Upload
chethan
View
215
Download
0
Embed Size (px)
Citation preview
8/8/2019 Development Small Sample Verfication
1/13
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 15, No. 4; August 2008
1070-9878/08/$25.00 2008 IEEE
1131
A New Dielectric Response Model for Water Tree Degraded
XLPE Insulation Part A: Model Development
with Small Sample VerificationAndrew J. Thomas and Tapan K. Saha
University of Queensland
School of Information Technology and Electrical Engineering
Brisbane, QLD 4072, Australia
ABSTRACT
Water tree degradation in underground XLPE insulated cables is a growing, world-
wide problem. This form of degradation is ultimately fatal for affected cables, and
therefore the detection of damaging trees in power cable insulation is vital for
distribution companies to avoid catastrophic failure. Dielectric responsemeasurements, in both the time and frequency domains, can generate valuable
information about the condition of the cable. However, the interpretation of how these
dielectric response measurements relate to water tree density and length is a difficult
task. This paper will present a new dielectric response model for water tree degraded
XLPE insulation. The model is based on finite element analysis to determine the
electrical behaviour of water tree degraded insulation. Preliminary simulations will
verify the model development by comparing the results to small sample Pulsed Electro-
Acoustic (PEA) measurements performed by other researchers. The importance of a
strong non-linearity mechanism for accurate modelling will also be elucidated.
Index Terms water trees, cross linked polyethylene insulation, finite element
method, dielectric measurements, space charge, conductivity, dielectric loss,
nonlinearities.
1 INTRODUCTION
WATER tree degradation is one of the most serious
afflictions that can occur within underground medium voltage
cross-linked polyethylene (XLPE) cables. Because of this
fact, many studies have been performed concerning how water
trees grow, how they bring about failure and how they can be
detected. Perhaps the best way to fully understand the
previous points is through modelling, which can analytically
describe the physical situation. Water tree growth is somewhat
inevitable in XLPE cables under certain conditions. Therefore,
it is a useful exercise to develop a model which can relatedielectric response measurements to the water tree degraded
condition of a cable, which aids in the cable diagnosis. This
will result in a more efficient and accurate diagnostic process.
A number of electrical behaviour models of water tree
affected XLPE cables have been developed in the past. Many
papers [1, 2] concentrate on power frequency (i.e.
capacitively) graded fields, with an emphasis placed on field
enhancement above the norm at the tree tip, with the general
aim of deducing the cause of water tree induced failure, which
of course is an important area of study. However, a water tree
model which is to be used in interpreting dielectric response
measurements will need to be dynamic, as dielectric response
measurements usually cover many decades of time or
frequency. In addition to this, the model will need to consider
non-linear effects with charging voltage, as this is an often
observed measurement phenomena.
The authors of [3] developed a model investigating a
mechanical non-linearity mechanism, which involved theopening and closing of conducting channels between voids
under Maxwell stresses. However, this model was developed
to prove that the mechanical non-linearity mechanism could
be responsible for the observed non-linearity, and is therefore
only local i.e. its solution domain is of a microscopic scale
around a number of voids and channels.
Another series of papers considers the dynamics of water
tree channels, where perhaps the best example is [4]. These
papers consider channel dynamics in the form of a Maxwell-
Wagner type equivalent circuit. The total capacitance andManuscript received on 11 April 2007, in final form 23 January 2008.
Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on April 24,2010 at 07:05:33 UTC from IEEE Xplore. Restrictions apply.
8/8/2019 Development Small Sample Verfication
2/13
A. J. Thomas and T. K. Saha: A New Dielectric Response Model for Water Tree Degraded XLPE Insulation Part A 1132resistance of the water treed region is calculated by allowing
the size of the water treed region to expand and contract to
simulate the filling of voids. The conduction in the region is
modelled by asinh dependence on the electric field, with the
strength of the sinh dependence governed by a value h. The
authors allow the size of the water treed region and the value h
to vary (along with the low field conductivity of the region,
which varies little) at different applied voltages in order to
match the simulated and measured data. However, allowing h
to vary along with the size of the water treed region with
voltage is perhaps physically unintuitive, as it is difficult to
find a physical reason as to why any change of this variable
should occur.
Therefore in order to expand beyond the scopes or
limitations of the previously mentioned papers, this paper will
detail the investigation towards a new electrical model for
water tree degraded XLPE. The goal of this model is to
macroscopically describe the electrical behaviour of water
trees in XLPE under certain applied electrical conditions, on
the scale of the water tree length itself. The overall goal of this
study is to enable a more refined understanding of the effect ofwater trees on dielectric response measurements, with the
actuation of this goal shown in the accompanying paper [5].
However, before such an application can be realised, an
electrical model for water tree degraded insulation must be
developed and verified, which is the content of this paper.
In Section 2, the details of a constructed one dimensional
finite element model will be given, which describes the time-
varying electrical behaviour of water tree degraded XLPE
insulation. Particular attention is payed to simulating the
space charge build up at the interface between the water tree
and the healthy XLPE, with this interface being modelled
diffusely in the investigation. In order to validate the model, in
Section 3 the results of the simulations will be compared tospace charge measurements performed by other researchers
[6] in small sample experiments. It will be seen that in order
for the model to accurately simulate the measurements, a
mechanical non-linearity mechanism based on Maxwell forces
is needed. Section 4 will contain discussion and conclusions
of the results of this study. This paper is a major expansion
of the preliminary/investigative work presented in [7].
2 MODEL DESCRIPTION
This section will address the technical details and
assumptions used to generate the model of electrical
behaviour of water treed insulation.
2.1 STRUCTURE OF WATER TREES
Water trees are generally considered to consist of water
filled micro-voids, which may or may not have
interconnecting, conducting channels. Numerous studies
[8-10] have been performed (to name only a few), using a
variety of techniques such as Transmission Electron
Microscopy (TEM) and Scanning Electron Microscopy
(SEM), to discover this. The largest voids generally found
within water tree channels are on the order of 5 m in
length [9]. The connecting channels between voids were
found to be on the order of >1 m in diameter at the foot of
a water tree, whereas the channels throughout the rest of the
tree are on the order of 10-100 nm in diameter [9].
The interface between a water tree and XLPE within an
insulation can easily be considered to be a rather
complicated structure. On a macroscopic scale (on the scale
of the water tree length itself), it is likely that there is a
decay of void and channel density near the tree tip, leading
to a change in conductivity and permittivity that is not
instantaneous, but diffuse. There is also the question of
field assisted opening and closing of water tree channels
through Maxwell forces, which will be investigated further
in later sections.
