297

PRACTICAL ASPECTS OF THE

DETECTION AND LOCATION OF
PARTIAL DISCHARGES IN POWER
CABLES

Task Force
D1.02.05

June 2006
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

PRACTICAL ASPECTS OF THE

DETECTION AND LOCATION OF PARTIAL
DISCHARGES IN POWER CABLES

Task Force
D1.02.05

E. Lemke, E. Gulski, W. Hauschild, R. Malewski,

P. Mohaupt, M. Muhr, J. Rickmann, T. Strehl,
F.J. Wester

Copyright © 2005
“Ownership of a CIGRE publication, whether in paper form or on electronic support only infers right of use for personal
purposes. Are prohibited, except if explicitly agreed by CIGRE, total or partial reproduction of the publication for use other
than personal and transfer to a third party; hence circulation on any intranet or other company network is forbidden”.
Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any responsibility, as to
the accuracy or exhaustiveness of the information. All implied warranties and conditions are excluded to the maximum extent
permitted by law”.

Page 2 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

Content

1. Introduction 3

2. Fundamentals 4

2.1 PD occurrence in power cables 4

2.2 Attenuation and dispersion of PD pulses in power cables 5
2.3 PD wave velocity in XLPE power cables 9

3. Experimental studies for evaluating the PD pulse attenuation 10

3.1 Test procedure 10

3.2 Test results 11

4. Practical aspects of PD fault location 16

4.1 Single PD source 16

4.2 Multiple PD sources 18
4.3 Mixed cable systems 19

5. Summary 21

6. References 22

Page 3 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

Abstract:

For enhancement of the reliable operation of power cables network preventive PD diagnosis tests are
increasingly performed. In order to assess the insulation condition on the basis of the obtained data,
fundamental knowledge on the PD occurrence is required. The presented brochure deals with both,
theoretical analysis and experimental studies on the wave propagation of PD pulses in power cables.
Furthermore, practical aspects of the detection and location of PD faults are discussed.

Index Terms:
Power cables, travelling waves, partial discharges (PD), PD diagnostics, PD fault location, PD mapping

1. Introduction
The conventional PD measuring
method specified in the relevant
IEC standards [1; 2] has well
been proven under laboratory
condition. Presently, this
technique is increasingly used
under on-site condition, in
particular for PD testing of
installed power cables [3]. One
issue of this paper is to develop
an approach which can simply
be handled for assessing the Fig. 1: Equivalent circuit for PD defects in power cables
sensitivity, which can be
achieved for the detection and location of PD defects in long power cables under on-site condition [4].
Furthermore, the PD occurrence is discussed from a practical point of view, such as the impact of the
superposition of PD pulses as well as the different wave propagation velocities in mixed cable sections
on the location uncertainty [5]. For better understanding first some fundamentals of the very complex
PD occurrence in power cables will be presented.

Page 4 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

2. Fundamentals

2.1 PD occurrence in power cables

As known from the literature,
partial discharges (PD) are caused
by insulation imperfections and
bridge only a part of the insulation
between the electrodes [6, 7].
Generally it can be distinguished
between internal discharges,
surface discharges and external
corona discharges. The occurrence
of internal discharges, which may
appear in gas-filled voids, cavities
and cracks, is usually explained
by the a-b-c model, where the test
object is substituted by a lumped
capacitor. In geometrically Fig. 2: Principle of PD pulse propagation in power cables
a) just after PD ignition, both pulses travel in the opposite
extended test objects, such as direction
power cables, power transformers b) after reflection at the “remote end”, both pulses travel in the
and rotating machines, the PD same direction
pulses propagate like travelling
waves. Under this condition the
classical a-b-c approach has to be
modified accordingly, as evident
from Fig. 1. Here the insulation
defect is also represented by the
capacitor Cc connected in series
with Cb which represents the
remaining healthy insulation, but
the lumped capacitor must be
substituted by the characteristic
impedance Z0. If the resulting
voltage across Cc becomes higher
than the breakdown voltage of the
defect, i.e. a puncture of the
parallel connected spark gap Sg
occurs, a current pulse of Fig. 3: Principle of the PD fault location using the time domain reflec-
tometry, cable length 1705 m, PD site from near end 1297 m
extremely short duration, usually in
the ns-range, is ignited. For such

Page 5 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

short pulses long power cables Inductance per unit length:

behaves like a travelling wave L´ = (µ0 / 2π) * [ln (d2 / d1)] (1)
guide and can therefore not be
Capacitance per unit length:
treated as a lumped capacitor
C´= 2π * ε0 * εr / [ln (d2 / d1)] (2)
Ca, but by the characteristic
Here are:
impedance Z0, as mentioned
d1 - the diameter of the inner cable conductor
previous. Due to the symmetrical d2 - the diameter of the outer cable conductor
configuration of power cables µ0 - the permeability constant
each PD event produces two ε0 - the permittivity constant
equal current pulses, i.e. the
εr - the relative permittivity of the extruded cable insulation.
pulse charge magnitude q / 2
travels away from the PD source Resistance per unit length due to the skin-effect:
in both directions, as R´= [kc / (π * d1)] * [ω * µ0 / (2 * κc)]1/2 (3)
schematically reflected in Fig. 2. Here are:
At the “remote end” of the cable Ω -the radian frequency (ω = 2πf)
section, which is generally open κc -the specific conductance of the inner cable conductor
during the PD measurement, a kc -a correction factor taking into account the losses in the outer
full reflection occurs. Therefore conductor
at the “near end” of the cable, Conductance per unit length due to the semiconducting layer:
where the PD coupling unit is G´= [ks * 4π * ds / (d1 * κs)] * {ε0 * εr * ω / [ln (d2 /d1)]}2 (4)
connected, not only the direct Here are:
pulse but also the reflected pulse κs - the specific conductance of the inner semiconducting layer
can be detected. This occurrence ds - the thickness of the inner semiconducting layer
is used for location the PD site ks - a correction factor taking into account losses in the outer
using the time domain semiconducting layer
reflectometry (TDR), as known Characteristic cable impedance:
from the relevant literature, as
for instance [3; 4; 5; 8; 9] and Zc = [R´+ jωL´ ]1/2 * [G´+ jωC´ ] -1/2 (5)
schematically shown in Fig. 3. Propagation constant:

2.2 Attenuation and γ = [(R´+ jωL´ ) * (G´+ jωC´ )]1/2 (6a)

γ = α + jβ (6b)
dispersion of PD pulses in
Attenuation constant:
power cables
{
α = (1/2)* ωL´* jωC´* {{[1 +(R´ / ωL´)2]*
For the analysis of the PD pulse
propagation in long power
}
[1 + (G´ / ωC´)2 ]}1/2 – 1} + (1/2)* R´* G´ 1/2 (7)
Phase constant:
cables, the travelling wave
theory must be used. This is {
β = (1 / 2)* ωL´* jωC´* {{[1 + (R´ / ωL´)2] *
based on the electrical line
model, already published at the
}
[1 + (G´ / ωC´)2 ]}1/2 + 1} - (1/2) * R´* G´ 1/2 (8)

beginning of the last century Table 1: Basic equations for analysis of travelling waves in single core
[10; 11]. The equivalent circuit power cables

Page 6 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

is reported in Fig. 4a, where the circuit elements L´, C´, R´ and G´ represent the inductance,
capacitance, resistance and conductance per unit length, respectively. Taking into account the greater-

a) b) c)

Fig. 4: Equivalent networks for analysing the travelling wave occurrence in power cables
a) classical circuit line model b) complex radial semicon layer loss model [12]
c) simplified semicon layer loss model

than-expected losses in power cables the radial semiconducting loss model shown in Fig 4b was
proposed by Stone and Boggs [12]. Here the circuit elements G1´, G2´, C1´ and C2´ represent the
conductance and the capacitance per unit length of the inner and outer semiconducting layer,
respectively. This approach has well been proven for frequencies up to 100 MHz and even higher [12].
For the detection and location of PD events in long power cables, however, only frequency modes
below 10 MHz are of interest [3]. Thus the very complex circuit shown in Fig. 4b can essentially be
simplified as reflected in Fig. 4c. Here Rc´ represents the equivalent resistance per unit length of the
inner and outer cable conductor, Cs´ the capacitance per unit length of the cable insulation and Rs´ the
equivalent resistance per unit length caused by the inner and outer semiconducting layer. The impact of
the parallel capacitances C1´ and C2´ reflected in Fig. 4b can be neglected for frequencies below 10
MHz.
As known from the electrical network theory, the series connection of Cs´ and Rs’ in Fig. 4c is
substitutable by a parallel circuit, composed by the elements C´ and G´, which is in principle equivalent
to Fig. 4a. Consequently, for further considerations the classical electric line model is also applicable
for power cables if the
characteristic elements R´ and G´
are modified accordingly, as will
be reported more in detail below.
For a single core cable the previous
mentioned elements L´, C´, R´ and
G´ are given by the well known
formulas (1) to (4) listed in Tab. 1.
For further estimations the basic
equations, required for analyzing of
travelling waves [10, 11], are also Fig. 5: Attenuation of a multiple reflected PD pulse in a power cable,
time base 2.5 µs/div.

Page 7 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

listed in Tab. 1 and additionally

in Tab. 2. Without going into Phase constant of loss-free cables:
details it shall be recalled, that β = ω (L´* C´ )1/2 (9)
travelling waves can generally
be considered as the Phase velocity:
superposition of a forward and a vp = ω / β (10)
backward propagating wave,
given by formula (12) in Tab. 2. Phase velocity of loss-free cables:
For the measurement and v0 = (L´* C´ )-1/2= (µ0 * ε0 * εr )-1/2 = c0 * (εr )-1/2 (11)
location of PD events in power
cables usually the forward Voltage wave dependent on the travelling distance:
travelling voltage wave is of Uw = U1 * exp (-γx) + U2 * exp (γx) (12)
interest. Reflecting formula (6b) exp (-γx) = exp (-αx) * exp (-jβx) (13)
which can also be expressed by
Forward travelling voltage wave of a PD pulse:
formula (13), the forward
propagating PD pulse is given Upd * exp (jωt) = u1 * exp (-αx) * exp (jωt - jβx) (14)
by equation (14). Here the first Loss-free cables:
right term reflects the
jωt – jβx = 0 → ω / β = x / t = v0 (15)
attenuation of the PD pulse
Upd * exp (jωt) = u1 * exp (-αx) (16)
magnitude and the second term
the dispersion. For loss-free Table 2: Equations for analysis of PD pulses travelling through single
cables, where the condition (15) core power cables
is fulfilled, the second right
term of equation (14) becomes the value 1. This means that the PD pulse magnitude is attenuated
exponentially with the travelling distance, as reflected by the formula (16).
The increasing attenuation of PD pulses dependent on the propagation distance is well known from
practical PD tests of power cables. As an example, Fig. 5 shows a typical PD pulse reflectogram
recorded for a XLPE power cable of 480 m length, where the PD defect occurred at a distance of about
450 m from the near end. Because of the relative high input impedance of the applied PD coupling unit,
subsequent pulse sequences occurred, due to multiple reflections at both cable ends. Additionally to the
attenuation a dispersion can be observed, i.e. the PD pulse width increases with the travelling distance.
This is equivalent to a reduction of the upper frequencies. That means the cable behaves like a low-pass
filter. This is also reported in
Fig. 6, which reflects the pulse vp = (L´* C´ * As )-1/2 = c0 * ( εr * As )-1/2 (17a)
response of a cable sample in As = [1 + (1 + Bs)1/2] / [2 * (1 + Bs)] (17b)
the frequency domain [13]. 2
Bs = (ω Cs´ * Rs´) (17c)
Fig 6a reveals that the injected
calibrating pulses are Es = [ln (d2 / d1)] / [ln [ (d2 - 2ds) / (d1 + 2ds)] (18)
characterized by a frequency vp = (L´* C´ * Es )-1/2 = c0 * (εr * Es )-1/2 (19)
spectrum up to about 10 MHz. -1/2 -1/2
vp = c0 * (εr * Rs ) = c0 * (2.4 * 1,35) = 167 m/µs (20)
Because the cable was open at
both ends, the arriving pulses
are full reflected and traveled Tab. 3: Equations for evaluation the travelling wave velocity

