Fulltext Vimp

Experimental Set-up for partial discharge
detection
Ingrid Roen Velo
Master i energi og milj

Innlevert: januar 2015
Hovedveileder: Frank Mauseth, ELKRAFT
Medveileder: Pl Keim Olsen, ELKRAFT
Norges teknisk-naturvitenskapelige universitet

Institutt for elkraftteknikk
Problem Description
Production of electric power from floating wind turbines will be a big oncoming chal-
lenge. Numerous Norwegian companies are working on different aspects of floating
wind power generation.
Power electronics used for HVDC converters will stress the cable with a DC voltage
with overlaid transients. The effect of these transients on the performance of the
polymeric cable insulation is yet not known and the main purpose of the project is to
investigate how transients can influence the life time of the cable insulation.
The work will mainly be experimental. In the first stage, a noise reduced test circuit
for AC voltage at higher frequencies must be designed and built. Further work will be
to measure partial discharges for AC voltage at different frequencies. Ultimately, the
test objects should be tested with a DC voltage with an overlaying AC to simulate the
voltage stress a HVDC cable will be subjected to close to the HVDC converters.
i
Preface
This report is a result of work with the master thesis conducted at the Norwegian
University of Science and Technology, NTNU, at the Department of Electrical Power
Engineering.
There are so many people I am grateful to for all the help during the work on this
master thesis. Most of all I would like to thank my co-supervisor and labparter; Pal
Keim Olsen. Without your previous work on the subject and your great spirit, none of
this would be possible in such a short amount of time. Further I would like to thank
my supervisor Frank Mauseth for the help with interpreting the results and giving
answers to the difficult questions that emerged throughout the thesis work.
I would like to thank the guys in the service lab, Aksel Andreas Reitan Hanssen and
Vladimir Klubicka for the help with the experimental set-up in the lab. Most of all
Bard Aalmaas, you have been a huge help with the security measures and the finalizing
of the experimental set-up. You always know what do, even when I do not really know
what I am asking.
To everyone in the workshop, thank you for the help with finding equipment and the
neccessary parts. A special thanks to Morten Fl a, always quick on the fix and Dominik
Hager for the help with making artificial cavities in the test samples.
Last but not least, thank you to all my fellow students, friends and family, especially
the guys at room G-220. Thank you for the great working environment, great expe-
riences and many laughs. Without you, being at shcool every day would not be the
same, and neither would the great memories of writing this master thesis.
iii
Abstract
There are many technological challenges related to the production of electrical power
from offshore wind farms. In offshore HVDC transmission systems converter stations
are installed between the AC and DC power grid. The resulting DC voltage from the
converter stations will contain transients originating from the switching of the power
electronic components. The AC ripple voltage on the DC side of the HVDC stations
can be in the range of 1-10% of the nominal DC voltage, depending on the size of
the engaged filter. Harmonic content of the ripple can range up to 30-100 times the
switching frequency, with the dominant components close to the switching frequency
which can be around 1-2 kHz.
The effect of the DC component and the overlaid transient on the ageing of polymeric
insulation is not yet known. An operational insulation system is one of the most
important prerequisite for functional high voltage system and partial discharge are
one of the main ageing mechanisms in a dielectric.
The main purpose of this master thesis has been to develop and build a low interference
experimental set-up for partial discharge detection that can simulate the conditions
in offshore cable insulation close to the converter stations. The experimental set-up
should be able to apply a combined DC voltage and high frequency AC voltage across
a test sample.
This master thesis is mainly experimental and have been performed in three sections;
sub-system test, system test and main test. The sub-system test was an analysis of
the AC voltage source and transformers to find a combination able to deliver a 2
kVrms sinusoidal voltage with little distortion across the frequency range 1-5 kHz. The
Behlman signal generator and Messwandler Bau transformer gave a THD below 1%
at all frequencies and is the AC voltage supply system used in the further tests.
The experimental set-up, based on a partial discharge straight detection circuit, is
built inside a faraday cage. During the system testing the noise and intereference
from the surrounding and circuit was tested and the nessesary alterations were done.
Resulting in a noise band of approximately 0.6 pC, allowing a detection limit of 1 pC.
The system test also included a benchmark test on the layered PET samples compared
to solid ones. The results indicaticated partial discharges in the interfaces between the
layers and the pressure was increased.
The results from the main test on the layered samples was very inconsistent when
applying the same voltage. The variation reduced the quality and credability of the
test making it difficult to interpret the results. The problem with the OMICRON
v
measuring system with interpreting the correct polarisation of the detected discharges
further complicate the intrepretation with reducing the ability to distinct the signal
from noise at unipolar applied fields.
The combined DC and AC voltage test show a clear distinction between the samples
with and without a cavity. The repetition rate during the cavity tests at 50 Hz AC
and DC voltage had a high initial value, decreasing down to a stable level. This is
consistent with the expectations that repetition rate is dependent of the conductivity
of the dielectric and the cavity. The 1 kHz AC and DC tests had a significantly
lower repetition rate, both initially and at the stable level, but a similar shape as the
repetition rate during the 50 Hz test. The pure DC voltage test is too inconsistent
and unreliable results to make any comparison to the results from the combined AC
and DC voltage tests.
The results indicate that more tests and further improvements must be done in the
experimental set-up to increase the reliability of the results.
vi
Sammendrag
Offshore HVDC overfringssystemer bruker omformerstasjoner mellom AC- og DC-
kraftnettet. Likespenning fra omformerstasjonene vil inneholde transienter. Disse lig-
ger i omr
adet 1 til 10 % av den nominelle likespenningen, avhengig av filteringen. Det
harmonisk innholdet i AC rippelen kan vre opptil 30-100 ganger bryterfrekvensen,
som kan vre rundt 1-2 kHz. Effekten av DC spenning overlagt en hyfrekvent AC
transient pa aldring av polymer isolasjoner er enn
a ikke kjent. Et velfungerende iso-
lasjonssystem er en av de viktigste forutsetningene for et driftssikkert hyspentanlegg,
og partielle utladninger er en av de mest krevende aldringsmekansismene man m a ta
hensyn til i et isolasjonsmateriale.
Hovedform alet med denne masteroppgaven har vrt a utvikle og bygge et eksperi-
mentelt oppsett for m aling av partielle utlandninger som kan simulere forholdene i
offshore kabelisolasjon nrt omformerstasjoner. Det eksperimentelle oppsettet skal
vre i stand til
a patrykke en kombinasjon av DC-spenning og hyfrekvent veksel-
spenning over et testobjekt.
Denne masteroppgaven er i hovedsak en eksperimentell oppgave og ble gjennomfrt
i tre seksjoner; sub-systemtest , systemtest og hovedtest. I sub-system testen ble en
analyse av en Behlman signalgenerator og ulike transformatorer gjennomfrt for a
finne en kombinasjon som var i stand til
a levere en 2 kVrms sinusformet spenning med
liten forvrengning over hele frekvensomradet 1-5 kHz. Behlman signalgeneratoren og
Messwandler Bau transformatoren ga en THD under 1% p a alle frekvenser og ble valgt
for videre testing.
Det eksperimentelle oppsettet ble bygget inne i et faradaybur. Systemtesten ble
gjennomfrt ved a styteste omgivelsene, oppsettet og komponentene. Det endelige
stybandet ligger p
a omtrent 0.6 pC, og en deteksjonsgrense p a 1 pC ble satt. Sys-
temtesten inkluderte en test pa de lagdelte prvene sammenlignet med en solid prve
for
a utelukke partielle utladninger i grensesjiktene. Resultatene indikerte at det
forekommer partielle utladninger i grenseflatene mellom lagene og trykket p
a prveobjektene
ble kt.
Resultatene fra hovedtesten p
a de lagdelte prvene var meget inkonsistente ved samme
p
atrykte spenning. Variasjonene reduserer kvaliteten og validiteten til resultatene som
gjr det vanskelig
a tolke resultatene. Problemet med OMICRON m alesystemet i
tolkningen av korrekt polaritet i sm
a utladninger gjr tolkningen av resultatene enda
vanskeligere.
Testresultatene fra den kombinerte AC- og DC-spenningen viser et klart skille mellom
vii
prvene med og uten hulrom. Repetisjonsraten i prvene med hulrom og p atrykt b
ade
50 Hz AC- og DC-spenning, hadde en hy initiell verdi som reduseres gradvis til et sta-
bilt niv
a. Dette er i overensstemmelse med forventningene om at repetisjonsfrekvens er
avhengig av isolasjonens og hulrommets ledningsevnen. Testene med kombinert 1 kHz
AC-og DC-spenning hadde et forlp lignende 50 Hz testen, men med en merkbar lavere
repetisjonsfrekvens, b ade i initiell verdi og stabilt niv
a. Den rene DC-spenningstesten
har for inkonsistente og up alitelige resultater til
a gjennomfre en sammenligning av
resultatene fra de kombinerte AC-og DC-testene.
Resultatene tyder p a at ytterligere forbedringer m
a gjres i det eksperimentelle oppset-
tet for
a ke p
aliteligheten av resultatene.
viii
Contents
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Transmission . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1.1 Converter Technology . . . . . . . . . . . . . . . . . . 1
1.1.1.2 Harmonics . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1.3 Cable insulation . . . . . . . . . . . . . . . . . . . . . 3
1.2 Aim of work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Theory 5
2.1 Partial discharges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Breakdown voltage of cavities . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.1 Pachens curve for breakdown in air . . . . . . . . . . . . . . . . 6
2.2.2 Partial Discharge under AC voltage . . . . . . . . . . . . . . . . 8
2.2.2.1 Voltage distribution . . . . . . . . . . . . . . . . . . . 8
2.2.2.2 Recurrence of discharges . . . . . . . . . . . . . . . . . 9
2.2.3 Partial Discharge under DC voltage . . . . . . . . . . . . . . . . 10
2.2.3.1 Voltage distribution . . . . . . . . . . . . . . . . . . . 11
2.2.3.2 Recurrence of discharges . . . . . . . . . . . . . . . . . 12
2.2.4 Partial discharge under combined AC and DC voltage . . . . . . 13
2.2.4.1 Voltage distribution within the dielectric . . . . . . . . 13
2.2.4.2 Repetition rate . . . . . . . . . . . . . . . . . . . . . . 13
3 Partial Discharge Detection 15

3.1 Detection circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.1 Sensitivity and noise . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.2 Partial discharge magnitude . . . . . . . . . . . . . . . . . . . . 16
3.2 Experimental set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2.1 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2.2 Test Cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2.3 Security Measures . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2.4 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4 Experimental 23
4.1 Sub-system test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.1.1 Experimental approach . . . . . . . . . . . . . . . . . . . . . . . 24
4.2 System test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3 Main test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
ix
4.4 Test Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.4.1 Manufacturing of the samples . . . . . . . . . . . . . . . . . . . 26
4.5 Test procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5 Results 29
5.1 Sub-system test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.1.1 Harmonics analysis . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.2 System test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2.1 Noise testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2.2 Combined 10 kVdc and 50 Hz 1 kVac,rms Voltage . . . . . . . . . 35
5.3 Main test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3.1 50 Hz 0.5 kVac,rms . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3.2 10 kVdc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.3.3 Combined 10 kVdc and 50 Hz 0.5 kVac,rms . . . . . . . . . . . . 47
5.3.4 Combined 10 kVdc and 1 kHz 0.5 kVac,rms . . . . . . . . . . . . 50
6 Discussion 53
6.1 Sub-system test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6.2 System test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6.2.1 Noise elimination . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6.2.2 Benchmarking . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
6.3 Main test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
6.4 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.4.1 Test samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
7 Conclusion 59
8 Further Work 61
Bibliography 63
A Experimental setup i
A.1 Experimental circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . i
A.2 Test Cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
B Hardware Data v
B.1 Transformers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
B.1.1 Petter Nodeland Designed Transformer . . . . . . . . . . . . . . v
B.1.2 Neon Transformer . . . . . . . . . . . . . . . . . . . . . . . . . . v
B.1.3 MESSWANDLER BAU Measuring Transformator . . . . . . . . vi
B.1.4 ARTECHE Measure Transformer . . . . . . . . . . . . . . . . . vi
B.2 Experminal setup equipment . . . . . . . . . . . . . . . . . . . . . . . . vi
B.2.1 DC Voltage Source . . . . . . . . . . . . . . . . . . . . . . . . . vi
B.2.2 AC Voltage Source . . . . . . . . . . . . . . . . . . . . . . . . . vii
C Sub-system test ix
x
C.1 Total harmonic distorition and FFT in Messwandler Bau Transformer . ix
C.1.1 Harmonic analysis at 1 kHz . . . . . . . . . . . . . . . . . . . . ix
C.1.2 Harmonic analysis at 2 kHz . . . . . . . . . . . . . . . . . . . . xii
C.1.3 Harmonic analysis at 3 kHz . . . . . . . . . . . . . . . . . . . . xv
C.1.4 Harmonic analysis at 4 kHz . . . . . . . . . . . . . . . . . . . . xviii
C.1.5 Harmonic analysis at 5 kHz . . . . . . . . . . . . . . . . . . . . xxi
D System test xxv

