OF5-01

TEMPERATURE DEPENDENCY OF WAVE PROPAGATION VELOCITY

IN MV POWER CABLE
1* 1 2 2
Yan LI , Peter A. A. F. Wouters , Paul Wagenaars , Peter C. J. M. van der Wielen and
1,2
E. Fred Steennis
1
Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, Netherlands
2
DNV KEMA Energy & Sustainability, P.O. Box 9035, Arnhem 6800 ET, Netherlands
*Email: <y.li.4@tue.nl>

Abstract: Propagation velocity of high frequency signals, e.g. from partial discharge, is a
vital parameter for time domain power cable diagnostic techniques. The propagation
velocity is mainly dependent on the permittivity of the insulation material, which can be
affected by external parameters like temperature or water ingress. This paper focuses on
the influence of temperature on the propagation velocity in medium voltage (MV) cables.
Laboratory scale tests are performed for both PILC and XLPE cable. Test results show
that the high frequency signal propagation velocity for XLPE will increase with the
temperature rise while PILC has opposite behaviour. The variation of propagation velocity
of XLPE is confirmed by data of a power cable subjected to strong load cycling monitored
over eight months.

1 INTRODUCTION power cable via 50 coaxial cable. The injected

In time domain reflectometry measurements, e.g.
in partial discharge diagnostic techniques for
power cable [1-3], propagation velocity should be
accurately known for precise defect location. On
one hand, the propagation velocity can be affected
by temperature. On the other hand, information on
variability in propagation velocity informs on
changing conditions as temperature variation by
e.g. cable loading or insulation ageing.
pulse and its reflections are recorded by an
oscilloscope. Heating is accomplished by a 10 kW
DC current source, capable of generating 600 A at
a maximum voltage of 16 V. It is connected to the
conductor from the second segment to the last
cable segment.

Two sets of laboratory scale test are presented to

evaluate the temperature effect on wave
propagation velocity in both XLPE (Section 2) and
PILC cables (Section 3). The effect is compared
after translating measured temperature to
temperature inside the insulation (Section 4). In
addition, propagation time data are extracted from
Smart Cable Guard systems [4] installed on live Figure 1: Test circuit for TDR on heated XLPE
cable connections with XLPE insulation (Section cable; the connector numbers correspond to the
5). The field data, with variation recorded over a connector types illustrated in Figure 2.
year is then compared with the load profile. The
observed temperature effect on propagation
velocity will be briefly further discussed in Section
6.

2 LABORATORY TEST ON XLPE CABLE

2.1 Test circuit

Six segments of 12/20kV XLPE single core
(aluminium) cable are connected to form the test
circuit. Each piece is about 12 m resulting in a
cable circuit of about 72 m. The test circuit is open
at both sides. Pulse signals are injected into the
cable via one open end and reflections are
recorded at the same end (time domain (1) (2) (3)
reflectometry, TDR). Due to space limitation, the
test circuit is half inside the test room and half Figure 2: Connector types applied for combining
outside in open air. Figure 1 illustrates the test XLPE cable segments; numbers correspond with
circuit. An 8 ns wide pulse is injected into the the connectors indicated in Figure 1

1861
 OF5-01

Temperature sensors are attached on the surface

of the cable outer jacket to measure the outer 0.1 14.4C,12.0C
sheath temperatures of indoor and outdoor cables.
20.4C,17.0C
These sensors are shown in Figure 1 as dots and 0.05
27.6C,21.9C

Amplitude [V]
labeled as T1 and T2. Letters A-G indicate the 0
reflection points of TDR measurements shown 37.3C,27.0C
later in Figure 3. Figure 2 shows the connector -0.05 48.7C,31.8C
types. Connector type 3 was improvised for the -0.1
present measurements. The cable earth screen
connections were realized with short wires parallel -0.15
to the connectors for all types. -0.2
0.95 1 1.05 1.1
2.2 Test results Time [s]

After heating for about 3 hours, the cable outer

(b)
jacket temperature has increased from 9 C to
48 C (indoor) and from 8 C to 30 C (outdoor).
TDR signals are recorded continuously. Results at 0.6
different temperatures are shown in Figure 3. All
reflected signals are labelled with symbols to 0.4
identify the reflections coming from the transition