In addition to this, the electro-osmosis water tree growth
theory, reviewed in [11], sensibly proposes that it is likely
that ionic species will diffuse, in an electrically assisted
manner, into the polymer beyond the tree tip. This diffusion
of impurities will likely increase the concentration of
hopping sites located close to the conduction band,
increasing the conductivity and permittivity somewhat
beyond the tree tip. This effect is also likely to result in a
macroscopic conductivity and permittivity profile that is
relatively gradual.
In order to model the electrical behaviour of water tree
degraded insulation, it is vital that this interfacial region is
considered properly, as this region will dominate the
electrical characteristics of the local material. These
considerations will be addressed in the next section.
2.2 PERMITTIVITY AND CONDUCTIVITYREPRESENTATIONS
In this investigation, the insulation properties will be
macroscopically considered in one-dimension only, with an
assumption of uniformity in the remaining two dimensions.
The water tree permittivity and conductivity are considered
to be uniform up to a certain point depending on the length
of the water tree. Beyond this region of relative uniformity,
there will be a transition from water tree electrical
properties to healthy XLPE electrical properties (i.e. a
substantial decrease in conductivity, and a less substantial
decrease in permittivity). In some previous water tree
modelling exercises, this change of electrical properties has
been considered as an ideal boundary, with an infinitely
sharp profile. However, for reasons stated previously, this
ideally sharp boundary at a macroscopic level is unlikely to
exist. Experimental evidence for a more diffuse boundary
will be shown in later sections.
The exact form of the interface, with regards to the
decreasing profile of the conductivity and permittivity is
practically impossible to measure directly. However
inferences or postulations can be made from available
information and these inferences can be tested through
less direct means for verification. As mentioned
previously, a major proposed water tree growth
mechanism is electro-osmosis, whereby hydrated ions
diffuse into the insulation beyond the water tree proper
with electric field assistance [11, 12]. Diffusion of
Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on April 24,2010 at 07:05:33 UTC from IEEE Xplore. Restrictions apply.
8/8/2019 Development Small Sample Verfication
3/13
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 15, No. 4; August 2008 1133
contaminants into XLPE has been studied previously [13,
14] with the contaminants in this case being sourced from
the semi-conductive layer. Figure 4 in [13] shows an
exponential-like decay in the FTIR absorbance level
(which is related to the concentration of species) of
contaminants away from the semi-conductor layer. It is a
reasonable assumption that the conductivity (and
permittivity) of an insulation are proportional to the
amount of contaminants it contains. In addition to this,the authors of [14] had some success in modelling the
space charge levels near the semi-conductive layer using
an exponential conductivity and permittivity profile.
It can also be noted at this point that according to the
electro-osmosis theory discussed in [12] a continual
recombination and resolvation process (the resolvation
owing to osmosis and perhaps dielectrophoresis) acting on
diffused ion pairs mechanically fatigues the insulation
around the ion locations. Therefore, it is also sensible to
propose that permanent mechanical damage of the polymer
is proportional to the ion/contaminant diffusion profile,
leading to the existence of proto-voids and channels in the
tree tip region, which may enable an electricalreconnection process through Maxwell stresses opening
electrically conductive paths up to these areas [3]. This
will be discussed further in Section 2.3.
While the contaminants and diffusion processes studied
in this literature is of a different form to the ionic
contaminants and electro-osmosis diffusion in the water
tree case, an exponential change of electrical properties
through space is nevertheless intuitive in light of this
information and a reasonably well informed hypothesis.
Therefore in this study, the conductivity and permittivity of
the dielectric mixture will change in an exponential manner
from the water tree electrical properties to the XLPE
electrical properties. This interfacial change will bemodelled by thesigmoidfunction, given by:
( )xe
xf+
=1
1(1)
Where is a shape parameter. In the limit of ,
equation (1) approaches the heavy-side step function. Using
equation (1) to express the conductivity and permittivity of
the modelled insulation (for a vented tree) is given by:
( ) ( )kxp
kx
w
eex
++
+=
11)(
(2)
( ))()( 11 kx
p
kx
w
eex
++
+=
(3)
In equations (2) and (3), k is a location parameter
determining the spatial location in the x-direction of
uniform water tree electrical properties. Also in equations
(2) and (3), w and w are the water tree channel
conductivity and permittivity respectively and p and p are
the XLPE conductivity and permittivity respectively. The
ranges of these values considered in this study are [15-17]:
Conductivity
Water tree: 1x10-11 1x10-7 S/m
XLPE: 1x10-18 1x10-16 S/m
Permittivity (relative)
Water tree: 2.3 5
XLPE: 2.3
It should be noted that the electrical behaviour of the
electrode/XLPE interface in this model has been neglected,
and therefore whatever limitations on accuracy that this
omission may make should be considered.
The above formulations for the conductivity and
permittivity are to be considered as the low-field
representations. As is well known, water tree degraded
cables exhibit dielectric response non-linearity. This non-
linearity must arise in changes with the charging voltage of
the electrical properties of the insulation. These changes
will be considered in the next section.
2.3 NON-LINEAR BEHAVIOUR OF ELECTRICALPROPERTIES
A change of the electrical properties in water tree
degraded insulation due to the electric field, at constant
temperature, can occur in two conceivable ways. The first
way is through electronic or hopping processes due to the
lowering of potential barriers between charge trapping
centres through high electric fields. The second way was
proposed in [3], and this is due to a proposed mechanical
alteration of the water tree degraded material through the
action of Maxwell forces. Both mechanisms will be
considered in this model.
Figure 1. Potential well model showing the hopping of an electron or
negative ion over a potential barrier, from site S1 to S2. Arrows show both
a classical jump over the potential barrier and a quantum mechanical
tunnelling through the barrier.
Figure 1 shows the mechanism of hopping or tunnelling
(electron) of a negative carrier over a potential barrier
lowered by an applied electric field. In solid insulation,
carriers do not move freely through the material, but rather
hop between localised trapping centres. In order for this
hop to occur, the carrier must gain thermal energy sufficient
to overcome the barrier height between adjacent trapping
centres. The energy required to traverse this barrier can be
significantly reduced by an applied electric field of
S1
W
S2
E
Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on April 24,2010 at 07:05:33 UTC from IEEE Xplore. Restrictions apply.
