21, rue d’Artois, F-75008 PARIS B1-209 CIGRE 2006

http : //www.cigre.org

Diagnostic Testing of High Voltage Water Treed XLPE Cables

Sverre Hvidsten*, Hallvard Faremo and Jan Tore Benjaminsen

SINTEF Energy Research
Norway

SUMMARY
In Norway it has been introduced an interruption cost based on estimates of energy not delivered.
These costs can become very high especially in case of electrical breakdown of high voltage
transmission grid components. Diagnostic testing of high voltage power apparatus is therefore
becoming increasingly important.

Water treeing has caused many premature failures in medium voltage (MV) cables. The commercially
available techniques for detection of water tree degradation in medium voltage cables are generally
based upon measurement of the dielectric response either in time or frequency domain. In case of the
old cables produced in the seventies, extensive examinations have been performed in order to establish
diagnostic criteria for predicting the ageing status of such old cables during on-site measurements.
From these examinations it has been shown that good assessments can be provided for this cable
group. In case of HV cables (≥ 36 kV), examinations of cables suffered from service failures have
revealed that water treeing can also occur for these cables. Generally, knowledge about water treeing
and condition assessment of high voltage cables (HV) is up to now insufficient or lacking. The
objectives of this study have therefore been to report results from on-site measurements followed by
laboratory investigations of HV XLPE cables.

The results show that the relative lengths of the water trees causing service failures in HV cables
decreases strongly with system voltage. For the old cables produced in the seventies, high densities of
vented water trees were observed similar to that observed for the MV cables. The results from the
diagnostic measurements in the laboratory as well as on-site, show that the diagnostic test equipment
initially designed for MV cables can be used to detect water tree degradation in such HV XLPE
cables. However, condition assessment of water treed HV cables with extruded insulation screens with
low densities of vented water trees remains to be verified.

The results also indicate that the different types of cable accessories do not have any impact on the
condition assessment of water treed XLPE high voltage cables.

KEYWORDS
High Voltage Cables - Diagnostic Testing - Condition Assessment - Dielectric Response - Water
Treeing.

[email protected]
1 INTRODUCTION

Interruption costs based on estimates of energy not supplied is introduced in Norway. These costs can
become very high, especially in case of breakdown of high voltage transmission grid components.
Condition assessment of high voltage power apparatus is therefore becoming increasingly important.

Water treeing in XLPE cables is a well-known phenomenon, and has caused serious problems for
utilities as a precursor for breakdown of medium voltage cables. Commercial available diagnostic
techniques for detection of water tree degradation in medium voltage cables are generally based upon
measurements of the dielectric response, either by measurements in the time - or frequency domain
[1]. Water treeing can also occur in high voltage cables. The objectives of this study have been to
report results from on-site measurements of dielectric loss factor and permittivity followed by
laboratory investigations of HV XLPE cables. Measurements of AC breakdown strength and degree of
water treeing were performed in the laboratory on the same cables.

High Voltage Cable Designs and Service Experience

In Scandinavia MV XLPE cables were introduced from about 1965 [2]. The first designs were steam
cured XLPE insulation with graphite painted insulation screen (see Figure 1a)). This was also the case
for the first high voltage XLPE cables. 36 kV cables were introduced around 1970 and the next
voltage classes up to 300 kV followed before 1980. However, in the second half of the seventies dry
curing were introduced. The other moisture prevention methods (PE sheath, axially and radial water
tightness, etc.) were first taken into use at the higher voltage levels. At the same time also the graphite
painted insulation screens were replaced with a fully bonded type.
6
100
Stress at the
a) b) conductor screen
Longest observed water trees [%]

80

Service stress [kV/mm]

4
60

40

Stress at the
insulation screen
20 2

0
12
24
36
52

72

145

Service voltage [kV]

Figure 1 a) Typical HV cable design of the seventies. b) Relation between the longest observed
water trees causing service failure and service voltage. The grey shaded area is the
corresponding service stress. The large range of service stresses is due to the large
differences in insulation thickness as well as conductor cross sections.

Figure 2 shows photographs of typical water treeing in MV and HV XLPE cables that have suffered
from service failure due to water treeing. As can be seen in Figure 2a), in case of 12 kV XLPE cables
(3.4 mm insulation thickness), it is typically found that long vented water trees can grow through the
insulation without causing an immediate breakdown. The relative length of the water trees causing
breakdown in HV XLPE cables is shorter than that found for MV cables. Figure 2d) shows an example
of water treeing in a 52 kV XLPE cable (10 mm insulation thickness) with insulation screens
consisting of graphite painting and semiconductive tapes. The lengths of the trees are typically 20% of
the insulation thickness.