Amplitude [V]
between coaxial cable and power cable (A), the 5
cable connectors (B-F) and the open far end (G). 0.2 14.4C,12.0C
Reflection C (from connector type 1), E (from 20.4C,17.0C
connector type 2) and G (from open end) are 0 27.6C,21.9C
zoomed in and shown in Figure 4. It is qualitatively 37.3C,27.0C
observed that with the increase of temperature, the 48.7C,31.8C
propagation time decreases, indicating that the -0.2
propagation velocity has increased. 1.3 1.35 1.4 1.45 1.5
Time [s]

(c)
10 10.8C,9.0C
8 20.8C,17.4C Figure 4: (a) depicts reflection C from connector 1;

30.9 C,23.4 C (b) shows reflection E from connector 2 and (c)
Amplitude [V]

6 illustrates reflection G from open end

40.8C,28.8C
4 48.8C,32.1C
C 2.3 Analysis
2 B F
DE
In order to analyse the measured result, two issues
0
A
have to be addressed. The first issue is the
G
-2 temperature correction, since the temperature
0 0.5 1 1.5 2 sensors are attached on the outer jacket of the
Time [s]
XLPE cable. The propagation velocity is mainly
determined by permittivity of the insulation.
Figure 3: Reflection patterns at different Therefore, the measured jacket temperature will be
temperatures; the legend indicates first the indoor converted to the temperature at the midpoint of the
temperature, next the outdoor temperature, letters insulation. The second issue concerns the
A-G correspond to the ones in Figure 1 quantitative calculation of the pattern shift at
different temperatures.

0.8 14.4C,12.0C Insulation temperature conversion: A thermal

20.4C,17.0C ladder network can be used to model the
0.6
27.6C,21.9C temperature distribution in radial direction of the
Amplitude [V]

0.4 37.3C,27.0C cable [6]. Each layer of cable can be modelled by a

thermal resistor and a thermal capacitor. The heat
48.7C,31.8C
0.2 source can be represented as a current source.
0 According to the thermal ladder network,
logarithmic temperature distribution across the
-0.2 dielectric material is assumed. Based on this
model, the time response of the temperature in the
0.75 0.76 0.77 0.78 0.79 0.8
Time [s] middle of dielectric material is shown in Figure 5.
The measured temperature of the outdoor cable
(a) jacket is then converted to outdoor temperature at

1862
 OF5-01

midpoint of the insulation, Table 1. The outdoor reflection from the far end. This might be caused
temperature is used since the propagation time by the indoor and outdoor temperature difference
between the connector 1 and open far end will be and lower signal to noise ratio. Directly past this
used for analysis in order to exclude the indoor peak, structures arise which travelled additional
and outdoor temperature difference. Tjx is the length along a not heated part. Later also the
jacket temperature of the XLPE cable, and Tdx is heated part is included resulting in a steeper time
the temperature at the middle of insulation of the shift again.
XLPE cable; t is the heating time according to
Figure 5.
0 12
1
-2 2 10
15
3
-4 8
4

time delay [ns]

amplitude [V]
-6 6
10 -8 4
T [C]

-10 2

5 -12 0

-14 -2

-16 -4
0 0.5 1 1.5 2 2.5 3 3.5
0 time [s]
0 0.5 1 1.5 2 2.5 3
t [h]
Figure 6: Time shift of reflection patterns of XLPE
Figure 5: Derived temperature difference between cable at different temperatures
the midpoint of XLPE insulation and the outer
jacket surface for single core XLPE cable 3 LABORATORY TEST ON PILC CABLE