D.1 Partial discharge under combined 50 Hz AC Vac,rms =1 kVRMS and DC
Vdc =10 kV voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxvi
E Main test xxix

E.1 Partial discharge under DC Vdc =10 kV voltage . . . . . . . . . . . . . xxx
E.2 Partial discharge under DC Vdc =20 kV voltage . . . . . . . . . . . . . xxxii
E.3 Partial discharge under combined DC Vdc =10 kV and 50 Hz AC Vac,rms =0.5
kVRMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxxiv
E.4 Partial discharge under combined 1000 Hz AC Vac,rms =0.5 kVRMS and
DC Vdc =10 kV voltage . . . . . . . . . . . . . . . . . . . . . . . . . . xxxviii
xi
xii
1. Introduction
1.1 Background
The market for HVDC transmission systems has increased substantially over the last
years, driven by intercountry connections, technological developments and new re-
newable energy sources. The application of wind energy and especially offshore wind
energy is a keystone in the policy of several European countries due to the higher wind
speeds and hence a greater production. However, the further out to sea and the higher
the power capacity, the greater is the challenge of transmission of the energy to the
mainland. [1]
1.1.1 Transmission
From the offshore wind turbine collection point, the transmission link to the shore
may be High Voltage Alternating Current (HVAC) or High Voltage Direct Current
(HVDC). Today the majority of power transmission from offshore wind farms are
conducted using HVAC, but as the offshore wind farms get a higher installed power
capacity and are located further from the onshore grid connection point, HVDC has
proven to have several advantages compared to HVAC. [1]
The submarine AC cables generate reactive current due to its capacitance charac-
teristics. This reduces the power carrying capability of the cable and reactive shunt
compensation may be needed. [1] The capacitive charging current limits the possible
lenght of the AC cable to a maximum practical length of about 50 to 100 km and may
cause system instability. Also the compensation alternatives are expensive and hard
to implement offshore. [2] In addition, the AC grid of wind turbines and the onshore
grid is synchronously coupled; all fault is propagated from one to the other. However
the AC solution is a well-etablished system with low investment costs compared to
HVDC. [1]
In DC no reactive power is generated and all of the line capacity is used for transmitting
active power. The main advantages with choosing DC is the low losses and no charging
current in the cable. There is virtually no limit on connection distance beyond practical
constraints of the manufacturers and the laying of the cable. [1]
1.1.1.1 Converter Technology
An HVDC system takes electrical power from an AC network via a transformer, con-
verts it to DC at a converter station and transmits it to the receiving point, where it
1
is converted back to AC and connected to the existing grid.
The HVDC connection using thyristor based line-commutated converters (LCC) is an
established and well documented technology that can transfer large amount of en-
ergy over a large distance with low losses. In spite of its excellent results onshore
it has proven to not be very suitable in offshore application. The converter stations
and auxiliary equipment have demanding space requirements, resulting in big offshore
converter platforms. Furthermore, the LCC technology is highly susceptible to AC net-
work disturbances, resulting in converter commutation failures which can temporarily
shut down the entire HVDC system. The main limitations of the classical HVDC are
the need for large converter stations onshore and offshore and a strong AC network
in both ends to control the thyristors, making the LCC technology an unfavourable
choice for radial connections such as offshore wind energy. [2] [3]
Figure 1.1: HVDC transmission scheme
The voltage source converter (VSC) HVDC system uses Insulated-Gate Bipolar Tran-
sistors (IGBT). A self-commutating VSC based HVDC technology with pulse width
modulation (PWM). A technique where the commutations do not require an AC volt-
age source and are practically instanteneous. VSCs are able to independently control
both active and reactive power exchange with the AC grid and may help with voltage
regulation and are able to operate in weak or even dead AC networks. Power reversal
can be accomplished by keeping the same voltage polarity. The major drawbacks of
the VSC technology is the high converter loss caused mainly by switching losses in the
converters and the high investment cost for the converter stations. Nevertheless the
VSC HVDC is considered more suitable for offshore application due to its increased
controllability and compact architecture. [4]
1.1.1.2 Harmonics
The resulting DC voltage from the converter stations will contain transients originating
from the switching of the power electronics components. The AC ripple voltage on
2
the DC side of the HVDC stations can be in the range of 1-10 % of the nominal DC
voltage, depending on the size of the filter employed. Harmonic content of the ripple
can range up to 30-100 times the switching frequency, with the dominant components
close to the switching frequency, which can be around 1-2 kHz. [6]
Figure 1.2: Harmonic DC voltage [5]
The high voltage transients will experience dampening along the cable due to the
capacitative properties. The insulation closest to the power electronic converters will
experience the most severe harmonics. [7]
1.1.1.3 Cable insulation
Oil-impregnated cables have been the leading techology for HVDC power transmis-
sion since its introduction. Recently however, much research and development has
been devoted to the development of extruded HVDC cables due to their advantages.
Such as the low material and production costs, as well as their dielectric strength,
electrical resistivity and their physical properties such as resistance to cracking and
moisture penetration. The main reason that prevents the use of polymer cables for
HVDC transmission lies in the degradation of polymers under high electric stress. This
phenomenon is associated with the formation of space charges. Space charge accumu-
lation in polymeric material poses major scientific and technological problems. The
continued development of the polymer material and the converter technology increase
the possibility of making extruded cables a feasible choice for HVDC transmission. [8]
3
1.2 Aim of work
Polymers is vulnerble to partial discharges, one of the most important ageing mech-
anisms. Repetition rate during pure DC voltage is very low, several order less that
at applied AC voltage at 50 Hz. However introducing a AC ripple will increase the
repetition rate and may affect the ageing of the polymeric insulation. The goal of
this assigment is to develop an experimental set-up for detecting and measuring par-
tial discharge at a combined DC and high frequency AC voltage. This is done to
simulate a stress similar to the one the cable insulation is subjected to close to the
converter stations in an offshore HVDC transmission system. This is to determine the
partial discharge characteristics during combined DC and high frequency AC voltage
to further ascertain its influence on ageing of polymeric material.
The work is mainly experimental and is divided into several different sections before
the circuit is built and the testing on partial discharge in a test object with a cavity
can begin. One part is to test the Behlman signal generator together with transformers
to generate a low distortion sinusoidal voltage across a frequency range, 1-5 kHz at 1
kVrms and 2 kVrms . As the hardware analysis is done, noise elimination/attenuation
tests should be preformed on the components of the circuit, and the set-up as a whole to
increase the sensitivity. When the sensitivity is at an acceptable level partial discharge
tests are preformed on test objects. The test objects may either be solid or with a
artificially manufactured cavity. Both AC, DC and Cobined DC and AC voltage
is applied across the sample to investigate the partial discharge activity to further
ascertain the capability of the experimental set-up.
4
2. Theory
2.1 Partial discharges

Partial discharges (PD) are one of the main fault and ageing mecanisms in an insu-
lation system. Partial discharges are small sparklike short circuits across parts of the
insulation, an electrical discharge that does not completely bridge the electrodes. Par-
tial discharges will gradually lead to degradation and ageing of the insulation. How
fast depends on the type of insulation system, the material as well as the repetition
rate and magnitude of the discharges. Consequently the occurrence of PDs give a good
indication of the state of the insulation system and have an impact on its lifespan. [9]
Partial discharges include internal discharges, surface discharges and corona discharges.
This report concentrates on partial discharges in cavities. Cavities occur during the
manufacturing and ageing of the dielectric. Due to the generally lower permittivity and
breakdown strength in the cavity, compared to the surrounding dielectric, a local field
enhancement is developed and partial discharges can occur. Partial discharges result
in a release of energy and cause degradation of the surrounding insualtion through
a combination of chemical, mechanical, thermal and radiative processes. Especially
the cavity surface is affected and by-products may form on the surface. This causes
local field enhancements within the cavity which leads to a concentration of the partial
discharges. This may cause inception of electrical trees and could eventually lead to a
breakdown of the insulation system. [10]
Figure 2.1: Different types of internal discharges, a) isolated cavity, b) cavity with electrode
contact, c) cavity developed to an electrical tree, d) cavity in interface between different
dielectric materials. [11]
Two criteria must be fulfilled in order for partial discharge to occur in cavities within
a dielectric; there must be free electrons to initiate an electron avalanche and a
suffieciently high electric field to start an avalanche and create a breakdown path.
These critera are both stochastical phenomenons explaining why partial discharges
are a very complicated phenomenon that often display chaotic behaviour. [6]
5
Free electrons can originate from volume or surface generation of electrons, such as
ionization by energetic photons from radiation, electron release by ion impact, detrap-
ping of electrons from traps at insulation surfaces or thermal ionization. The critical
field value is assumed to follow Paschens law, that is dependent of the type of gas,
its pressure and the size of the cavity. In a small cavity there will only rarely get an
exited atom/molecule, even though the field is of sufficient size. Partial discharge is
therefor a stochastical phenomenon, that will vary in both size and frequency from
time to time. [12] [13]
The actual field distribution which initiates the discharge is dependent of the shape
and amplitude of the applied voltage, the characteristics of the involved materials,
and the geometry of the cavity. The DC stationary field distribution is governed
by geometry and conductivity of the materials. The conductivity of the materials
is significantly influenced by temperature, amplitude of the electrical field and the
present ions. The AC field distribution is determined by geometry and permittivity of
the involved materials. The permittivity will vary with the frequency of the field, due
to different polarization mechanisms in the dielectric materials. Permittivity will also
vary with temperature, but not to the same extent as the conductivity. [14]
2.2 Breakdown voltage of cavities
2.2.1 Pachens curve for breakdown in air
The electrical resilience of a cavity, or its breakdown strength, is dependent of the type
and the pressure of the gas as well as the cavity dimensions. The breakdown voltage
in a homogeneous field is a function of the product of the gas pressure and size of the
cavity at a given temperature. This is Paschens law. [15]
U = (pd) (2.1)
6
Figure 2.2: Paschens curve for air at T=20 C [16]
A breakdown in a cavity bound by a dielectric occurs at approximately the same

voltage as between equally spaced electrodes. This voltage can be found in Paschens
curve. Considering a gap of a given size, as the pressure decreases so will the gas
density and the electron free path increases. Resulting in a reduced number of collisions
and a increase of the breakdown strength. Several factors may cause the breakdown
voltage to be lower than predicted. The walls of the cavity may be covered by static
charges produced when the cavity was formed or left after previous partial discharges.
Semiconducting layers may also occur on the cavity walls, more or less short circuiting
the cavity. This may cause a significant increase of the breakdown stress. [9]
The discharges will start at some critical field, Ep , given by Pachens law. As long
as heightradius of the amplitude, the amplitude of the electric field in the cavity is
given by:
Vc
Ec = (2.2)
h
The voltage across the cavity at which partial dicharges start to occur when the voltage
is increased is called the inception voltage, and the corresponding stress is called the
inception stress. When a partial discharge occurs at a voltage Vc,inception across the
cavity, the cavity is partly or completely discharged through a breakdown path in the
field direction. When the discharge reaches the surface of the dielectric it is followed
by discharges along the surface. After a few nano seconds the voltage across the void
has dropped to the extinction voltage, also known as the remanent voltage, Ve and
the PD is extinguished. A considerable part of the cavity surface is affected during the
7
PD and a surface charge is deposited on the surface of the cavity. Immediatly after
discharge the electric field in the cavity will have a size of only a fraction of the field
before the discharge occured. The cavity is once againg insulating until the voltage
across it exceed Vc,inception and a new partial discharge occur. [15] [14]
2.2.2 Partial Discharge under AC voltage
Figure 2.3: Internal discharge equivalent circuit under AC conditions [9]
Internal discharges at AC voltage can be described using the well-known abc equivalent
circuit illustrated in Figure 2.3. The intact part of the dielectric is described by a
capacitance Ca while the dielectric in series with the cavity is represented by Cb and
the cavity is represented by a capacitance Cc , which is shunted by a breakdown path. [9]
2.2.2.1 Voltage distribution
The AC voltage across a cavity can be expressed by the following equation:
Vc,ac = KAC Vac sin(t) (2.3)
ac is the amplitude of the AC voltage with an angular frequency of =2f applied