2
Figure 1b) shows the relation between typical lengths of the longest observed water trees in XLPE
cables that have suffered from service failure due to water treeing and service voltage. The critical
length of the water trees is strongly decreasing with increasing service voltage. This can be explained
by the significant higher service stress in HV cables as indicated in the same figure.

a) 12 kV b) 24 kV c) 36 kV d) 52 kV e) 145 kV

Figure 2 Photographs of typical water treeing in a) Um = 12 kV b) Um = 24 c) Um = 36 kV d) Um =

52 kV and e) Um = 145 kV (this cable only was equipped with the first type of extruded
insulation screen) XLPE cables that have suffered from service failure due to water
treeing. Um is defined according to IEC 60 183.

2 DIELECTRIC SPECTROSCOPY

Non-Destructive Diagnostic Test Method and Procedures

The non-destructive diagnostic test method used is during laboratory and on-site measurements is
based on measurements of dielectric response in frequency domain, and was initially developed for
condition assessment of medium voltage cables (12 and 24 kV) [1]. The system used during this work
has a maximum output voltage of 21 kV. The maximum frequency range is 100 Hz to 0.1 mHz.
However, during condition assessment on-site, the frequency range is most often limited to 10 to 0.1
Hz due to the capacitive loading of the measured cable (limiting the highest frequency) and available
time for doing the assessment (limiting the lowest frequency).

In order to examine the voltage dependence of the response, several measurements are performed at
different voltage levels up to service stress. For high voltage cables, measurements up to service stress
can only be performed on cables with a phase-to-ground voltage of 18 kV (Um = 36 kV) using this
method. For medium voltage cables, measurements above the service stress have also been performed
in order to increase the sensitivity of the methods [1]. During on-site testing, measurements at lower
frequencies were also performed, typically to 10 mHz. After the measurement at service stress, an
additional measurement at a lower voltage (in most cases 0.5Uo) was performed in order to examine
any hysteresis effect, which is an important criterion for condition assessment of medium voltage
cables with water trees.

Diagnostic Parameters Used During Condition Assessment

When applying a sinusoidal voltage to a dielectric material, the complex permittivity can be
determined from the resulting current [3];
σ
ε (ω ) = ε ′(ω ) − iε ′′(ω ) = ε r + χ ′(ω ) − i ( χ ′′(ω ) + ) (1)
ωε o
where χ’(ω) and χ’’(ω) is the real and imaginary parts of the dielectric susceptibility, εr is the relative
permittivity, ω is the angular frequency, σ is the conductivity and εo is the vacuum permittivity.

3
The relative permittivity is associated to the susceptibility, and the relation can be expressed as shown
in Equation 1. The last term in this equation is the contribution to the loss from the conductivity. In
frequency domain it is difficult to distinguish between the conductive and dielectric part of the losses.
However, at low frequencies the conductivity may dominate the dielectric losses (ε’’ω)) causing a
strong increase in the losses at low frequencies. Thus in such cases it may become possible to
distinguish between these losses from measurements performed at several frequencies.

The real and imaginary parts of the complex permittivity are linked by the dielectric loss factor [3];

tan δ (tan δ’(ω) = ε’(ω)/ ε’’(ω)) (2)

For diagnostic purpose several dielectric diagnostic parameters are established [4]. These are
summarised in Table 1.

Table 1: Description of diagnostic parameters used during condition assessment.

Diagnostic Parameter Description
ε’(ω) Real part of the permittivity
∆ε’(ω) Changes in the real part of the permittivity (∆ε’(ω) = χ’(ω))
ε’’(ω) The imaginary part of the relative permittivity (dielectric loss)
tan δ’(ω) The dielectric loss tangent (tan δ’(ω) = ε’(ω)/ ε’’(ω))
∆(∆ε’(ω)) Changes in the real part of the permittivity with voltage
∆ε’’(ω) Changes in the dielectric loss factor with voltage
∆tan δ’(ω) Changes in the dielectric loss tangent with voltage

Experience from medium voltage condition assessment of water treed XLPE cables, shows that the
real part of the permittivity do not change unless there is a significant change in the imaginary part of
the relative permittivity [3]. Therefore this parameter is rarely used as a diagnostic parameter, and
consequently the dielectric loss factor is directly proportional to the loss tangent independent of
frequency [3].

