Partial Discharge Pulse Propagation,

Localisation and Measurements in Medium

Voltage Power Cables
David Clark Ross Mackinlay Riccardo Giussani
University of Manchester HVPD Ltd HVPD Ltd
dave.clark@hvpd.co.uk ross.mackinlay@hvpd.co.uk Riccardo.giussani@hvpd.co.uk

Lee Renforth Roger Shuttleworth

HVPD Ltd University of Manchester
lee.renforth@hvpd.co.uk roger.shuttleworth@manchester.ac.uk

Abstract - This paper considers the measurement and transformer (HFCT) sensors attached to the cable earths
propagation of partial discharge pulses on distribution class allow the circuits to be tested without de-energisation,
power cable circuits, with the idea of attempting to determine whilst the circuit experiences normal operational stresses
discharge location sites on cables based on the parameters of and loads. Fig. 1 shows a typical temporary HFCT sensor
individual pulses. Single-ended discharge location techniques
attachment for measuring any PD pulses present within the
based on partial discharge pulse parameters and shape will
cable sheath as a result of PD activity between phases and
not be as accurate as time-domain reflectometry methods but
can be done on-line without the need for double-ended earth.
techniques. Power cables used for the transmission of 50/60
Hertz electrical power, are by design not intended to carry
transients or partial discharge pulses. The geometry and
construction of such power cables present a transmission line
that can heavily attenuate and distort the partial discharge
pulses, making their detection and discrimination all the
more difficult. Experimental and field work has been carried
out to develop basic knowledge rules to describe how the
time-domain parameters of individual pulses alter as a
function of the distance propagated from the discharge sites
for medium voltage power cables.

Index Terms—On-line, power cable, pulse broadening,

Time-domain Reflectometry, partial discharge, pulse
propagation.
I. INTRODUCTION
Partial discharge (PD) activity causes the gradual
breakdown of electrical insulation and can occur in any
medium voltage (MV) or high voltage (HV) plant or asset.
Common causes include mechanical damage to the
insulation system, microscopic voids or inclusions within
the insulation or the result of contamination or electrical
treeing [1]. At present there is still a majority of aged, 11
and 33 kV, paper-insulated lead covered (PILC) power
cables in-service over the UK utility distribution networks,
as well as cross-linked polyethylene (XLPE) type cables
[2]. Some of the various PD-related failure mechanisms
suffered by in-service power cables include delamination
of insulation layers and electrical treeing. On-line PD
testing of medium voltage, distribution class power cables
is a proven technique to assess the overall condition of the
insulation [3]. Non-intrusive high-frequency current
Fig. 1 High frequency current transformer sensor attachment on an 11
kV cable earth
However, measuring the cable PD activity forms only part
of the work. Locating a discharge site on the network can
ensure that a faulty section of cable can be replaced before
a complete breakdown of the insulation occurs. One of the
traditional methods of determining areas of discharges
within power cables relies on timing of pulses. The time-
domain reflectometry (TDR) technique is briefly reviewed
here.
A. Time-Domain Reflectometry
The TDR method can be carried out on-line or off-line
by injecting a calibration pulse or an edge (with a suitable
pulse width that exceeds the return time of the cable length
typically 0.1 – 10 µs) into the cable conductor or sheath.
Any impedance mismatches such as cable joints, fault