V
across the dielctric and KAC is given by the capacitances of the insulation and the
cavity illustrated in Figure 2.3:
Cb
KAC = (2.4)
Cb + Cc
The inception voltage for partial discharges measured over the terminals at AC voltage
is defined as:
8
1
VP DIV,AC = ( )Vc,inception (2.5)
KAC
[6]
2.2.2.2 Recurrence of discharges
Figure 2.4: Discharge recurrence during AC voltage [17]
Figure 2.4 illustrates the PD activity in during AC voltage, the voltage across the cavity
is given by v(t), the ignition voltage is Vi and the extingction voltage is denoted Vr .
The voltage across a cavity when no partial discharges occur is given by Equation 2.3.
After a partial discharge occur the voltage across the cavity drops and the voltage
immediatly after discharge is given by superposition of the main electric field and the
field of the surface charges at the cavity walls left by the PD. [9]
Vc,ac,pd = KAC Vac sin(t) U (2.6)
In this formula U=Vc,inception -Ve is the voltage drop across the cavity during a partial
discharge. A new PD will occur when the voltage in the cavity again reachs the
critical value, Vc,ac,pd Vc,inception . The same will occur in the negativ half period of
the alternating voltage. [15]
For Vac VPDIC,AC the discharge frequency is given by:
4f KAC Vac
fpd,ac = (2.7)
Vc,inception
[6]
The repetition rate of partial discharges is difficult to determine since partial discharge
inception and extinction voltage is strongly affected by the deposited discharge and the
shape of the voltage applied. Ripple on the voltage may induce partial discharges at
9
an applied voltage below the inception voltage and the waiting time for free electrones
may delay the discharge, letting the voltage increase to a level above breakdown voltage
before an avalanche occurs.
The partial discharge inception voltage have been obseved to be dependent on the
frequency of the AC voltage and increase at higher frequencies. The frequency depen-
dence can be described by certain parameters; the statistical time lag, charge transport
in the dielectric and the cavity surface decay. If the waiting time for a free electron
is much shorter than the period time of the voltage there is no frequency dependence
in the PD. However if the statistical time lag is in the same range as the period, the
partial discharges is shifted forward in phase and occur at higher voltage due to the
lack of free electrons. The result is fewer partial discharges per period and larger mag-
nitudes. The surface charges deposisted by previous partial discharges decay over time
due to conduction and recombination and if there are significant surface charge decay
the field in cavity decreases and hence decrease the number of partial discharges. If
there is a significant transport of surface charge in the cavity from shallow traps to
deeper traps it will reduce the surface emission of starting electrons. This effect is
intensified by increased frequency and will reduce the number of partial discharges per
period. [10] [18]
2.2.3 Partial Discharge under DC voltage
Figure 2.5: Internal discharge equivalent circuit under DC conditions

[9]
To illustrate partial discharge under DC conditions, the well known abc circuit from
AC partial discharge is extended with resistances, as seen in Figure 2.5. During AC
voltage the capacitive voltage determine the field but at stationary DC voltage it
is the resistive voltage that determine the field. During the increase and decrease
of the applied DC voltage one will often get a capacitive voltage distribution across
the insulation, and hence comparable conditions as during AC voltage. Ca and Ra
10
represent the properties of the entire dielectric between the cavity and the electrodes
and Cb and Rb represent the properties of the dielectric in series with the cavity, Cc is
the capacitance of the cavity and Rc represent the surface resitance of the void. [19] [12]
The time constant, , for charging of the cavity is given by:
Rb Rc KDC 0 b KDC
=( )(Cb + Cc ) = Rb Cb = (2.8)
Rb + Rc KAC b KAC
for a sylidrical cavity the time constant can be written:
D0 c + 0 b
= (2.9)
Dc + b
Where c and b is the relative permittivity of the cavity and the dielectric respec-
tively and c and b is the conducitivity cavity and the dielectric. D is given by the
dimentions of the cavity:
H h
D= (2.10)
h
2.2.3.1 Voltage distribution
The voltage distribution in the cavity under stationary DC conditions, assuming an

extinction voltage Ve =0 is given by:
t
Vc,dc = (1 e )KDC Vdc (2.11)
where Vdc is the applied DC voltage across the test sample and KDC is given by
Equation 2.12.
Rc
KDC = (2.12)
Rc + Rb
.
The partial discharge inception voltage during applied DC voltage is given by:
1
VP DIV,DC = ( )Vc,inception (2.13)
KDC
The inception voltage for DC is not well defined, because there will be no discharges
at the inception voltage, Vc,dc =VPDIV,DC . In practice, the inception voltage for partial
discharges under applied DC voltage is better defined as the voltage that will give a
certain number of discharges per time. It can be accepted that the DC voltage across
the cavity must be well above the inception voltage before discharges occur. [6]
11
2.2.3.2 Recurrence of discharges
When the DC is applied the field is capacitively graded. In this area the PD repetition
rate is high and follows the polarization current closely. After all the polarization
processes are completed, the field in the cavity is resistive and discharges only occur
infrequently. Because of the resistances located around the cavity is finite the surface
charge from the previous PD will decay and a new discharge will occur. The frequency
of which these discharges occur is given by the local resistances. [9] [19]
After a discharge occur the next one can take place after a recovery time tr . The
statistical timelag tL is assumed equal to zero, hence there is no waiting time for
free electrons. The equation below is an expression for the minimal time between
discharges. [19] [6]
VP DIV,DC
t0r = ln[1 ] (2.14)
VDC
for VDC > VPDIC,DC .

The repetition rate is given by equation 2.15 and is found using the first term in the
Taylor series.
1
n= (2.15)
t0r
1 Vdc
n (2.16)
VP DIV,DC
The repetition rate during DC voltage is proportional to Vc,dc and hence the external
voltage. It can be seen from Equation 2.16 that the repetition rate during DC voltage
is dependent of the time constant for the charging of the cavity. The charging time
is a function of both the permittivity and the conductivity of the involved materials.
Since the conductivity of a material is strongly dependent of the temperature and the
applied electric field, so is the repetition rate. The conductivity of a dielectric changes
by many orders of magnitude when the temperature is raised from a temperature of
20 C to a temperature between 60 and 90 C. The repetition rate is directly affected by
the increase in conductivity. If the conductivity is increased by two orders, so would
the repetition rate. [19] The repetition rate will also vary with time from the decrease
in the volume conductivity and reach a stable level after some time has passed. The
stable repetition rate is reached after a time dependent of the temperature. [9] [19] [20]
The repetiton rate under DC stress is several order of magnitude less than at AC
voltage. Only at very high stresses the repetition rate during DC reaches a level close
to the repetition rate at AC voltage. [19]
12
2.2.4 Partial discharge under combined AC and DC voltage
The abc equivalent for combined DC and AC voltage is the same as the one illustrated
in Figure 2.5 with resistances to take into account the stationary DC voltage and
capacitances for the alternating voltages.
2.2.4.1 Voltage distribution within the dielectric
A combined DC and AC voltage applied across the experimental set-up can be ex-
pressed as:
Vinput = Vdc + Vac sint (2.17)
ac is the amplitude of the AC voltage with an

where Vdc is the DC voltage and V
angular frequency of =2f.
The resulting peak voltage over the cavity, Vc , will be given by superposition of the
combined voltages across the cavity.
t
Vc = Vc,dc + Vc,ac = (1 e )KDC Vdc + KAC Vac sin(t) (2.18)
2.2.4.2 Repetition rate
The equation for the repetition rate under applied DC voltage, Equation 2.14 can be
modified to take into account the AC voltage. Since the frequency of the AC voltage is
much higher than the time constant given by Equation 2.8, it will appear as if the DC
voltage is offset by the peak voltage of Equation 2.3. The new time between discharges
can be written as: [6] [19]
VP DIV,DC Vc,ac
tr = ln[1 ( )] (2.19)
VDC Vc,dc
which can be rewritten as
VP DIV,DC VAC
tr = ln[1 ( )(1 )] (2.20)
VDC VP DIV,AC
.
The repetition rate n is the inverse of tr and can be approximated by ignoring all terms
after the first one in the Taylor series of n:
1
n= (2.21)
tr
13
1 Vdc 1
n ( )( ) (2.22)
VP DIV,DC 1 Vac
VP DIV,AC
Equation 6.2 is similar to the repetition rate during DC voltage with an additional
section to take into account the superimposed AC voltage. The repetition rate during
combined DC and AC voltage will have a similar shape as the voltage during DC
voltage, but will increase with an increased AC voltage. With an applied AC voltage
of half the inception voltage during AC voltage the repetition rate during a combined
DC and AC voltage will be twice the frequency at DC voltage. It also states that n
goes to infinity when V ac approaches VPDIV,AC . Due to Equation 2.7 this is clearly not
the case. [6]
Figure 2.6: Repetition rate illustration for combined DC and AC voltage
14
3. Partial Discharge Detection
3.1 Detection circuit
Figure 3.1: Equivalent circuit of straight detection circuit
Figure 3.1 illustrates a equivalent straight detection circuit. When discharges occur in
the test object Ct the voltage across the test object experience a sudden voltage drop
corresponding to the apparent charge. The coupling capacitor Ck in parallell with the
test object will deliver a current into the test object to counteract the voltage difference
that originate between them due to discharge, and create a low impedance circuit for
the high frequency signals. The measuring impedance can be placed in series with
either the coupling capacitor or the test object. The voltage drop during a discharge
will result in a small transient current and the measuring impedance will result in a
detectable voltage due to this current. The detection equipment is connected across
the measuring impedance and must be high frequency detection equipment due to the
short duration and high frequency of the voltage. [12] [15]
The coupling capacitor should be larger than the capacitance of the testobject to
get a good sensitivity. The measuringimpedance will affect the measured signal and
the data possible to get out of the measurement, hence the choice in design is very
important. [12]
3.1.1 Sensitivity and noise
During partial discharge measurements the sensitivity of the detection circuit must be
very good. Generally the aim is to maximize the signal to noise ratio by minimizing the
noise, maximizing the signal or a combination of both. The reliability and sensitivity of
15
detection circuits are strongly affected by interference. Interference can originate from
the main delivering the voltage to the equipment, the circuit itself, contact noise or
from the surrounding, since all wiring basically works as antennas when not connected.
[9] Due to the low repetition rate during pure DC measurements the signal to noise
ration should be very low to get a clear picture of the actual partial discharges occuring
in the test object.
A majority of the noise problems can be solved using electro-magnetical shields/fara-
day cage, careful design and testing all the equipment for PD separately, and hence
eliminating them as sources of partial discharge in the sample, since all types of dis-
charges have its own characteristic signature. Radio noise can be attenuated using
a faraday cage. Corona discharges can be eliminated using field grading spheres and
torids on sharp edges. Ground loops can be eliminated by careful design, grid inter-
ference can be attenuated using a filter on the main and floating objects can be pretty
much eliminated by having a neat and clean test cell. Choosing a frequency band
where the external noise is low and the signal from the defect high is important to
distinct PD from noise. [21]
3.1.2 Partial discharge magnitude
The magnitude of the detected discharges may vary in size. The magnitude depend
on where in the dielectric the discharges occur.
The apparent charge measured across the dielectric is not the same as the charge
released in the cavity due to a partial discharge. The model during partial discharge
detection is used to calculate the charge in the cavity rather than the one detected.
The capacitance of the dielectric is dependent of the permittivty and the dimentions
of the cavity and test object,
A
C = o r (3.1)
d
where o is the permittivity of vacuum=8.85410-12 and r is the reltive permittivity of

the material. A is the area and d is the height of the different parts of the dielectric.
Using the dimentions listet in table 4.2 one can calculate the capacitances of the test
objects using equation 3.1 getting Ca =20.28 pF, Cb = 0.4867 pF and Cc =0.278 pF.
The value of the coupling capacitor is Ck =800 pF.
The magnitude of the discharges across the entire cavity can be calculated. When a
discharge occur all the components is considered to be capacitances due to the high
frequency.
A charge Q is the product of the capacitance and the voltage across it.
Q=VC (3.2)
16
Assuming a remanent voltage (extinction voltage) equal to zero the charge resulting
from a partial discharge across the entire area of the cavity is found using Equation
3.3 where Vc is peakvalue of the breakdown voltage Vpaschen 950 Vpeak .
Qcavity = Vc Cc (3.3)
Resulting in a charge Qcavity =265.25 pC. The measured value for a discharge of this
size at the breakdown voltage is given in Equation 3.4. The value of Ceq is given by
the equivalent before discharge occur, and V is given by voltage division.
Qmeasured = Ceq V (3.4)
Cb Cc
Ceq = Ck + Ca + (3.5)
C b + Cc
Cb
V = Vc (3.6)
Cb + Ca + Ck
This gives a measured charge, corresponding to the cavity charge Qcavity , Qmeasured =452.3
pC.
17
3.2 Experimental set-up
The test circuit is based on a straight measuring circuit and is designed to able to apply
both high frequency AC and DC voltage across the test cell. An equivalent circuit for
the experimental set-up is illustrated in Figure 3.2. The experimental set-up is placed
inside a faraday cage to attenuate the noise from the radiowaves, and corona toroids
and spheres are connected at all the sharpe edges of the set-up to prevent corona
discharges during testing. Pictures of the experimental set-up can be found in section
A.1 in the appendix.
Figure 3.2: Equivalent circuit of the experimental set-up
Cac is a buffer capacitance to prevent the DC current from entering the AC voltage side
of the set-up. Rdc is a buffer resistance to prevent the AC current from entering the
DC voltage source. One of the key parameters in a test set-up like this, with both AC
and DC voltage source, is the buffer components and prevent creating a low impedance
path for the transient currents from the partial discharge activity in another branch
than through the measuring impedance.
Table 3.1: List of parameters in the experimental circuit
Parameter Definition Value