Page 8 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

thus several times through the cable. As evident from Fig. 6b the frequency spectrum is significantly
reduced, if pulses travelling twice through the cable. This becomes more pronounced after travelling
four times (Fig. 6c) and at least six times through the cable (Fig. 6d). Finally only frequencies below 2
MHz remained. This confirms the before mentioned low-pass characteristic of the cable, which is not
only caused by the attenuation but also by the dispersion, due to the frequency dependent phase
velocity.
That means, in real power cables each frequency mode travels at a specific velocity, as reflected by
equation (8). Consequently, the PD pulse width is extended as exemplarily reported in Fig. 7.

a) b) c) d)
Fig. 6: Frequency spectrum of pulses travelling several times through a power cable
a) direct injected pulse b) after twice the cable length
c) after four times the cable length d) after six times the cable length

a) b) c)
Fig. 7: Dispersion of a PD pulse travelling through a 52 m long XLPE cable
a) Pulse shape at the near end (left) and remote end (right), time scale 40 ns/div
b) Pulse shape at the near end, time scale 2 ns/div
c) Pulse shape at the far end, time scale 2 ns/div
These records refer to a PD pulse charge magnitude of about 9 pC occurring at one end of a 52 m long
XLPE cable sample. After a travel time of 320 ns the PD pulse appeared at the remote end, as can be
seen from the right pulse in Fig. 7a, which is equivalent to a wave velocity of 162 m/µs. Whereas the
origin PD pulse decoupled from the near end is characterized by a pulse width of only 4 ns (Fig 7b), the
PD pulse width measured at the remote cable end is nearly 20 ns (Fig. 7c), which is around 5-times
longer if compared to the initial value. Additionally the peak value is reduced substantially, i.e. from
originally 130 mV down to about 40 mV, which is less than 30% of the initial value.

Page 9 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

2.3 PD wave velocity in XLPE power cables

For location of PD sites the wave velocity vp is the most important parameter and should thus well be
known. For a theoretical analysis of vp the very complex equations (7) and (8) can strongly be
simplified if only frequency modes below 10 MHz are of interest. A useful approximation is given by
equation (17a). Because the factor As expressed by (17b) is always greater than 1, even if the factor Bs
given by (17c) becomes extremely low, it could be written: vp > c0 * (εr )-1/2. But this relation
contradicts the reality, because the velocity of loss-free XLPE cables amounts about 190 m/µs if
calculated according to equation (11) for an assumed relative permittivity of εr = 2.4. For real XLPE
power cables, however, the characteristic values range around 160 m/µs, which is nearly 20% below
the theoretically expected value. The main reason for this discrepancy seems to be a fictive
enhancement of the capacitance per unit length due to the reduction of the effective insulation thickness
between the inner and outer semiconducting layer. In order to verify this assumption the wave velocity
of a 24 kV XLPE cable could roughly be estimated on the basis of the following measured parameters:

Relative permittivity of the insulation: εr = 2.4

Diameter of the inner conductor: d1 = 14 mm
Diameter of the outer conductor: d2 = 27 mm
Thickness of the semiconducting layers: ds = 0.8 mm

The characteristic capacitance ratio Es can be expressed by equation (18) where equation (11) is
substituted by the equations (19) and (20) in order to get the phase velocity in the presence of the inner
and outer semiconducting layers. Inserting the previous reported cable parameters in formula (20) the
velocity becomes 167 m/µs, which is well comparable to the measured velocity of the here considered
cable sample, which was 164 m/µs.
In this context it should be mentioned that the very complex occurrence of the PD pulse propagation in
shielded power cables has been investigated comprehensively in the past. Exemplarily the references
[12 -19] should be mentioned. The published results, however, refer mainly to the frequency domain
and not to the time domain, which is required for an evaluation of the attenuation of the PD pulse
magnitudes in order to assess the PD detection sensitivity, which can be achieved in very long power
cables. Thus, experimental studies on the PD pulse attenuation in power cables were performed in the
time domain [4]. In the following some results will be reported more in detail.

Page 10 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

3. Experimental studies for evaluating the PD pulse attenuation

3.1 Test procedure

The experiments were performed on power cables of current production. Injecting artificial PD pulses
by means of a PD calibrator the pulse response was recorded by a digital scope. Because both, the
output impedance of the calibrator and the input impedances of the digital scope, were several orders
above the characteristic impedance of the power cables under investigation, the near and remote cable
end could be considered as “open”. Consequently, subsequent pulse sequences appeared at the
measuring site, due to multiple pulse reflections at both cable ends. As will be shown in the following,
this occurrence can be used advantageously for a fictive elongation of the investigated geometrical
cable length. This advantage, however, can only be used, if no losses are caused by each reflection. In
order to verify this requirement the following check has been performed:

a) b) c)
Fig. 8: Results of studies on the impact of pulse reflections on the attenuation and dispersion
a) Shape of the injected calibrating pulse, time scale 20 ns/div
b) Reflectogram for a x m long BNC cable, time scale 400 ns/div
c) Reflectogram for a y m long BNC cable, time scale 400 ns/div
First calibrating pulses shown in Fig. 8a were injected in a 200 m long measuring cable and the
occurring pulse sequences at both cable ends were recorded by a digital scope (Fig. 8b). After that the
cable length was reduced to 50 % of the previous one, i.e. down to 100 m. Consequently, the number of
the recorded pulses at the same sweep is increased by twice times (Fig. 8c). For comparison purposes
let us first consider the third pulse shown in the record of Fig. 8c, which is caused by the first reflection
at the remote end but recorded by the first channel connected to the near end, i.e. the total distance
traveled by the pulse amounts 2 * 100 m = 200 m.
Let us now consider the second pulse in Fig. 8b recorded by the second channel of the scope, which
refers to the remote cable end. Because the magnitude of the third pulse in Fig. 8c is equal to those of
the second one in Fig. 8b it can be concluded, that no additional attenuation is caused by the reflection.
This behavior is also confirmed if the magnitudes of the fifth and of the seventh pulse in Fig. 8c are
compared with those of the third and fourth pulse shown in Fig. 8b, respectively. Consequently, an
impact of multiple pulse reflections on an additional attenuation of the pulse magnitudes can be
excluded.

Page 11 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

For a quantitative assessment

of the PD pulse attenuation
also the following aspect has to
be taken into consideration. As
known from the travelling
wave theory for loss-free Fig. 9: Characteristic pulse reflectogram obtained for a 30 kV XLPE of
cables the second PD pulse due 45 m length, recorded at a time scale of 2 µs/div (left)
to the reflection at the remote and 0.5 µs/div (right)
end appears at a higher
magnitude compared to the Cable- Rated Voltage Length Virtual Wave Velocity
No. [kV] [m] Length [m] [m/µs]
first injected pulse. This can 1 24 448 5376 163
also be observed for short cable 2 33 45 1350 162
lengths, where the losses are 3 33 486 4860 163
comparatively low. 4 33 704 9856 161
A typical example, which 5 138 175 2450 168
refers to a 33 kV XLPE power Table 4: Survey on the investigated power cable samples
cable having a length of 45 m,
is reported in Fig. 9. The
records reveal the higher magnitude of the second pulse in comparison to the first injected pulse.
Because the cable losses can not be neglected, the peak value of the second pulse is not twice the
magnitude of the first one, as to be expected from a theoretical point of view. Due to this occurrence
only the peaks of the reflected pulses were evaluated in order to avoid any misinterpretations of the
experimental results. Consequently, the calibrating pulses were injected at the remote end.
As already mentioned, for the detection and location of PD events in power cables only a frequency
spectrum below 10 MHz is of interest. Therefore, the results reported in the following refer to a
limitation of the upper cut-off frequency down to about 10 MHz, which is equivalent to a step pulse
rise time of the complete measuring system of about 30 ns.

3.2. Test results

The experimental investigations
were executed on power cables
listed in Tab. 4. Here besides
the geometrical length also the
virtual length is presented,
which results from the
evaluation of subsequent
appearing pulse sequences due
to multiple pulse reflections at
both open cable ends, as Fig. 10: Time domain reflectogram obtained by injecting of artificial PD
already reported previous. pulses in a 448 m long XLPE cable
A typical measuring example is
reported in Fig. 10, which

Page 12 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

refers to the cable No. 1. Here the geometrical length was 448 m. Because the characteristic time
interval between the subsequent appearing pulses is 5.5 µs, which results from a travelling distance of
twice the cable length, the wave velocity amounts 163 m/µs, which is also listed in Tab. 5. In total 7
individual pulses could be evaluated from the recorded reflectogram (see Fig. 10). Therefore the virtual
length can be estimated by (7-1) x 2 x 448 m = 5376 m. In accordance to this calculation the virtual
lengths of all other investigated cables are also listed in Tab. 4. In order to evaluate the pulse
attenuation depend on the travelling time the peaks of the subsequent appearing pulses (see Fig. 10)
were analyzed accordingly.
A survey on the experimental gained data is presented in Fig. 11. Here the peak value of the first
recorded pulse is used as a reference, i.e. 100 %. With respect to a better resolution this value is not
plotted in the graphs.

a) Cable 1: T1 = 5 µs, T2 = 18 µs b) Cable 2: T1 = 3 µs, T2 =15 c) Cable 3: T1 = 3 µs, T2 = 18 µs

d) Cable 4: T1 = 3 µs, T2 = 15 µs e) Cable 5: T1 = 3 µs, T2 = 15 µs f) Cables 1 to 5: T1 = 4 µs, T2 = 18µs

Fig. 11 reveals that the measured data can well be fitted by an empirical approach, which is given by
equation (21). For comparison purposes additionally to the measured data also the calculated curves
using equation (21) are plotted in the graphs, displayed in Fig. 11a to 11e. Additionally, in Fig. 11f all
measured data are plotted together with a mean curve calculated by (21) using the characteristic time
parameters T1 = 4 µs and T2 = 18 µs. This graph reveals that equation (21) works quite well. Thus
approach (21) is used in connection with the characteristic time parameters for further analysis, where
the travelling time t is substituted in accordance to formula (22) by the travelling distance x. Here the