8/8/2019 Development Small Sample Verfication
4/13
A. J. Thomas and T. K. Saha: A New Dielectric Response Model for Water Tree Degraded XLPE Insulation Part A 1134sufficient magnitude. The number of charges released from
trap centres is a stochastic process based on Boltzmanns
statistics, therefore, an applied electric field increases the
probability of a site transition taking place and thus more
charges are released to contribute to charge flow.
Because conductivity is a function of the number of charge
carriers, it is easily seen that the prior explained process will
increase the conductivity of the material. A derivation of field
dependent conductivity from the jump probabilities in theforward and reverse directions (with respect to field) was
performed in [18], with the result being:
=
kT
eEa
eEa
kTE
2sinh
2)( 0 (4)
Where a is the distance between trap centres, e is the
charge of the carrier in Coulombs, k is Boltzmanns
constant and Tis the temperature in Kelvin.
A similar line of reasoning can be made to determine the
non-linearity with respect to applied voltage for the real and
imaginary parts of the permittivity. If, under an ac appliedvoltage, the carriers undergo a certain hop, or multiple hops
in the direction of the field, then a polarisation will be
created. The more carriers that undergo this process due to
field assisted de-trapping the higher the complex
permittivity will be. The derivation of the field dependent
permittivity expressions can be seen in Appendix I, and are
found as:
( ) 0''
2sinh
2 +
=
kT
eEa
eaE
kTE LF (5)
( )
= kT
eEaeaE
kTE LF2
sinh2'' '' (6)
Where the subscriptLFstands for low-field, with respect
to the permittivity. The above formulations complete the
proposed electronic or hopping non-linearity mechanism.
The second non-linear mechanism that will be
considered in this work is the mechanical mechanism. This
mechanism for non-linearity was proposed in [3]. The
mechanism is based upon the observed structure of water
trees consisting of voids with interlinking channels. The
authors of the referenced study consider these interlinking
channels to be closed, and therefore non-conducting under
low-field conditions, but with an opening of these channels,and conduction within them, under high field conditions.
The explanation given as to the source of the energy
required to open these channels is Maxwell forces due to
the electric field within the region. These Maxwell forces
are a pressure acting on a 2-dimensional surface in the
direction of the electric field, with this pressure being
related to the electric field and permittivity as:
( ) 22
1
2
1EnDEpr
r
rr
== (7)
Where nr
is the unit normal vector. If this pressure p is
considered to act on the many interfaces between the voids
and XLPE, it may act to open the channels between voids,
allowing them to fill with liquid and become conducting.
Over the tree cross section, this field induced pressure and
subsequent opening of channels should lead to an average
increase in conductivity and permittivity of the area.
Deducing the exact relationship between the Maxwell
forces (pressure) and the conductivity analytically (throughan additional model) is not within the scope of this study.
However, some experimental results can be examined to
give an indication of the empirical form of this relationship.
A number of Frequency Domain Spectroscopy (FDS)
measurements were performed on accelerated, wet aged
cable samples. Some of these cable samples were showing
typical LC (leakage current) responses as designated in
[19], which indicate long vented tree degradation. Figure 2
shows the loss response (imaginary part of the permittivity)
of one of these cables.
Figure 2. Typical measured leakage current response at differing applied
voltage levels. Uo -> rated voltage
These degraded samples had loss slopes inversely
proportional to frequency, which strongly suggests that the
conduction of the samples was completely dominated by
long vented trees. Therefore, assuming that the Maxwell
forces non-linearity mechanism, as proposed in [20], is the
dominant non-linearity mechanism, the square of the
applied voltage (and therefore applied field) versus the 0.01
Hz, low frequency loss will give the relationship between
the Maxwell forces and the conductivity of the long water
tree/s. The square of the applied voltage and the low
frequency loss for the cable samples showing leakage
current behaviour were found to be exponentially related,
with the exponential fitting havingR2 values in the range of
0.9-0.98, therefore signifying a likely exponential
relationship.
Therefore, in this study, we assume the relationship of
conductivity to pressure to be:
Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on April 24,2010 at 07:05:33 UTC from IEEE Xplore. Restrictions apply.
8/8/2019 Development Small Sample Verfication
5/13
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 15, No. 4; August 2008 1135
( )
>
8/8/2019 Development Small Sample Verfication
6/13
A. J. Thomas and T. K. Saha: A New Dielectric Response Model for Water Tree Degraded XLPE Insulation Part A 1136The function that governs the variance of the solution
variable is called an interpolation function. The brief
mathematical treatment of the FEM is based on explanations
given in the aforementioned references [21, 22].
If the general form of the differential equation is
considered, it can be represented by:
( ) 0= fL (13)
Where L is a general differential operator acting on
andfis the forcing function acting on (charge density in
equation (10) or equation(11)). This approximation of the
solution variable (the potential V in the solution of
Poissons equation) through the use of interpolation
functions can be considered mathematically by:
( ) ( ) ( )=
+=P
n
nn xFaxx1
(14)
Where P is the number of interpolation functions
defining the solution domain , an is a constant defining
the magnitude of the nth interpolation function, Fn is the
nth interpolation function and is a function that satisfies
the Dirichlet or essential boundary conditions. It is now
clear that the representation in equation (14) is no longer
exact for finite P, so now a residual, R, will exist in
equation (13), which is a measure of the inaccuracy of the
approximation, such as:
fLR
=
(15)
The objective of the finite element method is to minimizethis residualR. This minimization can be achieved through
the use of weighing functions, w, which are multiplied
throughout equation (13) to obtain an integerable solution
expression:
( ) 01
=
+
=
dfFaLLwP
n
nni (16)
The Garlekin method involves choosing the weighing
functions from the same set as the trial functions (while
keeping functions linearly independent) i.e.:
( ) ( )xFxw ii =
In the case of solving Poissons equation, this gives an
integrand of (from (16)):
00
=
+
dxdx
dV
dx
dw
L
(17)
Due to the linear nature of the interpolation functions
(and therefore the weighing functions in the Garlekin
method), equation (17) is now solvable forVthrough linear
algebraic methods, given appropriate boundary conditions
to make the solution unique. In this investigation, the
boundary conditions for each solution of equations (10) or
(11) are the potentials at the electrodes.
This ends a brief description of the method used to solve
for the potential, and also therefore the electric field. Thesolution domain , whose size is governed by the thickness
of the insulation, is segmented into 1000 elements for this
study. The valid solution domain can be considered to exist
within the proposed water tree structures, interface and
healthy XLPE segments in thex direction, and applicable in
the y and zdirections where approximate uniformity holds
in the electrical properties of the material.