Table 1: Insulation temperature for XLPE; the 3.1 Test circuit

outer jacket temperature Tjx is converted to the
The laboratory scale cable setup at DNV KEMA is
midpoint value Tdx of the insulation reached at time
utilized for PILC cable test. The setup consists of
t (see Figure 5)
50 m three core XLPE cable and 70 m three core
PILC cable (outdoor). 100 ns wide pulse is injected
Ref. 1 2 3 4 into the cable. Due to the setup limitation, the
Tjx (C) 12.0 17.0 21.9 27.0 31.8
current source for heating (same as for XLPE test)
is connected to the earth screen of cables, which is
t (h) 0.27 0.60 1.15 2.24 5.47 lead for PILC and copper for XLPE. Figure 7
shows the test setup. Since the resistivity of lead is
Tdx (C) 23.6 30.3 35.3 40.4 45.2
much higher than copper, heating power
Quantitative time shift: The minimum square distributes mainly along the PILC cable. Due to
error (MSE) [5] detection with time window is power source limitation, about 130 A current could
applied to get the time shift between two patterns. be loaded to the cable. Two temperature sensors
When time shift between pattern P (reference) and are attached on the surface of the PILC and the
Q (at elevated temperature) is calculated, a time XLPE cable separately.
window (length t) taken from the complete record
will be shifted along pattern Q back and forth to
find the best match in pattern P, i.e. where the
minimum error occurs. The time window length is
chosen to be 1500 ns to get a reasonable flat
result excluding error from noise. Figure 6 shows
the calculated time shift along the pattern between
heated cable and reference cable. The numbers in
the legend correspond to the patterns in Table 1.
The pattern shifts more as the temperature
difference increases. The shift becomes most
nd rd
apparent as from the 2 3 reflection (at 700 ns,
B-C in Figure 3). This is where the heated part of
the cable begins. The time delay increases after
each clear reflection from joint due to longer
propagation distance. After the steep part the time
delay slowly increases around and just after the Figure 7: Test setup for TDR on heated PILC
cable

1863
 OF5-01

3.2 Test results

1
Approximately 3 hours heating increased the
temperature of the PILC cable with about 25 C 0.8
and the XLPE with about 5 C. The measured

Amplitude [V]
0.6 PILC:23.5C XLPE:24.0C
result is shown in Figure 8. In the pattern, the
injected pulse is located around 2.1 s. The first 0.4
PILC:28.1C XLPE:24.9C
reflection near 2.6 s is from the coaxial cable PILC:33.1C XLPE:25.8C
connection to the power cable; the reflection 0.2 PILC:38.2C XLPE:26.1C
around 4.1 s arises from the transition joint, and PILC:43.0C XLPE:27.3C
the reflection around 5 s from the open end. The 0
PILC:48.1C XLPE:28.8C
later reflections are secondary or higher order 4.85 4.9 4.95 5 5.05 5.1 5.15
reflections. The transition joint reflection and the far Time [s]
end reflection are depicted enlarged in Figure 9. It
is observed that with the increase of temperature, (b)
the propagation time increases, indicating a lower
propagation velocity, which is opposite to the Figure 9: (a) enlargement of reflection from PILC
observations for XLPE insulation. to XLPE transition joint; (b) enlargement of
reflection from open end
6 PILC:23.5C XLPE:24.0C
3.3 Analysis
PILC:28.1C XLPE:24.9C

4

PILC:33.1 C XLPE:25.8 C Insulation temperature conversion: Applying a
PILC:38.2C XLPE:26.1C similar approach as for the XLPE cable heating
PILC:43.0C XLPE:27.3C test, the temperature difference between the
Amplitude [V]

2
PILC:48.1C XLPE:28.8C insulation midpoint and jacket of PILC cable is
shown in Figure 10. The measured outer jacket
0 temperature for PILC cable is converted to
temperature at the midpoint of the paper insulation
-2 as shown in Table 2.

-4 2.5
0 2 4 6 8 10
Time [s]

2
Figure 8: Reflection patterns at different
temperatures; the legend indicates first the indoor
1.5
temperature, next the outdoor temperature
T [C]

1
PILC:23.5C XLPE:24.0C

PILC:28.1 C XLPE:24.9 C
0 0.5
PILC:33.1C XLPE:25.8C
-0.1 PILC:38.2C XLPE:26.1C
Amplitude [V]

PILC:43.0C XLPE:27.3C
0
-0.2
0 0.5 1 1.5 2 2.5 3
PILC:48.1C XLPE:28.8C t [h]
-0.3
Figure 10: Derived temperature difference
-0.4 between the midpoint of paper insulation and the
-0.5 outer jacket surface for the three-core PILC cable
4.05 4.1 4.15 4.2 4.25
Time [s] Table 2: Middle of insulation temperature for PILC