R Series resitance 100
Cac,buffer Buffer capacitance 10 nF
Rdc,buffer Buffer resistance 100 M
Ck Coupling capacitor 800 pF
Rk Coupling resistance 280 M
Ct Testobject capacitance 200 nF
Rt Testobject resistance 1 T
18
The DC voltage across the test object can be expressed:
Rk
VDC,testobject = Vdc (3.7)
Rdc,buf f er + Rk
since Rt Rk .
The AC voltage across the tesobject:
Cac,buf f er
VAC,testobject = Vac (3.8)
Cac,buf f er + Ck + Ct
The resulting voltage across the testobject is:
Vtestobject = VAC,testobject + VDC,testobject (3.9)
3.2.1 Hardware
Generation
Behlman Signal generator a high frequency AC voltage source with a voltage

range of 0-270 VRMS and frequency range 45-10,000 Hz.
FUG DC voltage source with a voltage output range of 0-35 kV
Messwandler Bau Measuring Transformer
Some additional information on the hardware can be found in Section B.2 in the
appendix.
Measurement
OMICRON MPD600 measurement and analysis tool for partial discharges is used
for detection of the partial discharges. The OMICRON consist of a measuringshunt
connected in the detection circuit. The measuring shunt is connected to the measuring
unit that register and measures the size of the transient voltage signals. The analouge
signal is converted to digital signals and transmitted to the computer through fiber
optics for further treatment.
The detection limit will be limited by the sensitivity of the registrationinstrument and
the noiseband. The design of the experimental laboratory set-up is very important
with regard to interference reduction/elimination.
19
3.2.2 Test Cell
The set-up consist of a test cell with parallel plane Rogowski shaped electrodes with a
surface area of approximately 19.62 cm2 . During testing the cell is filled with degassed
mineral oil to prevent discharges at the edges of the electrodes. The oil is heated by a
heat coil, to maintain a temperaterature within 0.1% during testing. The pressure
applied to the sample is controlled by weights and are set to 50 kN/mm2 and 75
kN/mm2 . A picture of the test cell can be found in section A.2 in the appendix.
Figure 3.3: Test Cell
20
3.2.3 Security Measures
A door switch is connected to a connection box between the voltage sources

and the experimental set-up to eliminate the possibility of applying the voltage
without a shut door and cut the connection if the door is opened during tests.
A earthing rod is connected to the same security switch as the door. The rod
must be placed on the circuit before handling it to prevent shock from charged
components in the circuit.
A smoke detector is connected to the circuit to detect any development of smoke
from overheating in the oil.
Two temperature sensors in the test cell is programmed to terminate the test if
the temperate exceed 100 C.
Oil level measurer are set to terminate the test if large amounts of oil is leaking
from the test cell.
3.2.4 Calibration
The test circuit must be calibrated in order to measure the correct partial discharge
magnitudes. This is done using a calibrator and applying a known charge pulse across
the testcell with a test sample in place. The applied charge generates a current in the
circuit and the voltage across the measuring impedance due to the charge is detected
by the measuring equipment. The relation between the applied charge and related
voltage transient measured is the calibrationfactor. When this is know the OMICRON
detection program is able to convert the detected voltage signal to the correct charge
magnitude. Calibrating must be preformed each time a test object of different height
or material is used, since this changes the characteristics of the detection circuit and
hence the detected transient voltage.
In the experimental set-up two OMICRON units are placed in parallell to detect partial
discharges at two different gain levelse simultaneously. The two OMICRON MPD600
units in the experimental set-up is calibrated at 5 pC for the high gain unit and 50
pC for the low gain unit.
21
4. Experimental
The testing is performed to build a interference free experimental set-up illustrated in

figure 3.2 for partial discharge testing and is performed in sections where the further
work is based on the results from the previous test.
The testing on the experimental set-up is performed in a given order to eliminate
problems as they arise.
Sub-system test; testing on Behlman signal generator and transformers to find
a low noise generation system for high frequency AC voltage.
System test; testing of the external noise and interference originating from the
circuit to eliminate them, and preforme tests on the layered samples without a
cavity compared to solid samples to determine if discharge occur in the interfaces.
Main test; tests on the layered samples described in Section 4.4 at different
voltages and frequencies to further test the ability of the detection circuit by
interpreting the results.
4.1 Sub-system test

The experimental set-up should be able to deliver both DC and high frequency AC
across the test objects. The goal is to have a sinusoidal voltage with as little ripple as
possible since they may lead to partial discharges at a lower applied voltage than the
breakdown voltage. [9]. During the project assignment it was found that the Behlman
signal generator is a viable choice for generating high frequency sinusoidal voltage. It
is easy to regulate and can supply a wide range of frequencies with low distortion. The
results from the project did not have a sufficiently high frequency resolution and a new
test is nessesary to determine the distortion of the seemingly sinusoidal voltage. [22]
The voltage span of the Behlman signal generator is 0-270 VRMS and a transformer is
needed to reach the desired voltage for partial discharge testing.
The test is performed on both the Behlman Signal Generator and several transformers
to be able to find a transformer able to handle the high frequencies, 1-5 kHz, and still
reach a voltage of 2 kVRMS . The difficulty with high frequency transformers is the risk
of the core being saturated at higher frequencies distorting the output voltage. The
transformers to be tested:
The transformer design by Petter Nodeland [23] [24]
Neon Transformer
23
Messwandler Bau Measuring Transformer
Arteche Measure Transformer
Some additional information on the transformers and the Behlman signal generator
can be found in section B.1 in the appendix.
4.1.1 Experimental approach
The analysis to determine the quality of the output voltage were performed by connect-
ing the Behlman source and the transformer in series with a resitance on the primary
side of the transformer and a capacitor of 800 pF on the transformers secondary side.
The capacitor on the secondary side is ment to simulate a load similar to the one it is
subjected to during the partial discharge detection tests. The resistance is placed in
series with the primary side of the transformer to limit the current to the transformer,
get a higher voltage from the Behlman and a less sensitive controlsystem.
Initially a preliminary test is performed by observing the transformer ratios and the
voltage shape on the secondary side by placing a high voltage measurment probe across
the capacitor connected to a high resolution Yogogawa oscillioscope. The transformer
with the most promising results during the observations done with the oscillioscope is
investigated further with a fast fourier transform (FFT) of the signal.
The neccessary data for the FFT is collected using the yokogawa oscillioscope and the
corresponing computerprogram Xviewer for saving the collected data as csv files. The
text file is imported into a MATLAB script for plotting of the fast fourier transform
and finding the total harmoinc distorition (THD) of the signal. To get an satisfac-
tory analysis a suffiecient amount of data is needed to observe as much of the signals
components as possible. A high sampling frequency must be used to detect the har-
monic content of the voltages with high frequencies. The sampling where set to 62.5
M Samples and the aqusition time 20 ms at the yokogawa oscillioscope.
Tests are performed at 1 kHz, 2 kHz, 3 kHz, 4 kHz and 5 kHz at 1 kVRMS and 2
kVRMS .
4.2 System test

All the system tests is performed using the OMICRON measurement system.
Noise and frequency band
Noise measurement tests are performed to eliminate the sources of interference if pos-
sible. The tests are performed with individual parts connected both with and without
applied voltage to detect whether discharges occur. The testobject used during the
24
noise tests is a teflon plate to assure that no discharges occur in the sample and only
the set-up is tested.
A measurement on the noise band is performed to ascertain at what limit partial
discharges can be detected and not be strongly affected by underlaying noise. This is
done by turning all the circuit equipment on, with no applied voltage and using the
OMICRON program to detect the limit.
Finding the appropriate frequency band to preforme the tests are done using the
OMICRONS FFT function and is done in the same manner as detecting the noiseband,
observations and choosing an area where the noise content is low.
When the noise is at low level and all the segments of the circuit is tested and approved
the experimental circuit is completed and approved for further testing.
Benchmarking
The benchmarking will give an indication of the partial discharge behaviour in a layered
sample compared to a solid one to assure that no discharges occur in the interfaces of
the test objects decribed in section 4.4. The set-up tests are done using solid layered
test sample and a solid PET sample with a thickness of 250 m without cavity.
The tests are performed applying a combined DC and AC voltage, Vdc =10 kV and
VAC =1 kV at 50 Hz and a pressure of 50 kN/mm.
4.3 Main test

All the main tests is performed using the OMICRON measurement system.
The test are ment to be a frame of reference by observing the discharge magnitudes
and repetition rates at different applied voltages to be able to have some data to
compare to the expected results based on theory. The results is used for a comparison
to determine the capability and reliability of the experimental set-up.
The further testing is performed on the 3100 m samples with and without cavity
r= 1 mm. The tests are performed under equal conditions, a temperature of 60 C and
a pressure of 75 kN/mm.
Table 4.1: Main tests
Applied voltage Run time

Pure AC voltage, Vac =500 Vrms , f=50 Hz 24 hours
Pure DC voltage, Vdc =10 kVdc 24 hours
Combined AC and DC voltage, V= 10 kVdc +500 Vac,rms , f=50 Hz 24 hours
Combined AC and DC voltage, V= 10 kVdc +500 Vac,rms , f=1 kHz 12 hours
25
4.4 Test Objects
The test objects are created out of Polyethylene Terephthalatem (PET) sheets. The
choice of using PET sheet is due the consistent thickness, low amount of contami-
nants and constant dielectric properties. This makes the test objects simpler to create
without large variations in characteristics of the samples.
The test objects consist of three circular layers of the PET sheet glued together at the
edges. The middle sheet can be altered to either make a sample without or with a
cavity of a given radius. The thickness of the PET sheet is 100 m so the thickness of
the finished sample will be 300 m and the radius of the cavity can easily be changed
based on the desired size of the artificial cavity for the test.
Table 4.2: List of parameters for the test objects
Parameter Definition Value

b Relative permittivity of dielectric 3.5 [F/m]
c Relative permittivity of cavity 1 [F/m]
b Conductivity of dielectric 1.65e [S/m] at 80 C
-15
Conductivity of dielectric 1e-15 [S/m]

H height of test object 300 [m]
h height of cavity 100 [m]
R effective radius of electrodes 25 [mm]
r radius of cavity 1 [mm]
Figure 4.1: Test Object
4.4.1 Manufacturing of the samples
A stensil and press are used to create the circular objects from the PET sheet. The
middle layer is either solid or have a hole drilled in the middle to create a cylindrical
cavity filled with atmospheric air.
The samples with the drilled hole is inspected in a microscope in order to discard the
object with sharp edges or irregularities around the cavity that may lead to local field
26
enhancements. The samples are sorted into categories A, B and discarded, based on
the findings. This gives a more correct reading of the relevant partial discharges, e.i
the ones occuring due to the artificial air cavity.
Since partial discharges occur in the imperfections in a dielectric it is important to
manufacture the samples in a clean environment.
The three PET sheets are cleaned using isopropanol and dryed.
A catalyser (770) is applied along the edges and allowed to dryed before the sheets
are placed together.
The sheets are placed in a press and a pressure of approximately 3 bar is applied
When the objects is placed under pressure, glue (406) is applied between all the
layers along the edges.
After 2-3 hours the glue have sufficiently hardened and the test object is complete.
The final samples are visually examined for contaminants before they are approved
and numbered.
4.5 Test procedure

