the harmonic current emitted by the unit. A difficulty associated with this practice
is that the measured harmonic currents are often influenced by background distortion as well as grid
impedance at the test location. While the standard has helped to harmonise the format and detail of the
data, modelling the units as ideal current sources leads to gross overestimation of the installation’s
harmonic emission, as the internal damping of the generating units is being neglected (refer to section
4.2.1). Therefore, it is of utmost importance to model both generating units and other active sources
with a harmonic source (either voltage or current) accompanied by their internal harmonic impedance
plus any passive filter already considered in the initial design. A revision of the IEC standard has been
recently undertaken addressing the issues described above [142].
Another aspect to be considered is the phase angle of the distortion sources; while a simplified approach
applying the aggregation formula from the IEC 61000-3-6 [100] is always a possibility in the absence of
better information, considering more realistic data from manufacturers, in installations where there is a
great number of the same generating unit model, is expected to lead to harmonic cancellations between
the units [143]. This topic is discussed in detail in section 5.7 next.
The modelling of passive elements, such as MV cables, grid transformers and HV lines, between the
generating units and the connection point is of great importance, as these elements define how the new
installation interacts with the existing network. The modelling guidelines described in chapter 3 can be
applied here.
The first step is the modelling of the inter-array cabling that connects the generating units with one
another and to a step-up transformer. In an installation made of around a hundred units, the detailed
modelling of this MV cable system can be cumbersome. However, it has been seen that oversimplifying
this collector system will yield very pessimistic harmonic distortions [143]. There are a number of
aggregation methods described in the literature, such as [143] and [144], but they should be applied
with caution while understanding the associated risks and only when necessary. Validation of the
aggregated model by comparative assessment of frequency scans against the full detailed model is
recommended.
Once the MV collector system has been modelled, the remaining elements up to the PoC need to be
included, e.g. grid transformers and HV lines, either overhead or cables. While there are obvious
benefits in using detailed models that include the frequency-dependency of the equipment parameters,
this data is not usually available to the customer during the design stage. Even data at fundamental
frequency may not be fixed at the early stages of a project; thus, it is appropriate to analyse the sensitivity
of the harmonic results to changes in main parameters, such as short-circuit impedances of transformers
or capacitances of cables, within the tolerance limits established by applicable standards or in the
manufacturing contracts.
According to [145], the elements with greatest impact on the resulting distortion are the HV circuits,
especially with long underground cables, such as those of offshore wind farms. Here, even if no test
data is available, it may be beneficial to consider theoretical models for the frequency-dependency of
the cable’s resistance or take account of the parameter sensitivities described in section 3.2 of this TB.
An example of the effect of modelling frequency dependency of MV cables and transformers is illustrated
in Figure 5-11 below where (a) corresponds to “no frequency dependency” and (g) corresponds to
“frequency dependency”.
173
Figure 5-11 Impact of the Offshore Cable and Transformer Harmonic Losses Modelling on System
Resonance Damping at Offshore Substation (source [145])
Numerous publications (e.g. [69], [146], [147], [148]) have shown the dependency of harmonic distortion
to certain operational factors such MV collector network topology (e.g. radials in or out of service),
connection of WTGs, generation output, etc. Thus, it is recommended to incorporate these variables in
the studies in order to capture all practical combinations of resonance shifts and harmonic emissions
from the installation.
Once the customer’s installation has been accurately represented in the model, the external grid
information needs to be considered as well. Here, the customer will need to adapt to the amount of
information the network owner/operator can provide. However, it is worth noting that insufficient network
data will translate into a poor model and thus into a risky analysis. As an example, the study documented
in [141] demonstrates that the use of simplified grid impedance models based on short-circuit power
and X/R ratio can lead to very different amplitude and location of resonances when compared to a
detailed assessment using frequency dependent harmonic impedance envelopes. Therefore, it is
important to discuss with the relevant system owner/operator the need for accurate representation of
the external grid before investing in mitigation equipment.
Similar to the new installation, the external grid needs to be considered with all possible operation
scenarios that can have an impact on harmonic behaviour (see section 5.4 for a detailed discussion on
this topic).
It is demonstrated in TB553 (Appendix 2.1) [23] that the critical resonances between the external grid
and the customer’s installation always lie in the perimeter of the external grid harmonic impedance
envelope. Thus, it is generally acceptable to analyse interactions for Grid Code compliance considering
only the points in the perimeter of the grid impedance envelope. It should be noted that all points within
the perimeter need to be assessed, not only the vertices.
One part of a harmonic study is the evaluation of amplification of background harmonics caused by
interaction between the new installation and the external grid. As this is often more critical than harmonic
emission, it is important to consider realistic background distortion at the connection point. In cases
where representative measurements are not available (for example, the substation has not been built
yet), the system operator needs to estimate these values. Depending on the network topology and/or
available monitoring devices in the area, this may be difficult, and it is a topic requiring further research.
This is discussed in detail in section 5.8.
Harmonic studies are often performed to capture worst case operating conditions. As long as harmonic
limits are not exceeded this approach causes minimal workload and gives confidence in operation with
regards to harmonics. However, if limits are exceeded further considerations are needed to ascertain
whether mitigation measures are required. To justify this decision, a more detailed study on harmonics
would be beneficial. It may be the case that harmonic limits will only be exceeded if the new installation
is in an operating point which is rarely used. To perform detailed studies considering different operating
points, a full EMT model or frequency domain harmonic models considering different operating points
(e.g. with power steps of 10 %) are needed. This will normally result in a very large number of scenarios
to consider, but it becomes manageable when only applied to the areas of “non-compliance”. While
174
performing a detailed harmonic study, it is advisable to use accurate data for the phase angle of single
harmonics instead of general summation laws.
As it may be concluded from the paragraphs above, harmonic studies developed by a customer for a
new installation carry a certain amount of risk, mainly due to some data being quite preliminary. As a
result, it is easy to overestimate the emitted distortion and even to oversize mitigation solutions, such
as filters, that may even not be required later. While preliminary studies give a good indication of
potential harmonic issues, in some cases it is advisable to consider the level of uncertainties and to
discuss more detailed analysis or even a conditional connection with the system operator (as described
in Stage-3 of IEC 61000-3-6 [100]) to verify harmonic performance before committing to design a
particular mitigation solution.
5.7 Aggregation of Harmonic Sources

One of the challenges that arise when aggregating multiple non-linear devices relates to the fact that
the phase angles between the harmonic sources are usually not known. Depending on these angles,
the effects of multiple harmonic sources can either add or cancel out, therefore accurate estimation of
these effects is desirable to avoid over- or under-estimation of harmonic distortion levels leading to
equipment damage or unnecessary investment in mitigation.
When harmonic disturbances occur, the resultant harmonic voltage or current at the PoC is the vectorial
summation of the individual components of each harmonic source [149]. As is often the case, only the
harmonic amplitudes are recorded but the phase angle differences between the vectors is unknown, in
which case it is current standard practice to sum the amplitudes according to the general summation
law presented in the IEC 61000-3-6, shown in Equation 5.2 for harmonic voltages (and is analogous for
currents) [100].
𝛼
𝛼
𝑈ℎ = √∑ 𝑈ℎ𝑖 Equation 5.2
𝑖
Where
𝑈ℎ is the voltage magnitude at harmonic order ℎ
𝑈ℎ𝑖 is the magnitude of the various emission levels at order ℎ to be combined
𝛼 is an exponent as defined in Table 5-1
Table 5-1. IEC 61000-3-6 Summation Exponents for Harmonics [100]
Harmonic Order 𝜶
ℎ<5 1
5 ≤ ℎ ≤ 10 1.4
ℎ > 10 2
Key Point:
 The IEC 61000-3-6 general summation law may be applicable in many cases but is only
recommended for use when no better information is available. In addition, the alpha exponents
included in [100] were recommended based on information available at the time of publication,
mostly through analysis of rectifier bridges, and it is suggested that they may not be suitable
nowadays to capture the behaviour of modern harmonic sources [149].
The usefulness and practicality of the IEC 61000-3-6 general summation law is presented in [150], as
no phase angle information is required. Other contributions, such as [143] and [146], show that the use
of this summation law may yield very misleading results, and the application of the alpha exponent to
two harmonic sources with differing amounts of randomness, has been shown to be inconsistent with
measurement data. Furthermore, CIGRE TB 672 [79] concludes that if larger PV plants are built using
multiple individual PV inverters of the same model, the standard summation exponents (e.g. according
175
to IEC 61000-3-6) are not suitable and the harmonic currents of individual PV inverters should be added
up arithmetically independent of the harmonic order.
Results of extensive synchronised harmonic measurements performed in an off-shore wind farm (type
4 WTG) in Denmark are reported in [151]. It is shown that the use of the IEC general summation law
leads to incorrect results. In this particular case, harmonic current emissions between the 5 th and the
20th order are under-estimated at the PCC whereas harmonic currents between the 2nd and 4th and
above the 45th are over-estimated. A mismatch factor of 4 was observed for high frequency harmonic
currents. This work concludes that the use of the default exponents in the IEC general summation law
can lead to incorrect assessment of harmonic emissions with respect to Grid Code compliance and
recommends further work to better understand the phase angle behaviour of the power converter
harmonic emissions.
A recent contribution to the literature, [149], proposes that the general IEC 61000-3-6 summation law
be revised in order to adequately consider modern harmonic sources. This proposal suggests the use
of a uniform distribution of the difference in phase angles to calculate the probability that the magnitude
of the summation vector of two harmonics may exceed a given value. New alpha exponents are
recommended based on this work [149].
Another contribution [152] presents various methods for the summation and propagation of random
harmonic phasors, deriving analytical explicit expressions of currents and voltages for simple cases
where the injected currents exhibit uniform or Gaussian distributions and are independent. For more
complex cases, other techniques such as Monte Carlo simulation are required.
In summary, the application of the IEC 61000-3-6 general summation law is reiterated in [150] and [152],
with the advice that other exponents be used in specific cases where more detailed information is
available. A recommendation for a revision of the alpha exponents is included in [149]. For now, it is
recommended to apply caution and/or consult with converter manufacturers on adequate exponents to
be applied for the aggregation of harmonic emissions from multiple non-linear devices.
5.8 Background Harmonic Voltage Distortion

Background distortion is the existing harmonic content in the power system caused by the aggregated
emissions of non-linear devices at all voltage levels. This is a very important parameter that needs to
be accurately captured in harmonic studies, especially if amplification is expected. This amplification
could be the result of interactions between the customer facility and the grid, but it can also be caused
by changes in topology in the main grid (for example circuit outages, sectionalising stations, etc.) or the
connection of new long circuits. Typically, the process starts with a measurement campaign during a
representative period with as many monitoring devices as feasible in the area of interest. The main
challenge here is to determine the extent and duration of monitoring required in order to capture
representative conditions in the system. As is often the case, some nodes cannot be monitored (or have
not been built yet), therefore background estimation is needed. These aspects are discussed in the next
subsections.
5.8.1 Representative Period of Measurement

Measurements of harmonics are often performed before new assets are designed to evaluate the
harmonic distortion within the grid. Typically, these measurements serve as input to system or
component study either with the intention to design and rate these components or to predict the
harmonic distortion at other locations in the grid. Because of this, the current section gives
recommendations on analysis of measurements with focus on their use in harmonic studies. This
includes evaluation of harmonic levels, duration and impact of phase unbalance and of switching
operations in the grid.
5.8.1.1 Representative Period of Measurement

It is well known that harmonic levels change over time. Both short and periodic variations (e.g. changes
during summer and winter levels) are reported in the literature. It is important to take this variability into
account when performing harmonic studies. This is illustrated with the following examples.
Figure 5-12 shows the details of two different weeks of 7th harmonic measurement in a German
transmission substation. By analysing these curves, it is clear that the distortion is very different, and
that conclusions or requirements defined on the basis of either week may be wrong or not representative
of normal operating conditions. Table 5-2 gives the statistical analysis of the measured data. The 99th
percentile in week 1 is more than three times higher than in week 2.
176
Figure 5-12 Comparison of Two Different Weeks of 7th Harmonic Measurement in Germany
Table 5-2: Comparison of Different Statistic Values of Measurements of 7th Harmonic of Two Different
Weeks in Germany
Week 1 Week 2 Week 1/Week 2
Minimum 0.0247 % 0.0220 % 1.1246
Maximum 1.2753 % 0.3820 % 3.3388
Average 0.5042 % 0.1726 % 2.9210
Quantile 99 % 1.1221 % 0.3382 % 3.3184
Quantile 95 % 0.8937 % 0.2885 % 3.0982
Figure 5-13 shows the measurement of the 7th harmonic at the same location as the example above for
two different months. It illustrates that completely different results can be obtained. This can have
financial and performance implications, e.g. if design is based on measurements of the red curve
problems during operation may occur as equipment will be exposed to higher harmonic distortion than
specified or, alternatively, additional amplification caused by capacitive components may result in
breaches of statutory limits. Table 5-3 summarises the statistical analysis of the measured data. It clearly
indicates that the measured values in month 1 are much higher than in month 2. In extreme, the 99th
percentile of month 1 is 86 % higher than the equivalent for month 2 and thus could lead to inadequate
conclusions.
One observation can be made from this example: the statistical differences between the measured data
reduce as the measurement period increases.
177
Figure 5-13 Comparison of Two Different Months of 7th Harmonic Measurement in Germany
Table 5-3: Comparison of Different Statistical Values of Measurements of 7th Harmonic for Two Different
Months in Germany
Month 1 Month 2 Month 1/Month 2
Minimum 0.0149 % 0.0113 % 1.3200
Maximum 0.9590 % 0.5747 % 1.6687
Average 0.2989 % 0.1823 % 1.6394
Quantile 99 % 0.7829 % 0.4192 % 1.8675
Quantile 95 % 0.5996 % 0.3452 % 1.7370
The same trend is seen in long-term measurements. As an example, Figure 5-14 shows the phase-to-
ground 95th percentile values of the 11th and 13th harmonics in a 400 kV Danish substation. The
measuring period is 20 weeks and the 95th percentile levels are determined per week based on 10 min
aggregated values.
178
Figure 5-14 Phase-to-Ground 95th Percentile Values of the 11th and 13th Harmonics in a 400 kV Danish
Substation over 20 Weeks
The maximum and minimum weekly 95th percentile levels of harmonics are presented in Table 5-4.
Table 5-4 Maximum and Minimum 95th Percentile Levels of 11th and 13th Harmonics in a 400 kV Danish
Substation over 20 Weeks
Phase A Phase B Phase C
Minimum 11th 0.542 0.861 0.483
Maximum 11th 1.068 1.231 1.086
Minimum 13th 0.434 0.473 0.501
Maximum 13th 0.602 0.645 0.726
The ratios of the maximum and minimum weekly 95th percentile levels measured over the 20 weeks are
displayed in Table 5-5.
Table 5-5 Ratios of the Maximum and Minimum 95th Percentile Levels of 11th and 13th Harmonics in a 400
kV Danish Substation over 20 Weeks
Phase A Phase B Phase C
Ratio max. week/ min.

1.97 1.43 2.25
week 11th order
Ratio max. week/ min.