978-1-4799-3254-2/13/$31.00 ©IEEE
locations and the end of the cable can often be observed as
a partial pulse reflection [4]. Although the TDR method is
typically accurate to within one per cent of the cable length
[5], the technique requires the setting up of the trigger unit
and the transponder. More importantly the pulse must be
observable at the cable far end. Fig. 2 below shows the
typical arrangement for an on-line TDR test with the
trigger unit, the transponder and HFCT sensors for pulse
detection and injection.
Fig. 2 Time-domain reflectometry test set-up with transponder
Assuming a pulse reflection from the end of the cable
circuit can be seen, which may not always be possible for
particularly long cable circuits, then using (1) it is possible
to calculate the position of the discharge site as a
percentage of the cable length.
∆ R
% 1 100
Where
l cable length (m)
∆ time difference between consecutive pulses
To further understand how PD pulses become distorted as
they travel along power cables, a brief discussion of the
theory of PD pulse propagation is presented in the next
section.
II. PROPAGATION OF TRANSIENT PULSES IN CABLES
Fast fronted, short duration transients such as PD pulses
often consist of a broad band of frequencies up to the MHz
range. At the discharge location a PD pulse may be
considered to be Dirac-like in nature, with parameters such
as pulse width and rise time typically within the range of 1
– 10 ns. Such broad band pulses travel along a
transmission line with a group velocity which ( ) is
dependent upon the derivative of the angular frequency
and the wavenumber given in (2).
2
From (2) it can be deduced that is not equal to a
constant, and is in fact a function of frequency, which
results in distortion of the pulse shape with distance
propagated. Each frequency component within the pulse
travels at its own velocity, hence the higher frequencies
move away from the front of the pulse and the peak,
resulting in pulse attenuation and broadening. The second-
order differential equations, known as the telegraph
equations allow for dispersive, travelling wave solutions.
These describe how pulses are distorted on power cables
and general transmission lines and are derived from
Maxwell’s equations. These relate the dielectric properties
of a waveguide to the distortion seen by a pulse at some
distance, see [6] for further details. Pulse magnitude
attenuation, and broadening are described by the real, and
imaginary parts of the frequency-dependent propagation
constant γ(ω) which appears in the exponent of the transfer
function H(ω) associated with pulses measured at distances
l and x, given in (3)
3
4
Where
α(ω) pulse magnitude attenuation (Np/m)
β(ω) pulse dispersion (rad/m)
resistance (Ω/m)
1 L inductance (H/m)
G conductance (S/m)
C capacitance (F/m)
j √ 1
ω frequency (Hz).
Equations (3) and (4) can be used, along with the
distance the pulse has travelled (l) to compute the
amplitude attenuation and the pulse broadening, although
for a power cable these parameters are often difficult to
determine. As a pulse travels along a cable its shape
becomes distorted, which results in an increase in the pulse
rise and fall times and also the pulse width. It is also
possible to extend this and state that the pulse amplitude
and energy decreases with distance propagated, assuming
that reflections from any impedance mismatches are not
superimposed onto the pulse. Partial discharge recognition
and discrimination techniques often use the pulse rise time
and pulse width to distinguish them from interference and
noise [7], [4]. Typically PD pulses are far shorter duration
than noise and interference due to the electron avalanche
process that causes the insulation to breakdown. Thus
understanding how PD pulses distort as they propagate
through MV networks is important for f PD source
classification and also for the proposedd pulse shaped-
based, PD location techniques.
III. EXPERIMENTAL WOR
RK
Initially to study how transient pulses altter as they travel
along power cables, laboratory work was carried
c out using
RG223 co-axial cable and also a sevenn metre sample
length of XLPE, 11 kV-rated power cable. The aim of this
work is to build on the understandingg of PD pulse
propagation in power cables [8] andd parts of the
Electromagnetic Transient Program (EMTP P) work that has
been carried out [9] and develop a meaasurement-based
model of the phenomena that can be applieed to actual field
data. Equations derived in [10] for voltaage and current
surges as a function of distance have form med part of the
theoretical approach and been used as a a basis for
developing a measurement-based model. For this
experimental work a Lecroy Wavverunner, 44xi
oscilloscope was used with a bandwidth of o 350 MHz at a
sampling rate of 5 GHz/s. The pulse generaator used was an
Agilent 33250A with a bandwidth of 80 MH Hz.
A. Experimental Work on RG223 Co-axial Tyype Cable
A power cable generally consists off one or more
screened conductors surrounded by layerrs of insulation,
earth sheathing and a PVC extruded ovver-sheath. The
construction is effectively co-axial and hass been modelled
using RG223 cable, which has a 1 mm m² copper inner
conductor, solid polyethylene insulatioon, aluminium
braided outer screen, an impedance off 50 Ω and a
polyethylene over-sheath. Fig. 3 shows how w both the pulse
rise time and pulse alter over 3200 metres for a 5 V input
pulse with an initial pulse width of 10 ns annd an initial rise
and fall time of 5 ns.
Fig. 3 Pulse width as a function of distance propagatted for a 10 V Pulse
with a 5 ns initial width on an XLPE cable
Fig. 3 above shows that under laboratory conditions
c it was
possible to fit second-order polynomial cuurves to both the
pulse rise time and widtth as functions of distance
propagated.
B. Experimental Work on an XLPE
X Cable Sample
To extend the experimenntal work further off-cuts of
XLPE power cables were obtained.
o The particular cable
tested was a seven metre length of single-core, solid
conductor, XLPE, 11 kV rated, 300 mm², aluminium
conductor cable with innerr and outer semi-conductive
layers, a copper sheath and a PVC extruded over sheath.
Pulse injection tests were peerformed, the results of which
are presented here. A 1N41448 Schottky diode was used to
effectively contain the multiiple reflections from the open
circuit end of the cable. Partt of the test setup is shown in
Fig. 4 and a line diagram reepresentation of this in Fig. 5.
This indicates how with only a seven metre cable
specimen it was possible to use
u a Schottky diode to obtain
multiple pulse reflections froom the open-circuit end of the
cable. Hence for an open cirrcuit the reflection coefficient
(Г) is unity, resulting in a total
t reflection of the incident
pulse.
Fig. 4 Pulse Injection test setup for the XLPE cable Specimen
Fig. 5 Pulse Injection test setup linee diagram representation
Measurements of pulse rise time
t and pulse width were also
carried out and plotted againnst distance propagated along
the cable. This allowed reliable measurements of up to
1260 metres in steps of 14 metres
m (twice the cable length)
to be obtained, and plotted in
i Fig. 6 below. Thus when a
simulated PD pulse with ann intial 10 ns pulse width is
injected into the conductor the pulse rise time and width sections. Table I provides deetails of the calculated velocity
increase approximately linearly as a functtion of distance of propagation, which connfirms the cable lengths and
propagated positions of the HV connectoors.