The test object to be tested is cleaned with isopropanol and dryed to remove any
contaminants before it is lowered into oil in the testcell. Here it is moved around to
remove airbubbles before being placed on the bottom electrode. The upper electrode
must be partially lowered into the oil at an angle to let the airbubbles rise to surface.
When the airbubbles have risen the lid is put in place and the electrode is lowered
firmly in a vertical angle onto the sample to prevent the sample of sliding away. When
the electrode is in place with the sample in the correct position the pressure needed is
applied using the weights before connecting the electrode to the circuit and connecting
the corona toroid before closing the faraday cage.
If the equipment have been in off posistion for a long time the sources are turned
on, but with no voltage applied to reach operating temperature before applying the
voltage. When the equipment have been on for at least an hour the voltage is applied.
The DC voltage is tuned up slowly not induce any partial discharges.
27
5. Results
The results from the tests prefomed using the OMICRON measuring equipment a
digital filter is used on the raw data. The filter takes out symmetrical, high frequency
discharges out of the raw data. High frequency in this context is above 1 Hz, most of
the discharges detected have a repetition rate of 180 per hour and lower, i.e. with a
frequency well below 1/20 Hz. The filter was able to take out a small portion of the
discharges responsible for sudden jumps in the repetition rate.
5.1 Sub-system test

The preliminary testing of the Behlman source and the transformers gave som un-
expected results. The Petter Nodeland transformer showed a high current in the
transformer, resulting in a high temperature in the windings, measured to be 97 C, at
a voltage of 564 V at the secondary side of the transformer at 5 kHz.
The test on the neon transformer showed a transformer ratio of 1:100 in the 750 Hz
area but that the transformer ratio declined rapidly with increased frequency with a
1:3 ratio at 5 kHz. Due to the ratio, the transformer cannot be used to reach 2 kVrms
with the Behlmans maximum output of 270 Vrms .
During the preliminary tests on the Arteche Measure Transformer it is revealed that
the ratio between the primary and secondary side are very dependent on the frequency
applied. The ratio differs from 1:500 at around 1 kHz to 1:11 at 5 kHz. The voltage
looks good, but due the low ratio at higher frequencies it cannot be used to deliver
the desired voltage.
In the preliminary tests performed on Messwandler Bau Measuring Transformator
it became clear that both the voltage and the ratio looks good. Across the entire
frequency range of 1 kHz- 5 kHz the ratio ranges between approximately 1:100 and
1:400. The Messwandler Transformer is used for further testing.
29
Figure 5.1: Transformer ratio of the Messwandler Bau
5.1.1 Harmonics analysis
The total harmonic distortion (THD) of the Messwandler Bau measuring transformer
is calculated on harmonics of the voltage signal.
Table 5.1: Total harmonic distortion for the Messwandler Bau at 1 kV and 2 kV at frequencies
1 kHz - 5 kHz.
THD
Frequency 1 kVrms 2 kVrms
1 kHz 0.90 % 0.73 %
2 kHz 0.81 % 0.47 %
3 kHz 0.70 % 0.50 %
4 kHz 0.58 % 0.33 %
5 kHz 0.40 % 0.29 %
Figure 5.2 and Figure 5.3 on the two next pages illustrates the results from the FFT
of the signal on the secondary side of the transformer during the testing at 1 kHz and
1 kV and 2 kV, respectively.
The total harmonic distortion is below 1% for all the voltages and frequencies tested.
It is determined that a filter is not neccessary before continuing with the design of the
experimental set-up. The figures and additional data for the tests across the entire
tested range of frequencies and 1 and 2 kV can be found in appendix C.
30
FFT analysis plot 1 kHz and 1 kV
31
Figure 5.2: FFT plot 1 kHz and 1 kV

FFT analysis plot 1 kHz and 2 kV
32
Figure 5.3: FFT plot 1 kHz and 2 kV

5.2 System test
5.2.1 Noise testing
The tests on the different segments of the circuit were carried out and alterations were
done to lower the noise band. Resulting in a noise level of approximately 0.6 pC and a
very good starting point for further testing. Figure 5.4 and 5.5 show the noise content
and the FFT on the PD input, respectively. The frequency integrating range is set
in a low noise area 1.15 -1.45 MHz and is marked in purple in the FFT illustration.
The two peaks in the FFT plot, one small to the left of the frequency band and the
larger one to the right, originate from an external noise source and have a sporadic
occurence.
Figure 5.4: Noise content when all equipment is connected and in on state, but no voltage
applied
33
Figure 5.5: FFT and frequency band choice for the circuit. All equipment is connected and
in on state, but no voltage applied
34
5.2.2 Combined 10 kVdc and 50 Hz 1 kVac,rms Voltage
P = 50 kN/mm
Table 5.2: PD data for PET 2, Solid 250 m sample without cavity, 10 kVDC , 1 kVAC,RMS ,
f= 50Hz, P=50 kN/mm, t=18 hours, T=40 C
PD data for sample 2, without cavity

Total number of discharges 81
Maximum magnitude 5.4893 pC
Minimum magnitude -5.4565 pC
Mean discharge magnitude -1.3283 pC
Median discharge magnitude -1.1444 pC
Table 5.3: PD data for 3100 m layered samples without cavity, 10 kVDC , 1 kVAC,RMS , f=
50Hz, P=50 kN/mm, t=24 hours, T=60 C
PD data for layered samples, without cavity

Sample3 Sample 4 Sample 5
Total number of discharges 585 2595 4123
Maximum magnitude 7.0461 pC 75.6414 pC 84.7723 pC
Minimum magnitude -30.94842 pC -296.452 pC -466.121 pC
Mean discharge magnitude -1.1456 pC -11.9983 pC -21.3473 pC
Median discharge magnitude -1.3518 pC -2.8603 pC -5.1317 pC
The plots of the discharge magnitude and repetition rate of sample 2 and sample 3
can be seen on the next two pages in Figure 5.6 and 5.7, respectively. It should be
observed that the sampling time for sample 2 is 18 hours and the temperature is 40 C.
The plots for sample 4 and 5 can be found in Figures D.1 and D.2 in section D.1 in
the appedix.
From the data displayed in Table 5.3 above it can be seen that a high amount of
discharges were detected in the layered samples without a cavity. In addition the
discharges in the layered samples have a high magnitude compared to the discharges
observed in sample 2, displayed in Table 5.2. Since these samples does not include a
cavity these observations rises suspecion of discharges in the interfaces between the
layers. The weight on the samples were increased to a corresponing pressure of 75
kN/mm for the continued testing.
35
36
Figure 5.6: PET 2 plot. Solid 250 m sample without cavity, 10 kVDC , 1 kVAC,RMS , P=50 kN/mm, t=18 hours, T=40 C
37
Figure 5.7: PET 3 plot. 3100 m layered sample without cavity, 10 kVDC , 1 kVAC,RMS , f=50 Hz, P=50 kN/mm, t=24 hours, T=60 C
P= 75 kN/mm
Table 5.4: PD data for PET 2.2, Solid 250 m sample without cavity, 10 kVDC , 1 kVAC,RMS ,
f=50Hz, P=75 kN/mm, t=24 hours, T=60 C
PD data for sample 2.2, without cavity

Table 5.5: PD data for PET 6, 3100 m sample without cavity, 10 kVDC , 1 kVAC,RMS , f=
50Hz, P=75 kN/mm, t=24 hours, T=60 C

The plots of the discharge magnitude and repetition rate of sample 2.2 and sample 6
can be seen on the next two pages in Figure 5.8 and 5.9 respectively.
From the data diplayed in table 5.4 and table 5.5 above it can be suggested that increase
of the pressure did not have an effect worth mentioning. Only a marginal reduction
in magnitude and number of discharges for the layered sample 3 in the previos section
and 6 can be observed by the data. However, looking at the plot for sample 3, Figure
5.7, and for sample 6, Figure 5.9, a noticable reduction in the repetition rate after
some time has passed can be observed, and may indicate that the increased pressure
did have some effect.
It was decided to apply a lower AC voltage for the main test, since the ignition voltage
of the air cavity is approximately 1 kVac,rms and the further tests are applied a voltage
of 0.5 kVac,rms
38
39
Figure 5.8: PET 2.2, Solid 250 m sample without cavity, 10 kVDC , 1 kVAC,RMS , f=50 Hz, P=75 kN/mm, t=24 hours, T=60 C
40
Figure 5.9: PET 6, 3100 m layered sample without cavity, 10 kVDC , 1 kVAC,RMS , f=50 Hz P=75kN/mm, t=24 hours, T= 60 C
5.3 Main test
5.3.1 50 Hz 0.5 kVac,rms
The inception voltage of the cavity was found to be approximately 1 kVrms and the
applied voltage across the dielectric was reduced to 500 Vrms .
Table 5.6: PD data for PET 17, 3100 m without cavity, 0.5 kVAC,RMS , f=50Hz, P=75
kN/mm, t=24 hours, T=60 C

Mean discharge magnitude 11.1082 pC
Table 5.7: PD data for PET 16, 3100 m with cavity r=1 mm, 0.5 kVAC,RMS , f=50 Hz,
P=75 kN/mm, t=24 hours, T=60 C
PD data for sample 16, cavity r=1 mm

can be seen on the next two pages in Figure 5.10 and Figure 5.11, respectively.
At a applied voltage of 500 Vac,rms there should be no discharges in either sample, as
the applied voltage is below the inception voltage. The discharges in the sample with
a cavity have a low magnitude and occur in the first 2 hours of the measurements.
41
42
Figure 5.10: PET 17, 3100 m sample without cavity, 0.5 kVAC,RMS , f=50 Hz, P=75 kN/mm, t=24 hours, T=60 C
43
Figure 5.11: PET 16, 3100 m sample with cavity r=1 mm, 0.5 kVAC,RMS , f=50 HZ, P=75 kN/mm, t=24 hours, T=60 C
5.3.2 10 kVdc
The tests are preformed applying a pure DC voltage, Vdc =10 kV.
Table 5.8: PD data for PET 19, 3100 m without cavity, 10 kVDC , P=75 kN/mm, t=24
hours, T=60 C

Table 5.9: PD data for samples, 3100 m with cavity r=1 mm, 10 kVDC , P=75 kN/mm,
t=24 hours, T=60 C
PD data for samples with cavity r=1 mm

Sample 18 Sample 21 Sample 22
Maximum magnitude -1.159 pC 5.4882 pC 2.2333 pC
can be seen on the next two pages, in Figure 5.12 and Figure 5.13 respectively. The
plots for Sample 21 and 22 is found in figure E.1 and figure E.2 in the appedix.
As can be observed in both the tables above and the figures, there is a large variation.
No clear distinction between two of the samples with a cavity and the one without
a cavity can be observed. However, sample 21 have a high number of discharges. A
test were preformed to try to induce the same at 20 kVdc , data and plots from this
measurement can be found in the appendix in Table E.1 and Figure E.3. Even at a
voltage twice as high, the same occurence could not be induced to the same extent.
44
45
Figure 5.12: Pet 19, 3100 m sample without cavity, 10 kVDC , P=75 kN/mm, t=24 hours, T=60 C
46
Figure 5.13: PET 18, 3100 m sample with cavity r=1 mm, 10 kVDC , P=75 kN/mm, t=24 hours, T=60 C
5.3.3 Combined 10 kVdc and 50 Hz 0.5 kVac,rms
The tests are preformed during combined DC and AC voltage applied, Vdc =10 kV,
Vac =500 VRMS and f=50 Hz
Table 5.10: PD data for samples, 3100 m without cavity, 10 kVDC , 0.5 kVAC,RMS , f=50
Hz, P=75 kN/mm, t=24 hours, T=60 C
PD data for samples without cavity

Minimum magnitude -3.546 pC -3.6006 -13.3236 pC
Table 5.11: PD data for samples, 3100 m with cavity r=1 mm, 10 kVDC , 0.5 kVAC,RMS ,
f=50 Hz, P=75 kN/mm, t=24 hours, T=60 C