1.39 1.37 1.45
week 13th order
179
The variations in the levels of harmonics and therefore also the ratios between the maximum and
minimum 95th percentile values are highest for the 11th harmonic. The highest ratio over the 20 weeks
occur in phase C and is 2.25 and only 1.43 in phase B for the same harmonic. The 13th harmonic levels
are more constant over the 20 weeks with ratios changing from 1.37 to 1.45.
The reason for the large variation in level of harmonic is the dependency on operation conditions of the
power system, including changes in topology in the main grid (for example circuit outages, sectionalising
stations, etc.) and operating points of e.g. HVDC, wind power plants, arc furnace and production and
consumption schemes at distribution level. It cannot be guaranteed that the grid is operating under
representative or worst-case situations during the measuring period and it can be impossible to bring
the system to a known worst-case situation due to restriction in operating the grid during measurements,
for instance forced contingency situations. Therefore, knowledge of the system’s behaviour for instance
derived from use of simulation models combined with measurements is often required to ensure robust
designs.
Key Point:
 The levels of harmonics can vary significantly either over short time or according to season. For
some harmonics, the levels can be quite constant is relation to time and season where others
measured at the same location can vary significantly. Therefore, recommendation for a
“representative measuring period” is system dependent, but in general it can be stated that
measurements should be conducted for as long as possible, ideally for not less than three months,
including measurements of all three phases.
5.8.1.2 Evaluation of Harmonics

It is important to realise that there can be significant differences between for instance the 100 th percentile
value and 95th percentile values due to variation of harmonic levels over time. Both can be relevant
depending on the purpose. For instance, for harmonic system level planning studies or when issuing
limits for harmonics emissions, 95th percentile values are often used. For component rating the use of
the 100th percentile values may be necessary. To illustrate this point, an example of sorted 10 min
aggregated values measured over one week for the 11th, 13th and 23rd harmonics are presented for a
Danish 400 kV substation in Figure 5-15.
The 11th order 100th percentile is 20 % higher than the 95th percentile. At the 23rd harmonic; the level is
100 % higher. In Figure 5-16 the ratios between the 100th and 95th percentile levels evaluated for the
11th and 13th order harmonics measured over 20 weeks in a Danish 400 kV substation are shown.
As seen, the ratios can be very large indicating non-negligible difference between the 100th and 95th
percentile harmonic levels. The maximum ratio of 2.26 occurs for the 11th harmonic order in phase C in
week 9. The ratios at the 13th order harmonic are all around 1.2. It is noticed that the ratios change
significantly between phases and over time for the 11th order harmonic but are more constant at the 13th
order harmonic. For filter component rating where filters are tuned to certain harmonics this can have a
profound effect on the filter’s components.
Key Point:
 Harmonic 95th and 100th percentile levels can be significantly different and the ratio can vary
significantly over time. Depending on the type of study e.g. system planning or component rating
this should be considered.
180
Figure 5-15 Sorted 10 min. Aggregated Values Measured over One Week for the 11th, 13th and 23rd
Harmonics in a Danish 400 kV Substation
Figure 5-16 Ratios between the 100th and 95th Percentile Levels Evaluated for the 11th and 13th Order
Harmonics Measured over 20 Weeks in a Danish 400 kV Substation
181
5.8.1.3 Transmission Level Harmonic Voltage Measurements

Harmonic voltages obtained from measurements are often required as input for a harmonic study; extra
care must be taken when performing such measurements. The classical inductive or capacitive voltage
transformers are not capable of representing higher frequency components in the secondary of the
measuring circuit due to internal resonance in the voltage transformers. The inductive voltage
transformer can be used to some cut-off frequency; the lower the voltage level the higher the cut-off
frequency. The use of the capacitive voltage transformer for harmonic measurements is not
recommended at all. It is not within the scope of the document to discuss in detail how harmonic voltages
should be correctly measured at transmission level. The reader is advised to examine the details in
references listed in [153].
5.8.1.4 Impact of Network Topology Changes

The impact of network topology changes on the levels of harmonic distortion is illustrated in Figure 5-17,
which shows harmonic voltage distortion measured in five transmission stations in the North West of
Ireland over a period of seven consecutive days. These measurements show significant variations of
distortion in certain harmonic orders as a result of topology changes in the grid which, in turn, shift
harmonic resonant frequencies.
This first significant event (event #1) is the outage of a long cable circuit feeding Node#1. The
measurements show a significant step increase in distortion at the 7 th harmonic order in all other nodes,
suggesting a shift in resonance towards this harmonic order during this network topology. This event
resulted in the disconnection of Node #1 from the grid, hence the drop to zero in the measurements
during that period.
Event #2 is the disconnection of a short overhead line feeding into node #4. The measurements show
a drop in distortion at the 7th harmonic order and an increase at the 13th in all other nodes. Similarly to
Event #1, this event resulted in the disconnection of Node#4 from the grid, hence the absence of
measurements at that location.
182
Event #1 Event #2
NODE #1
NODE #2
NODE #3
NODE #4
NODE #5
Figure 5-17 Harmonic Voltage Distortion Measurements in five Substations in Ireland over a 7-
Day Period. Legend: 7th harmonic, 11th harmonic, 13th harmonic
A second example showing harmonic measurements at a wind farm located in the southern end of Brazil
is presented next. Figure 5-18 and Figure 5-19 show the 5th and 7th harmonic voltage measured over a
period of five months at the PCC for each of the three phases. The output power of the wind farm is also
shown for completeness. A sharp increase in distortion at both harmonic orders occurred on February
22nd, coinciding with the commissioning of a new 525 kV transmission line in the vicinity of this wind
farm. For the 7th harmonic the effect is more pronounced and it is probably caused by a system
impedance change and particularly by the “electrical proximity” of the non-linear loads from a
metropolitan and industrialized area to the PCC.
Other event impacting the harmonic distortion level at the PCC of the same wind farm is shown in Figure
5-20, covering a 9 hour period. The measurements in this figure show an abrupt reduction of the 5th
183
harmonic voltage at approximately 1:35 a.m., when two 525 kV transmission lines were disconnected
for voltage control. At 8:01 a.m., when the first of the two transmission lines was energized, a sharp
increase of voltage distortion at the 5th harmonic order was observed. The most interesting fact about
this event is that the two disconnected transmission circuits are about 1,000 km away from the wind
farm.
Figure 5-18 5th Harmonic: Voltage vs Power Figure 5-19 7th Harmonic: Voltage vs Power
Figure 5-20 Harmonic Voltage Distortion at a Wind Farm in Brazil
The two examples above illustrate the impact of switching operations over large areas of the power
system and highlights the difficulties in selecting a measurement period unaffected by these variations.
5.8.2 Estimation of Background Distortion

One of the most challenging aspects for System Owners/Operators when assessing the impact of new
harmonic generating customer connections is the estimation of the pre-connection background
distortion at the PoC and at any other nodes that could be affected by it. Unless the new customer
connects to a node where there are already reliable measurements (for a representative duration of time
– see section 5.8.1), the System Owner/Operator needs to estimate the background that the customer
will see at the PoC. It should be kept in mind that this is not an absolute value (as harmonic distortion
will dynamically change with system loading, generation operating points, system topology, etc) and a
clear distinction must be made between “representative” and “worst case” background distortion
expected at the PoC.
In practice, the connection of new customers (generation or demand) normally requires building new
circuits & stations and/or changing the topology of the grid. These topological changes can modify the
background distortion over a range of frequencies. Therefore, even if measurements are available, they
may not be representative for the future network topology.
184
Moreover, the problem becomes even more complex if other customer facilities have been recently
given emission limits but have not connected to the grid yet, or ramped to their contracted capacity,
hence their actual emissions are uncertain. In these cases, it is common practice to assume that the
new customers will eventually use their entire allocation of harmonic headroom. Therefore, the
“measured” background needs to be modified to capture the effect of the new contracted connections,
reducing the available headroom for future customers or network developments.
Accurate estimation of pre-connection background distortion is important for two reasons: (i) over-
estimation of background will result in very restrictive emission limits for the connectee, which may drive
unnecessary cost for mitigation (for example installation of harmonic filters), (ii) under-estimation of
background may result in over-relaxed emission limits to the new connectee that can cause breaches
in Planning Levels and damage to equipment, requiring a mitigation solution to be implemented by the
System Owner/Operator and possibly compensation for damages. Either way, there are costs
implications that are normally carried to the end customer.
The problem of prediction of background distortion can be divided into two parts:
1. Estimation of background in existing nodes where measurements are not available.
2. Prediction of harmonic distortion in future nodes or modification in existing nodes subject to significant
topology changes.
5.8.2.1 Nodes without Measurements

The estimation of background distortion in existing substations for which measurements are not
available is basically a problem of state estimation with a number of “known variables” (nodes with
harmonic measurements) and a number of “unknown variables” (nodes without harmonic
measurements). The challenge arises from the fact that the number of “unknown variables” is normally
significantly larger than the number of “known variables” and the problem becomes undetermined
system with more unknowns than equations. Various approaches can be found in literature ranging from
the conventional (and popular) “trial and error” approach where harmonic distortion is iteratively
calculated to fit measured values, to more sophisticated Harmonic State-Estimation (HSE) techniques.
State estimation techniques have been successfully applied to Power Systems in real time since the
late 1960’s to extract accurate estimates of variables with noisy measurements as well as to find
variables at non-monitored buses [154]. The basic idea is to “fine-tune” state variables by minimising
the error using statistical techniques. Power System State Estimation is normally based on the
assumption that the current and voltage waveforms are pure sinusoids with constant frequency and
magnitude. These techniques have been expanded to estimate state variables also at harmonic
frequencies (HSE) and several methods have been proposed [154] – [164]. HSE is a reverse process
of harmonic penetration studies, which analyse the response of the power system to given injection
current sources. The HSE uses the harmonic measurements at selected busbars to identify the location
and magnitude of harmonic sources. In addition, HSE is capable of providing information at locations
not monitored. In any HSE method, the quality of the estimates is a function of the number and location
of the observation points. Common to all is the requirement for an accurate harmonic model of the
system. An overview of some HSE methods described in literature follows.
A practical method for identifying harmonic sources is proposed and field tested in [155]. This method,
although successful, requires a large number of monitoring devices to be placed throughout the network.
An alternative method proposed in [156] uses neural networks to make initial estimates of harmonic
sources and, hence, requires less number of measurements. Seeking a balance between accuracy and
cost, some methods for optimising the number and location of the measurement devices are proposed
in [157]-[159].
In [160], the application of a HSE method with Global Positioning System (GPS) synchronised data is
illustrated. This paper shows the validation of the method using only partial and selected measured data
in Japan. This HSE method has also been used to successfully predict the harmonic voltage distortion
after a change of the actual power system configuration. Similarly, [161], describes the application of
HSE in the 132kV and 230kV network in Iran.
While many academic papers can be found in literature dealing with HSE, there is very little work
published describing practical applications by System Owners/Operators. A rare example comes from
the Spanish TSO (REE), where an integrated PQ monitoring and analysis system has been developed
[165]. This system integrates the following functionalities: (i) PQ Monitoring Equipment, (ii) Centralized
Power Quality Database, (iii) Built-in Data Analyser, (iv) Automatic Reporting and (v) Power Quality
185
State Estimator (ETAD). An overview of the system is shown in Figure 5-21. The State Estimator module
(ETAD) is described in detail in Appendix F.
Measurements Unknowns
Estimation
Figure 5-21 Power Quality Monitoring and Analysis System by Red Eléctrica de España (REE) [165]
From the literature review and TSO’s experience, it can be concluded that there are feasible solution
methods to tackle the problem of harmonic estimation in nodes without measurements. However, these
methods have not been widely adopted by System Owners/Operators, possibly due to the effort required
in implementing and integrating them with their standard harmonic analysis tools and models. Availability
of HSE functionality in commercial harmonic analysis packages would be a valuable step forward for
the accurate assessment of harmonic distortion in power systems and to assist in fair allocation of
emission limits to new customer connections.
5.8.2.2 Prediction of Harmonic Distortion in Future Nodes or Modification in