Non-linear Dielectric Response

Time and frequency domain dielectric response measurements on water treed medium voltage XLPE
cables have revealed that both vented and bow-tie trees cause the response to become non-linear [5].
The voltage dependence (non-linear dielectric response) of the complex permittivity due to water
treeing has been found to be particular useful during condition assessment of medium voltage cables
(12 and 24 kV), and is the main cause for performing measurements at several voltages [4]. Some
hypothesis of the cause for the non-linear response has been presented, explaining some of the features
of the dielectric response from water trees [5].

3 EXPERIMENTAL WORK
3.1 LABORATORY TESTING

Cable Samples
The cable samples characterised during the laboratory examinations are described in Table 2. All
cables had suffered from service failure, removed from service and shipped to the laboratory due to
poor service performance. The cables were produced in Scandinavia more than 20 years ago. The
XLPE insulated cables had the old design with extruded conductor screen and graphite insulation
screen with semi-conductive tapes. The outer sheath was made of PVC (non-water tight design). The
length and number of samples examined for the different cables were ranging from 5 to 90 metres.

Dielectric Spectroscopy
Non-destructive measurements of the dielectric response were measured according to the previously
described procedure prior to the AC breakdown voltage measurements and water tree examinations.

4
AC Breakdown Voltage and Water Tree Examinations
Measurement of the breakdown voltage was performed by applying an AC ramp voltage starting at Uo
(26 kV) increasing by a rate of 20 kV/min [1]. The water tree analysis was performed by microscopy
analysis of 0.5 mm thick microtomed slices stained in methylene blue dye solution. Only the longest
bow-tie and vented water tree was recorded as well as the tree density [1].

Table 2 Description of the cable samples characterised in the laboratory and types of examinations
performed.
Diagnostic Measurement
Service Total
Cable No. of Year of AC
Voltage Length Dielectric Water Tree
No. samples Production Breakdown
(Um) [kV] [m] Spectroscopy Analysis
Voltage
L1 36 9 90 1977 X X X
L2 52 6 78 1978 X X X
L3 52 4 56 1978 X X X
L4 52 1 5 1980 X X ---
L5 52 3 15 1980 X X X

3.2 ON-SITE TESTING

Cable Samples
A description of seven selected cables from the on-site testing is shown in Table 3. None of the
selected cables had previously suffered from service failure. Um ranged from 36 to 72.5 kV.

Table 3 Description of the cable samples measured on-site by dielectric spectroscopy.

Service
Total
Cable Voltage Year of
Length Description of Cable Accessories
No. (Um / Uo) Installation
[m]
[kV]
Four joints (tape and pre-moulded) and pre-moulded
O1 36 / 18 1600 1978
terminations.
O2 52 / 26 1960 1972 Six joints (pre-moulded) and pre -moulded terminations.
Type and number of joints unknown and pre-moulded
O3 52 / 26 5500 1980
terminations placed within oil-filled porcelain housing.
Type and number of joints unknown and pre-moulded
O4 52 / 26 260 1976
terminations placed within oil-filled porcelain housing.
Type and number of joints unknown and pre-moulded
O5 52 / 26 3350 1979
terminations placed within oil-filled porcelain housing.
Pre-moulded terminations (inside) and placed within oil
O6 72.5 / 36 100 1978
filled porcelain housing (outside).
Six tape and two pre-moulded joints, and pre-moulded
O7 72.5 / 36 3250 1978
terminations placed within oil-filled porcelain housing

The type of cable accessory is also described, as the dielectric response from especially short cables
has been found to be particular influenced by some types of joints and terminations [6]. All cables
were installed in the seventies except from Cable O3 that was installed in 1980. This cable was also
the longest cable having a total length of 5.5 km.

Dielectric Spectroscopy
Non-destructive measurements of the dielectric response were initially measured according to the
previously described procedures. However, in some cases the maximum voltage was reduced as well
as the maximum frequency.

5
4 EXPERIMENTAL RESULTS AND DISCUSSION
4.1 LABORATORY TESTING

Dielectric Spectroscopy
Example of measurements of tan δ and ∆ε’ on a sample of Cable L1 is shown in Figure 3.
Measurements on an unaged XLPE cable are added to Figure 3 a) for comparison. It can be seen that
the tan δ is strongly increased compared to that measured for an unaged cable. This increase could also
be due to the higher loss tangents of the accessories [6]. However, the relatively frequency
independent values (above 6 kV) and the hysteresis (higher loss tangents measured second time at 6
kV), and the strong voltage dependence (non-linearity) indicate that water treeing is the main cause for
the response [5]. This is also confirmed by the measurements of the ∆ε’. No non-linearity is observed
for the unaged cable.

