275 kV Cable Discharge Field Measurement and

Analysis of SVLs Chain Failure Using ATP

F. Ghassemi, S. Dennis, A. Ainsley, M. Haddad, S. Robson

Abstract—Results of field measurement of trapped charges adequate energy absorption capability for other transients on
on a 275 kV cable will be presented. The instrumentation and the cable such as energisation with no trapped charges,
calibration method used in the measurement will be discussed. It lighting on the relevant overhead lines connected to the cable
will be shown that if the cable is not force-discharged by means of and external faults, all of which had been considered during
earth switches, the trapped charges will decay very slowly. The
design stage.
atmospheric parameters such as humidity and temperature affect
the discharge time, which also depends on how much of the It is usually assumed that the trapped charges on an isolated
circuit is exposed to open air. cable are drained to earth within a short time of a few hours
The measurement results are used in EMTP-ATP simulation through the cable insulation losses. Very few test results of
to confirm that trapped charges may be sufficient to cause sheath cable charging and discharging can be found in the literature
voltage limiters (SVL) of the cable to fail. The number of SVL [4]. No guideline is provided by IEC or CIGRE on the
breakdown depends on the mode of SVL failure, i.e. whether it discharge time of extra high voltage (EHV) single phase
attains low impedance (short circuit) or open circuit when the cables. For this reason, National Grid, UK, performed a series
energy absorbed exceeds the rating. of cable discharge tests on a 275 kV pressurized-oil-filled
(POF) cable of 21 km length, to examine the discharge time of
Keyword: ATP-EMTP, cable trapped charges, cable discharge
time, cable sheath transients, SVL failure. the cable after being isolated and the effect of trapped charges
on the sheath voltage limiters (SVL) failure [3], [5]. Reference
I. BACKGROUND [6] investigated the effect of cable faults on SVL failure.
The paper describes the method of instrumentation used for
T HE phenomena of cable charging and discharging have
been a concern to cable engineers especially when
performing commissioning tests and during maintenance work
measuring dc charges on the cable and estimates discharge
time constant. Field mill devices were used to carry out non-
[1], [2]. Also, the charging/discharging phenomena have intrusive tests. The paper presents the result of calibration tests
received especial attention from utilities which are using high- in the laboratory that were necessary to confirm the linearity
voltage (HV) cables for voltage control during periods of light of transfer ratio of the probe for measuring the system voltage
load which requires nightly switching of cables. For this (50 Hz) and that due to trapped charges (a decaying dc signal),
particular activity the earth switches of the cable are not when the measurement probes are placed in the field.
usually used due to excessive wear and tear of switch contacts
and also, in accordance with their internal recommendations II. THE CABLE UNDER STUDY
for safety and security of supply, some utilities require that The cable circuit considered in the study consists of two
earth switches can only be operated if relevant staff is present parallel, three-phase, 275 kV, 1600 mm2, copper conductor,
in the substation. oil-filled paper insulated cables. Each circuit consists of three
A study was carried out to investigate the root cause of single phase cables. They connect substations (S/S) A and B.
excessive SVL failures in a number of EHV cables. The The cable detail is given in [3], which also illustrates the
EMTP programme was used to model a cable circuit with high crossbonding used in the cable installation considered in the
SVL failure rate. The results of the study were presented in study. The cores are transposed at every junction, which leads
IPST 2013 [3]. The study showed that if sufficient trapped to more balanced impedances in all three phases. The sheaths
charges remain on the isolated cable long enough, i.e. until the are cross-bonded as shown in [3].
next switching action some six to eight hours later, then there Fig. 1, illustrates the high level layout at S/S A and B.
was very high probability that the overvoltage surges on the Circuit 2 is used nightly for voltage control by switching it out
core and thus sheath were sufficient enough to cause SVL of service during the night usually between 22:00 and 7:00 the
failure. It was shown that this was the only condition that next day. The switching is done remotely from the National
would lead to failure and that the SVL would have had Control Centre. Although both circuits have earth switches at
both ends, they are only operated when the cables are taken
- out of service for a long period of time, e.g. maintenance. For
F. Ghassemi, S. Dennis, A Ainsley, are with National Grid Electricity regular voltage control switching, these switches are not
Transmission, National Grid House, Warwick Technology Park, Gallows
Hill, Warwick EC34 6DA, United Kingdom. operated because (i) they can only be operated locally at the
Email: forooz.ghassemi@nationalgrid.com substations for safety issues, and (ii) to avoid excessive wear
M. Haddad, S. Robson are with Cardiff University, Cardiff CF10 3XQ, and tear of isolator switches contacts. Both S/S A and B are
United Kingdom. unmanned.
The cables lengths for Circuit 1 and Circuit 2 are
Paper submitted to the International Conference on Power Systems respectively 21.22 km 21.74 km. The cables are placed in
Transients (IPST2017) in Seoul, Republic of Korea, June 26-29, 2017. trenches along the route with an average depth of 1 meter.
They enter tunnels in the vicinity of substations for a short
distance. The cable sealing ends at ground level are connected A series of tests were carried out to examine relationship
to a piece of overhead horizontal busbar with a length of about between the input, i.e. electrostatic field and the output
13.5 m as depicted in Fig. 2. The busbar is supported by two voltage. This was essential to ensure that the field mill has
vertical insulator poles, each with a vertical axial length (not constant transfer ratio between input and output signals for
creepage) of approximately 2.4 m. The isolator switches L21 both dc and ac signals, which in turn allows the field mill to be
and L22 are situated at the end of the busbar sections. calibrated in the field when it is installed under the live busbar
As can be seen from Fig. 1, the voltage transducers at both for 50 Hz voltage and measuring correctly when the cable is
ends of the cables (not shown in Fig. 2) are of capacitive switched out of service with trapped charges, e.g. dc voltage,
voltage transformer type with no dc transfer capability. This on the cable.
means that the trapped charges can only drain through the A metal plate was used as the source of electric fields. In
cable insulation losses and stray resistance of the outdoor order to increase the capacitance of the plate with respect to
insulation string at both ends. earth a 12 F capacitor was connected to the plate, which was
Circuit then energised to the mains voltage of 220 V, 50 Hz. The
To S/S A
Breaker Isolator neutral of the mains was earthed. The metal plate was raised
CCT1