Existing Nodes Subject to Significant Topology Changes
The estimation of harmonic distortion in future nodes or in existing nodes affected by network changes
is a trivial problem if a detailed network model with accurate representation of harmonic sources is
available. This is a natural follow up step from the estimation process described in the previous sub-
section. In this case, under the assumption that the harmonic sources are not affected by the topological
change, the network model can be modified and the new harmonic distortion can be calculated by
running harmonic penetration studies. Examples of network changes that can modify the background
distortion in a given node include the connection of cables, lines, reactive compensation devices,
harmonic filters, transformers, etc, or network topology changes such as sectionalising, loop-in new
circuits, decommissioning circuits, etc.
In cases where details of the harmonic sources are not available (or reliable), the “future” harmonic
distortion in existing or new nodes needs to be estimated. This can be done by calculating harmonic
impedance gains as a ratio between “existing” and “future” self-impedances at each node of interest
[166]. The harmonic impedances are obtained from frequency scans of the network model “before” and
“after” implementation of the network change. The calculated harmonic impedance gains are then
applied to the existing background distortion (measured or estimated) at the node of interest.
In case of future nodes radially connected to an existing node, background estimation is straightforward
using transfer gains derived from self and mutual impedances between the new node and the existing
node to which is directly connected to. In case where the future node is meshed or looped into an
existing circuit, the calculation becomes more complex since harmonic flows from multiple sources need
to be considered now along with phase angles and possible summation or cancelation effects. This
recent work [167] discusses the strengths, limitations and applicability of the transfer gains method for
various network topologies. It is concluded that the use of transfer gains produces accurate results in
radial network topologies, however it normally overestimates results in meshed circuits. The simplicity
186
of the method and the fact that results err on the side of safety makes this approach attractive for most
harmonic studies. However, careful consideration is recommended to any series resonance points in
the frequency scan which can lead to the computation of unrealistically large gain values.
A common method used to estimate the modification of background at a given node caused by the
connection of customer’s equipment is the “Constant Voltage behind Thévenin Equivalent” approach
[23], [168]. In this approach, the background harmonic distortion is assumed constant for any network
condition while the harmonic impedance of the system, seen at the PoC, is normally given as a range
of values (see section 5.4) to capture different operating conditions (including circuit outages). This
approach is illustrated in Figure 5-22 below, which represents the connection of an offshore wind farm.
In this example, V_back represents the pre-existing background distortion at the PoC (measured or
estimated), Z_Net represents the harmonic impedance of the transmission grid at the PoC (normally
given as a family of harmonic impedance envelopes for a range of frequencies) and Z_WF represents
the harmonic impedance of the customer facility and comprises all cables, transformers, reactive
compensation devices, etc at the customer facility beyond the PoC. Z_WF can also vary depending on
the operating conditions; for example the number of collector arrays and WTG in service.
Figure 5-22 Thévenin Equivalent Representation for Assessing Modification of Background Distortion
due to a Customer Connection [168]
The modification of the pre-existing harmonic voltage distortion at the PoC caused by the passive
components of the customer’s installation is given by Equation 5-3, where “h” denotes each harmonic
order and the impedances are given as complex numbers. The modification factor (kh) can, theoretically,
take any value depending on the relative values of the system and customer impedances. As such, the
effect of the customer installation on the distortion at the PoC could be: (i) amplification of pre-existing
background if kh > 1, (ii) reduction of pre-existing background if k h < 1 or (iii) no change in pre-existing
background if kh = 1.
𝑉ℎ−𝑏𝑎𝑐𝑘𝑔𝑟𝑜𝑢𝑛𝑑 (𝑝𝑜𝑠𝑡−𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛) 𝑍ℎ_𝑊𝐹
𝑘ℎ = = Equation 5-3
𝑉ℎ−𝑏𝑎𝑐𝑘𝑔𝑟𝑜𝑢𝑛𝑑 (𝑝𝑟𝑒−𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛) 𝑍ℎ_𝑊𝐹 + 𝑍ℎ_𝑁𝐸𝑇
The difficulties typically encountered with the “Constant Voltage behind Thévenin Equivalent” approach
relate to the fact that the network conditions leading to the highest amplification factors do not
necessarily coincide with the network conditions for which the background distortion was measured or
estimated. As a result, excessive (and sometimes unrealistic) “post-connection” background distortion
can be calculated leading to investment in mitigation solutions. Caution is advised in these situations to
avoid unnecessary costs. A possible approach, when feasible, can be to estimate the pre-connection
background distortion for each network condition and, therefore, provide a family of Thévenin
Equivalents with linked values of network harmonic impedance and associated background distortion.
This approach, while theoretically correct, may not be feasible in many cases due to the large volume
of data to be processed, high uncertainty in calculation assumptions or the need to reserve headroom
for future growth. A balance must be sought between practicality in the exchange of information between
customer and System Owner/Operator and the justification for costs in mitigation. In some cases, it
might be beneficial to re-examine critical conditions arising from worst-case assumptions.
187
5.9 Summary
Due to the proliferation of PE device connections in power systems, harmonic studies are gaining
importance as part of the planning and design process for new customer connections. In setting up the
simulation models, it is important to focus on the purpose of the analysis as well as to understand the
limitations of the methods, tools and models. The overall objective of the assessment is to define the
appropriate requirements for the non-linear connection, taking into account foreseeable conditions in
the power system as well as present and future uncertainties over the life time of the installation while
avoiding, insofar as possible, over-conservative requirements that may lead to investment in
questionable mitigation measures or over-design of harmonic filters. This chapter provides a general
discussion and practical guidelines on key aspects that bear consideration during harmonic impact
assessments of new customer connections. It is emphasised that no two power systems are the same
and regulatory and contractual arrangements may differ from country to country, therefore caution and
good engineering judgement is advised in implementation of the above.
An overview of the most common types of harmonic studies is provided, from a network owner/operator
perspective as welll as from a new customer or equipment manufacturer point of view. Key inputs and
outcomes of each type of analysis are highlighted.
Important considerations regarding the representation of the power system and setting up simulation
cases include the extent of the model and the number and type of scenarios to be considered.
General guidelines on the minimum number of nodes to be accurately represented as well as how to
derive accurate equivalents of the rest of the network are provided.
A discussion on the scenarios and contingencies that can be considered in the harmonic studies,
depending on the type and objective, is included in this chapter. It has been shown that the maximum
harmonic distortion does not always correlate with peak or valley system demand levels. Therefore, it is
imporant to incorporate intermediate demand levels in the studies in order to capture the changes in
system resonances and emissions from all non-linear devices. Outages of network components that
can affect the harmonic behaviour of the system at the PoC need to be captured. These can include
most circuits in the vicinity of the PoC as well as harmonic filters, capacitor banks, FACTS devices and
HVDC installations. Detailed guidelines are provided.
The need to capture all possible combinations of key variables than can affect the harmonic performance
of the system normally leads to unmanageable number of cases to run and analyse. Thus, automation
is normally adopted to perform the studies. Caution should be exercised to ensure that all scenarios
included in the assessment are realistic and comply with the utilty operational standards.
The concept of harmonic impedance loci and envelopes is explained and illustrated with a real example.
The most common types of envelopes used in industry are discussed, highlighting their strengths and
limitations. Practical considerations for creating harmonic impedance envelopes are provided.
It has been shown that the use of network equivalents derived from short-circuit impedance calculations
at power frequency are not adequate to reproduce the complex frequency dependent behaviour of
typical trasmission systems. Its use is strongly discouraged.
Some considerations for the representation of the customer installation when performing design and
compliance assessment studies are discussed. The need for detailed models of the power electronic
converters is emphasised while the use of “ideal harmonic current source” approach is shown to be
overly pessimistic. How to aggregate the harmonic emissions from multiple sources within the new
installation is a critical assumption that can have high cost implications in terms of non-compliance.
Detailed discussions with vendors are recommended to gain a better understanding of the prevailing
phase angle and expected cancelations. Furthermore, aggregation of multiple turbines, transformers
and MV cables into a single equivalent can be beneficial in some large installations to reduce
computational burden.
It is concluded that harmonic studies developed by a customer for a new installation carry a certain
amount of risk. As a result, it can be easy to overestimate the prospective emissions and even to
oversize mitigation solutions, such as filters, that may even not be required later on. While preliminary
studies give a good indication of potential harmonic issues, in some cases it is advisable to consider the
level of uncertainties and to discuss further detailed studies or special conditions for the connection with
the System Owner/Operator such as conditional connections.
A review of literature suggests that the commonly used IEC general summation law [100] may not be
adequate to capture the phase angle behaviour of emissions in modern power converters. Caution is
188
advised as well as consultation with converter manufacturers on adequate exponents to be applied for
the aggregation of harmonic emissions from multiple non-linear devices.
The topic of duration of background measurements is controversial. IEC documents recommend, as a
minimum, seven days of continuous operation. In principle, the objective is to measure for a long enough
period to derive a representative signature of the background distortion at that location. Since harmonic
distortion is continuously changing, depending on a large number of variables, pre-determining the
minimum duration requirements is a challenging (if not impossible) task. Therefore, recommendation for
a “representative measuring period” is system dependent, but in general it can be stated that measuring
should be conducted for as long as possible, ideally for not less than three months, including
measurements of all three phases.
As it is often the case, some nodes cannot be monitored (or have not been built yet), therefore
background estimation is needed. Various estimation methods are discussed, highlighting their
applicability and limitations.
During the discussions leading to the compilation of this chapter, the following areas where future work
is required were identified:
 Summation of harmonic sources: achieve a better understanding of the phase angle behaviour of
modern converters (deterministic vs stochastic), effects of converter control, modulation scheme,
operation point, background distortion, etc. The overall objective should be to gather enough
information, based on field measurements and theoretical analysis, to propose a robust and realistic
alternative to improve the IEC summation law currently adopted in most harmonics analysis.
 Develop practical methods for accurate aggregation of wind farm components (wind turbine
generators, transformers, cables, etc) into a single frequency dependent equivalent.
 Develop practical methods for the estimation of background distortion in meshed network topologies
where availability of measurements is limited.
189
190
6. Conclusions
Power systems globally are experiencing a transition towards decarbonisation of electricity production
through large-scale deployment of renewable energy sources (RES), which are gradually displacing
conventional thermal plant. This changing environment is seeing a proliferation of power electronic
converters connecting at all voltage levels in power systems, namely RES, FACTS devices, HVDC
systems, domestic load, etc. These devices are highly non-linear and emit harmonics at the point of
connection, but also modify pre-existing harmonics in the network. In addition, increased installation of
HVAC cables is creating system resonances at frequencies close to the characteristic emissions from
these non-linear devices. As a result, many power systems are experiencing an increase in harmonic
distortion. Power quality issues associated with harmonics in power systems are becoming more
pronounced and are driving a new focus towards the need to undertake detailed analysis at the planning
stages in order to ensure adherence to statutory limits.
This Technical Brochure has been compiled drawing expertise from a worldwide membership base and
provides comprehensive guidelines for practising power system engineers when they need to perform
harmonic distortion assessments. The document covers the modelling of the most common network
components and discusses key features that need to be considered in the assessments. The focus of
this Technical Brochure is on frequency-domain modelling for steady-state AC harmonic analysis in
power systems, typically in the range from power frequency up to the 50th harmonic (2.5 kHz in 50 Hz
systems or 3 kHz in 60 Hz systems), consistent with typical power quality assessments. The approach
and modelling guidelines provided are reasonably valid up to the 100 th harmonic if specialized studies
are required. The document is not intended to cover guidelines for analysing transient issues or
controller harmonic instability, although most of the guidelines for modelling network components
equally apply to those areas also.
The main focus of this Technical Brochure is on practical aspects of modelling for direct application in
the planning process of connecting a new customer (non-linear installation) to the power system, or
when introducing a change to the system as part of asset replacement or system expansion. These
guidelines will be valuable in the definition of harmonic performance specifications for new HVDC
converters, FACTS devices or other non-linear installations. They will also assist connectees when
modelling their installation to assess or demonstrate compliance with the emission limits provided by
the System Owner/Operator and to investigate and specify mitigation measures such as harmonic filters.
Furthermore, this Technical Brochure can also be used post-commissioning for any incident
investigation or to assist resolution of customer complaints via modelling and analysis.
By reviewing existing literature on modelling techniques and approaches for each network component,
complemented with comprehensive analysis and expertise from the members, this Technical Brochure
reflects the current state-of-the-art and best practice in network modelling for harmonic studies.
An overview of available solution methods for analysis of harmonics in power systems, including
frequency-domain, time-domain, harmonic-domain and hybrid methods is given. A comprehensive
analysis of the strengths and limitations of each method is provided. It is concluded that frequency-
domain solution methods provide adequate results for most applications, along with straight-forward
modelling of harmonic injections and frequency dependence, making this study domain accessible and
attractive to many engineers. Some cases where frequency-domain approaches are not suitable are
identified and recommendations are provided for the most appropriate alternative solution method.
Issues associated with asymmetry in network components are discussed and illustrated with an
example. Recommendations for three-phase modelling and application of unbalanced solution methods
are included.
A comprehensive review of models available to represent the most common passive network
components in harmonics studies is given. The relevant input data and level of detail required to
represent each component are presented and discussed. Different modelling options are assessed in a
benchmark model as well as in real system models to illustrate the effects and consequences of each
type of model in the context of system-wide studies. When available, some models have been validated
against measurement data and recommendations are provided. The importance of representing the
frequency dependency of resistive losses is stressed and illustrated with examples. Furthermore, the
importance and difficulties of accurate load modelling are highlighted. The general load composition and
its harmonic behaviour is normally one of the biggest unknowns when performing harmonic
assessments in power systems. Various approaches for load aggregation are presented and discussed,
however this is an area where further work is required to improve and validate load models considering
191
the on-going changes in load composition, i.e. the move towards more power electronics based, as well
as seasonal and daily variations.
A chapter is devoted to the accurate representation of harmonic generating equipment in the frequency
domain. The most common sources of harmonic distortion are presented, discussing the mechanisms
of harmonic generation for each device, aspects that influence their harmonic performance and
elements that need to be considered in the model. When possible, limitations of the frequency-domain
models are highlighted, and recommendations are provided for more sophisticated methods of analysis.
It is emphasized that the harmonic performance of these devices is generally complex and dependent
on many factors such as converter topology, control strategy, operating point, external grid, etc. As such,
manufacturer-specific models need to be adopted. Recommendations on the structure of the models
and the features that need to be captured are provided. Emphasis is placed on the need to capture not
only the harmonic current/voltage emissions but also the harmonic impedance of the devices. A common
model structure is proposed based on a Norton/Thévenin equivalent circuit. This approach aligns with
the recent IEC TR 61400-21-3 ED1 [62], applicable to the modelling of wind turbines. The same
principles can be extended to model all other non-linear devices, especially those based on power
electronic converters. In addition to the obvious advantages of standardisation, the proposed structure
offers the benefit of providing a comprehensive characterization of the non-linear device without the
need to disclose proprietary design features.
A final chapter is devoted to examining practical aspects, other than the modelling of each individual
component, that need to be considered when performing harmonic studies related to the connection of
non-linear devices to the power grid. In setting up the simulation models, it is important to focus on the
purpose of the analysis as well as to understand the limitations of the methods, tools and models. An
overview of the most common types of harmonic studies is provided, from a network owner/operator
perspective as well as from a new customer or equipment manufacturer point of view. Key inputs and
outcomes of each type of analysis are highlighted. For example, the overall objective of a typical system
owner/operator study is the definition of appropriate emission limit requirements for a new non-linear
connection, considering foreseeable conditions in the power system as well as present and future
uncertainties over the lifetime of the installation. The focus should be on avoiding, insofar as possible,
over-conservative requirements that may lead to investment in questionable mitigation measures or
over-design of harmonic filters while ensuring adequate system performance in all transmission and
distribution nodes, in adherence with statutory limits.
Recommendations on the extent of the model, scenarios and contingencies that can be considered in
the harmonic studies, depending on the type and objective, are provided. It has been shown that the
maximum harmonic distortion does not always correlate with maximum or minimum system demand
levels. Therefore, it is imporant to incorporate intermediate demand levels in the studies in order to
capture the changes in system resonances and emissions from all non-linear devices. The concept of
harmonic impedance loci and envelopes is explained and illustrated with a real example. The most
common types of envelopes used in industry are discussed, highlighting their strengths and limitations.
Practical considerations for creating harmonic impedance envelopes are provided. The use of network
equivalents derived from short-circuit impedance calculations at power frequency is shown to be not
adequate for meshed power systems and, therefore, this practice is strongly discouraged. Finally, the
background distortion at the point of connection is a key input in most harmonic studies.
Recommendations for an adequate period of measurements and data analysis are given. As is often
the case, some nodes cannot be monitored, therefore background estimation is needed. Various
estimation methods are discussed, highlighting their applicability and limitations.
The discussions and recommendations in this Technical Brochure address perspectives as seen from
(i) the system owner/operator; and (ii) the new connectee, with the overall objective of minimising risk
of equipment failure due to excessive harmonic distortion as well as avoiding unnecessary investment
in mitigation. While finding the right balance is never an easy task, it is hoped that the considerations
presented in this document will aid the engineer in making informed decisions.
Future work on the subject of harmonic modelling and analysis may expand the effort already performed
in the following directions:
 Load modelling: Characterisation and development of accurate aggregation and representation
methods. Special focus should be placed on capturing the transition in the general load
composition towards a more predominant power-electronic based type. Interactions between the
PE-based load impedance and the system harmonic impedance as well as characterisation of
harmonic emissions need to be accounted for in the enhanced load models. Validation through
lab tests and field measurements is recommended.
192
 Validation of power converter models with field measurements. Accurate methods to separate the
effect of system harmonic impedance and background distortion from the converter emissions.
 Summation of harmonic sources: achieve a better understanding of the phase angle behaviour
of modern converters (deterministic vs. stochastic), effects of converter control, modulation
scheme, operation point, background distortion, etc. The overall objective should be to gather
enough information, based on field measurements and theoretical analysis, to propose a robust
and realistic alternative to improve the IEC summation law currently adopted in most harmonics
analysis.
 Develop practical methods for accurate aggregation of wind or solar farm components (wind
turbine generators, PV converters, transformers, cables, etc.) into a single frequency-dependent
equivalent.
 Develop practical methods for the estimation of background distortion in meshed network
topologies where availability of measurements is limited.
193
194
7. Bibliography/references
[1] CIGRE JTF 36.05.02/14.03.03, “AC System Modelling For AC Filter Design – An Overview of Impedance
Modelling”, Electra 164, pp 133-151, February 1996.
[2] CIGRE WG CC-02 (CIGRE 36.05/CIRED 2), “Guide for assessing the network harmonic impedance”,
Electra No. 167, pp 97-131, July 1996.
[3] Arrillaga, J.; and Watson, N. R.; “Power System Harmonic Analysis“, 2 nd Edition, John Wiley & Sons, Ltd,
2003.
[4] IEEE Task Force on Harmonics Modelling and Simulation; Medina, A.; Segundo-Ramirez, J.; Ribeiro, P.;
Xu, W.; Lian, K.L.; Chang, G.W.; Dinavahi, V.; Watson, N.R., “Harmonic analysis in frequency and time
domain“, IEEE Transactions on Power Delivery, 28, (4), 2013.
[5] CIGRE TB 754 “AC Side Harmonics and Appropriate Harmonic Limits for VSC HVDC”, WG B4.67,
February 2019.
[6] Herraiz, S.; Sainz, L.; Clua, J., “Review of harmonic load flow formulations”, IEEE Transactions on Power
Delivery, vol. 18, pp. 1079-1087, July 2003.
[7] Testa, A.; Akram, M.F.; Burch, R.; Carpinelli, G.; Chang, G.; Dinavahi, V.; Hatziadoniu, C.; Grady, W.M.;
Gunther, E.; Halpin, M.; Lehn, P.; Liu, Y.; Langella, R.; Lowenstein, M.; Medina, A.; Ortmeyer, T.; Ranade,
S.; Ribeiro, P.; Watson, N.; Wikston, J.; Xu, W., "Interharmonics: Theory and Modelling," IEEE
Transactions on Power Delivery, vol.22, no.4, pp.2335-2348, Oct. 2007.
[8] IEEE Power Engineering Society Task Force on Harmonics Modelling and Simulation, “Tutorial on
Harmonics Modelling and Simulation“, TP-125-0, 1998.
[9] Dinh, N.; and Arrillaga, J.; “A salient-pole generator model for harmonic analysis “, IEEE Transactions on
Power Systems, 2001, 16, (4), pp. 609-615.
[10] Rittiger, J.; Kulicke, B., "Calculation of HVDC-converter harmonics in frequency domain with regard to
asymmetries and comparison with time domain simulations," IEEE Transactions on Power Delivery, vol.10,
no.4, pp.1944-1949, Oct 1995.
[11] Froebel, A.; Vick, R. "Chosen aspects for harmonic analysis in distribution networks", 22 nd International
Conference on Electricity Distribution, CIRED 13, June 2013.
[12] Harmonic Analysis in Frequency and Time Domain, IEEE Task Force, IEEE Power Delivery, July 2013
[13] J. Arrillaga and C. D. Callaghan, ''Three Phase AC-DC Load and Harmonic Flows,” IEEE Trans. on Power
Delivery, Vol. 6, No.1, January 1991, pp. 238-244.
[14] J.G. Mayordomo, L.F. Beites, R. Asensi, M. Izzeddine, A.H. Bayo, “INTAR, A Powerful Tool to Simulate
Harmonics under Balanced and Unbalanced Conditions”. International Conference on Electricity
Distribution. CIRED, 1997.
[15] IEEE Standard 1124, IEEE Guide for analysis and definition of DC side harmonic performance of HVDC
transmission systems
[16] Fuchs, E.F.; “Power Quality in Power Systems and Electrical Machines”, Elsevier Academic Press, 2008.
[17] C. F. Jensen; Ł. H. Koeewiak; Z. Emin; “Amplification of Harmonic Background Distortion in Wind Power
Plants with Long High Voltage Connections”, CIGRE Paris Session 2016
[18] C. F. Jensen; “Harmonic Background Amplification in Long Asymmetrical High Voltage Cable Systems”,
IPST 2017
[19] Lombard, X.; Mahseredjian, J.; Lefebvre, S.; Kieny, C., "Implementation of a new harmonic initialization
method in the EMTP," IEEE Transactions on Power Delivery, vol.10, no.3, pp.1343-1352, July 1995.
[20] Semlyen, A.; Medina, A.; "Computation of the periodic steady state in systems with nonlinear components
using a hybrid time and frequency domain methodology," IEEE Transactions on Power Systems, vol.10,
no.3, pp.1498-1504, July 1995.
[21] J. de Jesus Chavez, J.; Ramirez, A.I.; Dinavahi, V.; Iravani, R.; Martinez, J.A.; Jatskevich, J.; Chang, G.W.,
"Interfacing Techniques for Time-Domain and Frequency-Domain Simulation Methods," IEEE
Transactions on Power Delivery, vol.25, no.3, pp.1796-1807, July 2010.
[22] Arrillaga, J.; Medina, A.; Lisboa, M.L.V.; Cavia, M.A.; Sanchez, P., “The harmonic domain. A frame of
reference for power system harmonic analysis“, IEEE Transactions on Power Systems, 1995, 10, (1), pp.
433-440.
[23] CIGRE TB 553 “Special Aspects of AC Filter Design for HVDC Systems”, October 2013.
[24] J. Arrillaga, B.C. Smith, N.R. Watson, A. R. Wood, “Power System Harmonic Analysis”. John Wiley & Sons,
195
1997.
[25] E. Acha, M. Madrigal, Power Systems Harmonics – Computer Modelling and Analysis. John Wiley & Sons,
2001.
[26] CIGRE WG 36-05. “Harmonics, characteristic parameters, methods of study, estimates of existing values
in the network”. Electra 77, July 1981.
[27] IEEE Task Force on Harmonics Modelling and Simulation. Modelling and Simulation of the Propagation of
Harmonics in Electric Power Networks. Part I: Concepts, Models and Simulation Techniques. and Part II:
Sample Systems and Examples IEEE Transactions on Power Delivery, Vol 11, No 1, January 1996.
[28] W. T. Wiechowski, Harmonics in transmission power systems. Aalborg: Institut for Energiteknik, Aalborg
Universitet, 2006.
[29] J. Arrillaga, E. Acha, T.J. Densem, P.S. Bodger. Ineffectiveness of Transmission Line Transpositions at
Harmonic Frequencies. Proc IEE, 123C(2). 1986.
[30] Wenner, F., “A Method of Measuring Earth Resistivity,” Report No. 258, Bulletin of Bureau of Standards,
Vol. 12, No. 3, October 11, 1915
[31] Palmer, L. S., “Examples of geotechnical surveys,” Proceedings of the IEE, Paper 2791-M, vol. 106, pp.
231–244, June 1959
[32] Deri, A., Tervan, A. Deri, G. Tevan, A. Semlyen and A. Castanheira, "The Complex Ground Return Plane.
A Simplified Model for Homogeneous and Multi-Layer Earth Return," IEEE Power Engineering Review,
vol. PER-1, no. 8, pp. 31-32, Aug. 1981.
[33] H. W. Dommel, EMTP Theory Book, Prepared for Bonneville Power Administration, Portland, Oregon,
1995
[34] CIGRE TB 531 “Cable systems electrical characteristics”, 2013.
[35] CIGRE TB 556, “Power system technical performance issues related to the application of long HVAC
cables”, 2013.
[36] Ł. H. Kocewiak, B. Gustavsen, “Impact of Cable Impedance Modelling Assumptions on Harmonic Losses
in Offshore Wind Power Plants”, CIGRE Paris Session, August 2018. Paper C4-309.
[37] F.F. Da Silva, C.L. Bak, “Electromagnetic Transients in Power Cables”, Springer-Verlag London, 2013
[38] A. Ametani, “A general formulation of impedance and admittance of cables”, IEEE Transactions on Power
Apparatus and Systems, vol. PAS-99, no.3, 1980.
[39] Ł. H. Kocewiak, “Harmonics in large offshore wind farms,” PhD Thesis, 2012, pp. 332, 978-87-92846-04-
4.
[40] M.H.J. Bollen, S.M. Gargari, “Harmonic resonances due to transmission cables”, in CIGRE Belgium
Conference, Brussels, 2014.
[41] A. Pagnetti, “Cable modelling for electromagnetic transients in power systems”, PhD thesis, Université
Blaise Pascal – Clermont II, Clermont-Ferrand, France, 2012.
[42] A. Ametani, T. Ohno, N. Nagaoka, “Cable System Transients: Theory, Modelling and Simulation”, Wiley
IEEE Press, 2015.
[43] U. R. Patel and B. Gustavsen and P. Triverio, “An Equivalent Surface Current Approach for the
Computation of the Series Impedance of Power Cables with Inclusion of Skin and Proximity Effects”, IEEE
Transactions on Power Delivery, 2013, volume 28, number 4, pp 2474-2482.
[44] C. H. Chien and R. W. G. Bucknall, “Harmonic Calculations of Proximity Effect on Impedance
Characteristics in Subsea Power Transmission Cables”, IEEE Transactions on Power Delivery, 2009,
volume 24, number 4, pp 2150-2158.
[45] U. R. Patel and B. Gustavsen and P. Triverio, “Proximity-Aware Calculation of Cable Series Impedance
for Systems of Solid and Hollow Conductors”, IEEE Transactions on Power Delivery, 2014, volume 29,
number 5, pp 2101-2109.
[46] CIGRE WG B1.03, TB 272, “Large cross-sections and composite screens design”, Electra, June 2005.
[47] Ł. H. Kocewiak, B. Gustavsen and Andrzej Hołdyk, "Wind Power Plant Transmission System Modelling for
Harmonic Propagation and Small-signal Stability Analysis", International Workshop on Large-Scale
Integration of Wind Power into Power Systems, and Transmission Networks for Offshore Wind Farms,
Berlin, 2017.
[48] Benato and S. D. Sessa, "A New Multiconductor Cell Three-Dimension Matrix-Based Analysis Applied to
a Three-Core Armoured Cable," in IEEE Transactions on Power Delivery, vol. 33, no. 4, pp. 1636-1646,
Aug. 2018.
196
[49] IEEE Std. 399, IEEE Recommended Practice for Industrial and Commercial Power Systems Analysis,
1997.
[50] G. Funk, T. Hantel: Frequenzabhängigkeit der Betriebsmittel von Drehstromnetzen, etz Archiv Band 9 Heft
11, 1987.
[51] Ivan Arana Aristi, “Switching overvoltages in offshore wind power grids. Measurements, modelling and
validation in time and frequency domain”. PhD Thesis, Technical University of Denmark. November 2011.
[52] CIGRE WG 13-05, “The Calculation of Switching Surges. Part II: Network Representation for Energisation
and Re-Energisation Studies on Lines Fed by an Inductive Source”. Electra No 32, pp 17-42, 1974.
[53] Bjørn Gustavsen, “A Filtering Approach for Merging Transformer High-Frequency Models With 50/60-Hz
Low-Frequency Models”. IEEE Transactions on Power Delivery, Volume: 30, Issue: 3, June 2015, PP1420-
1428.
[54] CIGRE TB 566 “Modelling and Aggregation of Loads in Flexible Power Networks”, WG C4.605, February,
2014.
[55] IEEE Task Force on Harmonic Modelling and Simulation, “Impact of Aggregate Linear Load Modelling on
Harmonic analysis: A comparison of Common Practice and Analytical Methods”, IEEE Transactions on
Power Delivery, Vol 18, No. 2, April 2003
[56] Capasso et al “Representation of large asynchronous Load Areas for Harmonic Penetration studies”,
International Conference on Harmonics in Power Systems, 1992.
[57] R. Lamedica, A. Prudenzi, E. Tironi, D. Zaninelli, “A Model of Large Load Areas for Harmonic Studies in
Distribution Networks”, IEEE Transaction on Power Delivery, Vol. 12, Nº 1, January 1997.
[58] Zhang, X.P.; and Handschin, E.; “Frequency-dependent simple harmonic model of synchronous
machines“, IEEE Power Engineering Review, May 2000, pp. 58-60.
[59] Powertech Labs 1994 report 267D766
[60] Hydro-Québec, Direction MAINTENANCE DES ÉQUIPEMENTS, ET SÉCURITÉ DES BARRABES,
SERVICE ESSAIS ET ÉTUDES TECHNIQUES, DIVISION ÉTUDES TECHNIQUES - TRANSPORT,
Omer Bourgault, et al, September 1995.
[61] J.B. Ward “Equivalent Circuits for Power Flow Studies”, AIEE Winter General Meeting, 1949.
[62] IEC TR 61400-21-3 ED1: Wind energy generation systems – Part 21-3: Wind turbine harmonic model and
its application.
[63] CIGRE TB 719 “Power Quality and EMC issues with future electricity networks”. JWG C4.24/CIRED,
March 2018.
[64] Snyder, M. "Development of Simplified Models of Doubly-Fed Induction Generators (DFIG): A contribution
towards standardized models for voltage and transient stability analysis", Master Thesis, Chalmers
University of Technology, 2012.
[65] Muller, S.; Deicke, M.; De Doncker, R.W., "Doubly fed induction generator systems for wind
turbines," IEEE Industry Applications Magazine, vol. 8, no. 3, pp.26-33, May/Jun 2002.
[66] Larose, C.; Gagnon, R.; Prud’Homme, P.; Fecteau, M.; Asmine, M.; “Type-III Wind Power Plant Harmonic
Emissions - Field measurements and aggregation guidelines for adequate representation of harmonics”,
Proceedings 11th International Workshop on Large-Scale Integration of Wind Power into Power Systems
as well as on Transmission Networks for Offshore Wind Power Plants (Wind Integration Workshop),
Lisbon, 2012.
[67] Bradt, M.; Badrzadeh, B.; Camm, E.; Mueller, D.; Schoene, J.; Siebert, T.; Smith, T.; Starke, M.; Walling,
R., "Harmonics and resonance issues in wind power plants," Transmission and Distribution Conference
and Exposition (T&D), 2012 IEEE PES, vol., no., pp.1-8, 2012.
[68] Lui, S.Y.; Pimenta, C.M.; Pereira, H.A.; Mendes, V.F.; Mendonca, G.A., "Aggregated DIFG wind farm
harmonic propagation analysis", Anais do XIX Congresso Brasileiro de Automática, CBA 2012.
[69] King, R.; Ekanayake, J.B., "Harmonic modelling of offshore wind farms," IEEE Power and Energy Society
General Meeting, July 2010.
[70] Hernandez, E.; Madrigal, M., "A Step Forward in the Modelling of the Doubly-fed Induction Machine for
Harmonic Analysis" IEEE Transactions on Energy Conversion, vol.29, no.1, pp.149-157, March 2014.
[71] E. Guest, T. Rasmussen, K. H. Jensen, “An Impedance-Based Active Filter for Harmonic Damping by
Type-IV Wind Turbines”. 17th Wind Integration Workshop. 17 - 19 October 2018. Stockholm, Sweden.
Submission-ID WIW18-20.
[72] M. Lehmann, M. Pieschel, M. Juamperez, K. Kabel, Ł. H. Kocewiak, S. Sahukari, “Active Filtering with
Large-Scale STATCOM for the Integration of Offshore Wind Power”. 17 th Wind Integration Workshop. 17
197
- 19 October 2018. Stockholm, Sweden. Submission-ID WIW-61.