Figure 3 c) shows that the non-linearity, using the increase of the response at 14 and 6 kV, is relatively
frequency independent, making the diagnostic parameters independent of the applied frequency. This
is in many cases fortunate, as e.g. for long cables such high voltages can only be applied reducing the
frequency. This feature of the dielectric response was also observed for the other cables. The non-
linearity in the ∆ε’ is higher than for the tan δ. This is also observed for medium voltage cables with
similar design [4].
-2
1x10 14 kV -2
1x10
-2
1x10 6 kV 2nd
14 kV ∆(∆ε')
Dielectric loss tangent, tan δ

Change in permittivity, ∆ε'

6 kV
6 kV 2nd Degree of non-linearity
-3
1x10
6 kV 1x10-3
1x10-3 ∆(tan δ)
1 kV
1 kV

1x10-4 1x10-4 1x10

-4

Unaged

-5 1x10-5 1x10-5
1x10 0.1 1 10 0.1 1 10
0.1 1 10 Frequency [Hz] Frequency [Hz]
Frequency [Hz]

a) tan δ b) ∆ε’ c) ∆tan δ and ∆(∆ε’)

Figure 3 a) Example of laboratory measurement of the dielectric loss tangent (tan δ) of the 36 kV
XLPE Cable L1. b) The corresponding change in permittivity, (∆ε’(ω). c) Degree of non-
linearity as a function of the frequency using the diagnostic parameters described in Table
1.

As shown in Table 4, the tan δ of Cable L2 to L5 was all higher than 0.001 at voltages close to service
stress, which is approximately one decade higher than that measured for unaged cables. This also
applies to the ∆(∆ε’). However the non-linearity was rather low for Cable L4.

Water Treeing
Typical water treeing in the five high voltage cables is shown in Figure 4. For Cable L1 water trees
were growing both from the conductor as well as from the insulation screen. The longest observed
water tree was growing from the conductor screen. For cable L2 to L5 water trees were typically
growing from the insulation screen only. Also some short bow-tie trees were detected.

6
 a) L1 b) L2 c) L3 d) L4 e) L5

Figure 4 Typical water treeing detected in the different high voltage cables. a) Cable L1 with water
trees growing both from the conductor as well as the insulation screen (part of the
breakdown channel after destructive laboratory testing can also be seen). b) Cable L2
showing a high density of water trees growing from the insulation screen. c) Cable L3 also
with water trees growing from the insulation screen (lower density than L2). d) Cable L4
with vented trees from the insulations screen (lower density as well as less “needle shaped”
vented water trees). e) Cable L5 with vented trees growing from the insulation screen. The
breakdown channel can also be seen, partly covering the water tree causing the electrical
breakdown during testing in the laboratory.

Relation Between Destructive and Non-Destructive Data

Table 4 shows a summary of the results from the spectroscopy as well as the destructive measurements
on the five cables. The non-destructive diagnostic parameters measured on the cable sample with the
longest water trees or the lowest ACBD voltage is shown.

Table 4 A summary of the results from the laboratory testing. The diagnostic
parameters are taken from the cable sample with the lowest AC breakdown
voltage or longest observed water tree. The values are taken at 0.1 Hz unless noted.
Cable No. ref. [4] L1* L2 L3 L4 L5
ε’’(ω) | Uo < 0.2 1.2 6.7 3.9 2.5 9.0
Diagnostic
tan δ’(ω) | Uo < 0.1 0.5 2.9 1.7 1.1 3.9
Parameter
(x10-3) ∆(∆ε’(ω)) < 0.01 1.0 2.7 1.7 1.5 2.3
∆tan δ’’(ω) < 0.01 0.3 1.2 1.2 0.1 2.2
l max [%] --- 43 14 15 16 19
Destructive Density [cm ] -2
--- 0.1 50 2.7 --- 75
Analysis
UACBD [kV] / > 300 / 83 104 / 104 / 104 /
---
UACBD/Uo > 12 /4.6 4.0 4.0 4.0
lmax: Maximum length of the longest observed water tree
UACBD: Lowest measured breakdown voltage
Uo: Service stress of the cable
* Data was taken at 1 Hz due to noise at 0.1 Hz

The ∆ε' as well as the tan δ varied significantly between the different samples of Cable L1. The
sample with the lowest breakdown voltage had also the lowest tan δ values, which is in contradiction
to that measured on medium voltage cables with similar design. This is likely due to the low density of
vented water trees of this cable compared to the other cables [5]. The effect of the density of the
vented water trees on the dielectric response is also demonstrated by comparing the results from Cable
L2 and L3, having the same length of the longest observed water trees as well as breakdown voltage.
Both the ∆(∆ε’) as well as the tan δ were significantly higher. However, the non-linearity of the tan
δ was about the same.