Page 13 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

velocity vp is approximated by a mean value of 160 m/µs. Under this condition the time constants T1
and T2 can be expressed by
equivalent lengths, as evident Ut = U0 * exp [(- t / T2 ) - t / (t + T1 )] (21)
from the equations (23a) and
Ut - the pulse magnitude appearing after the travelling time t
(23b). Recalling equation (16)
U0 - the peak value of the first pulse, i.e. the 100 % reference value
the PD pulse attenuation
T1 - a first time constant, ranging between 3 µs and 5 µs.
dependent on the travelling
distance is also given by T2 - a second time constant, ranging between 15 and 20 µs.
equation (25). Because the t = x / vp ; x = t * vp (22)
attenuation in power cables is
T1 = X1 / vp ; X1 = T1 * vp = 4 µs * 160 m / µs = 0.64 km (23a)
generally expressed in dB/m or
in dB/km, the equation (24b) T2 = X2 / vp ; X2 = T2 * vp = 18 µs * 160 m / µs = 2.88 km (23b)
shall be expressed accordingly, Ux = U0 * exp [(-x /X2 ) - x / (x + X1)] (24a)
using equation (26).
In order to evaluate the Ux = U0 * exp [(-x /2.88 km) - x / (x +0.64 km)] (24b)
frequency dependent attenuation
Tab. 5: Equations for evaluation the PD pulse attention in extruded
equation (7) can be simplified, if power cables (I)
only frequencies below 10 MHz
are taken into consideration. Under this condition the equations (27a) and (27b) are generally fulfilled.
Thus the attenuation constant reflected by equation (7) can be approximated by (28a). After inserting
the equations (3) and (4) we get the equation (28b).
Here are:
d3 = d1 + 2 ds
αx = (1/2.88 km) + 1/(x + 0.64 km) (25)
d4 = d2 – 2 ds
αd = 20 log (Ux / U0) = 8.686 * αx
kcs - a correction factor, = 8.686 * [(1/2.88 km) + 1 / (x + 0.64 km)] (26)
which takes into account the
additional losses in both, the R´/ jωL´ << 1 G´/ jωC´ << 1 (27a)
outer cable shield as well as in {[1 + (R´/ωL´)2 ] * [1 + (G´/ωC´)2 ]}1/2 – 1 = 0 (27b)
the outer semiconducting layer.
Based on formula (28b) the αd = (R´* G´/ 2)-1/2 (28a)
attenuation αd in dB/km as αd = kcs * {(2ds /κs) * [µ0 / (2 * κc)]1/2*
function of the frequency f =
{ε0 * εr / [d1* ln (d4 /d3)]}2 }1/2 * ω5/4 (28b)
ω/2π has been calculated for a 24
kV XLPE cable, with the
following cable parameters:d1 = αd = (αd10 / f0 ) * fm = (2.125 dB/km) * fm (29a)
14 mm, d2 = 27 mm
ds = 0.7 mm, εr = 2.4 αd2 = (αd20 / f0 ) * fm = (2.375 dB/km) * fm (29b)

κc = 3.54 * 107 S/m Tab. 6: Equations for evaluation the PD pulse attention in extruded
power cables
κs = 2 * 10-2 S/m, kcs = 1.7

Page 14 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

The obtained graph is displayed

25
in Fig. 12 and indicated by
“C03”. It seems of interest, that
20
this calculated curve is quite well

Attenuation [dB/km]
C02
covered by the approach of Stone C01

and Boggs [12] derived from the 15

radial semiconductive loss model

(see Fig. 4b). For a 28 kV XLPE 10
C03
cable they calculated for a
frequency of 80 MHz an 5
attenuation of αd10 = 0.17 dB/m
and measured a value of αd20 = 0
0 2 4 6 8 10 12
0.19 dB/m. Furthermore they
Frequency [MHz]
found that the attenuation
dependent on the measuring Fig. 12: Calculated attenuation of travelling waves in XLPE cables
frequency can roughly be dependent on the frequency
C01- equation (29a); C02 - equation (29b); C03 - equation (28b)
approximated by a linear
function, given by the equations (29a) and (29b). The curves calculated from these relations are
indicated in Fig. 12 by “C01” and “C02”.
Based on the here obtained results it can be concluded, that the approach according to equation (28b) is
in principle applicable for a quantitative assessment of the attenuation in power cables. In this context it
has to be stressed, that this approach refers only to the frequency domain and not to the time domain,
but only the latter ensures an
estimation of the attenuation 1000
Detectable PD pulse charge [pC]

of the PD pulse magnitude as

function of the travelling
distance in order to evaluate 100
the PD detection sensitivity
for very long MV power
cables. Thus the better
approach is equation (26) as 10

will be reported now more in

detail.
Taking into account the
1
background noise level under 0 1 2 3 4 5 6 7 8 9 10 11
on-site condition, the worst Power cable length [km]
case arises, if the PD source is
located at the remote cable Fig. 13: PD detection sensitivity dependent on the cable length calculated
on the basis of a reference value of 10 pC achievable for a power
end, for which the travelling cable of 1 km length
distance x of PD pulses is

Page 15 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

equal to the whole cable length xc. PD events can only be recognized if the peak value of the PD
pulses, which arrive at the near cable end, exceed significantly the ambient noise level Un. Recalling
equation (24b) the condition (30) must be fulfilled. Because the cable acts as a low-pass filter, as
already mentioned previous, a quasi-integration of the propagating pulses takes place. Therefore, after
longer travelling distance, say xc > 1 km, the peak value appearing at the near cable end can be
considered as a measure for the U0 * exp [(-xc /2.88 km) - xc / (xc +0.64 km)] > Un (30)
apparent charge. Thus equation
(30) can also be expressed by qmin * exp [(-xc /2.88 km - xc / (xc + 0.64 km)] > qn (31)
equation (31).
For better understanding let us qn < (10 pC) * exp [(-1) * (1 /2.88 + 1 / 1.64)] = 3.8 pC (32)
first consider a power cable
having a length of xc = 1 km as qmin = (3.84 pC) * exp [(xc /2.88 km) + xc / (xc + 0.64 km)] (33)
a reference. Additionally, an Here are:
apparent charge of qmin = 10 qmin - the apparent charge created by a PD source located at the remote
pC, produced by a PD source cable end (x = xc)
at the remote end, should be qn - the ambient noise level expressed by an equivalent charge
detectable using a PD Tab. 7: Equations for evaluation the PD detection sensitivity
measuring circuit in
compliance to the relevant IEC standards [1; 2]. Reflecting equations (27 a) and (27 b) we get from
equation (32) that the origin PD magnitude of 10 pC is attenuated at the near end down to an equivalent
magnitude of qmeas = 3.8 pC. This value can only be detected, if the ambient noise level is
significantly lower. Assuming that the ambient noise level will not change, if power cables longer than
1 km are tested, the formula (33) applies. Based on these assumptions the minimum detectable apparent
charge qmin as function of the cable length xc has been calculated and plotted in Fig. 13. This graph
reveals that PD events below 100 pC could never be detected for cable lengths above 7 km. For cable
lengths above 10 km the critical PD threshold level exceeds 300 pC.
Finally it should be mentioned, that with respect to simplified calculations only a homogenous single
core cable was considered. An
installed power cable line,
however, is generally composed
of several cables. Due to
different wave impedances of
the cable sections and the joints
each connecting point causes
partial pulse reflections, as
exemplarily reported in Fig. 14.
This is equivalent to an energy
loss and causes thus an
enhancement of the PD pulse Fig.14: Time domain reflectogram of a power cable showing partial pulse
attenuation compared to a reflections caused by six joints distributed along a power cable
homogenous cable structure. section