[73] M. Sascha, J. Meyer, P. Schegner, "Characterization of small photovoltaic inverters for harmonic
modelling." Harmonics and Quality of Power (ICHQP), 2014 IEEE 16th International Conference on. IEEE,
2014.
[74] Enslin, Johan HR, and Peter JM Heskes. "Harmonic interaction between a large number of distributed
power inverters and the distribution network." IEEE transactions on power electronics 19.6 (2004): 1586-
1593.
[75] Bosman, A. J. A., Cobben, J. F. G., Myrzik, J. M. A., & Kling, W. L. (2006, September). ”Harmonic modelling
of solar inverters and their interaction with the distribution grid”. In Universities Power Engineering
Conference, 2006. UPEC'06. Proceedings of the 41st International (Vol. 3, pp. 991-995). IEEE.
[76] Wang, F., Duarte, J. L., Hendrix, M. A., & Ribeiro, P. F. (2011). “Modelling and analysis of grid harmonic
distortion impact of aggregated DG inverters”. IEEE Transactions on Power Electronics, 26(3), 786-797.
[77] Limsakul, C., et al. "An impedance model of a PV grid-connected system." Photovoltaic Specialists
Conference, 2008. PVSC'08. 33rd IEEE. IEEE, 2008.
[78] Aprilia, E. C., Ćuk, V., Cobben, J. F. G., Ribeiro, P. F., & Kling, W. L. (2012, October). “Modelling the
frequency response of photovoltaic inverters”. In Innovative Smart Grid Technologies (ISGT Europe), 2012
3rd IEEE PES International Conference and Exhibition on (pp. 1-5). IEEE.
[79] CIGRE TB 672 “Power Quality Aspects of Solar Power”, by JWG C4/C6.29. December 2016.
[80] CIGRE TB 139 “Guide to the Specification and Design Evaluation of AC Filters for HVDC Systems”, by
WG 14.30, April 1999.
[81] CIGRE WG 14.03, “AC Harmonic Filter and Reactive Compensation for HVDC with Particular Reference
to Non-Characteristic Harmonics”, Complement to the paper published in Electra No. 63 (1979), CIGRÉ
Technical Brochure No. 65, June 1990.
[82] CIGRE TB 65. “AC harmonic filters and reactive compensation for HVDC with particular reference to non-
characteristic harmonics,” WG 14.03, June. 1990.
[83] CIGRE TB 143 “Cross-Modulation of Harmonics in HVDC Schemes”. WG 14.25, June 1999.
[84] IEC/TR 62001-1/4:2016 “High-voltage direct current (HVDC) systems - Guidebook to the specification and
design evaluation of A.C. filters. Part 1: Overview. Part 2: Performance. Part 3: Modelling. Part 4:
Equipment”.
[85] Arrillaga, J.; Liu, Y.H.; and Watson, N. R.; “Flexible Power Transmission: The HVDC Options”, John Wiley
& Sons, Ltd, 2007.
[86] Arrillaga, J.; “High Voltage Direct Current Transmission”, IET, 1998.
[87] CIGRE TB 25 “Static Var Compensators”, prepared by WG 38-01, Task Force no 2 on SVC, 1986.
[88] L.J. Bohmann and R.H. Lasseter, “Equivalent Circuit for Frequency Response of a Static VAR
Compensator”, IEEE Trans. PWRS, 1(4), 68-74,1986.
[89] CIGRE TB 123 “Thyristor Controlled Series Compensation”, prepared by WG 14.18, December 1997.
[90] CIGRE TB 144 “Static Synchronous Compensator (STATCOM)”, prepared by WG 14.19, August 2000.
[91] CIGRE TB 663 “Guidelines for the Procurement and Testing of STATCOMs”, prepared by WG B4.53,
August 2016.
[92] CIGRE TB 223 “Active Filters in HVDC Applications”, Working Group 14.28. April 2003.
[93] M. Lehmann, M. Pieschel, Ł. H. Kocewiak, M. Juamperez, S. Sahukari, K. Kabel, “Active Filtering with
Large-Scale STATCOM for the Integration of Offshore Wind Power,” in Proc. The 17th International
Workshop on Large-Scale Integration of Wind Power into Power Systems as well as Transmission
Networks for Offshore Wind Farms, , Energynautics GmbH, 17-19 October 2018, Stockholm, Sweden.
[94] H. Wu ; X. Wang ; Ł. Kocewiak ; L. Harnefors, “AC Impedance Modelling of Modular Multilevel Converters
and Two-Level Voltage-Source Converters: Similarities and Differences,” in Proc. the 19th Workshop on
Control and Modelling for Power Electronics, IEEE, 25-28 June 2018, Padova, Italy.
[95] Ł. H. Kocewiak, B. Laudal Øhlenschlæger Kramer, O. Holmstrøm, K. Høj Jensen, L. Shuai, “Resonance
Damping in Array Cable Systems by Wind Turbine Active Filtering in Large Offshore Wind Power
Plants,” IET Renewable Power Generation, Institution of Engineering and Technology, 6 July 2017,
Volume 11, Issue 7, Page(s) 1069-1077.
[96] S. K. Chaudhary, C. Lascu, R. Teodorescu, Ł. H. Kocewiak, “Voltage Feedback based Harmonic
Compensation for an Offshore Wind Power Plant,” in Proc. IEEE International Conference on Power
Electronics, Drives and Energy Systems (PEDES), IEEE Press, 14-17 December 2016, Trivandrum, India.
198
[97] Ł. H. Kocewiak, M. Gautschi, L. Zeni, B. Hesselbæk, N. Barberis Negra, T. Stybe Sørensen,