S/S from the surface to a height of 0.7 meter by using non-metallic

Bus bar Earth support. The field mill was placed under the plate. Fig. 3,
B Switch CVT DNO-TO
boundary shows the test set up. The voltage of the metal plate was
measured directly by Channel 1 of a digital oscilloscope and
the output of the field mill was measured by Channel 2. The
plate, and the connected capacitor, was energised by closing
DNO-TO
boundary the switch. After waiting for a few minutes, the voltage was
removed. The capacitor was then discharged through the input
To S/S A resistance of the oscilloscope of 10 M. The capacitor losses
CCT2
also contribute to the discharge time constant.
Isolator Fig. 4 and 5, show the plate/capacitor voltage measured
switch L22 directly by the oscilloscope and by the field mill respectively
for disconnection at nearly zero and peak of the voltage. As it
is expected there is a 90 phase shift between the direct
measurement, i.e. the voltage of the plate and the
measurement by the field mill which is proportional to the
electric field. It can also be deduced that the transformation
ratio between the input and output of the field mill is constant
for the ac part of the signal, i.e. when the plate is connected to
the mains, and after the voltage is removed and the plate
voltage very slowly discharges. Thus, it was confirmed that
the field mill can be calibrated using the 50 Hz voltage by
S/S placing it under the live busbar. When the voltage is removed
A the field mill correctly measures the dc voltage on the cable
connected to the exposed busbar.

Fig. 1. High Level Substation A and B Layout.

Fig. 3. Field Mill Test Setup

Fig. 2. Cable Connection to the Air Busbar.

III. MEASUREMENT METHOD

A non-intrusive measurement system was devised. The
main component of measurement system was field mill. These
devices give an out voltage in proportion to the electric field
that is imposed on them [7].
Fig. 4. Test Rig Calibration, Switching off at Nearly Zero Voltage
 time constant is about 6 hours increasing to about 13 hours at
around the time of the cable re-energisation. It can be seen
from the record that the temperature was in fact increasing and
humidity reducing, leading to increase in the time constant.