3 SMALL SAMPLE INVESTIGATION
In order to access the accuracy of the model, or its ability
to reproduce physical situations faithfully, it was decided
that the model would be tested on a small sample, spacecharge measurement study performed in [6]. In this way,
the space charge profiles produced by the model could be
compared to measurement results on a small scale
experiment, before being extended for use onto large scale
cable specimen geometries.
3.1 MEASUREMENT DETAILS
This subsection will briefly address the measurement
details as given in [6]. The researchers used the Pulsed-
Electro Acoustic (PEA) method to measure volume space
charge densities within the bulk of water tree aged, small
sample specimens. The samples measured in the
aforementioned study were 1mm thick, laboratory aged
specimens. An FeSO4 solution, with concentration of 0.5
M/L, was applied to the sample to encourage initiation and
growth of water trees. After 800 h of 7 kV (peak), 5 kHz
voltage application, water trees were found uniformly over
the voltage application electrode area, with the trees being
of sufficient length to reach the centre of the sample.
During measurement of the space charge profile, a 7 kV
(peak) voltage was applied at varying frequencies, 50 Hz,
0.1 Hz, 0.01 Hz and dc. Two dc measurements of note
were made. One of these dc measurements consisted of the
application of the dc voltage for 1 h, followed by a shorting
of the sample for a further hour, with space charge
measurements being made sporadically during theseintervals. The other of these dc measurements consisted of
the application of a dc voltage, followed immediately by
the space charge measurement.
Seven different samples are measured in the referenced
study. Four of these samples are labelledA,B,CandD, and
were aged for 300, 400, 800 and 1200 h, respectively,
under the previously mentioned conditions. The other three
samples W, D1 and D2 were each aged for 800 hours, but
with moisture contents of 1310 (parts per million) ppm, 480
ppm and 50 ppm respectively.
Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on April 24,2010 at 07:05:33 UTC from IEEE Xplore. Restrictions apply.
8/8/2019 Development Small Sample Verfication
7/13
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 15, No. 4; August 2008 1137
3.2 PEA MEASUREMENT DETAILS ANDIMPLICATIONS
This section will address briefly certain considerations
that need to be taken into account when interpreting
Pulsed-Electro Acoustic measurement results. The PEA
measurement method involves the application of short
voltage pulses to a measurement sample. These voltage
pulses exert a force on any charges within the sample,
and these charges in turn launch acoustic waves
throughout the sample to be measured by a pressure
transducer. Comprehensive descriptions of the PEA
method can be found in [23, 24].
Of particular interest in this study is the attainable
resolution of the PEA systems whose results are being
analysed in this paper. The theoretical space resolution
attainable within a PEA system is given by [23]:
tvl = (18)
Where l is the space resolution, v is the velocity of
propagation of the acoustic wave (approximately 2000
m/s in polyethylene [23]) and t is the time-width of the
applied voltage pulse. This finite resolution will act to
broaden in space any actual space charge profile in the
measurement results. In addition to this, attenuation and
dispersion effects can also act to broaden the space
charge profiles, although in [25] it was found that LDPE
(and presumably by extension XLPE) was a non-
dispersive material. In [23], PEA measurements were
performed on a 2 mm thick LDPE specimen, using a
voltage pulse width of 30 ns. The theoretical space
resolution, using equation (18), is therefore
approximately 60 m. However, the authors of the study
found a resolution of 100 m was the minimum
achievable, which is a 67% increase on the theoretical
resolution.
The implications for this on interpreting PEA
measurement results is that the actual space charge
distributions may be somewhat narrower in space than
the measured distribution. Within the resolution range
however, only the net charge density can be obtained
[23]. Therefore, in order to effectively compare model
simulations (which of course produce actual space
charge distributions) with those obtained by PEA
measurements, the finite resolution must be applied to
the simulated results. This can be estimated by applying
a moving average filter to the simulated distribution,with a window size equal to the resolution. This is
supported by a mathematical study of the PEA
measurement system performed in [26], which found that
the PEA signal is a weighted running average of the
space charge distribution.
A demonstration of this adjustment of the simulated
space charge profile can be observed in Figure 3,
whereby the model is used to produce a very localised
space charge profile in the same geometry and applied
voltage as that in [6], by creating a very sharp change in
conductivity and permittivity between the water
tree/XLPE interface. The resolution in this case is set to
100 m.
Figure 3. Comparison between simulated space charge profile andresolution adjusted profile
It can be observed in Figure 3 that the resolution
adjustment can have a very significant effect on the charge
profile; however the effect isnt quite as significant for
broader simulated profiles. The area under both displayed
profiles is the same (net charge density), so it can be seen
that the resolution simulation operation manipulates the
profile in the correct fashion. All simulated space charge
profiles to be compared with PEA measurements in the
following sections will be resolution corrected.
3.3 COMPARISON OF RESULTS
This section will compare the measurements of the
previously described experimental setup with the
simulations from the model. However, first it is pertinent to
examine the experimental evidence for the hypothesis that
the water tree/XLPE interface is macroscopically diffuse,
instead of an ideal boundary. If the interface does
approximate an ideal boundary, the space charge
distribution due to the interfacial polarization should ideally
be a surface charge and approach an impulse function. A
possible example of an approximation of such a distribution
can be observed in Figure 3, with regards to the actual
space charge profile. After resolution is taken into account,it can be observed that the high and narrow space charge
density is reduced in magnitude and broadened. The spatial
extent of the resolution corrected distribution is 100-130
m. Therefore, it can be said that any space charge
distribution due to a water tree/XLPE interface that is
significantly greater than the PEA measurement system
resolution has some spatial extent. A space charge profile
with spatial extent at such an interface strongly suggests a
relatively gradual electrical properties profile if equation
(12) is considered.
Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on April 24,2010 at 07:05:33 UTC from IEEE Xplore. Restrictions apply.
8/8/2019 Development Small Sample Verfication
8/13
A. J. Thomas and T. K. Saha: A New Dielectric Response Model for Water Tree Degraded XLPE Insulation Part A 1138If Figure 4 is carefully examined along with Figure 6
from [6], it can be observed that the spatial extent of the
space charge profile due to the water tree/XLPE interface
is approximately 330-350 m. The theoretical resolution
for the PEA measurement system used in [6] is 60 m
(pulse width of 30 ns), so assuming that the same
mechanisms occur as in [23], the resolution can be
assumed to be equal to 100 m. Therefore, it is clear that
the water tree/XLPE interface in [6] cannot be consideredto be an ideal boundary. To give further support to this
generally, the results of a higher resolution system [27]
can be analysed.