B. Blaumeiser, S. Vogelsanger, “Power Quality Improvement of Wind Power Plants by Active Filters
Embedded in STATCOMs,” in Proc. The 15th International Workshop on Large-Scale Integration of Wind
Power into Power Systems as well as Transmission Networks for Offshore Wind Farms, Energynautics
GmbH, 15-17 November 2016, Vienna, Austria.
[98] E. Guest, T. W. Rasmussen, K. H. Jensen, “An Impedance-Based Active Filter for Harmonic Damping by
Type-IV Wind Turbines,” in Proc. The 17th International Workshop on Large-Scale Integration of Wind
Power into Power Systems as well as Transmission Networks for Offshore Wind Farms, Energynautics
GmbH, 17-19 October 2018, Stockholm, Sweden.
[99] Davor Vujatovic, Kah Leong Koo and Zia Emin: “Methodology of Calculating Harmonic Distortion from
Multiple Sources”, International Conference on Power Systems Transients (IPST2015) in Cavtat, Croatia
June 15-18, 2015.
[100] IEC T/R 61000-3-6, Electromagnetic Compatibility (EMC) Part 3: Limits – Section 6: Assessment of
emission limits for distorting loads in MV and HV power systems – Basic EMC publication. First edition,
1996-10.
[101] Bernhard Hemmer, Andrea Mariscotti and Dieter Wuergler: “Recommendations for the Calculation of the
Total Disturbing Return Current From Electric Traction Vehicles”, IEEE Transactions on Power Delivery,
Vol. 19, No. 3, July 2004.
[102] D. Arlt, “Power Supply of Electric Arc Furnaces and Requirements of the Supply network”; 12th
International Seminar “Electrical Engineering of Arc Furnaces”; Steel Academy, Duesseldorf May 2015.
[103] D. Arlt, “Seminar electroheat; University of Applied Sciences Duesseldorf”.
[104] CIGRE TB 237: “Static synchronous compensator (STATCOM) for arc furnace and flicker compensation”,
prepared by WG B4.19. December 2003.
[105] H. M. Mahmoud, E. El-Din Sharouda, “Improvement of power system harmonics level generated from the
electric arc furnaces (EAF) to the acceptable level by using shunt passive filters”. CIGRE Paris session
2018. Paper C4-107.
[106] J-P. Hasler, J. Sneed, M. Holmberg, J. Lund, M. Naslund “Power quality analysis and IEC standard
evaluation using measurements and simulations in a STATCOM application”. CIGRE Paris session 2018.
Paper C4-114.
[107] C. Lazaroiu, D. Zaninelli , “DC arc furnace modelling for Power Quality”, U.P.B. Sci. Bull., Series C, Vol.
72, Iss. 1, 2010.
[108] G. Carpinelli, F. Iacovone, A. Russo, and P. Varilone, “Chaos-Based Modelling of DC Arc Furnaces for
Power Quality Issues”, IEEE Transactions on Power Delivery, Vol 19, Issue. 4, pp 1869 – 1876. October
2004.
[109] W. Mack Grady, “Understanding Power System Harmonics”, on-line course notes on power quality, Baylor
University, Waco, Texas, April, 2012.
[110] D.E. Rice, “A Detailed Analysis of Six-Pulse Converter Harmonic Currents”, IEEE Transactions on Industry
Applications, 1994, Vol. 30, No. 2, pp. 294-304.
[111] S. Hansen, L. Asiminoaei, F. Blaabjerg, “Simple and Advanced Methods for Calculating Six-Pulse Diode
Rectifier Line-Side Harmonics”, Record of the IEEE Industry Applications Conference, 38th IAS Annual
Meeting, 2003, Vol. 3, pp. 2056-2062.
[112] M. Malinowski, Ph.D. Thesis, “Sensorless Control Strategies for Three-Phase PWM Rectifiers”, Institute
of Control and Industrial Electronics, Warsaw, Poland, 2001.
[113] A.von Jouanne, B. Banerjee, “Assessment of Voltage Unbalance”, IEEE Transactions on Power Delivery,
Vol. 16, No. 4, October 2001.
[114] A. Kamenka, «Six tough topics about harmonic distortion and Power Quality indices in electric power
systems», Schaffner Group, 2014
[115] T. Makris, «Harmonics demystified», Ecolibrium, pp.32-38, 2010
[116] EPECentre (Electric Power Engineering Centre), «Outcomes from the EPECentre Workshop on Power
Quality in Future Electrical Networks», University of Canterbury, 2009
[117] D.P. Manjure, E.B. Makram, “Impact of unbalance on power system harmonics”, 10th International
Conference on Harmonics and Power Quality, 2002, Vol. 1, pp. 328-333.
[118] K. Lee, T.M. Jahns, W.E. Berkopec, T.A. Lipo, “Closed-Form Analysis of Adjustable-Speed Drive
Performance Under Input-Voltage Unbalance and Sag Conditions’, IEEE Transactions on Industry
Applications, Vol. 42, No. 3, May/June 2006, pp. 733-741.
[119] K. Lee, G. Venkataramanan, T.M. Jahns, “Source Current Harmonic Analysis of Adjustable Speed Drives
199
Under Input Voltage Unbalance and Sag Conditions:, IEEE Transactions on Power Delivery, Vol. 21, No.
2, pp. 567-576.
[120] Y. Sun, C. Dai, J. Li, J. Yong, “Frequency-domain harmonic matrix model for three-phase diode-bridge
rectifier”, IET Gener. Transm. Distrib., 2016, Vol. 10, Iss. 7, pp. 1605–1614
[121] T. Alexander, S.M. Lingham, T.S. Davies, P. Iwanciw, “Experience With 3-Phase Sinusoidal Regenerative
Front-Ends”, Fifth International Conference on Power Electronics and Variable Speed Drives, 1994, pp.
246-250.
[122] J. Kikuchi, T.A. Lipo, “Three-Phase PWM Boost-Buck Rectifiers With Power-Regenerating Capability”,
IEEE Transactions on Industry Applications, Vol. 38, No. 5, September/October 2002, pp. 1361-1369.
[123] L. Moran, J. Espinoza, M. Ortiz, J. Rodriguez, J. Dixon, “Practical Problems Associated With the Operation
of ASDs Based on Active Front End Converters in Power Distribution Systems”, Record of the IEEE
Industry Applications Conference, 39th Annual Meeting, 2004, Vol. 4, pp. 2568-2572.
[124] J.R. Rodriguez, J.W. Dixon, J.R. Espinosa, J. Pontt, “PWM Regenerative Rectifiers: State of the Art”, IEEE
Transactions on Industrial Electronics, Vol. 52, No. 1, Feb. 2005, pp. 5-22.
[125] B.K. Bose, “Modern Power Electronics and AC Drives”, 2002, The University of Tennessee, Knoxville,
ISBN 0-13-016743-6.
[126] L. Li, D. Czarkowski, Y. Liu, P. Pillay, “Multilevel Selective Harmonic Elimination PWM Technique in Series-
Connected Voltage Inverters”, IEEE Transactions on Industry Applications, Vol, 36, No. 1,
January/February 2000.
[127] M. Chomat, L. Schreier, J. Bendl, “Control of Active Front-End Rectifier in Electric Drive Under Unbalanced
Voltage Supply in Transient States”, PRZEGLᾼD ELECTROTECHNICZNY (Electrical Review), ISSN
0033-2097, R. 88, No. 1a, 2012
[128] X.H. Wu, S.K. Panda, J.X. Xu, “Design of Plug-In Repetitive Control Scheme for Eliminating Supply-Side
Current Harmonics of Three-Phase PWM Boost Rectifiers Under Generalized Supply Voltage Conditions”,
IEEE Transactions on Power Electronics, 2010, Vol. 25, Issue 7, pp. 1800-1810
[129] A.V. Stankovic, K. Chen, “A New Control Method for Input-Output Harmonic Elimination of the PWM Boost-
Type Rectifier Under Extreme Unbalanced Operating Conditions”, IEEE Transactions
[130] A.V. Stankovic, T.A. Lipo, “A Novel Control Method for Input Output Harmonic Elimination of the PWM
Boost Type Rectifier Under Unbalanced Operation Conditions”, IEEE APEC 2000, Feb. 6-10, pp. 413-418.
[131] B. Wu, J. Pontt, J. Rodriguez, S. Bernet, S. Kouro, “Current-Source Converter and Cycloconverter
Topologies for Industrial Medium-Voltage Drives”, IEEE Transactions on Industrial Electronics, Vol. 55,
No. 7, July 2008, pp. 2786-2797
[132] K.A. Puskarich, W.E. Reid, P. Hamer, “Harmonic Experiences With a Large Load Commutated Inverter
Drive”, IEEE Paper No. PCIC-99-18
[133] B. Ozpineci, L.M. Tolbert, “Cycloconverters”, An on-line tutorial for the IEEE Power Electronics Society,
2001
[134] A. Symonds and M. Laylabadi, “Cycloconverter Drives in Mining Applications”, IEEE Industry Applications
Magazine, November/December 2015, pp 36-46.
[135] CIGRE TB 568. “Transformer Energization in Power Systems: A Study Guide”, WG C4.307, February
2014.
[136] Y. Vernay, B. Gustavsen, “Application of Frequency-Dependent Network Equivalents for EMTP Simulation
of Transformer Inrush Current in Large Networks”. 2013 International Conference on Power Systems
Transients (IPST2013) in Vancouver, Canada July 18-20, 2013.
[137] B. Gustavsen and A. Semlyen, “Rational approximation of frequency domain responses by vector fitting”,
IEEE Trans. Power Delivery, vol. 14, no. 3, pp. 1052-1061, July 1999.
[138] B. Gustavsen and A. Semlyen, “Enforcing passivity for admittance matrices approximated by rational
functions”, IEEE Trans. Power Systems, vol. 16, pp. 97-104, Feb. 2001.
[139] G. Alvarez-Cordero, A. Bachiller Soler, A. Gomez-Exposito, J. A. Rosendo Macias, C. Gomez-Simon, “A
Methodology for Harmonic Impedance in Large Power Systems. Application to the Filters of a VSC”.
CIGRE Paris Session 2012. Paper C4-112.
[140] R. de Groot, F. van Erp, K. Jansen, J. van Waes, M. Hap, L. Thielman. “Method for Harmonic and TOV
Connection Impact Assessment of Offshore Wind Power Plants – Part I: Harmonic Distortion”. 17th Wind
Integration Workshop. 17 - 19 October 2018. Stockholm, Sweden. Submission-ID WIW18-124.
[141] V. Myagkov, L. Petersen, S. Burutxaga Laza, F. Iov, L. H. Kocewiak, “Parametric Variation for Detailed
Model of External Grid in Offshore Wind Farms”. Proceedings of the 13 th International Workshop on Large-
scale Integration of Wind Power Into Power Systems As Well As on Transmission Networks for Offshore
200
Wind Power Plants (wiw2014). Berlin, November 2014.

[142] B. Andresen, P. E. Sørensen, F. Santjer, and J. Niiranen, "Overview, status and outline of the new revision
for the IEC 61400 -21 – Measurement and assessment of power quality characteristics of grid connected
wind turbines," in 12th International Workshop on Large-Scale Integration of Wind Power into Power
Systems as well as on Transmission Networks for Offshore Wind Power Plants, London, 2013, pp. 399-
404.
[143] K.L. Koo, Z. Emin, “Comparative evaluation of power quality modelling approaches for offshore wind
farms”, 5th IET International Conference on Renewable Power Generation (RPG) 2016, 2016 p. 52 (7).
[144] F. Ghassemi and K.L. Koo, “Equivalent Network for Wind Farm Harmonic Assessments”, IEEE
Transactions on Power Delivery, Vol. 25, Issue 3, PP 1808 - 1815 , July 2010.
[145] L. H. Kocewiak, B. Gustavsen and A. Holdyk. “Wind Power Plant Transmission System Modelling for
Harmonic Propagation and Small Signal Stability”, Proceedings in 16 th Wind Integration Workshop,
Berlin/Germany, from 25 to 27 October 2017.
[146] L. Shuai ,Ł. H. Kocewiak, K. H. Jensen, "Application of Type 4 Wind Turbine Harmonic Model for Wind
Power Plant Harmonic Study", International Workshop on Large-Scale Integration of Wind Power into
Power Systems, and Transmission Networks for Offshore Wind Farms, Vienna, 2016.
[147] A. Holdyk, J. Holboell, E. Koldby, A. Jensen, “Influence of offshore wind farms layout on electrical
resonances”. CIGRE Paris session 2014. Paper C4-310.
[148] K. Leong Koo, Z. Emin, “Harmonic Specification for Offshore Wind Farm Connections – Determination,
Issues and Recommendations”. Dublin CIGRE Symposium, June 2017. Paper 96.
[149] Xiao, Y.; Yang, X., "Harmonic Summation and Assessment Based on Probability Distribution," IEEE
Transactions on Power Delivery, vol.27, no.2, pp.1030,1032, April 2012.
[150] Wikston, J. “Harmonic summation for multiple arc furnaces” in IEEE Power Engineering Society Winter
Meeting, pp.1072-1075 vol.2, 2002.
[151] M. Eltouki, T. W. Rasmussen, E. Guest, L. Shuai, Ł. Kocewiak. “Analysis of Harmonic Summation in Wind
Power Plants Based on Harmonic Phase Modelling and Measurements”. 17th Wind Integration Workshop.
17 - 19 October 2018. Stockholm, Sweden. Submission-ID WIW18-164.
[152] Baghzouz, Y. ; Burch, R. F. ; Capasso, A. ; Cavallini, A . ; Emanuel, A. E. ; Halpin, M. ; Langella, R. ;
Montanari, G. ; Olejniczak, K. J. ; Ribeiro, P. ; Rios-Marcuello, S. ; Ruggiero, F. ; Thallam, R. ; Testa, A. ;
Verde, P.; "Time-Varying Harmonics: Part II—Harmonic Summation and Propagation" IEEE Transactions
on Power Delivery, vol.17, no.1, January 2002.
[153] CIGRE TB 596, “Guidelines for Power Quality Monitoring: Measurement Locations, Processing and
Presentation of Data”. Joint Working Group CIGRE/CIRED C4.112. October 2014.
[154] A. P. S. Meliopulos, F. Zhang, S. Zelingher, “Power System Harmonic State Estimation”, IEEE
Transactions on Power Delivery, Vol 9, No 3, July 1994. PP 1701-1709.
[155] G. T. Heydt, "Identification of harmonic sources by a state estimation technique," IEEE Transactions on
Power Delivery, vol. 4, no. 1, pp. 569-576, January 1989.
[156] R. K. Hartana, G. G. Richards, “Harmonic Source Monitoring and Identification Using Neural Networks”,
IEEE Transactions on Power Systems, Vol 5, No 4, November 1990, pp 1098-1104.
[157] J. E. Farach, W. M. Grady, A. Arapostathis, “An Optimal Procedure for Placing Sensors and Estimating
Locations of Harmonic Sources in Power Systems”. IEEE Transactions on Power Delivery, Vol 8, No 3,
July 1993, pp 1303-1310.
[158] C. Madtharad, S. Premrudeepreechacharn, N. R. Watson, R. S. Udom, “An optimal measurement
placement method for power system harmonic state estimation”. IEEE Transactions on Power Delivery,
Vol 20, No 2, April 2005, pp 1514-1521.
[159] M. S. Rad, H. Mokhtaru, H. Karimi, “An Optimal Measurement Placement method for Power System
Harmonics State Estimation”, 2012 International Conference and Exposition on Electrical and Power
Engineering (EPE 2012). 25-27 October, Iasi (Romania)
[160] N. Kanao, H. Yamashita, H. Yanagida, M. Mizukami, Y. Hayashi, J. Matsuki, “Power System Harmonic
Analysis Using State-Estimation Method for Japanese Field Data”. IEEE Transactions on Power Delivery,
Vol 20, No 2, April 2005. Pp 970-977.
[161] M. Moghadasian, H. Mokhtari, A. Baladi “Power System Harmonic State Estimation using WLS and SVD:
A Practical Approach”. 14th International Conference on Harmonics and Quality of Power - ICHQP 2010.
26-29 Sept. 2010, Bergamo (Italy).
[162] L. F. Beites, M. Alvarez and A. Diaz García, "Sensor optimum location algorithm for estimating harmonic
sources injection in electrical networks," in International Conference on Renewable Energies and Power
201
Quality (ICREPQ’14) Renewable Energy and Power Quality Journal (RE\&PQJ), 2014.
[163] Z.-P. Du, J. Arrillaga and N. Watson, "Continuous harmonic state estimation of power systems," IEE
Proceedings- Generation, Transmission and Distribution, vol. 143, no. 4, pp. 329-336, July 1996.
[164] K. Yukihira, “Development of a Program for Estimating Harmonic Current Sources”, CRIEPI Report
T89043, May 1999.
[165] A. Diaz García, L. F. Beites, M. Alvarez and L. Soto Cano, “Power Quality Monitoring and Assessment in
the Spanish Transmission System” CIGRÉ Paris, 2016. C4-108.
[166] Y. Fillion, S. Deschanvres, “Background harmonic amplifications within offshore wind farm connection
projects”, presented at the International Conference on Power Systems Transients (IPST2015) in Cavtat,
Croatia June 15-18, 2015.
[167] D.H. Mills, Z. Emin, D. O’Brien, M. Val Escudero and C. F. Jensen, “Calculation Method Selection for
Harmonic Voltage Distortion Gains”. CIGRE Symposium in Aalborg (Denmark) 4th – 7th June 2019. Paper
65.
[168] A.J. Hernandez M., S. Wijesinghe, A. Shafiu, “Hamonic Amplification of the 576 MW Gwynt-y-Môr Offshore
Wind Power Plant”, presented at the 11th International Workshop on Large-Scale Integration of Wind
Power into Power Systems, Lisbon, Portugal, 2012.
202
Appendix A. Component representation for