Fig. 5. Test Rig Calibration, Switching off at Nearly Peak Voltage

IV. FIELD MEASUREMENT

The objective of the measurement in 2015 and 2016 was to
monitor slow changing dc voltage on the core of the cable,
hence the frequency response of the field mill probe was not
critical. The device’s output is an analogue voltage that was
then fed onto an analogue to digital converter with a sampling
rate of one second and storage capability for longer than four Fig. 6.a. Three Phase Voltages for 8/7/2015
weeks. The data was later downloaded to a permanent storage.
Note that since the sampling rate of the recorder is 1 s, the 50
Hz signal is not sampled accurately. However, measurement
of 50 Hz signals or transients associated with switching were
not part of the scope of the measurement. Thus, a low pass
filter was used on the sampled data to reduce the noise due to
aliasing.
The field mills, one for each phase were placed under the
section of the busbar between the first insulator string nearest
to the cable sealing end and the second insulation string
closest to the switch, see Fig. 2. The vertical distance between
the field mill and the busbar was approximately 3 m.
Atmospheric parameters namely, temperature, relative
humidity and pressure (THP) were also recorded at both S/S A
and B close to the air busbar where the field mills were placed. Fig. 6.b. THP for 8/7/2015
The recording was performed at the sampling rate of 30 s.
Recording of THP was essential to verify the dependency of
the discharge through exposed part of the circuit, e.g. insulator
strings, to the atmospheric parameters, whose effect on the
insulation losses of a trenched cable is very low.
Fig. 6.a, and Fig. 6.b, illustrate respectively the three
phase voltages and temperature, relative humidity and
pressure (THP) for a day in the second week of July 2015. The
highest voltage at the time of switch off reached 1.7 per unit
(pu) with respect to the peak of the nominal phase voltage
√ଶ
(275 =224.5 kV) which decayed to about 0.75 pu at the
√ଷ
time of cable re-energisation. Higher than nominal voltages at
the time of de-energisation can be expected in cables and
capacitive switchings due to restrike in circuit breakers. The
maximum and minimum temperatures during the Fig. 7. Rate of Change of Voltage for 8/7/2015
measurement period were about 17 and 11 C respectively
with the relative humidity remaining below 70%. Fig. 8.a and 8.b, illustrate the voltages and atmospheric
Logarithmic linearization was used to determine the parameters records during another period starting 21/7/2015.
discharge time constants. The decay time constant is defined The same effects can be observed. From Fig. 8.a the
as the time for an exponentially decaying curve to reduce from slowdown in the discharge is clearly visible at around the time
its peak to 1e or 37% of the peak. Fig. 7, illustrates the rate of the cable is about to re-energised. This very well correlates
with the temperature and humidity when the former increased
change of voltage as the linear lines at different times during and stabilised and the latter reduced markedly at around the
discharge. Two distinct time constants can be identified, one is same time. This is very well visible in Fig. 9, that shows the
when the cable was switched out for about two hours when the rate of change of discharge when it changed at the same time
as changes in temperature and humidity. At the time of re- about 3.5 hours changing to shorter time constant of about 1.5
energisation about 13% voltage was still on the red phase of hour. The cable is virtually discharged in about 5 hours.
the cable reduced from about 128% of voltage at the time of
disconnection.