Page 16 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

4. Practical aspects of PD fault location

4.1 Single PD source

If a PD source is active in a
power cable line, the occurring
pulses will travel away from
the PD source towards both
cable ends, as already reported
in the figures 2 and 3. For
location of PD defects
generally the time domain Fig. 15: Practical example for measuring the wave velocity of PD pulses in
reflectometry (TDR) is used. a XLPE cable
Whereas the direct PD pulse
appears at the time t1, the
reflected pulse appears at the
time t2. Therefore the
characteristic time difference
given by equation (34) can be
considered as a measure for the
distance of the PD source from
the far end, given by equation
(35 a).
Based on this relation the PD
fault distance from the near
Fig. 16: Location of the PD site in a power cable by evaluation the time
end, where the PD pulses are difference between the direct pulse (A) and the reflected pulse (B)
decoupled, can be expressed
by equation (35 b). ∆ t = t2 – t1 (34)
Here l represents the total power cable length. As evident from x21 = vp * ∆ t /2 (35 a)
this formula, the uncertainty in PD fault location is not only
governed by the true determination of the characteristic time x12 = l – (vp * ∆ t ) / 2 (35 b)
difference ∆t but also by the true knowledge of both, the Tab. 8: Equations for calculation
propagation velocity vp and the cable length l. the PD site
As already mentioned previous, the wave velocity of PD pulses in real power cables is substantially
lower as the theoretical one given by equation (11). Therefore, vp has to be determined experimentally.
This is done by the injection of calibrating pulses into one cable end and measuring the time difference
between the first (direct) and second (reflected) pulse. A typical time domain reflectogram obtained by
the calibration of a 480 m long XLPE cable is reported in Fig. 15. The time difference between the first
and second pulse amounts 5.9 µs, thus the propagation velocity can be predicted by the simple relation
vp = 2 * 480 m / 5.9 µs = 163 m/µs.

Page 17 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

Figure 16 refers to a practical

example of on-site PD location,
where the power cable was
subjected to a damped AC test
voltage [3]. Whereas the first
detected pulse (A) arrived direct
from the PD source at the near
cable end (direct pulse), the
second one (B) arrived delayed
due to the reflection at the
remote cable end (reflected
pulse). Because the input Fig. 17: Time resolved PD pulses obtained from the “near” (upper) and
the “remote” termination
impedance of the PD coupling
unit exceeds substantially the cable impedance, the direct pulse is also reflected at the detection side
and appears hence again (A’) after travelling twice through the cable length. From the time difference
between the direct (A) and the reflected pulse (B), together with the calibrated propagation velocity,
the PD site can simply be calculated by equation (35 b).
In principle it could happen that the time difference between the direct pulse (A) and the reflected
pulse (B) is equal to the time difference obtained by the calibration procedure. This special case
appears, if the PD´s are located in either the near or the remote cable termination. To determine which
of both terminations are affected by PD´s, the analysis of the pulse shape is a useful tool, because the
pulse width increases with increasing propagation time, due to the already discussed dispersion. As an
example, Fig. 17 reflects a typical record, where the upper trace refers to the termination, where the PD
defect is originated and where also the PD coupling unit is connected. The shorter duration of the first
(direct) pulse is well visible compared to the pulse from the lower trace, obtained after connecting the
PD coupling unit to the opposite cable termination.
Another typical situation might occur by the superposition of the direct and reflected PD pulses, as
exemplarily shown in Fig. 18.
This will happen if the time
interval between the both
characteristic pulses is
comparable to the duration of
the direct pulse. If, for instance,
a PD pulse width of 1 µs is
measured at the near end, such
superposition effects may
appear for a critical distance of
the PD site from the remote end
of less than 80 m due to the
characteristic value of Fig. 18: Superposition of the direct end reflected PD pulse
vp / 2 = 80 m / µs.