frequency-domain studies
It is well known from power frequency system studies that a three-phase network can be decomposed
into its equivalent sequence network to ease the calculation burden. This method can also be used at
higher frequencies and is hence equally effective for harmonic propagation studies. If the power system
component under study is symmetrical, meaning equal self- and mutual impedance between all three
phases at the h’th harmonic, the resulting system sequence impedance matrix Z 012(h) takes the following
form:
𝑍0 (ℎ) 0 0
𝑍012 (ℎ) = [ 0 𝑍1 (ℎ) 0 ] App Equation A.1
0 0 𝑍2 (ℎ)
where 𝑍0 (ℎ), 𝑍1 (ℎ), and 𝑍2 (ℎ) are the zero, positive and negative sequence impedances at the h’th
harmonic.
The decoupled sequence representation shown in Equation A.1 is the preferred choice for the
representation of electrical components in larger networks for many utilities.
No component in the transmission network is, however, perfectly balanced. Depending on the design of
the component, the self- and mutual impedances between phases can be significantly different. For
such a system, a full phase representation is needed and will have the following form:
𝑍𝑎𝑎 (ℎ) 𝑍𝑎𝑏 𝑍𝑎𝑐

App Equation A.2
𝑍𝑎𝑏𝑐 (ℎ) = [ 𝑍𝑏𝑎 𝑍𝑏𝑏 (ℎ) 𝑍𝑏𝑐 ]
𝑍𝑐𝑎 𝑍𝑐𝑏 𝑍𝑐𝑐 (ℎ)
Only in the case where the three self- and six mutual impedances of Equation A.2 are equal to each
other, the off-diagonal elements in the sequence impedance matrix will be zero and a decoupled
sequence representation can be used for modelling with no loss of accuracy.
The off-diagonal terms in Equation A.2 are related to the coupling between sequences (inter-sequence
coupling). For the decoupled sequence representation, the inter-sequence coupling is either neglected
by choice or naturally zero because of a balanced system. Therefore, an applied positive sequence
voltage can result in the flow of positive sequence current only. Conversely, by using a phase
representation (or a coupled sequence representation), the inter-sequence coupling is taken into
consideration and therefore an applied positive sequence voltage can result in the flow of both negative
and zero sequence current. It can be concluded that when the component under study is no longer
balanced, a decoupled sequence model is no longer sufficient to correctly represent the true electrical
behaviour of the component. The effect of inter-sequence coupling is especially pronounced at and
around harmonic resonance.
Only the positive, negative and zero sequence impedances are needed to construct a model in the
decoupled sequence domain. This is very beneficial for systems where specific component data is
unavailable. To construct a true phase-domain transmission system model, a geometrical description of
all transmission lines and their electrical properties must be known, as well as the phase impedances
of transformers, loads and others equipment. Depending on the desired level of detail, specific
information pertaining to individual line section lengths, cable system bonding, OHL conductor
transposition and substation grounding are some of the parameters that may be required. Such
requirements for data availability and the high level of detail can make this type of study very time
consuming which in some cases causes the engineer to choose decoupled sequence modelling despite
its apparent limitations. The system response is very sensitive to component parameters near
resonance. Therefore, the mutual- and hence inter-sequence coupling is highly sensitive as well. Using
phase-domain models, the influence of specific component parameters can be examined which is a
benefit as some parameters are more prone to change than others. For instance, specific design
parameters vary within bands and it is only after the component has been constructed that specific data
can be delivered. In the case of planning studies assumptions must therefore be made.
203
Further to the example presented in Section 2.2.2.6, the sensitivity of the per-phase amplification on the
single-core 220 kV cable system is examined by varying the relative permittivity of the main insulation
and the system length by ± 5 % (expected variation under design). The results are shown for the three
phases in Figure A.1 (a) and (b).
App Figure A.1 Phase Amplification as Function of the Harmonic Order and (a) the Relative Permittivity
and (b) the Cable System Length
The results show that near resonance even a minor change in some of the many component parameters
can have a large influence on the harmonic amplification. The shunt capacitance of the cable is linear
dependent on the relative permittivity of the main insulation material. Therefore, this parameter has a
strong influence on the resonance frequencies of the cable system. The cable system length determines
the resonance frequencies due to the influence on all electrical cable parameters. It is therefore
recommended to undertake a sensitivity analysis if one wants to ensure that resonance points are not
missed if only a single fixed parameter value is used.
The long, unloaded radial is subjected to strong inter-sequence coupling due to low damping and
resonance behaviour in the frequency range of interest. This phenomenon is also found in a loaded
meshed grid. In Figure A.2 the harmonic voltages obtained at two Danish 400 kV buses is represented
by their phase- and sequence voltage quantities after injection of positive sequence harmonic current
of equal magnitude at all harmonics from the 2 nd to the 20th harmonic. The system is at low load (45%)
but under normal operating conditions (n-0).
204
App Figure A.2 Phase and Sequence Voltage in an Inter-Sequence Coupled Meshed Transmission
Network
The frequency scans obtained at Substation 1 and 2 are presented in Figure A.3. The frequency scans
indicate unbalanced conditions at the harmonics where Figure A.2 shows strong inter-sequence
coupling. This is the case around the 16th harmonic at Substation 1 and at the 18th harmonic at
Substation 2. The results show that the highest phase voltage at the 16th harmonic (measured at
Substation 1) is 41% higher compared to the positive sequence voltage, while the difference is 84% at
the 18th harmonic for Substation 2.
App Figure A.3 Frequency Scans in Inter-Sequence Coupled Meshed Transmission Network at Two
Unique Substations
205
206
Appendix B. CIGRE 14-bus benchmark system

Transmission Network Topology
The 14-bus benchmark system developed for use throughout this Technical Brochure is shown in Figure
B.1 and is based on the IEEE 14-bus system [27]. However, the circuit data been modified based on
the standard circuit geometry and parameters used in the Irish transmission grid to better illustrate the
model development in a practical scenario. The purpose of this model is to serve as a “base-case
topology” for sequential expansion including other network components such as loads and generators.
All transmission circuits are OHLs. The system frequency is 50Hz.
App Figure B.1 220/110 kV Transmission Network Topology
Transmission OHL Parameters

The physical layout and conductor details for the 220kV and 110kV OHL circuits are included in Table
B.1 and Table B.2 below.
App Table B.1 220 kV OHL Parameters
Phase Conductors (600ACSR Curlew)

Conductors per phase 1
DC Resistance 0.05527 Ω/km
GMR 12.75733 mm
Outer radius 15.8115 mm
Inner radius 4 mm
Shield Wires
Shield Wires NO
207
Circuit Geometry
Phase A co-ordinates (-8.05, 16.4) m
Phase B co-ordinates (0, 16.4) m
Phase C co-ordinates (8.05, 16.4) m
Phase Transposition Yes (assumed perfect symmetry)
Soil Resistivity
Soil Resistivity 400 Ω.m
App Table B.2 110 kV OHL Parameters
Phase Conductors (430ACSR Bison)

Conductors per phase 1
DC Resistance 0.0757 Ω/km
GMR 10.91496 mm
Outer radius 13.5 mm
Inner radius 3 mm
Shield Wires
Shield Wires NO
Circuit Geometry
Phase A co-ordinates (-4.5, 12.7) m
Phase B co-ordinates (0, 12.7) m
Phase C co-ordinates (4.5, 12.7) m
Phase Transposition Yes (assumed perfect symmetry)
Soil Resistivity
Soil Resistivity 400 Ω.m
220/110 kV Transformers
The basic parameters of the 220/110kV transformers shown in Figure B.1 are included in Table B.3
below.
App Table B.3 220/110kV Transformers Parameters
Rated Power 250 MVA

HV rated voltage 220 kV
LV rated voltage 110 kV
Vector group YN-YN
Short Circuit Impedance, Zsc 15.57%
X/R 66.53
208
Appendix C. CIGRE 14-bus network model

development examples
Example 1: Development of 220 kV and 110kV Network Model
This section illustrates the development of a model for the benchmark transmission network shown in
Figure B.1 using several steps. The model is developed incrementally, by adding one overhead line
circuit at a time, and the changes in harmonic impedance observed at a selected node are highlighted.
The network topology and circuit parameters are presented in the previous section. The next
subsections will illustrate the incremental model development and the changes in harmonic impedance.
Bus 5 has been selected as the constant reference node to monitor these changes. Results are
presented for two line model options (lumped and distributed) in order to illustrate the limitations of the
lumped parameter representation.
220 kV System Model Development

The incremental development of the 220 kV system is included in this section. For illustration purposes,
the harmonic impedance is calculated with lumped parameter and distributed parameter models and
the harmonic impedance results are compared.
The sequential steps and harmonic impedance calculation set-up are shown in Figure C.1.
STEP-1: One Circuit STEP-3: Three Circuits

I = 1 (f) Amp
(pos. sequence)
1
5 I = 1 (f) Amp
250 km
(pos. sequence)
V5-SC(f) 1
5 4
250 km
100 km
V5-SC(f)
175 km
STEP-2: Two Circuits

I = 1 (f) Amp
(pos. sequence)
1 2
5 4
250 km
100 km
V5-SC(f)
STEP-4: Four Circuits STEP-5: Five Circuits

I = 1 (f) Amp
(pos. sequence) I = 1 (f) Amp
1 (pos. sequence)
5 4 1
250 km
100 km 5 4
250 km
100 km
V5-SC(f)
175 km V5-SC(f)
175 km
250 km
250 km
180 km
2 3
2 3
STEP-6: Six Circuits I = 1 (f) Amp STEP-7: Seven Circuits

I = 1 (f) Amp
(pos. sequence) (pos. sequence)
1 1
5 4 5 4
250 km 250 km
100 km 100 km
V5-SC(f) V5-SC(f)
120 km 175 km 120 km 175 km
250 km 250 km
180 km 180 km
200 km
2 3 3
2
App Figure C.1 Sequential Steps followed in the 220 kV Network Model Development
Figure C.2 shows the calculated impedance, as seen from Bus 5, for each step of model development.
It can be seen that the frequency response of the system becomes more complex as more circuits are
added to the model. In particular, the number of resonances increases with the number of circuits, as
209
there are more components interacting with each other. These interactions lead to addition or
cancellation of resonances depending on the characteristics of the circuits and their lengths. The most
important observation from this figure is that the lumped parameter model (blue traces) fails to reproduce
the complex frequency behaviour of this basic system. In the best of the cases illustrated, the lumped
model approximates the first and/or second resonant peaks, but the high-frequency responses are
always missed. This observation supports the recommendation to always use a distributed parameter
model.
210
Step 1 - One Circuit

18000 100
16000 80
14000 60
Sending End Voltage [V]

40
Phase Angle [deg]

12000
20
10000
Harmonic Order
0
8000
0 5 10 15 20 25 30 35 40
-20
6000
-40
4000
-60
2000
-80
0
0 5 10 15 20 25 30 35 40 -100
1 OHL - LUMPED MODEL 1 OHL - DISTRIBUTED MODEL Harmonic Order 1 OHL - LUMPED MODEL 1 OHL - DISTRIBUTED MODEL
Step 2 - Two Circuits

10000 100
9000 80
8000 60
7000 40
Phase Angle [deg]

6000 20
5000 Harmonic Order
0
0 5 10 15 20 25 30 35 40
4000
-20
3000
-40
2000
-60
1000
-80
0
0 5 10 15 20 25 30 35 40 -100
2 OHL - LUMPED MODEL 2 OHL - DISTRIBUTED MODEL 2 OHL - LUMPED MODEL 2 OHL - DISTRIBUTED MODEL
Harmonic Order
Step 3 - Three Circuits

7000 100
80
6000
60
5000
40
Phase Angle [deg]

4000 20
Harmonic Order
0
3000 0 5 10 15 20 25 30 35 40
-20
2000 -40
-60
1000
-80
0
0 5 10 15 20 25 30 35 40 -100
Step 4 - Four Circuits

4500 100
4000 80
3500 60
40
Phase Angle [deg]
3000
20
2500 Harmonic Order
0
2000 0 5 10 15 20 25 30 35 40
-20
1500
-40
1000
-60
500
-80
0
0 5 10 15 20 25 30 35 40 -100
Step 5- Five Circuits

4000 100
80
3500
60
3000
40
Phase Angle [deg]
2500
20
Harmonic Order
2000 0
0 5 10 15 20 25 30 35 40
1500 -20
-40
1000
-60
500
-80
0
-100
0 5 10 15 20 25 30 35 40
Step 6- Six Circuits

4000 100
80
3500
60
3000
40
Phase Angle [deg]
2500
20
2000
0
Harmonic Order
0 5 10 15 20 25 30 35 40
1500 -20
1000 -40
-60
500
-80
0
0 5 10 15 20 25 30 35 40 -100
Step 7- Seven Circuits

3000 100
80
2500
60
2000 40
Phase Angle [deg]
20
1500
Harmonic Order
0
0 5 10 15 20 25 30 35 40
-20
1000
-40
500
-60
-80
0
0 5 10 15 20 25 30 35 40 -100
App Figure C.2 Harmonic Impedance at Bus-5 under Incremental 220kV Network Model Development
211
Addition of 110 kV System