TABBLE I
VELOCITY OF PROPA
AGATION CONSTANTS

Cable length (m) Returrn time (µs) Return speed (m/µs)

498 6.26 81.15
798 10.04 80.47
1398 17.45 80.69

The calculated values weree in agreement with figures

Fig. 6 Pulse width as a function of distance propagatted for a 10 V Pulse
with a 5 ns initial width
The main causes of pulse attenuation andd dispersion for
the XLPE cable specimen are the presennce of the semi-
conductive layers and skin-effect losses which
w increases
high-frequency losses [8].
quoted by other authors forr this type of cable [5]. Pulse
injection tests were carried out
o with a range of initial pulse
widths to try and understand how the XLPE cables and the
HV connectors alter the parrameters of the simulated PD
pulses, and to try and understtand the ‘natural dispersion’ of
the cable i.e. the highest freqquency component present for
a given cable length. Pulses were injected at the first HV
connector and measured at a various points upon the
network. An example of a siimulated pulse propagating on
the network is shown below in i Fig. 7.

IV. PROPAGATION TESTS ON AN ARTIFICIAL

NETWORK
Pulse injection experiments and PD tessts were carried
out on some XLPE power cables. The arrtificial network
consisted of 1398 metres of single-core, 180 mm², XLPE,
copper conductor/sheath cables laid in a trefoil 0 10 20 30 40 50 60 70 80 90 100

arrangement, interconnected with highh-voltage (HV) Time (µs)

Fig. 7 Injected 50 V Pulse propagating 498 metres between HV
connectors and a step-up transformer. Thee HV connectors
connector 1 and HV connector 2
are located at 498, 798 and 1398 metrees, which allow
direct measurements with a voltage proobe and also a To compare parts of the experimental work that was
HFCT sensor. A discharging joint was located at 248 carried out on RG223 co-axxial type cables, simulated PD
metres from one of the HV connectors. Measurements
M of pulses consisting of differennt pulse widths (with the rise
pulse propagation were made with a range of pulse shapes and fall times, and amplitude kept constant) were injected
to assess the frequency response of the cables and the and measured upon the artifficial MV network, the results
effect of the HV connectors upon the pulsee characteristics. of how these parameters off interest alter as function of
The system is realistic in the sense that most of the distance propagated throughh the network are plotted next.
components are used in real distribution systems, but of Fig. 8 shows how the artificiial PD pulse rise time (10 – 90
course are easy to access for any measuurements. Pulse per cent of the maximum puulse amplitude) increases as a
propagation tests were carried out by launnching PD type function of distance propagatted.
pulses with an Agilent 33250A 80 MHz funnction generator.
This allowed pulses with amplitudes of 10 V with rise and
fall times of 5 ns to be used as a propagatiion source. The
pulse generator was normally applied at the t input of the
first HV connector; this corresponds to the start of the
circuit. The other HV connectors can thhen be used as
measuring points along the cable. Pulsee characteristics
with distance data were then constructed frrom this, as well
as velocity of propagation constants for thee different cable
 increased as a result [11]. Ass with the pulse rise times, the
pulse widths of the simulatedd PD pulses were measured as
a function of distance propaggated, these are plotted in Fig.
10. Shown in Fig. 10 is pulse width gainst distance
propagated plots with secondd order (in length) polynomial
curves fitted to the data.