Figure 4 Time resolved space charge profiles under 7 kV 50 Hz ac voltage
application, sample W.
The resolution given in [27] is 10 m, however this
corresponds to the theoretical resolution of the PEA
system using equation (18) (pulse width of 5 ns).
Therefore, in order to be conservative, 40 m is added to
give an assumed resolution of 50 m (in [23], the actual
resolution is 40 m greater than the theoretical
resolution). The sample to be analysed in [27] is a 0.5 mm
thick XLPE specimen, with water trees grown to the
centre of the sample. If Figure 5 (a) in [27] is carefully
examined, it can be observed that the space charge profile
due to the water tree/XLPE interface has a spatial extent
of approximately 170 m. This again supports the
hypothesis that the water tree/XLPE interface is gradual
in its electrical properties. The two examples given
previously are for samples with likely numerous trees
uniformly reaching the centre of the samples, however the
same diffuse interface can be observed for single trees,
see Figure 4 in [28].
From the previous analysis, it is clear that the diffuse
water tree/XLPE interface hypothesis has been confirmed.
Therefore, the rest of this section will be dedicated to
comparing the model results and space charge
measurements in detail. Figures 4 and 5 show the time
resolved space charge profile under ac voltage excitation
for sample W, and the maximum space charge density of
samples W, D1 and D2 for differing applied frequencies,
respectively [6].
Figure 5 Maximum space charge profiles for varying frequency applied
voltages (50 Hz, 0.1 Hz, 0.001 Hz and dc) for samples W, D1 and D2.
Figure 6 shows the model space charge densitysimulation for a tree bridging half of the 1 mm thick
insulation, with an applied voltage of 7 kVpeak, 50 Hz at
the phases of 108o and 288o, which are the maximum
space charge profiles. The space charge profiles are
resolution corrected, and the water tree conductivity is
equal to 5x10-8 S/m.
By examining Figure 6, it can be observed that the
simulated space charge density corresponds very well to
the space charge density as shown in Figure 4, at its
maximum values of 108 and 288 degrees. It can be
observed in Figure 4 that the spatial extent of the space
charge density is approximately 350 m, with a maximum
space charge density of 1.6 C/m3. The maximum spacecharge density follows the applied voltage with a slight
lag, as in Figure 5, with the maximums occurring at 108
and 288 degrees. Figure 7 shows the total charge versus
the applied voltage, with the total charge being calculated
by integrating the charge density in Figure 6 and
multiplying it by the electrode area given in [6]. It can be
observed that the total charge behaves linearly with
applied voltage, and is practically identical to (A-1),
Figure 5 in [6].
Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on April 24,2010 at 07:05:33 UTC from IEEE Xplore. Restrictions apply.
8/8/2019 Development Small Sample Verfication
9/13
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 15, No. 4; August 2008 1139
Figure 6. Simulated space charge density for a water tree bridging half the
insulation thickness, at phases of 108o and 288o.
Figure 7. Total integrated charge at 108 degrees versus applied voltage
The nature of the space charge build up at the interface of
the water tree and XLPE is due to interfacial polarisation, asargued for convincingly in [6]. This interfacial polarisation is
due mainly to the large decrease in conductivity at the water
tree/XLPE interface. As the relatively large conductivity of
the water tree is likely due to the presence of ionic carriers
[16], the space charge distribution at the tree tip is almost
certainly comprised of ions. It is also possible that beyond
the tree tip, due to the electro-osmosis hypothesis of water
tree growth, the diffusion of ionic impurities and therefore
the likely existence of impurity states relatively close to the
conduction band, that ionic hopping between traps/impurity
sites occurs.
The dc behavior of the water trees can also be
examined with the model. Figure 8 shows the spacecharge measurements for samples A, B, Cand D in [6]
under 7 kV dc voltage application. The measurements
were made immediately after the dc voltage application
[6]. By examining Figure 7 in [6], it can be observed that
trees bridging increasingly greater percentages of the
thickness of the insulation moving from sample A to D.
Figure 9 shows the simulated space charge density after 1
s of voltage application (in an attempt to replicate the
measurement being made immediately after the dc
voltage application).
It can be observed that the simulated space charge profiles
in Figure 9 match well with the measurements shown in
Figure 8. The conductivity of the water treed region in Figure
9 is 5x10-11 S/m, which is within the range of reported
conductivities in the literature. As can be observed, the water
tree conductivity used in Figure 9 is much below that of
Figure 6. This reduction is supported by the results shown in
Figure 5. It can be observed in Figure 5 that the samples that
were dried (D1 and D2) have a reduced maximum spacecharge magnitude (compared to sample W), and that this
magnitude takes considerably longer to saturate (if it does at
all within the applied voltage time frame, this is hard to tell
visually) under dc voltage conditions.
Figure 8. 7 kV dc space charge measurement on four aged samples, A, B,
CandD [6].
Figure 9. Simulated space charge profiles for 7 kV dc voltage application
(one second after voltage application) with 4 different water tree lengths.
If the space charge process is considered via a simple
Maxwell-Wagner interfacial model, the time constant for
the build up of the interfacial space charge can be
determined essentially by the ratio of the dual permittivity
and the water tree conductivity (XLPE conductivity can be
neglected in this simple model due to its small magnitude).
If the charge build up process can be roughly approximated
Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on April 24,2010 at 07:05:33 UTC from IEEE Xplore. Restrictions apply.
8/8/2019 Development Small Sample Verfication
10/13
A. J. Thomas and T. K. Saha: A New Dielectric Response Model for Water Tree Degraded XLPE Insulation Part A 1140by this mechanism, than a slower build up of charge (longer
time constant) can be explained through a reduced water
tree conductivity. Because the magnitude of maximum
space charge density in Figure 8 is equal to approximately 1
C/m3, which is somewhat close to the magnitude of samples
D1 and D2 in Figure 5 (which almost certainly have a
reduced water tree conductivity due to drying), a lower
conductivity for the simulation of Figure 8 was deemed
necessary. According to the model results, which clearlycan reproduce the profiles of Figure 8 accurately, it is
unlikely that the space charge profiles of Figure 8 have
reached a steady state value.