7
 4.2 ON-SITE TESTING

Results from dielectric spectroscopy measurements on Cable O1 (Um = 36 kV) is shown in Figure 5.
Figure 5 a) show the dielectric loss tangent (tan δ) at different voltages in the frequency range of 10 to
0.1 Hz and voltages up to 18 kV. The tan δ is voltage dependent indicating that water trees are present
within the insulation. However, data from measurement of dielectric spectroscopy on medium voltage
cables also indicate that a voltage dependent increase in the tan δ can occur, and can possible cause
misleading interpretation of the data. Such increase is particular clear at low frequencies below 0.1 Hz,
but can contribute to the response at higher frequencies. In such cases it is very important to be able to
measure at several frequencies in order to separate the response from the accessories and the cable
(water trees) [6].

1x10-2 1x10-2
Dielectric loss tangent, tan δ

Change in permittivity, ∆ε'

-3 18 kV -3 18 kV
1x10 1x10

12 kV
9 kV 12 kV

3 kV
9 kV
1x10-4 1x10-4
Unaged XLPE cable

-5
1x10 1x10-5
0.1 1 0.1 1
Frequency [Hz] Frequency [Hz]

a) tan δ b) ∆ε∋ c) Termination

Figure 5: a) Example of on-site measurements of the dielectric loss tangent (tan δ) of one of the
phases of the 36 kV XLPE Cable O1. b) The corresponding change in permittivity,
(∆ε’(ω)). c) Picture of the termination where the measurement were performed.

-2 -2
1x10 1x10
Dielectric loss tangent, tan δ

Change in permittivity, ∆ε'

-3 -3
1x10 1x10 18 kV

18 kV
12 kV

6 kV
12 kV
-4 -4
1x10 1x10
6 kV

1x10-5 1x10-5
0.01 0.1 0.01 0.1
Frequency [Hz] Frequency [Hz]

a) tan δ b) ∆ε∋ c) Termination

Figure 6: a) Example of on-site measurement of the dielectric loss tangent (tan δ) of one of the
phases of the 52 kV XLPE Cable O3. b) The corresponding change in permittivity,
(∆ε’(ω)). c) Picture of the termination where the measurement were performed.

8
 1x10-2 1x10-2
Dielectric loss tangent, tan δ

Change in permittivity, ∆ε'

-3 -3
1x10 1x10

3-21 kV 3-21 kV

1x10-4 1x10-4

1x10-5 1x10-5
0.1 1 10 0.1 1 10
Frequency [Hz] Frequency [Hz]

a) tan δ b) ∆ε∋ c)
Figure 7: a) Example of on-site measurement of the dielectric loss tangent (tan δ) of one of the
phases of the 72.5 kV XLPE Cable O6. b) The corresponding change in permittivity,
(∆ε’(ω)). c) Picture of the termination where the measurement were performed.

For the old cables with insulation screens of graphite painting and semi-conductive tapes the tan δ is
approximately frequency independent. Increasing the voltage from 3 to 18 kV, causes the tan δ to
increase from 0.00025 to 0.0015. These results also indicate the importance to measure at several
voltages up to the service stress, as this increases the sensitivity of the measurements and improves the
condition assessment. Figure 5a) shows the corresponding increase in the permittivity (or
susceptibility) as a function of voltage. The relative increase with voltage is about the same as that
measured for the tan δ.

Figure 6 shows the measurement of Cable O3 (Um = 52 kV). As for Cable O1, the tan δ as well as the
∆ε’(ω) are voltage dependant. However, the tan δ at the lowest subjected voltage (6 kV) is very low
as for a new cable. This demonstrates the necessity to use a sufficient high voltage during testing. It is
likely that this cable would have been assesses as good if voltages up to 6 kV only had been used.
Anyway, the ageing of this cable is likely not severe, but the results strongly indicate that water trees
are present within the insulation.