Fig. 10.a. Three Phase Voltages for 30/7/2015

Fig. 8.a. Three Phase Voltages for 21/7/2015

Fig. 10.b. THP for 30/7/2015

Fig. 8.b. THP for 21/7/2015

Fig. 11. Rate of Change of Voltage for 30/7/2015

Fig. 9. Rate of Change of Voltage for 21/7/2015 Fig. 12.a, illustrates the voltages in red and yellow phases.
The blue phase measurement is missing. This was due to flat
Fig. 10.a, illustrates the voltages for a de-energisation in battery in the field mill and associated recorder, which was
30/7/2015. It can be seen that the voltages reached nearly 2 pu detected after measurement was done. It can be seen that the
and decayed to a very low level by the time the cable was re- voltage in red phase reaches 1.3 pu at the time of de-
energised. Fig. 10.b shows the change in atmospheric energisation reducing to about 0.3 pu the next day at around 7
parameters. It can be observed that the relative humidity was am when it starts to increase again. The temperature was not
very high reaching 100% while the temperature dropped to as particularly low and it starts to increase at around 4 am on
low as 6 C. The duration of high humidity was relatively 1/8/2015 reaching 18 C at the time of the switching.
long. Two distinct time constants can be observed, first was Likewise, the humidity reduces sharply in the morning. Fig.
13, shows the rate of change of voltage on the cable. The voltage and temperature of the cable (core/insulation) [5]. As
positive rate of change of voltage just before energisation is can be seen from all the records shown in this paper the initial
shown by the broken straight line. The increase in the phase rate of change of voltage on the cable changes within a couple
voltages can be attributed to the coupling with the live of hours after de-energisation. This is influenced not only by
adjacent busbar and increase in the system loading, which in the atmospheric parameters but also the voltage magnitude on
turn induces higher voltage in the cable. This increase offset the core.
the discharge of the cable through the insulator strings.
Therefore, depending on the substation layout, the V. MAINTENANCE RECORDS
disconnected cable may in fact be charged up due to induction, The cable has 18 junction boxes, of which 14 contain
if it is not earthed. SVLs, hence a total of 42 SVLs limit the sheath voltage.
Which phase SVL fails most appear to be random but usually
at least two phases SVLs fail in each junction box. Almost
always, the SVLs close to the substation when switching is
done have failed.
Two maintenance reports will be presented in this section.
The objective is to check whether any correlation between
atmospheric parameters and SVL failures exists.
In March 2011, there was fire in one of the junction boxes.
All three SVLs in the junction box had catastrophically failed
(destroyed). The subsequent check on earthing and
connections in the junction box showed that all were
according to the guidelines. The SVLs were changed and the
cable was put back in service. Thermovision tests showed that
no hotspot existed in the junction box and all connections,
Fig. 12.a. Three Phase Voltages for 31/7/2015
including earthing, were satisfactory.
The weather record for spring 2011 revealed that March
2011 was exceptionally warm and dry month. It was one of
the warmest on record with much below average rainfall
(http://www.metoffice.gov.uk/climate/uk/interesting#y2011).
On 18/06/2015, the cable was de-energised for routine
maintenance. Six SVLs on the route had to be changed. The
test to assess the SVL condition involves applying a test
voltage and measuring the leakage current. If the current
exceeds the declared value by the manufacturer the SVL is
discarded. At this point of time all SVLs were healthy.
The cable was put back in service and then taken out on
18/7/2015. In total 13 SVLs were found to have failed. Note
that, network loading reduces to a minimum of about 36% of
the maximum load in winter, therefore the cable is used more
regularly for voltage control. Inspection of the weather records
Fig. 12.b. THP for 31/7/2015
revealed that June /July of 2015 were one of hottest period on
record. The temperature on 1/7/2015 set a new record
(http://www.metoffice.gov.uk/climate/uk/interesting#y2015).
The maintenance and SVL failure records very well
correlate with the weather condition and findings of this paper.

VI. EMTP-ATP SIMULATION

Reference [8] showed that if the trapped charges observed
on 8/7/2015, shown in Fig. 6.a., were considered in the model
two SVLs in the first junction box would fail. Note that the
other trapped charge levels at the time of cable energisation
measured and presented in this paper were not high enough to
cause any SVL failure.
Fig. 13. Rate of Change of Voltage for 31/7/2015 The same model described in [3] and [8] is used here to
examine the chain failure of SVLs at different junction boxes.
Reference [5] provides details for estimating the time Note that [8] assumed that at the time a SVL’s absorbed
constant of discharge of a cable when it is not earthed. The energy exceeds its rating of 31 kJ the SVL impedances
atmospheric parameter variations should be taken into account reduces to zero and it becomes like a short circuit. It was also
when there is an exposed part of the cable circuit. Note also shown that the secondary transient generated by the collapse
that the rate of discharge is also a function of the initial of voltage at the terminal of the first failed SVL was
responsible for the breakdown of the adjacent SVL on the The next pair of SVLs to fail is that in junction box
other phase in the same junction box [8]. The purpose of this JB2148, some 8 km away from the closing circuit breaker.
section is to examine whether the SVLs in other junction Fig. 17 shows the energy absorbed by the SVLs in junction
boxes would also breakdown when the failed SVLs no longer box JB2148. Note that the SVLs in JB2154 had been replaced
are operating correctly. The earthing impedance of the with short circuit. These are the last SVLs to fail because the
junction boxes is assumed to be 15  for all results presented next simulation with the failed SVLs in JB2148 replaced with
hereafter [3], [5], [8]. very small resistances showed that the energy absorbed by the
For all the tests that follow, the trapped charges at the SVLs in JB2136 did not exceed the rating.
instant of cable energisation are considered to be 0.75, -0.75
and 0.55 pu of the peak of the nominal voltage in phases a, b
and c respectively. The energisation is assumed to occur at the
positive peak of the nominal voltage on Phase b.
First it is assumed that the failed SVLs change their
impedance to a short circuit when they fail. The energy
absorbed by the SVLs in the first junction box, JB2174, 1.547
km away from the closing circuit breaker is shown in Fig. 14.
As can be seen, Phases b and a are the first and second SVLs
to fail. The SVL on Phase a fails as a result of the secondary
transient produced by the Phase b SVL breakdown [8]. No
other SVL fails.