Page 18 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

4.2 Multiple PD sources

During HV energizing of a
cable section in the field,
several PD sources distributed
along the cable can also become
active simultaneously. As a
result, in just a small time
interval, PD pulses can be
detected from different
locations along the cable length.
Although the time differences
between the different PD Fig. 19: Multiple PD events occurring during a time span of only 45µs
activities may become less than 10µs they are generally still distinguishable because of the
comparatively high propagation velocity. A typical example for this is shown in Fig. 19, which refers
to three different PD activities originated from two defects, recognized in a 270 m long 3-phase paper
insulated cable (PILC). Analyzing the time differences between the direct pulses (Ai, Bi, Ci) and the
reflected pulses (Ar, Br, Cr) it can be stated, that the PD events A and C are originated from the same
source, whereas B is due to another source. Because the received direct pulses travel back and forward
again, also the complementary pulse sequences (Ai’, Bi’, Ci’) appear in Fig. 19. In this measuring
example the relevant time difference of about 2.4 µs is equal for the PD occurrences A and C. Using
equation (35) the distance of the PD source from the remote end can be calculated as 192 m if a mean
velocity of vp /2 = 80 m / µs is assumed.
For the characteristic PD event B the time difference is about 1µs. Thus the PD source is located about
80 m from the remote end. That means, in the time interval between the succeeding PD events A and
C, which is about 33 µs, the PD event B occurred in another location along the cable.
Partial discharges may also occur, which originate from different defects in the power cable within a
time interval less than the characteristic propagation time, as reported in Fig. 20a. The two significant
PD events are indicated by S and T. In the location analysis the direct and the reflected PD pulses are
running through each other, where the direct pulse of PD event T (Ti) is detected before the ref
lected pulse of PD event S (Sr).
Fig. 20b shows an example
obtained for a 610 m PILC
section. The indicated PD
events Y and Z are originated
from the same defect, because
the time differences ∆tY and ∆tZ
between the direct and the
reflected pulse are equal. In
general the overlapping pulse
diagrams for different PD events a) b)
Fig. 20: Examples for PD location diagrams with overlapping PD pulses
are dependent on the following from different PD sources (a) and from the same PD source (b)
three factors:

Page 19 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

1. Number of PD defects
2. Length of the power cable
3. Frequency of the test voltage

If the number of defects in a

cable sample increases, the
number of PD events will
naturally increase. The time
interval, where the matching PD
pulse may occur, is proportional
to the length of the cable sample. Fig. 21: Schematic view of the propagation of PD pulses in mixed power
cable sections
Thus the probability for
overlapping reflectograms will increase for longer power cables.

4.3 Mixed cable systems

vAB = 2 * (lA + lB) / ∆t (36)
Due to changes from
impregnated paper insulation to ∆t = 2 * lA / vA+ 2 * lB / vB (37)
plastic insulation for power
vAB = (lA + lB) / ( lA / vA+ lB / vB) (38a)
cables, more and more
situations will occur where vAB = (lA + lB) * vA* vB / ( lA / vB + lB / vA) (38b)
different insulation types are
mixed in cable networks. 1. ∆t < ∆t ,
xi xB
Because of the different i.e. the PD fault is originated in the part B:
propagation velocities for the xiB = (lA + lB) – vB * ∆txi / 2 (39)
different insulation types,
errors and misinterpretations 2. ∆txi > ∆txB ,
may occur in the PD fault i.e. the PD fault is originated in the part A:
location. Fig. 21 shows a xiA = lA – vA * (∆txi - ∆txB ) / 2 (40)
schematic view of a mixed
cable section, existing of two Tab. 8: Equations for evaluation the PD site in mixed cables
cable parts of different
insulation types A and B, each with a different propagation velocity vA and vB , respectively.
If a PD occurs at location xi, the direct and the reflected PD pulse will travel different lengths trough
both cable parts. Therefore, another approach for the location of the PD source is necessary than for the
single cable type location, reflected by equation (35).
Based on the classical calibration procedure, the wave propagation velocity resulting from the time
difference between the direct and reflected pulse can be considered as an “average” velocity of the
mixed cable section given in Tab. 8 by the equation (36). The measured time difference ∆t depends on
the ratio between the lengths of different cable types and their characteristic propagation velocities, as
expressed by equation (37). By combination of (36) and (37) we get finally the equations (38 a) and
(38b). For analysis of the PD origin locations in mixed cable sections, again the time difference

Page 20 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

between the direct and the reflected PD pulse is measured. The next step is, however, to find out from
which cable part the PD pulse is coming from. Therefore, the measured time difference ∆txi is
compared with the travelling time difference ∆txB to be expected for the cable part B, which is the
remote part from the detection side.
For ∆txi < ∆txB the PD
source must be located in
the remote cable part B,
whereas for ∆txi > ∆txB
the PD source must be
originated in the near cable
part A. For ∆txi = ∆txB the
PD occurs exactly in the
cable joint connecting both
cable parts A and B.
As a result, after the
distinction from which
cable part the PD pulse is
originated, the location of
the discharge source can be
calculated according to the
conditions (39) and (40):
Figure 22a reflects a
practical measuring
example for a possible
location error, due to a
mixed insulated power
cable section. The
investigated cable sample
was a combination of a
XLPE cable having a length
of 465 m with a PILC cable
of 561 m length. Both PD
Fig. 22: Practical example for errors in PD mapping of mixed power cables
maps show the individual (XLPE + PILC)
PD locations along the a) Using conventional single cable location technique based on
power cable length for all equation (35 b)
three phases, where the b) Using modified location technique based on equations (38) to (40)
black dots reflect the
accessories.
The upper PD map (a) is created by the conventional location technique valid for non-mixed cables,
which reveals four PD concentrations at the locations:

A = 0 m, B = 500 m, C = 645 m, D = 955m

Page 21 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

If the measured data of the same cable sample are analyzed by means of the equations (38) to (40),
where the characteristic velocities vA = 140 m/µs and vB = 160 m/µs are used, another PD map is
obtained, as reflected in Fig. 22b. The measured PD concentrations are qualitatively well comparable
to the plots of Fig. 22a, but the determined PD sites along the cable differ significantly.
As expected, the PD concentration for A = 0 m did not shift, but the other three concentrations shifted
as follows:
o B shifted from 500m to 465m, which is 35 m lower
o C shifted from 645m to 620m, which is 25 m lower
o D shifted from 955m to 949m, which is 6 m lower

The here presented practical measuring examples underline the very complex matter of the PD location
technique, which is currently increasingly applied for on-site PD testing of MV power cables [4]. In
principle also other aspects have to be taken into consideration, such as the already mentioned
enhanced PD pulse attenuation due to numerous joints in a cable section, which causes partial pulse
reflections. Also a bad condition of the cable shield due to corrosion may contribute to an increased
attenuation of the propagating PD pulses, which may essentially reduction the PD detection sensitivity.
Such effects have also to be taken into consideration if installed MV power cables are PD tested, but it
is not the aim of this paper to present all PD occurrences which may happen under practical conditions.