(a)
Ref. 1 2 3 4 5

Tjp (C) 23.5 28.1 33.1 38.2 43.0 48.1

t(h) 0.08 0.35 0.60 1.05 1.83 3.13

Tdp (C) 25.0 30.1 35.3 40.4 45.2 50.3

Quantitative time shift: The previously described

MSE method is applied to the PILC cable patterns
to find the time shift. The results are shown in

1864
 OF5-01

Figure 11 for five measurements. The time shift of the PILC cable is affected more by temperature.
increases with propagation time and temperature Also this figure suggests that the variation is not
difference. The steep rise observed directly after linear and becomes more pronounced at higher
the reflection from the injection cable is caused by temperatures.
the time window beginning to cover the reflection
from the PILC to XLPE transition point, which gives
2.5
a clear shift. The increase in time delay directly -XLPE
after the first reflection is where the heated cable PILC
starts. 2

1.5

time shift [%]

100 10
1
2 1
3
4 0.5
time delay [ns]

amplitude [V]
50 0 0
0 5 10 15 20 25

temperature difference [ C]

Figure 12: Propagation time variation comparison;

0 -10
0 2
Figure 11: Time shift of reflection patterns of PILC
cable at different temperatures
4 COMPARISON OF THE TEMPERATURE
EFFECT OF PILC AND XLPE CABLE
Test results show that the XLPE and PILC cable
have an opposite temperature effect on
propagation velocity. With the conversion from
measured jacket temperature to conductor
temperature, it is possible to compare both cables.
For XLPE, the time shift is calculated from
reflection C (from connector 1 in Figure 1) to G
(from open end) utilizing the outdoor part
(consistent with the temperature data in Table 1).
For PILC, the time shift is derived from the
reflections at both ends of the PILC cable. The
comparison between the XLPE and PILC cables
propagation time variation is shown in Figure 12.
The deviations caused by different time window
lengths are indicated in the error bar. The velocity
500
Load [A]
the absolute values are indicated
5 FIELD DATA
4 6 8 10
time [s]
Propagation time data from installed Smart Cable
Guard system recorded over a year is compared
with load profile data over the same time period.
The propagation time and load of a XLPE cable
st
section (4.8 km long with 6 RMUs) from May 1 ,
th
2010 to January 14 2011 is shown in Figure 13. It
shows that though the load does not have a clear
summer winter cycle, the propagation time is lower
at summer time and higher at winter time. This
indicates a clear ambient temperature variation. On
top of the ambient temperature effect, the load also
influences the propagation time. The peak load
th
around July 9 , 2010 corresponds to a lower
propagation time at that period. A weekly trend is
observed. The propagation time variation is
opposite compared to the load current variation.
Besides, the variation within each day is observed.
The relatively high load on Thursday and Monday
correlates with the relatively large drop in
propagation time. Further, there is a delay between
the increase in current and the thermal response.
500 12:40
Saturday
Propagation time [s] Load [A]

00:40
0 0 0 0 0
0
1 8: 25:0 0 5: 05:0 1 6: 10:0 0 5: 35:0
28

29

01

03

05

2010 2010 2010 2010

04-06- 13-08- 21-10- 30-12-
6-

6-

7-

7-

7-
-0

-0

-0

-0

-0
s]