Figure 7 shows the results from the on-site measurements of Cable O6 (Um = 72.5 kV). In this case no
voltage dependence was observed for up to 21 kV in the frequency range of 10 to 0.1 Hz. This applies
to the tan δ as well as for the ∆ε’(ω). This indicates that no severe water treeing is present in this
cable. However, this cable was installed in 1978, and it is a possibility that a different design was used
for this cable (extruded insulation screen). This can cause a decrease of the water tree density (and also
length), and consequently on the diagnostic parameters [2]. This complicates the condition assessment,
and it is possible that long water trees will not be detected during the on-site testing if the density is
very low. In addition, up to now the knowledge about water treeing for cables with Um higher than 52
kV is insufficient, and more data should be gained in order to improve the condition assessment.

Note the different frequency ranges used for the different cables. For the short cable (Cable O6),
frequencies from 10 Hz could easily be used. For the long cable (Cable O3) of 5.5 km, only
frequencies below 0.1 Hz could be used. This demonstrates the necessity to be able to measure the
dielectric response even at frequencies below 0.1 Hz.

The results from a selection of eight high voltage cables assessed on-site is summarised in Table 5. It
can be seen that Cable O2 which has been longest in service has the highest tan δ and ∆ε’ as well as
degrees of nonlinearity.

9
 Table 5 Summary of results from the on-site testing. The diagnostic parameters
are taken at 0.1 Hz using values from the phase with the highest tan δ or
non-linearity of the response. The voltage dependence was calculated using
values at service stress (Uo) and half the service stress (0.5Uo).
Diagnostic
Parameter ref. 1 2 3 4 5 6 7
(x10-3)
tan δ’(ω) < 0.1 1.1 7.1 0.5 2.6 0.12 0.3 0.3
∆(∆ε’(ω)) < 0.01 1.1 15.4 0.6 1.5 0.01 0.4 0.2
∆tan δ’’(ω) < 0.01 1.0 5.2 0.4 1.2 0.02 0.03 0.1
∆ tan δ (0.1 Hz) = tan δ Uo - tan δ 0.5Uo
∆(∆ε’) (0.1 Hz) = ∆ε’Uo - ∆ε’ 0.5Uo

The measurement showed therefore that a severe water tree ageing had occurred in this cable. Cable
O5 - O7 had low loss tangents and low degree of non-linearity, and no severe water treeing is expected
in these cables. The service performance of these cables will be recorded, and new measurements
performed after about 5 years in order to detect any changes of the diagnostic parameters.

5 CONCLUSIONS

From the laboratory and on-site measurements on water treed high voltage XLPE cables it can be
concluded that;

- Water treeing does occur in high voltage cables.

- The relative length of the critical water trees causing service failures decreases strongly with
system voltage.
- Water treeing in XLPE cables with the old design used in the seventies with insulation screens
of graphite painting and semi-conductive tapes is characterised with a high density of trees
growing from the insulation screen.
- The non-destructive method based on dielectric spectroscopy initially designed for medium
voltage XLPE cables (12 and 24 kV), can be used for condition assessment of water treeing in
old high voltage XLPE cables with insulation screens of graphite painting and semi-
conductive tapes.
- The different types of cable accessories do not have any impact on the condition assessment of
the water treed high voltage XLPE cables.

BIBLIOGRAPHY

[1] S.Hvidsten et.al.: “Condition Assessment of Water Treed Service Aged XLPE Cables by Dielectric
Response Measurements”, Session 2000 CIGRE, Paper 21-201, Paris, France, August 2000.
[2] H.Faremo et.al.: “Service Experience for XLPE Cables Installed in Norway – From Graphite Painted
Insulation Screens to Axially and Radially Water Tight Cable Constructions”, 14th International
Conference and Exhibition on Electricity Distribution (CIRED), Birmingham, pp. 3.2.1 – 3.2.5.
[3] A.K.Johnscher: “Dielectric Relaxation in Solids”, Chelsea Dielectric Press, London 1983.
[4] P.Werelius et.al.: “Dielectric Spectroscopy for Diagnosis of Water Tree Deterioration in XLPE Cables”,
IEEE Transaction on Dielectrics and Electrical Insulation, Vol. 8, No.1, February 2001.
[5] S.Hvidsten: “Nonlinear Dielectric Response of Water Treed XLPE Insulation”, Dr. ing. Thesis, NTNU,
Trondheim, Norway, 1999. ISBN 82-471-0433-4.
[6] A.Avellan et.al.: “Frequency Domain Response of Medium Voltage XLPE Cable Terminations and its
Influence of Cable Diagnostics”, IEEE International Symposium of Electrical Insulation, Anaheim, USA,
April 2-5, 2000.

10