The previous discussion is an effective segue to the
discussion of the clear steady state behaviour of sample
W under dc voltage application as shown in Figure 5. For
sample W, within 60 s the maximum space charge density
has reached its highest value, approximately 1.5 C/m3. It
was found that, in replicating this saturation of the
maximum space charge density with the model that the
Maxwell forces based non-linearity mechanism plays an
important part. When no non-linear mechanisms are being
considered, the only explanation for a steady state spacecharge density for sample W is a close to ideal boundary
between the water tree and XLPE interface. Because of
the very slow time constant of such an interface, the
movement of charge into a close to ideal boundary is
likely to be beyond any practical measurement time.
However, as was discussed previously, the measured
space charge profiles in the literature, and the successful
simulations shown in this section, strongly suggest a
diffuse boundary, with a relatively gradual change of
electrical properties. The simulation results of a diffuse
interface, with electrical properties the same as that which
produced Figure 6, without the inclusion of a non-linear
mechanism, under 7 kV dc voltage application can beobserved in Figure 10.
By examining Figure 10, it can be observed that the
space charge profile, while initially resembling those in
Figure 4 at short times, continues to grow in magnitude and
shift to the right (towards to the negative electrode). It is
clear that this behaviour of the space charge is in gross
contradiction with the dc space charge behaviour of sample
W, shown in Figure 5. This behaviour can be understood in
two ways. First, the build up in magnitude of the total
charge (positive and negative) can be explained through the
simple, parallel plate, charge-capacitance relationship. As
the distance between the plates (in this case, the distance
between the active space charge region and the opposing
electrode) decreases, in order to keep the voltage constant,
the total charge on the plates must increase.
The shift to the right over time of the charge grouping,
can be explained through an analysis of Poissons and the
continuity equation. By examining equation (12) it can be
seen that for a steady state charge density to prevail in a
conductively inhomogeneous medium, the electric field
must redistribute in such a way as to cancel the rate of
change, in space, of the conductivity.
Figure 10. Simulated space charge profiles for 7 kV dc voltage
application, non-linearity mechanism omitted.
This redistribution of electric field is performed by a
redistribution, or general build up, of charge. This build up
of charge takes time, depending on the time constant of the
dielectric region, therefore giving a different chargedistribution at different times as seen in Figure 10. By
including equation (12) in Poissons equation (10), the
following expression can be derived (see Appendix II):
tgradJ
=
r
(19)
It can be seen in equation (19) that at steady-state (with
regards to charge density), the static charge density will
depend on the spatially uniform current density and the
spatial gradient of the ratio of the permittivity to the
conductivity. Therefore, after a steady state develops, at
any gradient of the ratio of the permittivity to the
conductivity will yield a time-independent charge density.
In order to cease this movement of charge through the
interface, and therefore cease its build up in magnitude so
that it reaches a steady state on the order of 1.5 C/m3 as in
Figure 5, it is required that there is a negligible gradient
of the ratio of the permittivity to the conductivity beyond
the observed charge magnitude in the centre of the
sample.
The introduction of the mechanical non-linearity
mechanism solves this problem. Figure 11 shows the
evolution of the space charge density in the sample when
the mechanical non-linearity mechanism is enabled, with an
A value of 350 Pa in equation (8). It should be noted that all
simulations shown previously, apart from Figure 10, have
had the non-linear mechanism enabled with an A value of
350 Pa. Figure 12 shows the corresponding simulated
change in the conductivity profile over time.
A steady-state condition in terms of the space charge
profile is achieved after approximately 50 ms. This result
can be compared to that seen in Figure 10, whereby the
steady state condition has not developed after 5 s and with a
space charge magnitude in gross contradiction with the
Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on April 24,2010 at 07:05:33 UTC from IEEE Xplore. Restrictions apply.
8/8/2019 Development Small Sample Verfication
11/13
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 15, No. 4; August 2008 1141
results seen in Figure 5. The behaviour seen in Figure 11 is
emergent after the introduction of a mechanical non-
linearity mechanism. This mechanism acts to flatten the
conductivity (and therefore permittivity) profile in the
region beyond the space charge peak, as can be observed in
Figure 12. It can be observed that the maximum space
charge in Figure 11 reaches a steady state value on the
order of 1.5 C/m3, which agrees with the maximum dc
space charge values shown in Figure 5 for sample W.
Figure 11. Space charge evolution in small sample simulation, with a tree
bridging roughly 50% of the insulation.
Figure 12. Simulated conductivity profile evolution over time under the
action of the proposed mechanical non-linearity mechanism.
The enhanced, or non-zero, space charge profile beyond
the space charge peak in Figure 11, towards the negative
electrode, is due to the proposed mechanical non-linearity
mechanism resulting in a slight slope of the conductivity in
the x~0.6-0.9x10-3 region, instead of a completely flat
response. Unfortunately, the static dc space charge
distribution due to water tree degradation which has been
simulated in Figure 11 has not been displayed in [6] or
elsewhere in the available literature to the authors
knowledge. It has already been discussed that the dc space
charge distributions shown in Figure 8 are unlikely to be
static distributions, being measured immediately after the dc
voltage application and having the characteristics of a low
water tree conductivity. Therefore, the non-linearity
mechanism which results in Figure 11 is difficult to confirm
directly. However, it was found that without such a
mechanism, the steady state maximum charge density under
dc voltage application for sample W in Figure 5 was
impossible to replicate. This and the successfulness of the
non-linear enabled model in reproducing the measured space
charge profiles in Figures 6 and 9 give confidence to the
postulated mechanism. In addition to this, the accompanying
paper [5] extends this model and enables it to predict the
dielectric response of water tree degraded XLPE cables. In
that paper, it is shown that the model can accuratelyreproduce the non-linear behaviour of degraded cable
samples measured using Frequency Domain Spectroscopy.
4 DISCUSSION AND CONCLUSION
A point of note can be made at this point about the
previous simulations and the mechanical non-linearity
mechanism. As was stated previously, it was found through
extensive simulations that without the proposed
mathematical form of the mechanical non-linearity
mechanism, a steady state maximum space charge density
as displayed for sample W in Figure 5 was not possible to
obtain. Therefore, the preceding results give supportingevidence to the general, non-analytical, mechanical non-
linearity mechanism as proposed in [3]. The electrical
mechanism of non-linearity described by equations (5) and
(6) was also considered. Without the mechanical non-
linearity mechanism, it was found that the electrical
mechanism caused some very slight increases in
conductivity and permittivity. However, when coupled with
the proposed mechanical non-linearity mechanism, the
effects of the electrical mechanism was negligible.