10

10

10

10

10

31.7 31.8
Propagation time [

31.6
31.6
31.5

31.4 31.4
28

29

01

03

05

02: 43 05:44 16: 35 10: 40

6-

6-

7-

7-

7-

2010 2010 2010

30-05- -2010 01-10- 28-11-
-0

-0

-0

-0

-0

01-08
10

10

10

10

10

Figure 13: Propagation time and load of XLPE cable for year and week cycle

1865

This thermal response time of typically 3 hours is
more apparent when the cable load starts
compared to when the load is switched off
probably because the temperature rise due to
higher load is faster than the temperature
decrease caused by lighter load. It should be noted
that the load profile is not the same over the
complete cable section. The measured current
represents the highest loaded part of the section.
Unfortunately, at present there is no correlated
PILC data available for propagation time and load
current. From the laboratory tests a clearer effect is
expected here.
6 DISCUSSION
The velocity change of the XLPE cable with
temperature can be attributed to the real part of the
permittivity of the XLPE. The decrease of velocity
in percentage is in agreement with the decrease
in the measured temperature range [7,8], which is
about 0.7-3% in the temperature range of 20 C to
60 C. For PILC, increases with temperature [9];
however quantitative analyses on the temperature
dependence of in high frequency range is scarce
in literature. Furthermore, it is acknowledged that
the insulation parameters of paper insulation can
vary a lot depending on manufacturer, production
year, country and even from cable to cable. Further
study is needed to demonstrate the velocity
dependence of for PILC cable in live circuits.
Another factor that may affect the velocity is the
change in dimension of the cable. This may have
an effect on the pressure on the dielectrics.
7 CONCLUSION
Experiments show that high frequency signals
propagates faster with higher temperatures for
XLPE cable and slower for PILC.
It is observed that the temperature effect on
propagation velocity is more significant for PILC
than for XLPE.
The temperature dependence of the real part of
permittivity () dominates the velocity change for
XLPE cables. For PILC, further work is needed to
explain the origin of the velocity dependency with
temperature.
8 ACKNOWLEDGMENTS
The authors would like to thank DNV KEMA
Energy and Sustainability, Enexis, Alliander and
Locamation for their financial support.
1866
OF5-01
9 REFERENCES
[1] G. M. Hashmi, R. Papazyan, M. Lehtonen:
Comparing Wave Propagation Characteristics
of MV XLPE Cable and Covered-Conductor
Overhead Line using Time Domain
Reflectometry Technique, 2007 International
Conference on Electrical Engineering (ICEE
'07), pp.1-6, 11-12 April 2007
[2] G. M. Hashmi, R. Papazyan, M. Lehtonen:
Determining wave propagation characteristics
of MV XLPE power cable using time domain
reflectometry technique, 2009 International
Conference on Electrical and Electronics
Engineering (ELECO 2009), pp.I-159-I-163, 5-
8 Nov. 2009
[3] S. Markalous, T. Strehl, C. Herold, T. Leibfried:
Enhanced signal processing for conventional
and unconventional PD measuring methods:
Wavelet de-noising, automatic detection
algorithms and averaging for arrival time-based
PD location in transformers and power cables,
2008 International Conference on Condition
Monitoring and Diagnosis (CMD 2008),
pp.1115-1118, 21-24 April 2008
[4] Peter C.J.M. van der Wielen and E. Fred
Steennis: Risk-controlled application of current
MV cable feeders in the future by intelligent
continuous diagnostics, in 2011 IEEE Power
and Energy Society General Meeting, pp.1-6,
24-29 July 2011
[5] F. Censi, G.Calcagnini, M. D Alessandro, M.
Triventi, P. Bartolini: Comparison of alignment
algorithms for P-Wave coherent averaging,
Computers in Cardiology, 2006 , pp.921-924,
17-20 Sept. 2006
[6] George J. Anders: Rating of Electric Power
Cables in Unfavorable Thermal Environment,
pp.1-75, May 2005, Wiley-IEEE Press
[7] V. Dubickas, H. Edin: On-line time domain
reflectometry measurements of temperature
variations of an XLPE power cable, 2006 IEEE
Conference on Electrical Insulation and
Dielectric Phenomena, pp.47-50, 15-18 Oct.
2006
[8] C. Fanggao, G. A. Saunders, R. N. Hampton,
S. M. Moody and A. M. Clark: The effect of
hydrostatic pressure and temperature on the
permittivity of crosslinked polyethylene,
Seventh International Conference on Dielectric
Materials, Measurements and Applications,
pp.267-270, 1996
[9] I. Mladenovic, C. Weindl, C. Freitag:
Comparison of parametric partial discharge
and dissipation factor characteristics of MV
PILC cables, Conference Record of the 2012
IEEE International Symposium on Electrical
Insulation (ISEI), pp.319-322, 10-13 June 2012