Minimum magnitude -506.375 pC -132.17 -183.937 pC
can be seen on the next two pages, in Figure 5.14 and Figure 5.15 respectively. The
additional plots for sample 12, sample 20, sample 8 and sample 15 can be found Section
E.3 in the appendix.
It can be observed in both the tables and the corresponing figures that there is a sig-
nificant difference in both the total number of discharges and their magnitude between
the solid samples and the ones with a cavity. The repetition rate for the samples with
a cavity decrease from the moment voltage is applied, and after about 3-4 hours the
discharges rate reaches a approximate stable level of discharges per hour. This steady
level however varies from sample to sample.
47
48
Figure 5.14: PET 13, 3100 m sample without cavity, 10 kVDC , 0.5 kVAC,RMS , f=50 Hz, P=75 kN/mm, t=24 hours, T=60 C
49
Figure 5.15: PET 9, 3100 m sample with cavity r=1 mm, 10 kVDC , 0.5 kVAC,RMS , f=50 Hz, P=75 kN/mm, t=24 hours, T=60 C
5.3.4 Combined 10 kVdc and 1 kHz 0.5 kVac,rms
The tests are preformed under combined DC and AC voltage, Vdc =10 kV, Vac =500
VRMS and f=1000 Hz for 12 hours.
Table 5.12: PD data for PET 26, 3100 m without cavity, 10 kVDC , 0.5 kVAC,RMS , f=
1kHz, P=75 kN/mm, t=12 hours, T=60 C

Table 5.13: PD data for samples, 3100 m with cavity r=1 mm, 10 kVDC , 0.5 kVAC,RMS ,
f=1 kHz, P=75 kN/mm, t=12 hours, T=60 C

Maximum magnitude 53.11 pC 15.4275 pC -1.0078 pC
can be seen on the next two pages, in Figure 5.16 and figure 5.17 respectively. The
additional plots for sample 27 and 25 can be found section E.4 in the appendix.
The tests at a combined DC and high frequency AC voltage have been run for 12
hours. The results have a large degree of variation. The results for sample 24 and 27
showed a similar shape of the repetition rate as during the 50 Hz combines test, but
at a lower level.
50
51
Figure 5.16: PET 26, 3100 m without cavity, 10 kVDC , 0.5 kVAC,RMS , f=1 kHz, P=75 kN/mm, t=12 hours, T=60 C
52
Figure 5.17: PET 24, 3100 m with cavity r=1 mm, 10 kVDC , 0.5 kVAC,RMS , f=1 kHz, P=75 kN/mm, t=12 hours, T=60 C
6. Discussion
6.1 Sub-system test

A number of transformers were tested. The results from the sub-system test proved
three of the four transformer to have either a to low transform ratio or to much voltage
distortion on the secondary side. At higher frequencies several of the transformers
had difficulty reaching the 2 kVRMS . The Messwandler Bau transformer proved to
be the most promising transformer. It has a relatively high transformation ratio,
which made it possible to deliver a voltage up to 2 kVrms across the entire frequency
spectrum of 1 kHz to 5 kHz. The transformer ratio may indicate that resonance occur
at approximately 4 kHz. The FFT analysis results showed the harmonics from the
source and transformer to be within a reasonable limit without extensive filtering, all
have a THD below 1%. However the goal of reaching a voltage of 5 kVRMS cannot be
delivered with this equipment.
6.2 System test
6.2.1 Noise elimination
The initial testing of noise in the components in the experimental set-up was a time
consuming part of the experimental work. The system test was conducted with sections
of the set-up connected to the OMICRON to eliminate or attenuate the noise and a
number of alterations were done.
The coupling resistance and the buffer resistance were increased from 10 M and
28 M to 100 M and 280 M because noise from the DC source was detected
and an error on the regulating controls of the DC source due to a induced voltage
caused by the AC current across the buffer resistance.
The control system of the heat coil were alterered due to noise.
A part of the security system were moved, due to noise caused by movement.
Ground loops due to the security system were eliminated.
The main supply can be a source of noise, but it is passed through a filter. Another
source of noise, not solved in this task, is the voltage sources situated outside the
faraday cage. Having all current carrying wiring inside the faraday cage eliminates the
wiring picking up noise from the surroundings and creating noise in the circuit.
53
The FFT on the PD input signal shown in Figure5.5 has two peaks. These originate
from noise from the surroundings and appear at sporadic times of the day and the
frequency band chosen for testing is placed in the lowest range possible between these
peaks. The noise measurements show a distinct noise band at approximately 0.6 pC.
However, this is the noise band measured in one time window. Not the entire time
test is run and does not necessarily give a clear image of the noise during the whole
day.
6.2.2 Benchmarking
A large amount of discharges was detected in the layered sample compared to the
solid sample during the benchmarking with an applied pressure of 50 kN/mm. In
addition the partial discharges in the layered samples had a higher magnitude. The
repetition rate for two of the samples decreases after the voltage has been applied
and stabilized on a relatively constant level, for sample 3 at approximately 20/hour
and for sample 5 at approximately 120/hour. Since these samples do not include a
cavity these observations raises the suspicion of discharges in the interfaces between
the layers. Sample 4 have a large variation in repetition rate with an increase to a
maximum of approximately 840 discharges per hour after about 2.5 hours.
The results from the benchmarking at 75 kN/mm indicate a decrease in magnitude
and a reduction in the number of discharges. The results suggested that an increase
of the pressure did have an effect since the results in the solid and layered sample are
similar. Comparing the results to the data from the 50 kN/mm test show a small
reduction in magnitude and number of discharges for the layered samples. However, a
noticeable reduction in the repetition rate after some time has passed is observed and
may indicate that the increased pressure did have an effect.
The benchmarking results showed some aspects which can be interpreted as interfer-
ence. The limited control of the DC voltage amplification and a too low waiting time
could be a contributor to the large repetition rates in the first hours of the tests. The
results from samples with an increased weight may still indicate a problem with the
applied pressure on the sample. The design of the electrodes does not distribute an
even pressure across the entire sample and the pressure along the edges of the elec-
trodes could be a problem. The applied AC voltage in the system tests of 1 kVrms is
the inception voltage of the air cavity and the voltage is decreased to 500 Vrms for the
main test.
6.3 Main test