A further point to be made is regarding the acceptable
range of the mechanical non-linearity turn on constant A.
This constant was varied through various values, however it
was found that only within the range of 300-500 Pa was asteady state maximum space charge density attainable that
still allowed a good comparison between the space charge
profile simulations and Figures 4 and 8. In order to be
conservative with the magnitude of the mechanical non-
linearity mechanism, the highest value in this range, 500 Pa,
has been chosen to be used in the second part of this study
[5]. The relatively low value of this constant may bring into
question the physicality of the model. However, it should be
noted that this is a macroscopic simulation, and that stresses
due to microscopic features such as geometrically sharp
voids, may act to enhance these pressures.
In conclusion, this paper has detailed the development of a
dynamic electrical finite element model for water treedegraded insulation. The model uses a diffuse boundary
between the water tree and XLPE, along with an analytical
mechanical non-linearity mechanism based on the developed
Maxwell forces within the material. By doing this, the
simulated results incorporating the mechanical non-linearity
mechanism can attain a close match to the small sample PEA
measurements performed in [6], without the need for any
applied physical variables which change with applied
voltage. Due to the success of the model, a second half of the
study was undertaken to extend the model in a practical role
Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on April 24,2010 at 07:05:33 UTC from IEEE Xplore. Restrictions apply.
8/8/2019 Development Small Sample Verfication
12/13
A. J. Thomas and T. K. Saha: A New Dielectric Response Model for Water Tree Degraded XLPE Insulation Part A 1142of interpreting diagnostic response measurements. This work
is presented in the accompanying paper [5].
APPENDIX I
Polarisation can be expressed both macroscopically and
microscopically:
uEP == )(0
'(20)
Where u is the number of dipoles operating at fieldEand
is the average dipole moment. If the simplistic 2-
dimensional potential well case is considered, as illustrated
in Figure 1, then the average dipole moment from the
hopping of carriers can be expressed as:
=+
==
=
kT
eEa
kT
eEaea
eafff
eafff
u
dq
rfe
rfe
N
i
ii
2cosh
2sinh
)(
)(1 (21)
Whereff is the probability of a hop in the direction ofthe field for an negatively charged carrier, given
thermal emission of a carrier, fr is the probability of an
hop in the opposite direction, given thermal emission of
a carrier, fe is the probability of therm-e of the carrier
and d is the distance between the dipolar charges. As
can be seen from equation (21), the simplification has
been made that all potential wells are a distance a apart
(hence d = a), and all the carriers are single charged
(hence q = e). From equations (20) and (21) then, the
polarisation for the orientation processes described
above is:
( )EkT
eEaea
kTWN
P 0'
2sinh
2
exp
=
= (22)
WhereNis the total number of possible dipoles in a dielectric
(Nis independent ofE, whereas u is dependent). Rearranging
(22), an expression for the permittivity can be obtained:
( ) 0'
2
2sinhexp
+
=E
kT
eEaea
kT
WN
E (23)
Equation (23) can be expressed without the N term byusing the low field permittivity:
( ) 0''
2sinh
2 +
=
kT
eEa
eaE
kTE LF (24)
Where the low field permittivity is:
kT
aekT
WN
LF4
exp 22
'
(25)
If the real part of the permittivity is converted to
susceptibility:
10
'' =
(26)
The constancy between the real and imaginary parts of
the susceptibility, through the Universal Relaxation Law
[29], can be evoked to give equation (6).
APPENDIX II
Poissons equation can be expressed as:
EDrr
== (27)
The conductive current density can be defined as:
EJr
= (28)
Therefore, equation (28) can be re-expressed as:
Jr
= (29)
Using the identity:
( ) ( ) ( ) AfAfAfrr
+=
Equation (29) can be re-expressed as:
JJrr
+=
(30)
Substituting the continuity equation (12) into equation(30) gives equation (19).
ACKNOWLEDGEMENTS
The authors would like to acknowledge Ergon Energy
and the Australia Research Council for the funding that
made this study possible. The authors would also like to
acknowledge the project members from the Queensland
University of Technology (QUT) for their support. In
particular, the authors would like to acknowledge Frith
Foottit, of QUT, for the development and continual
maintenance of the accelerated ageing experiment, Bolarin
Oyegoke, of QUT, for his experimental advice and support
and Prasanna Wickramasuriya, of Ergon Energy, for hissupport during the measurements.
REFERENCES[1] P. V. Notingher, I. Radu and J.C. Filippini, "Electric field
calculations in polymers in the presence of water trees", IEEE 5 th
International Conference on Conduction and Breakdown in Solid
Dielectrics (ICSD), Leicester, United Kingdom, pp. 666-670, 1995.
[2] I. Radu, J.C. Filippini, P.V. Notingher and F. Frutos, "The effect of
water treeing on the electric field distribution of XLPE.
Consequences for the dielectric strength", IEEE Trans. Dielectr.
Electr. Insul., Vol. 7, pp. 860-868, 2000.
Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on April 24,2010 at 07:05:33 UTC from IEEE Xplore. Restrictions apply.
8/8/2019 Development Small Sample Verfication
13/13
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 15, No. 4; August 2008 1143
[3] S. Hvidsten, E. Ildstad, J. Sletbak, and H. Faremo, "Understanding
water treeing mechanisms in the development of diagnostic test
methods", IEEE Trans. Dielectr. Electr. Insul., Vol. 5, pp. 754-760,
1998.
[4] S. Nakamura, T. Osaki, N. Ito, I. Sengoku and J. Kawai, "Dynamic
behavior of interconnected channels in water-treed polyethylene
subjected to high voltage", IEEE Trans. Dielectr. Electr. Insul., Vol.
9, pp. 390-395, 2002.
[5] A. J. Thomas and T.K.Saha, "A new dielectric response model for
water tree degraded XLPE insulation Part B: Dielectric response
interpretation", IEEE Trans. Dielectr. Electr. Insul., Vol. 15, pp. ,2008 (Paper 1660 in this issue).
[6] Y. Li, J. Kawai, Y. Ebinuma, Y. Fujiwara, Y. Ohki, Y. Tanaka, and
T. Takada, "Space charge behavior under ac voltage in water-treed
PE observed by the PEA method", IEEE Trans. Dielectr. Electr.
Insul., Vol. 4, pp. 52-57, 1997.
[7] A. J. Thomas and T.K.Saha, "A theoretical investigation for the
development of a water tree dielectric response model", IEEE Conf.