Fig. 8 Pulse rise time as a function of distance proppagated for an 10 V

Pulse with initial widths from 10 – 100 ns

On-line PD testing is often carried out with inductive,

ferrite-core, HFCT type sensors that cann be temporally
installed upon cable earths. Fig. 9 below shows how the
rise time of a simulated PD pulse, with an initial pulse
width of 10 ns compares with measurements carried out
with a HFCT sensor. Fig. 10 Pulse width as a function off distance propagated for a 10 V Pulse
with initial widths froms 10 – 100 ns

A consequence of dispersioon is pulse broadening, which

Fig. 9 Pulse rise time as a function of distance proopagated for a 10 V
Pulse with a 10 ns initial width when measured witth the HFCT sensor
and the voltage probe
increases the pulse width (50 percent amplitude points on
the rising and falling edges of the pulse) as it propagates
along the cable. At the dischharge site PD pulses are short
in duration (typically < 10 1 ns) the effects of pulse
broadening are more pronouunced as there exists a greater
group delay between the freequencies, this can be seen as
the spectrum of the pulse iss broader. Fig. 11 shows how
the pulse width of an intial 10 ns wide pulse alters over the
1398 metres upon the test network
n when measured with
the voltage probe and the HF FCT sensor.

Using the HFCT sensor to make measuurements has the

advantage that the circuit need not be de-energised,
although the flat frequency response off the sensor is
limited to a bandwidth of approximately 200 MHz at the -3
dB points. Such limitations are based on thhe type of ferrite
used in the toroid, the magnetic circuit leength (and area)
and the number of turns upon the secondaary coil. Fig. 11
shows that over the 1398 metres the pulse p rise time
increases linearly as a function of distannce propagated.
Differences in pulse rise times measured with
w the voltage
probe and the HFCT sensor may be due thhe configuration Fig. 11 Pulse width as a function off distance propagated for a 10 V Pulse
of the HV connectors and the inclusion off the sensor. To with a 10 ns initial width when measured with the HFCT sensor and the
voltage probe
allow the pulses to travel through the HV V connector re-
configuration of the network was required. This was a
Measurements performedd directly using the voltage
condition which did not apply to the direcct measurements
probe showed a nonlinearr response as a function of
where signals were measured at open points p upon the
distance propagated. In com
mparison to the pulse width as
network. Inclusion of the series inductaance of the HV
measured using the HFCT seensor which showed the pulse
connector shows that the rise time off the pulses is
width to increase approximaately linearly as a function of
the distance propagated. The final part of the work was to
compare the results with some of the measured PD pulse
parameters, with the system energised at 14 kV Phase-to-
earth. An example of a PD pulse from the discharging
cable section is shown in Fig. 12.
A 25
20
mp 15
10
lit
Volts (mV)
5
0
ud
-5
-10
-15
e -20
-25
( 5
Time (µs)
6 7
Fig. 12 Partial discharge pulse as measured at HV connector 1, 250
metres from the PD Location – PD 1
The PD pulse in Fig. 12 was measured at 250 metres from
the discharging cable section at HV connector 1. From Fig.
11 the predicted pulse width and rise time as functions of
cable length (l) was given by the linear approximations (5)
and (6) respectively
0.1468
, 0.0705
Parameters as measured with the HFCT sensor for the PD
pulse are provided in Table II. These equations were used
to calculate the predicted PD pulse parameters shown in
Table II.
TABLE II
MEASURED AND PREDICTED PD PULSE PARAMETERS
Pulse parameter Measured (ns)
Pulse width 76
Pulse rise time 22.41
For this particular PD pulse, and a number of those