The results during the pure AC voltage tests showed a low number of discharges, all
situated in the first 2.5 hours. There should be no discharges at this voltage level. In
the no cavity sample the discharge also has a higher magnitude than the sample with
54
a cavity. The discharges may be due to startup noise or small contaminants on the
electrode surface. The high discharge detected in the no cavity sample may be noise,
but no noise at this level has been detected.
The results from the pure DC test were very varying and discussing the results with
the data collected during these tests, is difficult. The test on the sample without a
cavity and two of the samples with a cavity had a low number of discharges. The third
sample with a cavity had a high number of discharges. The sample with the 20 kVrms
show a similar repetition rate. A low repetition rate during the first hours before the
repetition rate increased. The discharge during applied DC voltage are all small in
magnitude. Since the field is unipolar, it is more likely that charge accumulate on the
cavity surface, creating a field in the opposite direction and a smaller voltage drop to
the extinction voltage. [26] The small magnitude and a typically low repetition rate
will make the DC test results more prone to noise. Another problem with the small
magnitudes is the problem with the OMICRON, the detection system have proven to
have some issues determining polarity at low magnitudes. The polarity of discharges
during DC voltage should all be of the same polarity. [19] Without additional data it
is impossible to explain the partial discharge occurence in the DC tests.
The results from the combined DC and AC voltage test at 50 Hz illustrated a sig-
nificant difference between the samples with and without a cavity. The solid layered
samples experienced a low number of discharges with a small magnitude. The results
from the test performed with a cavity showed an increase in both the magnitude and
repetition rate. The repetition rate is high in the initial hours before it decrease to
a relatively stable level. The initial repetition rate and the stable level is different
for the samples. Sample 9 stabilize at approximately 125 discharges/hour, Sample 8
stabilize at a level of approximately 75 discharges per/hour while sample 15 have a
stable level of approximately 20 discharges/hour. The repetition rates for the samples
with a cavity follow an expected shape indicated by the equation for the repetition rate
during a combined DC and AC voltage. It is dependent of the time constant which
is a function of the conductivity of the cavity and the dielectric which approaches a
stable level after a time dependent of temperature.
D0 c + 0 b
= (6.1)
Dc + b
1 Vdc 1
n ( )( ) (6.2)
VP DIV,DC 1 Vac
VP DIV,AC
The stable value of the repetition rate is affeced by the conductivity of the air in the
cavity. This value is unknown since there is no information on the humidity, possible
by-products or other contributing factors. The placement of the discharge detected by
the low gain unit can be seen in Figure 6.1. It shows that discharges are concentrated
near the peak value of the applied voltage.
55
Figure 6.1: Picture from Omicron illustrating occurence of discharges, high gain for sample
8, 3100 m with cavity r=1 mm, 10 kVDC , 0.5 kVAC,RMS , f=50 Hz, P=75 kN/mm, t=24
hours, T=60 C
The results from the combined DC and AC voltage with a frequency of 1000 Hz gave
some variations. The sample without a cavity gave a low number of discharges at
low magnitude. The tests on the samples with cavity had a large variation in the
number and magnitude of the detected discharges. It should, however, be observed
that the repetition rate was overall lower at the high frequency test compared to the
50 Hz test. The stable level is approximately 10 discharges/hour for sample 24 and
40 discharges/hour in sample 27. The reduced number of discharges can be due to an
alteration of the capacitance of the test object at higher frequencies because of the
permittivity. A decrease of the capacitances result in a lower stress across the cavity
and a lower repetition rate. But additional data must be collected to know the effect
of the frequency on the permittivity and partial discharge behavior. An error source
in the comparison between the 1 kHz tests and the 50 Hz tests is the fact that the
Behlman source was not tested at 50 Hz, and might have a higher THD at 50 Hz, than
at 1 kHz. In Figure 6.2 one can observe a slight shift in the placement of the voltage
phase compared to Figure 6.1 for the 50 Hz comined voltage test. This may be due
to the measuring shunt. The RLC measuring shunt might result in a phase shift at
higher frequencies. It could also be explained by the statistical time lag.
56
Figure 6.2: Picture from Omicron, high gain for sample 27, 3100 m with cavity r=1 mm,
10 kVDC , 0.5 kVAC,RMS , f=1 kHz, P=75 kN/mm, t=12 hours, T=60 C
The results from the discharge tests show a large variation in the number of discharges
and magnitude throughout the samples with the same characteristics and voltage.
There are a number of phenomenona that may contribute in explaining the variations
in the number and magnitude of the discharges. The accumulation of space charge due
to the resitive field may alter the field across the cavity. Production of by-products
from the discharges in the cavity or on the cavity walls may alter the characteristics
of the cavity, creating large varations in the breakdown strengt and the number of
discharges.
Some observations have been made in several samples.
Partial discharge with a high positive magnitude when a unipolar field is applied.
This might indicate a flashback in the cavity or noise.
Partial discharge magnitude above the calculated value for a discharge across the
whole cavity area might be explained by the electrostatical forces on the test
object, altering the dimentions and increasing the area. It might also be due
to a too low pressure on the sample making the discharges bigger due to the
intersections.
High repetition rate in the first hours may be due to an insuffient waiting time
for the voltage sources, or a too rapid increase of the DC voltage.
Noise. The presence of noise is hard to determine, espesially when the OMICRON
has difficulty determining the polarity of small discharges. Inconsistencies may
also be caused by contaminants or airbubbels on the electrode surfaces.
Both the quantity and the quality of the results must be increased to get a clear image
of the partial discharge activity in the samples. The inconsistencies in the results
suggest that further improvements must be done with the experimental set-up to get
a more reliable result.
57
6.4 Procedure
6.4.1 Test samples
The manufacturing of the test samples is in theory a simple procedure, but there are
many possible error sources that can be caused by poor manufacturing. Oil leakage
has been a problem that has occured during the testing and several samples have been
discarded. It is very important to create them using a sharp blade to be able to get the
glued surfaces to curve in the same direction to maximize the probability of successful
glueing and hence minimize the chance of oil leakage into the sample.
The oil ingress can in most cases be observed as the samples are lowered into the
mineral oil in the test cell, but in some instances the ingress had some kind of delay
and cannot be discovered until the test finished and the whole measurement must be
discarded. In the beginning the process of creating the samples took a minimum of
12 hours for the glue to hardened sufficiently before removing it from the press. In
combination with the oil leakage problem the set-up stood still for days. Due to the
limited time this proved to be an issue. The problem with the sample production time
was however considerably reduced (to about two hours) by applying the catalysator
770 before the glue, 406. The problem with the oil leakage still occured, but with a
significantly lowered production time it would only lead to a 2 hour delay rather than
12.
In addition to the problem with oil seeping into the samples, another issue to be ad-
dressed is the possible contaminants. Since a contaminant can lead to field enhance-
ment it is a important to work in a clean environment when assembling the samples.
The environment is reasonably clean, but still the occurrence of dust and other parti-
cles can easily occur. In some completed samples it has be observed imperfections and
hence, the sample is scrapped. The assumption that samples are contaminated with
small imperfections without it being discovered, is reasonable.
58
7. Conclusion
The main objective with this master thesis was to develop an experimental set-up able
to detect partial discharge activity during a combined DC voltage superimposed a AC
voltage. The sub-system tests showed that both the Behlman signal generator and the
Messwandler Bau transformer functions to a satisfactory degree with a total harmonic
distortion of 1 % in the voltage range 1-2 kVrms and frequency range 1-5 kHz.
An experimental setup based on a straight detection circuit for partial discharge de-
tection is built with the Behlman and Messwandler bau connected to generate the AC
voltage.
The resulting noise band of 0.6 pC was found after a number of alterations in the system
test, and the detection limit were set at 1 pC. The benchmarking of the layered test
object indicated what could be interpreted as partial discharge in the intersection in
the sample at a pressure of 50 kN/mm. The pressure is increased to 75 kN/mm before
the main tests were conducted. But some observation may still indicate a too low or
uneven pressure across the area of the electrodes.
The experiments conducted during the main tests were preformed for several voltages
and test objects both with and without a cavity. The result from the main test varied
to a large extent. The problem with the OMICRON and the issue with polarity
determination makes the interpretations even harder. The AC test voltage was below
the inception voltage and still discharges occured, implicating noise. The results during
the DC tests varied to a large extent, making the results unreliable. Making any
comparison between the tests with combined voltage and DC test is not possible with
these results.
The combined DC and AC voltage test show a clear distinction between the samples
with and without a cavity. The results during the cavity tests at 50 Hz AC and DC
voltage are more reliable. The repetition rate had a high initial value, decreasing to a
stable level. This is consistent with the expectations that repetition rate is dependent
of the conductivity of the dielectric and the cavity. The 1 kHz AC and DC tests had
a significantly lower repetition rate, both initially and the stable level, but the same
shape is observed.
The statistical quality and the quantity of data are always of great importance when
discussing experimental results. The results from the testing gave varied results with
a large difference in both magnitude and repetition rate in similar samples at the
same applied voltage. More test must be performed to get a better fundation for
interpretation of the results. The varying results indicate some possible issues with
noise and the design of parts in the set-up.
59
8. Further Work
There is a number of exciting possibilities for further work in the field of partial
discharge detection and interpretation.
The results from the partial discharge tests indicated some sources of noise, and the
high repetition rate during the first hours could be due to a low waiting time or to a
rapid increase of the DC voltage. A control system should be developed to increase
the reliability of these parameters. The possible noise originatinating from having the
voltage sources outside the faraday cage should be addressed. A control system for
both the DC voltage source and the Behlan signal generator must be made to be able
to move all the current carrying parts of the set-up inside the faraday cage.
The problem with the OMICRON and determining the polarity of small partial dis-
charges should be solved to increase the reliability. Changing the measuring shunt to
an RC circuit, rather than an RLC circuit, could reduce the oscillations and possibly
reduce the problem.
The discharge with a magnitude higher than the calculated value of a complete break-
down of the cavity surface indicated discharges being affected by the interfaces due to
an unsufficient pressure or electrostatical forces. A new design of the testcell should
be investigated to assure an even pressure. The problems with oil ingress into the
test objects and indications of discharge activity in interfaces suggest that it should
be researched to use a different approach for production of the test objects. Using
Uglestad spheres or injecting air bubbles in an epoxy resin might be a direction worth
looking at.
As stated above, a problem with the interpretation of the result is due to the limitations
of the OMICRON. One possibility is to determine where the discharges occur can
be done by modelling the partial discharges detected to look at the characteristic
signatures. All the different types of partial discharges have a distinct characteristic.
It has been done some research on analysing partial discharge and model the discharges
based on their characteristics. This gives the opportunity to eliminate the discharges
occurring in the surroundings, and focus on the discharges in the cavity. [9] [27] [13]
To determine whether the discharges occur in the sample or originate from external
sources it is aslo possible to use a balanced circuit rather than a straight used in this
thesis. The balanced circuit is also known to have a higher sensitivity.
A natural path for further work is to continue the testing to increase the reliability of
the results before continuing at additional voltage and frequency levels as well as inves-
tigating the temeprature dependence. The quantity and quality of the experimental
results are the most critical parameters for laboratory reseach, and further studies are
encouraged to validate and increase the credibility of the experimental results in the
report.
61
Bibliography
[1] ABB, ABB Review, The corporate technical journal of the ABB Group, vol. 22,
March 2008.
[2] E. Spahic and G. Balzer, Offshore Wind Farms - VSC-based HVDC Connection,
Power Tech IEEE Russia, pp. 16, June 2005.
[3] N. Mohan, T. M. Undeland, and W. P. Robbins, Power Electronics, Converters,
Applications, and Design. John Wiley and Sons, Inc, 3 ed., 2003.
[4] P. Bresesti, W. L. Kling, R. L. Hendriks, and R. Vailati, HVDC Connection of
Offshore Wind Farms to the Transmission System, IEEE Transactions on Energy
Conversion, vol. 22, March 2007.
[5] C. H. Chien and R. W. G. Bucknall, Analysis of Harmonics in Subsea Power
Transmission Cables Used in VSC-HVDC Transmission Systems Operating Under
Steady-State Conditions, IEEE Transactions on Power Delivery, vol. 22, October
2007.
[6] P. K. Olsen, F. Mauseth, and E. Ildstad, The effect of DC superimposed AC
Voltage on Partial Discharges in Dielectric Bounded Cavities, International Con-
ference on High Voltage Engineering and Application, 2014.
[7] F. Mauseth, S. Hvidsten, H. H. Sternes, and J. Aakervik, Influence of DC
stress Superimposed with High Frequency AC on Water Tree Growth in XLPE
Insulation, Nordic Insulation Symposium, June 2013.
[8] W. Choo and G. Chen, Electric Field Determination in DC Polymeric Power
Cable in the Presence of Space Charge and Temperature Gradient under DC
Conditions, International Conference on Condition Monitoring and Diagnosis,
April 2008.
[9] F. R. Kreuger, Partial Discharge Detection in High Voltage Equipment. Butter-
worths and Co., 1989.
[10] C. Forssen, Modelling of Cavity Partial Discharges at Variable Applied Fre-
quency, Doctorial Thesis in Electrical Systems, 2008.
[11] K. D. Hammervoll, Ikke-destruktiv tilstandskontroll av kabelskjter i dis-
tribusjonsnett, Master i energi og milj, NTNU, June 2010.
[12] L. Lundegaard, Partielle utlandninger, Begreper, maleteknikk of mulige anven-
delser for tilstandskontroll, SINTEF Energy Research, pp. 150, March 1996.
63
[13] L. Niemeyer, A Generalized Approach to Partial Discharge Modeling, IEEE
Transaction on Dielectrics and Electrical Insulation, vol. 2, pp. 510528, August
1995.
[14] P. K. Olsen, F. Mauseth, and E. Ildstad, Modelling of Partial Discharges in Poly-
meric Insulation Exposed to Combined DC and AC Voltage, Nordic Insulation
Symposium, June 2011.
[15] E. Ildstad, TET 4160 High Voltage Insulation Materials. NTNU Department of
Electric Power Engineering, 2012.
[16] E. F. Bere, Innvirkning av fuktighet pa dielektriske tap og risiko for partielle
utladninger i innstpte endeviklinger, Master i energi og milj, NTNU, June
2014.
[17] H. Suzuki, K. Aihara, and T. Okamoto, Complex behaviour of a simple partial-
discharge model, Europhysics Letters, vol. 66, pp. 2834, February 2004.
[18] D. Malec and T. Lebey, Effects of the Voltage Waveform on Partial Discharges
Activity in a Vented Sample(HDPD, Conference on Electrical Insulation and
Dielectric Phenomena, IEEE, 2000.
[19] P. H. F. Morshuis and J. J. Smit, Partial Discharges at DC Voltage: Their
Mechanism, Detection and Analysis, IEEE, pp. 328340, October 2005.
[20] B. Salvage and N. R. Steinberg, Discharge Repetition in an Air Filled cavity
in a Solid Dielectric under Direct Voltage Conditions, Electronic Letters, vol. 2,
pp. 432433, November 1966.
[21] S. Hvidsten, Power cables, Partial discharge measurements, ELK-30, Condition
assessment of high voltage apparatur, 2013.
[22] I. R. Velo, Test setup for measuring partial discharge in polymer insulatin sub-
jected to hvdc superimposed by high frequency ac, Prosjekt i energi og milj,
NTNU, June 2014.
[23] P. I. Nodeland, Design av en hyfrekvent transformer for testing av hvdc kabler
for flytende vindturbiner, Prosjekt i energi og milj, NTNU, Desember 2009.
[24] P. I. Nodeland, Utvikling av testoppsett for spenningsprving av hvdc pex-
kabler, Master i energi og milj, NTNU, June 2010.
[25] S. Beg and B. Salvage, Discharge Repetition in an Air Filled Cavity in Polythene
under High Direct Electric Stresses at Elevated Temperatures, Electronic Letters,
vol. 5, pp. 118120, March 1969.
[26] U. Fromm, Partial Discharges and Breakdown Testing at High DC Voltage,
Doctorial Thesis Delft University of technology, september 1995.
[27] U. Fromm, Interpretation of Partial Discharges at DC Voltages, IEEE Trans-
action on Dielectrics and Electrical Insulation, vol. 2, pp. 761770, October 1995.
64
A. Experimental setup
A.1 Experimental circuit
Figure A.1: Picture of the experimental setup from outside the faraday cage with the voltage
sources, top: DC source, bottom: Behlman
i
Figure A.2: Picture of the experimental setup inside the faraday cage with the testcell in
the center
ii
A.2 Test Cell
Figure A.3: Picture of the test cell
iii
B. Hardware Data
B.1 Transformers
B.1.1 Petter Nodeland Designed Transformer
Transfromer was designed for 20 kHz - but has proven to work at its best at 15 kHz.
Vpri = 350 VRMS
Vsec = 10 kVRMS
Npri = 25 turns
Npri = 724 turns
Primary Side Secondary side
Frequency f 20 kHz 20 kHz
Voltage V 350 VRMS 10 kVRMS
Current I 5 ARMS 0.18 ARMS
Windings N 25 turns 724 turns
Conductor crossection A 3.17 mm2 0.12 mm2
Conductor diameter d 1.78 mm 0.34 mm
[24] [23]
B.1.2 Neon Transformer
An alternative transformer was located in Frank Mauseths office- a neon transformer.