5. Summary
Based on theoretical analysis and experimental studies of the PD occurrence in long power cables the
following conclusions can be drawn:

1. For assessment the attenuation of PD pulses in principle the classical travelling wave theory based
on the electric line model is applicable.

2. Because for the measurement and location of partial discharges in MV power cables only a
frequency spectrum below 10 MHz is of interest, the radial semiconducting loss model proposed by
Stone and Boggs [12] can essentially be simplified.

3. The theoretical analysis can further be simplified if the calculation refers only to the inner
semiconducting layer, whereas the impact of the outer semiconducting layer is taking into account by a
correction factor, which ensures a comparatively simple estimation of the frequency dependent
attenuation and dispersion.

4. It is supposed, that the excessive reduction of the travelling wave velocity below the theoretically
expected value is caused by a fictive enhancement of the capacitance per unit length due to a fictive
reduction of the insulation thickness by the inner and outer semiconducting layer.

Page 22 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

5. The simplified approach has successfully been proven for quantitative assessment of the wave
attenuation in the frequency domain. For evaluation of the PD detection sensitivity, however, the
analysis must be done in the time domain. In order to get the necessary parameters comprehensive
practical studies on the PD wave propagation in long power cables have been performed.

6. Using an instrumentation of comparatively high input impedance the injected pulses were full
reflected both, at the near and the far end. Thus results could be obtained for virtual cable lengths much
longer than the investigated geometrical cable lengths.

7. Based on experimental data an empirical approach was derived to ensure a simple estimation of the
attenuation PD pulses dependent on the travelling distance. This allows an assessment of the PD
detection sensitivity, which can be achieved for very long power cables.

8. If on-site PD tests of power cables are executed, besides the attenuation also superposition
phenomena of the direct and reflected pulses have to be taken into consideration. This may happen if
the PD site is close to the far end as well as if multiple PD sources, distributed along the cable length,
ignite simultaneously.

9. The uncertainty in PD location may be enhanced for mixed cable sections because of the different
wave propagation velocities in different cable types, such as XLPE and PILC.

6. References
[1] High-voltage test techniques – Partial discharge measurement. IEC 60270 (2000)

[2] Testing of cables, wires and flexible cords – Partial discharge measurement. IEC 60885-3 (1988)

[3] E. Gulski, E. Lemke, M. Gamlin, E. Gockenbach, W. Hauschild, E. Pultrum: Experiences in

partial discharge detection of distribution power cable systems. ELECTRA 35 (2003) No. 208

[4] E. Lemke, T. Strehl: Möglichkeiten und Grenzen der TE-Fehlstellenortung in kunststoffisolierten

Energiekabeln unter Vor-Ort-Bedingungen. ETG-Fachtagung Köln (2004) pp. 209-213

[5] F.J. Wester, Condition Assessment of Distribution Power Cables using PD Diagnosis at Damped
AC Voltages, PhD Thesis TU Delft (2004), ISBN 90-8559-019-1

[6] A. Gemant, W. Philippoff: Die Funkenstrecke mit Vorkondensator. Z. f. Physik 13 (1932) 9,

pp. 425-430

[7] F. H. Kreuger: Discharge detection in high-voltage equipment. Temple Press, London (1964)

Page 23 /23
CIGRE WG D1.33 HV Measurements
TF 05 Partial Discharge Measurements

[8] E. Lemke: A new method for PD measurement of polyethylene insulated power cables. 3rd. ISH
Milan (1979) paper 43.13

[9] M. Beyer, W. Kamm, H. Borsi, K. Feser: A new method for detection and location of distributed
partial discharges in high voltage cables under external interference.
IEEE Trans. PAS 101 (1982) 9, pp. 3431-3437

[10] L. Binder: Über Einschaltvorgänge und elektrische Wanderwellen. ETZ 35 (1914)

pp. 177 and 203

[11] R. Rüdenberg: Elektrische Schaltvorgänge. Springer-Verlag Berlin (1933)

[12] G.C. Stone, S.A. Boggs: Propagation of partial discharge pulses in shielded power cables.
Proceedings of Conference on Electrical Insulation and Dielectric Phenomena. IEEE Publication
82CH1773-1(1982) paper V-6, pp. 275-280

[13] P. Osvath, G. Biasutti, et. al. Bulletin ASE/UCS 78/9

[14] W.L. Weeks, Yi Min Diao: Wave propagation characteristics in underground power cable. IEEE
Trans. PAS-103 (1984) 10, pp. 2816-2826

[15] J.P. Steiner, P.H. Reynolds, W.L. Weeks: Estimation the location of partial discharges in cables.
IEEE Trans. On El. Insulation 27 (1992) 1, pp. 44-59

[16] Partial discharge detection in installed extruded cable systems. Report by CIGRE WG-21-16
(July 2000)

[17] G. Mugala, R. Papazyan, P. Nakov: High frequency characteristics of medium voltage cables
using time domain reflectometry techniques. 17th NORD-IS (2001) pp. 211-218

[18] R. Papazyan, R. Eriksson: Calibration for time domain propagation constant measurements on
power cables. IEEE Trans. on Instrumentation and Measurement 52 (2003) 2, pp. 415-418

[19] S. Boggs: The case of frequency domain PD testing in the context of distribution cable. IEEE
Electrical Insulation Magazine 19 (2003)4, pp. 13-18

Page 24 /23