This section illustrates the effect of expanding the model to include the 110kV system. All 220 kV lines
are represented with the accurate distributed parameter model. This example examines the impact of
representing some 110kV circuits with a lumped parameter model (i.e. ignoring long line effects) on the
harmonic frequency response observed at 220 kV. For scenarios are compared:
1. All 110 kV circuits are represented with lumped parameters model (ignoring long line effects).
2. All 110kV circuits up to one bus away from the observation point (bus 5) are represented with
distributed parameter model (with long line effects). The rest of 110kV circuits are represented with
lumped parameter model.
3. All 110kV circuits up to two buses away from the observation point (bus 5) are represented using
the distributed parameter model (with long line effects). The rest of the 110kV circuits are
represented using the lumped parameter model.
4. All 110 kV circuits are represented with distributed parameters model (capturing long line effects).
The harmonic impedance has been calculated as shown in Figure C.3. Frequency scan results for bus
5 are included in Figure C.4.
13 14
100 km
80 km
110 kV 60 km 85 km
12
11 10
70 km 50 km
35 km 45 km
6 9
I = 1 (f) Amp
220 kV (pos. sequence)
1
5 V5-SC(f) 4
250 km
100 km
120 km 175 km
180 km 250 km
200 km
2 3
App Figure C.3 Setup for Calculation of Harmonic Impedance at Bus 5 including 110 kV System
The following observations can be made for this example:

 The harmonic impedance of this system (as seen from Bus 5) seems to be dominated by the
impedance of the 220kV OHL network; i.e. there is very little difference between step 7 in Figure
C.2 (which only captures 220kV circuits) and Figure C.4 (which captures 220kV and 110kV OHL
circuits).
 The choice of 110kV line model (lumped or distributed) has very little impact on the harmonic
impedance seen at the selected 220kV node (Bus 5) for a wide range of frequencies. The only
significant differences can be appreciated above the 26th harmonic order.
It should be noted that the conclusions derived from this example are only applicable to the system
shown in Figure B.1. It is stressed that the frequency response of a system is determined by its topology
and its individual components, therefore these results cannot be extrapolated to a real power system.
Instead, this example must be interpreted as a methodology to assess the impact of a portion of a
system, e.g. lower voltage level, on the frequency response in the area of interest. This information can
be used, for example, to decide where to concentrate modelling efforts.
212
1800
1600
Voltage at Bus #5 [V]
1400
1200
1000
800
600
400
200
0
0 5 10 15 20 25 30 35 Harmonic Order 40
All 110kV circuits with lumped model One 110kV bus away distributed Two 110kV bus away distributed All 110kV circuits as distributed model
100
80
60
40
Phase Angle [ deg]
20
0
0 5 10 15 20 25 30 35 40
-20
-40
-60
-80
-100
Harmonic Order
All 110kV circuits with lumped model One 100kV b awasy as distributed Two 110kV bus away as distributed All 110kV circuits as distributed
App Figure C.4 Harmonic Impedance at Bus-5 with 220kV and 110kV Network Models
213
214
Appendix D. Cable sensitivity analysis

Sensitivity analysis was performed using a representative study case to identify the parameters that
affect the frequency response of an underground power cable. This type of analysis is important due to
the presence of parameter uncertainty in the planning stage of a cable connection. The types of
uncertainties considered were length, design characteristics, layout, bonding scheme and model type.
Description of the Cable System

The parameters of the 380kV underground power cable used in this study is provided in Table D.1.
App Table D.1 Cable Parameters
U0 = 230 kV Diameter or thickness Relative Relative

Resistivity (Ω.m)
1600 mm² Cu (mm) permittivity permeability
Conductor (copper) 50.3 (diameter) 2.86×10-8 (at 90°C) - 1

Inner SC 2 - - 1
Insulation (XLPE) 26 - 2.3 1
Outer SC 2.8 - - 1
Metal sheath (aluminium) 2.5 2.84×10-8 (at 20°C) - 1
Outer sheath (PE) 5 - 2.3 1
The cable comprises six layers as illustrated in Figure D.1.
PE outer sheath r3
r2
Aluminium sheath b
Outer semiconducting layer

a
Core r1
XLPE insulation layer
conductor
Inner semiconducting layer
App Figure D.1 Cable Layers
To model the given cable, some commercial software requires that corrections be made, as described
below.
Conductor resistivity: In some software it is not possible to model stranded conductors hence a solid
copper conductor was used with corrected resistivity value (taken from IEC 60228) to account for the
spaces between the stranded/segmented wires of the conductor.
215
𝜋𝑟12 App Equation D.1

𝑅𝑐 = 𝜌𝑐
𝐴
Where 𝜌𝑐 is the resistivity of a solid conductor and 𝑟1 is the radius of the conductor.
Insulation relative permeability: The semi-conductive layers are considered as part of the insulation
and the permittivity is corrected according to Equation D.2 so that the total cable capacitance remains
the same.
𝑟
ln 2 App Equation D.2
𝑟1
𝜀𝑖𝑛𝑠 = 𝜀𝑋𝐿𝑃𝐸
𝑏
ln
𝑎
Where, 𝜀𝑋𝐿𝑃𝐸 is the XLPE insulation permittivity, 𝑟2 the radius over the insulation including the
semiconductor layers, 𝑟1 the radius over the conductor, 𝑏 the outer radius of the insulation and 𝑎 the
inner radius of the insulation.
The trefoil formation shown in Figure D.2 was chosen for the base case in the analysis.
App Figure D.2 Base Case Cable Layout
The cable had a length of 100km and was cross-bonded as shown in Figure D.3. It comprised 15 major
sections, each of which was divided into three minor sections. The sheaths were cross-bonded between
minor sections (cross-bonding joints) and grounded at the end of the major sections (earthing joints).
The minor sections should have a similar length to keep the system as balanced as possible.
App Figure D.3 Cable Cross-Bonding Configuration
Cable Impedance
Frequency scans were performed on the base case cable configuration with the ends open-circuited or
short-circuited. The results obtained were the magnitudes and phases of the sequence impedances. In
Figure D.4 and Figure D.5 both magnitude and phase of the positive and zero sequence impedances
are shown for both configurations.
It is evident that a duality exists between open-circuited and short-circuited cable terminations where
the series resonance peaks in the open-circuit case become parallel resonance peaks in the short-circuit
case and vice versa. For the base case, the first resonance peak occurs at 255Hz, very close to the fifth
harmonic, and for frequencies up to 2500Hz, twelve resonant points can be observed (six parallel and
six series resonances).
For the remainder of the sensitivity analysis the short-circuited termination was used.
216
App Figure D.4 Positive-Sequence Impedance: Cable Termination Open- or Short-Circuited
App Figure D.5 Zero-Sequence Impedance: Cable Termination is Open- or Short-Circuited
217
Harmonic Amplification
App Figure D.6 Harmonic Propagation for Long EHV Cable

For calculating the voltage amplification, variable frequency voltage sources were connected to the three
phases of the sending end of the cable and the receiving end was left open-ended (Figure D.6). To have
no voltage amplification at the fundamental frequency, the cable was fully compensated using shunt
reactors at both ends. The injected harmonics were in the range of 50Hz to 2500Hz with a 50Hz step
for all three sequences (positive, negative and zero) separately. The sequence voltages at the sending
and receiving end were measured and the amplification ratio was calculated as the division of these two
voltages for each harmonic.
The amplification ratio of the positive and negative sequence harmonics was the same and is illustrated
in Figure D.7. As expected, high voltage amplification is observed near the parallel resonance
frequencies of the cable. The amplification ratio of the zero sequence harmonics can be seen in Figure
D.8 where similar behaviour is observed.
App Figure D.7 Positive-Sequence Voltage Amplification
App Figure D.8 Zero-Sequence Voltage Amplification
218
Cable Length
The harmonic impedance for five different cable lengths (100km (base case), 30km, 50km, 70km and
150km) can be seen in Figure D.9 and Figure D.10. As the cable length increases both the frequency
and the magnitude of the resonances decrease and the number of resonances within the frequency
range increases. Similar behaviour is observed in both the positive- and zero-sequence impedance.
Key point:
 Cable length plays a significant role in the frequency, magnitude and number of harmonic resonance
points in both the positive- and zero-sequence harmonic impedance.
App Figure D.9 Positive-Sequence Harmonic Impedance Comparison for Different Cable Lengths
219
App Figure D.10 Zero-Sequence Harmonic Impedance Comparison for Different Cable Lengths
Cable Design Characteristics

The thicknesses of each of the six cable layers are provided either from datasheets or from cable sample
measurements. However, the nominal thickness of the various layers as stated by manufacturers can
differ from the actual (design) thickness. Therefore, geometrical data from the manufacturer can be
inaccurate when used for cable parameter calculations. Variations of ±10% in the conductor radius and
insulation thickness were investigated while the thickness of the other cable layers remained constant.
 Conductor Radius
The harmonic impedance comparison for variations in the conductor radius is presented in Figure D.11.
The thicknesses of all other layers remain fixed (i.e. representing an outer radius change). As the
conductor radius increases so does the cable capacitance, resulting in a downward shift (decrease) in
resonant frequencies. This variation decreases the inductance of the cable and hence the overall
frequency shift is small. The positive-sequence impedance magnitude is unchanged. Zero-sequence
impedance resonant frequencies are hardly affected; however the zero sequence impedance
magnitudes are reduced as slightly more damping is introduced with increasing conductor radius. This
is shown in Figure D.12.
Key point:
 Conductor radius has a minor influence on the frequencies at which resonances occur in the
positive-sequence and has a minor impact on the zero-sequence impedance magnitude.
220
App Figure D.11 Positive-Sequence Harmonic Impedance Comparison for Variations in the Cable
Conductor Radius
App Figure D.12 Zero-Sequence Harmonic Impedance Comparison for Variations in the Cable Conductor
Radius
221
 Insulation Thickness
The harmonic impedance comparison for deviations in the insulation thickness is presented in Figure
D.13 and Figure D14. For the positive sequence impedance, as the insulation thickness decreases, the
cable capacitance increases resulting in resonance peaks shifting to lower frequencies. In this example,
the magnitude slightly decreases for smaller insulation thickness. For the zero-sequence impedance,
only the resonance magnitudes are affected. When the insulation thickness increases, so does the
impedance magnitude.
Key point:
 Insulation thickness affects resonances seen in the positive-sequence harmonic impedance in terms
of frequency and magnitude, and in the zero-sequence only the magnitude is affected.
App Figure D.13 Positive-Sequence Harmonic Impedance Comparison for Variations in the Cable
Insulation Thickness
222
App Figure D.14 Zero-Sequence Harmonic Impedance Comparison for Variations in the Cable Insulation
Thickness
Cable Layout
Cable installation and layout refer to where and how individual cables are positioned with respect to one
another. Example layouts include trefoil, flat and triangle as shown in Figure D.15. In the planning stage
of a cable connection the exact layout is unknown, making it important to study how different choices of
cable layout can affect the harmonic impedance.
Trefoil Flat Triangle Touching trefoil
App Figure D.15 Examples of Cable Layout
The harmonic impedance comparison for four cable layouts (trefoil, flat, triangle and touching trefoil) is
illustrated in Figure D.16 and Figure D.17. Variations in the cable layout greatly affect both the frequency
and the magnitude of the positive-sequence impedance resonance peaks. More specifically, the flat and
triangle formations present a greater number of resonance peaks compared with the base case where
the trefoil formation was used. This is due to the higher asymmetry present in the first two layout
configurations compared to the trefoil formation. For the touching trefoil formation, a shift of the
resonance peaks to higher frequencies is observed. When the distance between the cables decreases,
the mutual inductance increases, and the total positive-sequence inductance decreases. The zero-
sequence impedance is identical in all four cases.
Key point:
 Cable layout significantly affects the frequency, magnitude and number of resonances seen in the
positive-sequence harmonic impedance.
223
App Figure D.16 Positive-Sequence Harmonic Impedance Comparison for Different Cable Layouts
App Figure D.17 Zero-Sequence Harmonic Impedance Comparison for Different Cable Layouts
224
Sheath Bonding
The bonding configuration as well as the number of major sections used can greatly influence the
frequency response of a cable connection. Generally, for long (several km and more) cables, to prevent
large currents from circulating in the sheaths giving rise to high losses, whenever the two ends of the
sheaths are grounded, cross-bonding must be used.
 Type of Bonding
The base case described in Figure D.3 where the cross-bonding configuration with 15 major sections
was used, is compared with a solidly-bonded 100km cable. The different bonding configurations will not
result in changes to the shunt admittance matrix (as no alteration is made to the distance between the
conductor and the sheath) but the series impedance matrix may differ [37].
The harmonic impedance comparison for these two bonding configurations is shown in Figure D.18 and
Figure D.19. The positive sequence impedance is significantly affected by the type of cable bonding
used with the cross-bonded cable yielding higher magnitudes and lower resonant frequencies than for
the solidly-bonded cable. In the case of solid bonding there is a 177 Hz frequency shift to higher
frequencies at the first parallel resonance and the magnitude is approximately ten times lower.
The type of bonding has only a negligible effect on the zero-sequence impedance magnitudes as shown
in Figure D.19. The zero-sequence impedance is largely independent of the type of bonding used, as
approximately no current should flow to ground at the grounding points.
Generally, for large cable connections (several km) the only viable option (aside from transposition of
phases) is sheath cross-bonding due to the large circulating sheath currents and associated higher
losses when solid bonding is used. However, the result of this comparison is valuable for the simulation
of cable systems where it is evident that the use of a simplified solid bonding configuration can result in
a high deviation of the harmonic impedance.
Key point:
 The type of bonding significantly affects the positive-sequence harmonic impedance but has only a
minor effect on the zero-sequence harmonic impedance.
App Figure D.18 Positive-Sequence Harmonic Impedance Comparison for Different Cable Bonding
Configurations
225
App Figure D.19 Zero-Sequence Harmonic Impedance Comparison for Different Cable Bonding
Configurations
Number of Major Sections

The harmonic impedance comparison for different numbers of major sections is shown in Figure D.20
and Figure D.21. For the positive-sequence impedance, the frequency, magnitude and number of
resonant peaks are affected when different numbers of major sections are used. This is more evident
when only one major section is used where deviations are observed even at the first parallel resonance.
The cable with five major sections present the same frequency and magnitude for the first parallel
resonance, which starts to deviate from the second resonance peak. Similarly, the cable with ten major
sections differs from the base case after the second parallel resonance. For the same reason given in
the previous section, the zero-sequence impedances are not affected by the number of major sections,
as shown in Figure D.21.
Key point:
 The number of major sections affects the frequency, magnitude and number of resonant peaks in
the positive-sequence.
226
App Figure D.20 Positive-Sequence Harmonic Impedance Comparison for Different Numbers of Major
Sections
App Figure D.21 Zero-Sequence Harmonic Impedance Comparison for Different Numbers of Major
Sections
227
Cable Model
In this section, the use of single or cascaded PI -sections are compared with the frequency-dependent
(FD) distributed parameter phase model. Since modelling of the sheath bonding is not possible when
the PI model is used and the positive- and zero-sequence impedances from the line constants routine
are calculated for a single cable section (and not for the complete cross-bonded cable) the comparison
between the PI and FD models was performed when the latter was configured to use solid bonding.
The harmonic impedance comparison of the models can be seen in Figure D.22. When a single PI
section was used, only the first parallel resonance is present which is shifted to a lower frequency. As
the number of PI sections increases, more resonance peaks are represented, and their frequencies
increasingly shift towards those of the FD model. However, due to the representation of the multiple PI
sections at 50Hz, the magnitude of the resonance peaks is significantly higher than those of the FD
model.
Key point:
 Model type should be selected carefully depending on the study. In the case of the PI model, the
number of PI sections used influences the number of resonance points able to be modelled.
App Figure D.22 Positive-Sequence Harmonic Impedance Comparison for Different Cable Models
228
Appendix E. Example of load models