Data:
NEONTRAFO TYPE
N 50/4 S
PRIM
UP = 230 V, 50 Hz
IP = 1.13 A
cos = 0.55
SEC
US = 4000 V - PE
IS = 50 mA
PP = 260 VA
v
IP00 ta40c
19.01.96
NEONCOMP OY
TURKU - FINLAND
B.1.3 MESSWANDLER BAU Measuring Transformator
Labnr.: B01-0967
Type: TRAFO M ALE
Modell: REL 10
Serienr.: 89/932436
Fabrikat: MESSWANDLER BAU
Sp. min (P): 11 KV

Sp. max (S): 110 V
Ampere (S): 2 A
Effekt (P): 60 VA
Merkeverdi: 11KV - 110V 3A 60VA
B.1.4 ARTECHE Measure Transformer
Labnr.: B01-0800
Type: TRAFO M ALE
Modell: UCN-36
Serienr.: 10012441 / 6
Sp. type: AC
Fabrikat: ARTECHE
Sp. min (P): 36 KV

Sp. max (S): 120 V
Effekt (P): 200 VA
Merkeverdi: 36KV - 0.5% - 200VA
B.2 Experminal setup equipment
B.2.1 DC Voltage Source
Labnr.:B02-0405
The DC voltage source is a high voltage DC source from Fug. It has the possibility of
vi
a voltage output range from 0-35 kV and a current range of 0-4 mA. The source can
be controlled externally.
B.2.2 AC Voltage Source
Labnr.:B02-0584
The AC voltage source is a Behlman signal generator. The Behlman ACM-3000 OSN-
1-45/10000 is a high quality variable frequency AC power supply that produces low
distortion sinusoidal wave power over a wide range of output frequencies and voltages.
The Behlman voltage source available in the NTNU inventory is build up out of an
ACM- 3000 and an OSN-1-45/10000. It has a maximum voltage range 0- 270 VRMS
and a frequency bandwidth of 45- 10,000 Hz and the rated power is 3000 VA.
vii
C. Sub-system test
C.1 Total harmonic distorition and FFT in Messwandler Bau

Transformer
C.1.1 Harmonic analysis at 1 kHz
Table C.1: Harmonic magnitudes and THD at 1 kHz, 1 kV and 2 kV
1 kHz
1 kV 2 kV
Harmonic dB V[e+03] f[e+04 Hz] dB V[e+03] f[e+04 Hz]
1st 60.0628 1.0073 0.1000 66.0274 2.0016 0.1001
2nd 3.3643 0.0015 0.2002 6.4260 0.0021 0.2002
3rd 18.4392 0.0084 0.3001 21.9879 0.0126 0.3002
4th -2.4498 0.0008 0.4001 3.3423 0.0015 0.4002
5th 7.3337 0.0023 0.5002 15.3291 0.0058 0.5003
6th -3.3104 0.0007 0.6003 4.0278 0.0016 0.6003
7th -1.7252 0.0008 0.7003 1.7468 0.0012 0.7004
8th -5.0208 0.0006 0.8003 -0.3070 0.0010 0.8001
9th -8.8534 0.0004 0.9005 -2.8641 0.0007 0.9005
10th -10.5893 0.0003 1.0004 -6.6331 0.0005 1.0005
11th -9.0920 0.0004 1.1005 -1.1117 0.0009 1.1006
12th -12.2572 0.0002 1.2005 -5.4013 0.0005 1.2006
13th -17.3637 0.0001 1.3005 -6.5293 0.0005 1.3005
THD -40.8945 -42.7277
ix
1 kV
x
Figure C.1: FFT plot 1 kHz and 1 kV

2 kV
xi

2 kHz
1 kV 2 kV
1st 60.0237 1.0027 0.2002 66.0288 2.0019 0.2003
2nd 8.8042 0.0028 0.4004 0.6618 0.0011 0.4005
3rd 16.7532 0.0069 0.6006 18.1822 0.0081 0.6008
4th 0.8089 0.0011 0.8008 3.2071 0.0014 0.8010
5th 4.2506 0.0016 1.0012 7.8490 0.0025 1.0013
6th -5.0607 0.0006 1.2011 -0.9325 0.0009 1.2015
7th 6.2978 0.0021 1.4015 3.6681 0.0015 1.4018
8th -3.9066 0.0006 1.6014 0.6519 0.0011 1.6020
9th 2.7576 0.0014 1.8019 -3.1889 0.0007 1.8023
10th -2.0617 0.0008 2.0021 -8.0325 0.0004 2.0026
11th -5.5485 0.0005 2.2022 0.1525 0.0010 2.2030
12th -14.2950 0.0002 2.4029 -11.2394 0.0003 2.4034
13th -12.8451 0.0002 2.6030 -9.9830 0.0003 2.6037
THD -41.6788 -46.5906
xii
1 kV
xiii

2 kV
xiv

3 kHz
1 kV 2 kV
1st 60.0028 1.0003 0.3001 66.0335 2.0030 0.3001
2nd 5.7832 0.0019 0.6002 5.0493 0.0018 0.6002
3rd 14.6357 0.0054 0.9003 15.5450 0.0060 0.9003
4th -1.4349 0.0008 1.2005 0.1174 0.0010 1.2006
5th -7.0954 0.0004 1.5011 10.6539 0.0034 1.5006
6th -2.3237 0.0008 1.8011 -0.7087 0.0009 1.8007
7th 8.5450 0.0027 2.1009 -2.5797 0.0007 2.1012
8th -1.1953 0.0009 2.4008 -1.8146 0.0008 2.4010
9th 1.2636 0.0012 2.7011 -6.2540 0.0005 2.7010
10th -4.1652 0.0006 3.0013 -5.0623 0.0006 3.0016
11th -1.6701 0.0008 3.3014 -2.7722 0.0007 3.3015
12th -14.5575 0.0002 3.6008 -7.3699 0.0004 3.6015
13th -13.3737 0.0002 3.9011 -4.7645 0.0006 3.9017
THD -43.1474 -47.9638
xv
1 kV
xvi

2 kV
xvii

4 kHz
1 kV 2 kV
1st 60.0060 1.0007 0.4003 66.0445 2.0055 0.4003
2nd 7.5078 0.0024 0.8005 0.3899 0.0010 0.8006
3rd 10.7621 0.0035 1.2009 11.2780 0.0037 1.2008
4th 0.3700 0.0010 1.6012 3.6066 0.0015 1.6011
5th 4.4964 0.0017 2.0013 12.1923 0.0041 2.0014
6th -4.4983 0.0006 2.4016 -2.7910 0.0007 2.4017
7th 7.2793 0.0023 2.8020 -2.4285 0.0008 2.8021
8th -1.0200 0.0009 3.2023 -0.8208 0.0009 3.2023
9th 4.7833 0.0017 3.6025 2.5330 0.0013 3.6027
10th -3.1726 0.0007 4.0028 -4.9000 0.0006 4.0028
11th -2.6680 0.0007 4.4030 -1.3152 0.0009 4.4032
12th -24.9168 0.0001 4.8030 -9.3174 0.0003 4.8035
13th -8.9505 0.0004 5.2036 -11.0249 0.0003 5.2038
THD -44.7482 -49.5494
xviii
1 kV
xix

2 kV
xx

5 kHz
1 kV 2 kV
1st 60.0297 1.0034 0.5001 66.0284 2.0018 0.5002
2nd 5.7813 0.0019 1.0002 1.4528 0.0012 1.0005
3rd 2.9841 0.0014 1.5005 3.4332 0.0015 1.5005
4th 1.7235 0.0012 2.0005 4.7544 0.0017 2.0006
5th -2.1606 0.0008 2.5007 11.0826 0.0036 2.5010
6th -6.8642 0.0005 3.0010 -1.4851 0.0008 3.0012
7th 4.0719 0.0016 3.5010 4.3206 0.0016 3.5013
8th -0.5190 0.0009 4.0012 0.4570 0.0011 4.0014
9th 3.8102 0.0016 4.5014 1.0204 0.0011 4.5017
10th -2.3777 0.0008 5.0015 -3.4791 0.0007 5.0019
11th -5.0659 0.0006 5.5015 -0.4978 0.0009 5.5020
12th -26.1117 0.0000 6.0026 -7.6613 0.0004 6.0022
13th -18.1499 0.0001 6.5020 -5.1565 0.0006 6.5024
THD -47.8961 -50.7887
xxi
1 kV
xxii
Figure C.9: FFT analysis plot 5 kHz and 1 kV

2 kV
xxiii
Figure C.10: FFT analysis plot 5 kHz and 2 kV

D. System test
xxv
D.1 Partial discharge under combined 50 Hz AC Vac,rms=1 kVRMS and DC Vdc=10 kV voltage
P=50 kN/mm
xxvi
Figure D.1: PET 4, 3100 m layered sample without cavity, 10 kVDC , 1 kVAC,RMS , f= 50Hz, P=50 kN/mm, t=24 hours, T=60
xxvii
Figure D.2: PET 5, 3100 m layered sample without cavity, 10 kVDC , 1 kVAC,RMS , f= 50Hz, P=50 kN/mm, t=24 hours, T=60 C
E. Main test
xxix
E.1 Partial discharge under DC Vdc=10 kV voltage
xxx
Figure E.1: PET 21.2, 3100 m layered sample with cavity r=1 mm, 10 kVDC , P=75 kN/mm, t=24 hours, T=60 C
xxxi
Figure E.2: PET 22, 3100 m layered sample with cavity r=1 mm, 10 kVDC , P=75 kN/mm, t=24 hours, T=60 C
E.2 Partial discharge under DC Vdc=20 kV voltage
Table E.1: PD data for PET 22.2

xxxii
With cavity r=1 mm
xxxiii
Figure E.3: PET 22.2, 3100 m layered sample with cavity r=1 mm, 20 kVDC , P=75 kN/mm, t=24 hours, T=60 C
E.3 Partial discharge under combined DC Vdc=10 kV and 50 Hz AC Vac,rms=0.5 kVRMS
xxxiv
Figure E.4: PET 12, 3100 m layered sample without cavity, 10 kVDC , 0.5 kVAC,RMS , f= 50 Hz, P=75 kN/mm, t=24 hours, T=60 C
xxxv
Figure E.5: PET 20, 3100 m layered sample without cavity, 10 kVDC , 0.5 kVAC,RMS , f=50 Hz, P=75 kN/mm, t=24 hours, T=60 C
xxxvi
Figure E.6: PET 8, 3100 m layered sample with cavity r=1 mm, 10 kVDC , 0.5 kVAC,RMS , f=50 Hz, P=75 kN/mm, t=24 hours, T=60 C
xxxvii
Figure E.7: PET 15, 3100 m layered sample with cavity r=1 mm, 10 kVDC , 0.5 kVAC,RMS , f=50 Hz, P=75 kN/mm, t=24 hours, T=60 C
E.4 Partial discharge under combined 1000 Hz AC Vac,rms=0.5 kVRMS and DC Vdc=10 kV voltage
xxxviii
Figure E.8: PET 25, 3100 m layered sample with cavity r=1 mm, 10 kVDC , 0.5 kVAC,RMS , f=1 kHz, P=75 kN/mm, t=12 hours, T=60 C
xxxix
Figure E.9: PET 27, 3100 m layered sample with cavity r=1 mm, 10 kVDC , 0.5 kVAC,RMS , f=1 kHz, P=75 kN/mm, t=12 hours, T=60 C

Fulltext Vimp

Uploaded by

Copyright:

Available Formats

Fulltext Vimp

Uploaded by

Document Information

Original Description:

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

Fulltext Vimp

Uploaded by

Copyright:

Available Formats

Experimental Set-up for partial discharge

Ingrid Roen Velo

Master i energi og milj

Norges teknisk-naturvitenskapelige universitet

3 Partial Discharge Detection 15

D System test xxv

E Main test xxix

1.1.1.1 Converter Technology

Figure 1.1: HVDC transmission scheme

Figure 1.2: Harmonic DC voltage [5]

1.1.1.3 Cable insulation

2.1 Partial discharges

2.2 Breakdown voltage of cavities

2.2.1 Pachens curve for breakdown in air

A breakdown in a cavity bound by a dielectric occurs at approximately the same

2.2.2 Partial Discharge under AC voltage

Figure 2.3: Internal discharge equivalent circuit under AC conditions [9]

2.2.2.1 Voltage distribution

The AC voltage across a cavity can be expressed by the following equation:

Vc,ac = KAC Vac sin(t) (2.3)

ac is the amplitude of the AC voltage with an angular frequency of =2f applied

2.2.2.2 Recurrence of discharges

Figure 2.4: Discharge recurrence during AC voltage [17]

Vc,ac,pd = KAC Vac sin(t) U (2.6)

2.2.3 Partial Discharge under DC voltage

Figure 2.5: Internal discharge equivalent circuit under DC conditions

2.2.3.1 Voltage distribution

The voltage distribution in the cavity under stationary DC conditions, assuming an

for VDC > VPDIC,DC .

2.2.4.1 Voltage distribution within the dielectric

Vinput = Vdc + Vac sint (2.17)

ac is the amplitude of the AC voltage with an

2.2.4.2 Repetition rate

which can be rewritten as

Figure 2.6: Repetition rate illustration for combined DC and AC voltage

3.1 Detection circuit

Figure 3.1: Equivalent circuit of straight detection circuit

3.1.1 Sensitivity and noise

3.1.2 Partial discharge magnitude

where o is the permittivity of vacuum=8.85410-12 and r is the reltive permittivity of

Qmeasured = Ceq V (3.4)

Figure 3.2: Equivalent circuit of the experimental set-up

Table 3.1: List of parameters in the experimental circuit

Parameter Definition Value

The resulting voltage across the testobject is:

Vtestobject = VAC,testobject + VDC,testobject (3.9)

Behlman Signal generator a high frequency AC voltage source with a voltage

Figure 3.3: Test Cell

A door switch is connected to a connection box between the voltage sources

The testing is performed to build a interference free experimental set-up illustrated in

4.1 Sub-system test

4.1.1 Experimental approach

4.2 System test

Noise and frequency band

4.3 Main test

Table 4.1: Main tests

Applied voltage Run time

Table 4.2: List of parameters for the test objects

where o is the permittivity of vacuum=8.85410-12 and r is the reltive permittivity of