Fig. 15. Energy Absorbed by SVL in JB2167

Fig. 16. Junction Box Locations

Fig. 14. Energy Absorbed by SVL in JB2174

The same amount of trapped charges are now considered

but with the failed SVLs, in Phase b and Phase a in JB2174
replaced with very low resistance element (0.001) to
represent short circuited SVL. All other parameters in the
model remain the same. Fig. 15 illustrates the energy
absorbed by the SVLs in the next junction box, JB2167. As
can be seen the SVLs in Phases a and b have failed in a much
shorter time. This may be attributed to the fact that the
Fig. 17. Energy Absorbed by SVL in JB2148
opposite polarity waves on Phase a and b drive much larger
current in the loop between junction boxes JB2174, JB2160
The above results show that the failure of the first SVL may
and JB2167 when the SVL in the former are replaced with
lead to a chain of failures if the trapped charges are high
short circuit. This high current and the clipped voltages of 20
enough and the SVLs impedance becomes very low (short
kV across the SVLs in Phases a and b in JB2167 with opposite
circuit mode) after it breaks down. With the first two SVLs in
polarity cause high current to flow in the loop limited mostly
junction box JB2174 failed, the SVLs in JB2167 will also fail
by the sheath impedance between the junction boxes and the
with trapped charges at 60%, −60% and 45% of the nominal
non-linear resistance of SVLs. Note that the junction box on
respectively in Phases a, b and c which would not happen if
the other side of JB2167 (JB2160), as shown in Fig. 16, is a
the SVLs in JB2174 were intact, i.e. healthy.
major point and therefore has physical short circuit to earth,
If it is assumed SVLs adopt a short circuit resistance of
thus making Phases a and b in JB2167 between two earthed
0.01  instead of 0.001, then the SVL in JB2154 in Phase B is
points with relatively short distance.
the last SVL to fail. If this SVL is replaced with a resistor of
If the failed SVLs in JB2167 are replaced with the short
circuit and study is run with all other parameters including the 0.01  then no other SVL fails, indicating the low impedance
trapped charges the same as before, the SVLs in the next path created by failed SVLs reduces the transient voltages in
junction box, JB2154, also fail, but in a longer time after the sheath. This test shows that the failure mode of SVL and
switch closing.
the impedance they adopt when they fail is important in the Further simulation showed that if SVL breakdown leads to
subsequent SVL failure. complete open circuit in the SVL then only one SVL in
The actual failure mechanism of MOV-based SVLs is not JB2174 and subsequently one in JB2167 respectively in Phase
clearly understood. The maintenance team, who test every b and a would fail, for the same amount of trapped charges. In
failed SVL offline, advised that all SVLs exhibit low this scenario SVL failure causes an open circuit in the SVL
resistance after the failure. However, it is not known what and then for subsequent energisation the failed SVL was
value of resistance exactly a failed SVL adopts under voltage replaced by an open circuit. This is similar to the condition
stress in the field. with 100  linear resistor in series with 1H inductance to
Using the model, tests were conducted to examine the chain represent a breakdown in SVLs.
failure mechanism considering two levels of resistance for
SVLs after they failed. First it was assumed that the SVL’s VII. CONCLUSIONS
non-linear characteristic changes to a linear 100  resistance Field measurement results were presented to confirm that if
in series with a 1 H inductance at the instant its energy a cable is not connected to earth after isolation, for example by
exceeds the declared rating of 31 kJ. Fig. 18 illustrates the means of earth switches, a charge can remain on the cable for
energy absorbed by the SVLs in JB2174, the nearest junction a long period of time. Furthermore, it was shown that the rate
box to the closing circuit breaker. The model is similar to that of decay of the trapped charge is influenced by the
used to obtain the results shown in Fig. 14, except for the atmospheric parameters such as humidity and temperature and
failure model. As can be see only SVL in Phase b has failed the length of the exposed part of the circuit.
and the transient generated by the failure is not enough to SVL failure data was presented together with the weather
cause breakdown of the SVL in Phase a. Replacing the failed information during two periods of time when a large number
SVL in Phase b of JB2174 with linear R-L element with the of SVLs had failed. Good correlation was observed between
above values, the cable was energised again with all the failure rate and warm/dry weather, confirming that enough
parameters including the trapped charges remaining the same. trapped charges may remain on the cable until the next
No other SVL in JB2174 failed. In the next junction box, energisation and that they were the root cause of high number
JB2167, the Phase a SVL failed. Fig. 19 shows the energy of SVL failure.
absorbed by the SVLs in JB2167. No other SVLs failed after Using the voltage of trapped charges obtained from
the faulty SVLs in JB2174 and JB2167 were replaced with measurements at the instant of cable energisation, it was
linear R-L circuit to represent failure. shown through simulation that the mode of SVL failure
influences the chain breakdown of SVLs for subsequent
energisation. The worst mode is when a SVL’s relatively high
impedance changes to a short circuit when it fails. It was
shown that SVLs up 8 km may be affected. If however it is
assumed that a failed SVL becomes open circuit at the instant
of break down then the subsequent failures will be less and
only SVLs in shorter distance from the closing circuit breaker
are affected, for the same amount of trapped charges.