Example Network
An example network is developed in this appendix to illustrate the impact of representation of loads at
different voltage levels and by considering different types of load models. This example is based on a
typical network in Brazil but could be extended to other countries as well. In Brazil, transmission system
studies are carried out with an accurate grid representation for voltages up to 69 kV. The database for
load flow, short-circuit etc. includes the transmission and sub-transmission network parameters for
voltages up to 69 kV. In some cases, it is possible to represent the grid from 69 kV to 13.8 kV, but this
does not apply to the distribution grid (1 kV and below), where most of the loads are concentrated. The
databases are also used as the starting point for transmission system harmonic studies. Therefore, it is
recommended to adopt some methodology to represent the distribution loads concentrated at some 69
kV and 13.8 kV buses, through equivalents.
This example considers several representations of a 100 MVA load supplied by a 230/69 kV transformer,
located at the border of the transmission/distribution system as illustrated in Figure E.1. The initial
simulations did not model the resistance correction with frequency. The harmonic impedance as seen
from the 230 kV bus was determined for different load representations.
App Figure E.1 Simple Configuration for Sensitivity Analysis
 Load: 95 + j 31 MVA
 Transformer: 230 / 69 kV (Y / Δ), 150 MVA, X = 10%, QF=50
The load models selected for the studies are (Figure E.2):
 CIGRE model described in Section 3.4.1.3
o Similar to Method 1 (CIGRE JTF 36.05.02/14.03.03); Passive Load – domestic
o Illustrated in Figure 3-37
o Referred to in this Appendix as “Rs X”
o It should be noted that this model uses the following expressions to define R and X:
𝑈2
𝑅=𝑃
𝑃2 + 𝑄2
𝑈2
𝑋=ℎ𝑄 2
𝑃 + 𝑄2

o Method 2 (CIGRE WG CC02)
o Referred to in this Appendix as “R p X”
 CIGRE/EDF model described in Section 3.4.1.1

o Method 3 (CIGRE WG 36.05)
o Referred to in this Appendix as “(R s X) p X”
229

o Method 3 (CIGRE WG CC02)
o Referred to in this Appendix as “Motor”
Rs X Rp X (Rs X)p X Motor
Rs Rs Rm
Rp Xp Xp
Xs Xs Xm
App Figure E.2 Load Representations
In case of reactive power compensation, a shunt capacitor is added in parallel to the load.
The following cases were studied for load represented at 230 kV, 69 kV, 220 V and 220 V through a
system connection equivalent of 69 kV and 13.8 kV systems. Results are provided in the next pages.
Key Points:
 The study example demonstrated that the load representation at 220 V with a detailed
representation of 69 kV and 13.8 kV network and through a system connection equivalent can
predict the system resonance. The choice of load model did not significantly change the system
response, as seen at the 230 kV bus.
 Load representation at 230 kV and 69 kV (that did not include the downstream impedance) do not
capture resonance.
 Results confirmed the importance of connecting the load model through an equivalent downstream
impedance.
230
Load represented at a 230 kV Bus

The following cases were studied:
 RsX
 RpX
 (R s X) p X
The associated frequency scan is shown in Figure E.3.
Load at 230 kV (95 MW + 31 Mvar)
App Figure E.3 Frequency Scan at 230kV Bus - Load Represented at a 230 kV Bus
231
Load represented at a 69 kV Bus

 RsX
 RpX
 (R s X) p X
 40 % R p X + 60% motor
 40% (R s X) p X + 60% motor
 R p X + capacitor
 40% R p X + 60% motor + capacitor
Load at 69 kV (95 MW + 31 Mvar)
App Figure E.4 Frequency Scan at 230kV Bus - Load Represented at 69 kV Bus
232
Load Represented at a 220 V Bus through a 69 kV and 13.8 kV System

Connection
 (R s X) p X + capacitor
 40% (R p X) + 60% motor + capacitor
App Figure E.5 System Connection Configuration from 230 kV to 220 V
App Figure E.6 Frequency Scan at 230kV Bus - Load Represented at a 220 V with Explicit
Representation of 69kV and 13.8kV Systems
233
Load at a 220 V Bus Represented through a 69 kV and 13.8 kV System

Connection Equivalent
 (R s X) p X + capacitor
 40% (R p X) + 60% motor + capacitor
App Figure E.7 System Connection Equivalent at 69 kV
App Figure E.8 Frequency Scan at 230kV Bus - Load Represented at a 220 V through a Reduced
Equivalent Representation of the 69kV and 13.8kV Systems
234
Appendix F. Synchronous generator model

validation
The HQ model presented in Section 3.5.1.3 was validated against field measurements at 315 kV, as
illustrated in Figure F.1.
3 x 370 MVA Machine 1650 MVA
13.8 kV HVDC
+
C15 735 kV GND
GND
315 kV
0.25uF 6nF
C4 5nF B6P_1
+
0.01472,0.1472Ohm
C6
0,5.888Ohm
+
GND 6-pulse bridge
GND
+
350 MVA CVT
Electra 167 735/315/15 3-phase
RL2
+
80.96
2
R1
RL1
1 YgYgD_np1
+
C16
+
315/120
BUS5 15.5 km 2
+
+
C8
200nF Electra 167 YD_1 4.7nF 1
+
2 1 3 C3
YYD_1 gates -
4.7nF
GND
4.7nF
a FDline1 a
+
C12 C11 GNDGND
10nF 3nF b + FD b GND 3 B6P_2
315/120 c Electra 167
GND GND c 735/315/15 735/315/15 6-pulse bridge
GND
Electra 167 YD_2 + 2
+
C14
2 1 1 3-phase
+
C13 C10
10nF 3nF
+
3 C2
YYD_2
GND 4.7nF
GND
+
+
C17 C9
+
0.25uF 4.7nF C1 GND C7 gates -
4.7nF GND 5nF
+
C5
350 MVA
0.01472,0.1472Ohm
0,5.888Ohm
GND GND 6nF
CVT CVT GND

GND
GND
RL4
+
80.96
Z(f) measured by
R2
RL3
1650 MVA
+
C18
+
+
200nF white noice injection

in HVDC controller
315/120
Z(f)
Electra 167 YD_3
2 1
+
C19 C22 C21

0.25uF 10nF 3nF
GND
0.01472,0.1472Ohm
0,5.888Ohm
GND GND GND

RL6
+
80.96
R3
RL5
350 MVA
+
C20
+
+
200nF
App Figure F.1 HQ Field Measurement Setup to Measure Z(f) Seen at 315 kV
Measurement data can be seen in Figure F.2 [60] and corresponding simulation results in Figure F.3.
App Figure F.2 Measurement Data of Z(f)
235
App Figure F.3 Simulated Data of Z(f)
Measurements of the machine impedance of a 50 MVA 13.8 kV hydraulic, salient pole machine (X’’d=0.3
pu and X’d=0.4 pu) over a range of frequencies are shown in Figure F.4.
App Figure F.4 Measurement Data of a 50 MVA 13.8 kV Machine
In the range of 60 to 2500 Hz, linear interpolation of the data gives:

Z(f) = (-0.11 + 0.0216 * f) where Z is in Ohm and f is in Hz.
Phi(f) = (82.7 – 0.0014 * f) where Phi is in degrees and f is in Hz.
Table F.1 shows the variation of R and X variation vs. frequency of this machine.
236
App Table F.1 Measured Parameters of a 50 MVA Machine vs. Frequency
Frequency (Hz) h R (Ohm) X (Ohm) X/R
60 1 0.153 1.18 7.71

120 2 0.322 2.46 7.64
180 3 0.497 3.75 7.54
240 4 0.674 5.02 7.45
300 5 0.856 6.31 7.37
360 6 1.040 7.60 7.31
420 7 1.245 8.98 7.22
660 11 2.023 14.00 6.92
1500 25 5.275 31.86 6.04
It is worth noting that T’’d should normally be used to estimate the resistance of the machine at power
frequency (IEEE assumes X/R at 60 Hz as being around 10 which is somewhat high for 50 MVA
machines but low for 370 MVA). This resistance is then increased by the factor ℎ. The Electra model is
more pessimistic because it considers √ℎ.
The harmonic impedance of a 120MVA hydraulic machine (based on measurements and manufacturer
design data) is presented in Figure F.5. In conjunction with Table F.1, this shows that the harmonic
reactance can either increase or decrease with frequency.
App Figure F.5 Harmonic Impedance of a 120 MVA Hydraulic Machine
Sensitivity Analysis
A sensitivity analysis using 13.8 kV machines (illustrated in Figure F.6) shows the importance of
modelling of high frequency (HF) losses which have an impact on the magnitude of resonant peaks.
237
App Figure F.6 Sensitivity Analysis: Machine Losses
Model Simplification
This case study shows that once the model has been validated against measurement data, it is possible
to represent the power station from the HV side as a simplified Rs-Ls||Rp equivalent. This is illustrated
in Figure F.7. The validation shows that the curves are almost identical, proving that the approximation
is highly accurate.
App Figure F.7 Model Simplification following Validation
238
Appendix G. Harmonic state estimation in Spain

Red Eléctrica de España (REE), the Spanish TSO has several active research programmes in the fields
of harmonic estimation and observability of Power Quality (PQ) parameters in the Spanish system. A
PQ estimation algorithm has been recently implemented, allowing REE to infer harmonic voltages and
current injections associated with perturbations without measurements.
The Spanish peninsular system (with a peak of 45GW) has installed capacities of 23GW and 4GW for
wind and photovoltaic generation respectively. In addition, a new HVDC VSC between Spain and France
was commissioned in 2015. To cope with these potentially perturbing loads, REE has been installing
new PQ monitor devices, especially in conflicting buses with large distorting loads and generators.
In order to shed some light on the behaviour of PQ parameters, a Power Quality State Estimator software
tool (called ETAD) has been developed. With the aid of this program it is possible to evaluate the voltage
distortion (with some accuracy) of buses where no PQ monitoring device is installed. This allows for the
PQ estimation of the whole network.
ETAD uses information stored in the Centralized Power Quality Database and network models obtained
from the state estimator of REE’s Control Centre. Furthermore, the program takes into account the type
of loads connected to the transmission system.
Methodology of the PQ Estimator

One of the main objectives of the software tool is harmonic voltage distortion estimation.
The idea of a PQ estimator arose from the difficulty of installing measurement equipment at all suspect
nodes.
The estimation of harmonics is based on the classical procedure of a harmonic penetration study.
Harmonic-producing loads are substituted by harmonic current sources, which, once propagated in the
network, create the harmonic voltages. Each harmonic order uses its corresponding impedance matrix.
When a sufficient number of nodes are measured, state estimation techniques can be used to identify
harmonic sources. But in a real power system, only partial measurements are made, or these
measurements are made at buses at which no current source is present. Furthermore, measurement
errors contribute to the limitation on the number of measurements, which makes harmonic detection a
highly challenging endeavour. The most common outcome is an undetermined system (more unknowns
than equations).
Harmonic current sources are the outcome of the estimation. Once these have been estimated, a current
penetration process allows the calculation of harmonic voltages in all the system. The solution adopted
in this project is to determine the currents by means of the minimum Euclidean norm solution of the
associated undetermined system of linear equations.
The network of study has n buses, where q of them are considered as sources of currents and only p
buses are measured. For each harmonic order, the impedance matrix of the network is calculated. The
nodal equations are V=Z·I, with Z the submatrix nxq corresponding to the columns of the buses in which
the current sources may exist.
If there are p<n measurement devices, selecting the rows corresponding to these p buses, the following
set of equations can be set: Zm·I=Vm. The most common situation is to have p<q, so the system is
undetermined. That is, there are more possible current sources than measurement devices. When the
range of matrix Zm is p, the estimation of sources I is given by: I = I0 + Nh (h ϵ Cq-p ). Being I0 a particular
solution and N any q x (q-p) matrix of range (q-p) which respond to: ZmN=0.
A suboptimal solution of this problem is shown below:
- From the Euclidean norm of the current vector, defined by:
ǁIǁ2 = ǀI1ǀ2 + … + ǀIqǀ2
- The minimization problem is:
min ǁIǁ2 restricted by Zm·I=Vm
- When the range of Zm is p, the solution is:
239
Ie = Z’m (ZmZ’m)-1Vm
It is numerically more efficient to use the QR factorization of Z’m : Z’m = QR

The solution of this system of equations is: R’x = Vm Ie = Q x
(Where A’ is the conjugate transpose of A )
The corresponding estimated voltages of the network are obtained, for each Ie , by means of the current
propagation equation : V = Z· Ie .
Validation
To check the accuracy of this method, a set of random harmonic current injections has been simulated
on an REE transmission network model. For a small subset of buses, the calculated voltages are
considered as measured voltages. The previously-described method is then applied to these harmonic
voltages, the network and the network data in the model. The results are shown below:
App Figure G.1 Accuracy of the Estimation Method
Even when the harmonic current sources are poorly estimated, the values of the resulting harmonic
voltages are more precise. The application of this method to a large number of different scenarios can
increase its accuracy.
An example of the results for the estimation of the 5th harmonic voltage using real scenarios and real
measurements for October and November 2015 is shown below:
App Figure G.2 Example of Results of the Harmonic State Estimation Method
Another important issue is the location of the new PQ monitoring devices. These should maximize the
visibility of the PQ estate of the whole network. It is also important to analyse the location of new devices
in which the errors could be minimal. More PQ monitoring devices will result in the program yielding
more accurate results.
240
CIGRE
21, rue d'Artois
75008 Paris - FRANCE
© CIGRE
ISBN : 978-2-85873-468-9

Technical Brochure: Network Modelling For Harmonic Studies

Uploaded by

Copyright:

Available Formats

Technical Brochure: Network Modelling For Harmonic Studies

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

Technical Brochure: Network Modelling For Harmonic Studies

Uploaded by

Copyright:

Available Formats

C4/B4

Network modelling for harmonic

M. VAL ESCUDERO, Convenor G. LIETZ, Chapter Lead DE

2. Study domain and modelling aproaches ............................................................... 21

3. Classical network element models ......................................................................... 31

3.7 Network Equivalent for Harmonic Studies .......................................................................................... 103

4. Power electronic based network element models ................................................105

5. General considerations for harmonic studies ......................................................159

Appendix A. Compoment representation for frequency-domain studies……….203

Figures and Illustrations

Figure 4-26 SVC Model for System-Wide Studies .............................................................................................. 136

App Figure D.2 Base Case Cable Layout ........................................................................................................... 216

App Table B.1 220 kV OHL Parameters ............................................................................................................. 207

Equation 3.44 ........................................................................................................................................................ 96

App Equation A.1................................................................................................................................................. 203

Acronyms and abbreviations

ACS Average Cold Spell

UMEC Unified Magnetic Equivalent Circuit

2. Study domain and modelling aproaches

2.2 Frequency-Domain Methods

Balanced Unbalanced Balanced Unbalanced

Figure 2-1 Overview of Frequency-Domain Methods

2.2.1 Frequency Scan

[𝑌ℎ ][𝑉ℎ ] = [𝐼ℎ ] Equation 2.1

2.2.2 Harmonic Penetration

2.2.2.1 Direct Method

2.2.2.2 Iterative Method

2.2.2.3 Applications and Limitations

2.2.2.4 Balanced Harmonic Penetration

2.2.2.5 Unbalanced Harmonic Penetration

 Unbalanced Harmonic Penetration based on a Balanced Fundamental Load Flow

 Unbalanced Harmonic Penetration based on an Unbalanced Fundamental Load Flow

2.2.2.6 Selection of Balanced or Unbalanced Penetration Method

Example: Difference between a balanced and an unbalanced propagation study on a

Figure 2-3 Frequency Scan of the Apparent Impedance of Phases A, B and C

This material is further addressed in Appendix A.

2.2.3 Harmonic Load Flow

2.3 Time-Domain Methods

2.3.1 Application and Limitations

2.4 Hybrid Methods

2.5 Harmonic Domain

Table 2-1 Summary of Study Domains and their Applications

Advantages Disadvantages Applications

frequency installations of nearby harmonics

on specific and linear models correctly converters (including saturated,

 Incorporates non-  Time consuming  Currently, applications of this method

 Benefit from use of cases where models are

3. Classical network element models

3.1.1 Modelling Considerations

Figure 3-1 Overhead Line Nominal-PI Model

Figure 3-2 Overhead Line Equivalent-PI Model

3.1.2 Consideration of Skin Effect

𝑅(𝜔) 0.035 ∙ 𝑀2 + 0.938 𝑀 < 2.4

Table 3-1 Corrections for Skin Effect in OHLs

Company Voltage (kV) Harmonic Order Resistance

400, 275 3.45ℎ2

ℎ > 7 ⋅ 76 𝑅1 (0.267 + 0.485√ℎ)

4<ℎ<8 𝑅1 (0.864 − 0.024√ℎ

8<ℎ 𝑅1 (0.267 + 0.485√ℎ)