recorded the parameters were in good agreement with
those predicted from the plots shown in Fig. 9 and Fig. 11.
Although there was variation between the pulses, this may
be due to reflections causing distortions of the pulse shape,
or limitations of the measurement bandwidth.
V. CONCLUSION
Location of PD pulses on MV cable networks is most
accurately done using TDR and transponder methods.
However, for location using only single ended data,
corrections for PD charge, can be attempted using the
parameters of the PD pulse. The cable propagation
parameters must be known for this to be useful, and in
general, this will be the main problem for the method.
However, it would allow simple corrections of charge and
locations on cables to be made with data which is regularly
collected during simple PD measurements. Experimental
pulse injection tests carried out on RG223 co-axial cables,
and MV power cables has shown that pulses distort as they
propagate. Understanding the change in the pulse
parameters as a function of the distance propagated makes
single-ended PD locations more conceivable for MV
power cables. Although both the pulse rise time and pulse
width become longer with distance propagated along the
cable, using the pulse width as a locating parameter may
prove more accurate as it appears to show more stable
results as a function of distance. Work on XLPE cables has
shown encouraging results, where PD locations can work
providing the propagation parameters can be estimated for
each cable type.
REFERENCES
[1] Renforth, L, Mackinlay, R, Seltzer-Grant, M, Clark, D and
Shuttleworth, R, “Experiences from over 15 years of On-line Partial
Discharge Testing of In-service MV and HV Cables, Switchgear,
Transformers and Rotating Machines”, in IEEE IX Latin American
Conference on Automatic Control and Industry Applications
(LARC)1-4th October, 2011.
40 5 [2] M. Seltzer-Grant, D. Denissov, R. Mackinlay, F. Petzold, L.
Renforth and H. Schlapp, “On-Line Continuous Monitoring for In
Service Distribution Class Cables and Switchgears,” Proceedings of
5 6 the 21st International Conference on Electricity Distribution
(CIRED), Frankfurt, 6-9 June 2011.
[3] D. Clark, R. Mackinlay, M. Seltzer-Grant, L. Renforth, J.
McWilliams and R. Shuttleworth, “On-Line Partial Discharge
Testing of some of the Worst Performing Circuits on a Utility
Distribution System,” INSUCON, 2013.
[4] Mackinlay, R, Michel, M and Walton, C, W, “Tools for Partial
Discharge Testing MV cables and Plant”, 18th International
Conference on Electricity Distribution (CIRED) Turin,6-9th June,
2005.
[5] B. Howarth, M. Coates, and L. Renforth, “Fault Location
Techniques for One of the World ’ s Longest AC Interconnector
Cables,” in IEE ACDC International Conference on AC and DC
Predicted (ns) Power Transmission, 2006, pp. 14–18.
76.7 [6] J. D. Glover, M. S. Sarma and T. J. Overbye, “Power System
Analysis and Design,” 5th Edition, 2011.
22.62 [7] R. Mackinlay and C. Walton, 2001, “Diagnostics for MV cables and
switchgear as a tool for effective asset management", Proceedings
CIRED June 2001.
[8] S. Boggs, A. Pathak, and P. Walker, “Partial discharge. XXII. High
frequency attenuation in shielded solid dielectric power cable and
implications thereof for PD location,” IEEE Electrical Insulation
Magazine,, vol. 12, no. 1, pp. 9–16, 1996.
[9] I. Cotton, V. O’Donnell, and N. Christofides, “Limitations in the
application of on-line and off-line PD measurement systems,” in
CIRED International Conference and Exhibition on Electricity
Distribution, 2005, pp. 1–5.
[10] A. P. Morando and S. Leva, Note di teoria dei CAMPI
VETTORIALI. Progetto Leonardo BOLOGNA, 1998.
[11] Wagenaars, P, “Integration of Online Partial Discharge Monitoring
and Defect Location in Medium-Voltage Cable Networks”. Ph. D
Thesis, Technische Universiteit Eindhoven, 2010.