VIII. REFERENCES
[1] IEC 60141-1 :1993 : Tests on oil-filled and gas-pressure cables and their
accessories - Part 1: Oil-filled, paper or polypropylene paper laminate
insulated, metal-sheathed cables and accessories for alternating voltages
up to and including 500 kV, 1993.
Fig. 18. Energy Absorbed by SVL in JB2174 with Linear Resistance- [2] CIGRE WG C4.502, "Power System Technical Performance Issues
Inductance Representing Failed SVL Related to the Application of Long HVAC Cables," CIGRE Technical
Brochure, No. 556, Oct. 2013.
[3] F. Ghassemi.: Effect of trapped charges on cable SVL failure, IPST
2013, Paper ID-219, Vancouver (Canada) June 2013.
[4] M. Takaoka, “The potential distribution of direct current oil-filled
cable”, IEEE Trans. Power Apparatus and Systems, vol. PAS-90, no. 6,
pp. 2622-2630, 1971.
[5] I. Lafaia, F. Ghassemi, A. Ametani, J. Mahseredjian, S. Dennis,
A. Haddad, S. Robson.: Experimental and theoretical analysis of cable
discharge, IEEE Trans. on PRWD, DOI:
10.1109/TPWRD.2016.2602361, August 2016.
[6] A. Ametani, C. T. Wan.: Sheath overvoltages due to a fault on an EHV
cable, IPST-1995, pp. 87-92, 1995.
[7] P. Tant, B. Bolsens, T. Sels, D. Van Dommelen, J. Driesen, R. Belmans:.
Design and Application of a Field Mill as a High-Voltage DC Meter,
IEEE Trans. Instrumentation and Measurement, Vol. 56, Issue: 4, 2007.
[8] F. Ghassemi, S. Dennis, A. Ainsley, A. Haddad, S. Robson.:
Measurement of Cable Trapped Charges and Discharge, EEUG 2016
Fig. 19. Energy Absorbed by SVL in JB2167 with Linear Resistance- Conference,, Paper 104, Birmingham, UK, September 2016.
Inductance Representing Failed SVL