Electr. Insul. Dielectr. Phenomena, Kansas City, USA, pp. 369-372,
2006.
[8] E. Moreau, C. Mayoux, C. Laurent, and A. Boudet, "The structural
characteristics of water trees in power cables and laboratory
specimens", IEEE Trans. Electr. Insul., Vol. 28, pp. 54-64, 1993.
[9] R. Ross and J. J. Smit, "Composition and growth of water trees in
XLPE", IEEE Trans. Electr. Insul., Vol. 27, pp. 519-531, 1992.
[10] S. M. Moody, V. A. A. Banks, and A. S. Vaughan, "A preliminary
study of the relationship between matrix morphology and watertreeing in medium voltage polymeric insulated cable", IEEE Conf.
Electr. Insul. Dielectr. Phenomena, pp. 352-360, 1991.
[11] R. Ross, "Inception and propagation mechanisms of water treeing",
IEEE Trans. Dielectr. Electr. Insul., Vol. 5, pp. 660-680, 1998.
[12] L. A. Dissado, S. V. Wolfe, and J. C. Fothergill, "A Study of the
Factors Influencing Water Tree Growth", IEEE Trans. Electr. Insul.,
Vol. 18, pp. 565-585, 1983.
[13] J. Bezille, H. Janah, J. Chan, and M. D. Hartley, "Influence of
diffusion on some electrical properties of synthetic cables", IEEE
Conf. Electrical Insulation and Dielectric Phenomena, Victoria,
Canada, pp. 367-372, 1992.
[14] J. Hjerrild, J. Holboll, M. Henriksen and S. Boggs, "Effect of
semicon-dielectric interface on conductivity and electric field
distribution", IEEE Trans. Dielectr. Electr. Insul., Vol. 9, pp. 596-
603, 2002.
[15] T. Toyoda, S. Mukai, Y. Ohki, Y. Li, and T. Maeno, "Conductivityand permittivity of water tree in polyethylene", IEEE Conf. Electr.
Insul. Dielectr. Phenomena, Austin, USA, pp. 577-580, 1999.
[16] R. Ross and M. Megens, "Dielectric properties of water trees", IEEE
Intern. Conf. on Properties and Application of Dielectric Materials,
Xi'an, China, pp. 455-458, 2000.
[17] G. Katsuta, A. Toya, L. Ying, M. Okashita, F. Aida, Y. Ebinuma, and
Y. Ohki, "Experimental investigation on the cause of harmfulness of
the blue water tree to XLPE cable insulation", IEEE Trans. Dielectr.
Electr. Insul., Vol. 6, pp. 887-891, 1999.
[18] J. J. O'Dwyer, The Theory of Electrical Conduction and Breakdown
in Solid Dielectrics. Oxford: Clarendon Press, 1973.
[19] P. Werelius, P. Tharning, R. Eriksson, B. Holmgren, and U. Gafvert,
"Dielectric spectroscopy for diagnosis of water tree deterioration in
XLPE cables", IEEE Trans. Dielectr. Electr. Insul., Vol. 8, pp. 27-42,
2001.
[20] S. Hvidsten, E. Ildstad, B. Holmgren, and P. Werelius, "Correlation
between AC breakdown strength and low frequency dielectric loss of
water tree aged XLPE cables", IEEE Trans. Power Delivery, Vol. 13,
pp. 40-45, 1998.
[21] J. N. Reddy, An introduction to the finite element method, 3rd ed.
New York: McGraw Hill, 2006.
[22] V. N. Kaliakin, Introduction to Approximate Solution Techniques,
Numerical Modeling, and Finite Element Methods. New York:
Marcel Dekker, Inc., 2002.
[23] Y. Li, M. Yasuda, and T. Takada, "Pulsed electroacoustic method for
measurement of charge accumulation in solid dielectrics", IEEE
Trans. Dielectr. Electr. Insul., Vol. 1, pp. 188-195, 1994.
[24] T. Takada, Y. Tanaka, N. Adachi, and X. Qin, "Comparison between
the PEA method and the PWP method for space charge measurement
in solid dielectrics", IEEE Trans. Dielectr. Electr. Insul., Vol. 5, pp.944-951, 1998.
[25] Y. Li, K. Murata, Y. Tanaka, T. Takada, and M. Aihara, "Space
charge distribution measurement in lossy dielectric materials by
pulsed electroacoustic method", IEEE Intern. Conf. on Properties and
Application of Dielectric Materials, Brisbane, Australia, pp. 725-728,
1994.
[26] Y. Zhu, D. Tu;, and T. Takada, "Mathematical analysis and
interpretation of pulsed electro-acoustic system", IEEE Intern. Conf.
on Properties and Application of Dielectric Materials, Xi'an, China,
pp. 63-66, 2000.
[27] S. Mukai, Y. Ohki, Y. Li, and T. Maeno, "Time-resolved space
charge observation in water-treed XLPE", IEEE Conf. Electr. Insul.
Dielectr. Phenomena, Atlanta, USA, pp. 645-648, 1998.
[28] Y. Ohki, Y. Ebinuma, and S. Katakai, "Space charge formation in
water-treed insulation", IEEE Trans. Dielectr. Electr. Insul., Vol. 5,
pp. 707-712, 1998.[29] A. K. Jonscher, Universal Relaxation Law. London: Chelsea
Dielectrics Press, 1996.
Andrew J. Thomas (S04) is a Ph.D. candidate at
the University of Queenslands School of
Information Technology and Electrical
Engineering. He graduated from the Queensland
University of Technology in 2003 with honors and
worked in Connell Wagners power systems group
for one year. During that year, 10 months were
spent at Connell Wagners Advanced Technology
Centre in Newcastle, Australia within the high
voltage testing division. His current research
interests include insulation condition assessment
techniques, specialising in the diagnosis of water tree deteriorated XLPEcables.
Tapan K. Saha (M93-SM97) was born in
Bangladesh and immigrated to Australia in 1989. Dr.
Saha is a Professor of Electrical Engineering in the
school of Information Technology and Electrical
Engineering, University of Queensland, Australia.
Before joining the University of Queensland in 1996,
he taught at the Bangladesh University of
Engineering and Technology, Dhaka, Bangladesh for
three and half years and then at James Cook
University, Townsville, Australia for two and half
years. He is a Fellow of the Institution of Engineers,
Australia. His research interests include power systems, power quality and
condition monitoring of electrical plants.