Lic Thesis

THESIS FOR THE DEGREE OF LICENTIATE OF ENGINEERING
Application of Dielectric Spectroscopy for

Estimating Moisture Content in Power
Transformers
by
CHANDIMA EKANAYAKE
Department of Electric Power Engineering
CHALMERS UNIVERSITY OF TECHNOLOGY
Gteborg, Sweden 2003
ii
Application of Dielectric Spectroscopy for Estimating Moisture
Content in Power Transformers
CHANDIMA EKANAYAKE
CHANDIMA EKANAYAKE, 2003
Technical report no. 465L
Department of Electric Power Engineering
Chalmers University of Technology
SE-41296 Gteborg
Sweden
Telephone: +46 (0)31 772 1641
Fax: +46 (0)31 772 1633
E-mail: chandima.ekanayake@elkraft.chalmers.se
Chalmers Bibliotek, Reproservice
Gteborg, Sweden 2003
iii
Abstract
Moisture in oil-paper insulated power transformers has an immense effect on
their performance. Therefore, early identification of moist insulation is vital
for avoiding failures in transformers. Although chemical analyses of oil
provide direct information on water content, they have to be performed under
laboratory conditions. In contrast, estimations of moisture content using
on-site electrical measurements allow for obtaining this information faster.
The study presented in this report aims at further investigating the
possibilities of using dielectric response measurements for diagnostics of
power transformer insulation and special attention is paid to frequency
domain spectroscopy (FDS) measurements. FDS measurements were
performed on various transformers, both under field and laboratory
conditions. In addition, the same measurements were carried out on
pressboard samples containing different moisture contents. The results were
interpreted by means of a so called X-Y model, and the sensitivity of this
model was also evaluated. Based on the observations of this analysis a
simplified model, called the X model, was introduced. In the X model, the
presence of spacers in transformer insulation has been neglected. On the
other hand, an additional measurement of the oil sample was introduced for
estimating oil conductivity. Derived moisture contents from the X model
were compared with the corresponding moisture contents obtained from the
Karl Fischer analyses of oil samples.
It was revealed that FDS measurements can be used as a reliable tool for
estimating moisture content in transformer insulation when a consistent
database on dielectric responses of well-defined pressboard samples is
available. A reasonable agreement between the estimated and the measured
moisture contents was found. However, the estimates for distribution
transformers were higher than the corresponding values derived from the
chemical analyses. Results of the measurements on field installed
transformers belonging to the Ceylon Electricity Board (Sri Lanka) revealed
that most of the transformers used in this study were wet. Therefore,
immediate precautions were suggested for avoiding further ageing of their
insulation.
Key words: oil-paper insulation, Frequency domain spectroscopy
measurements, modelling transformer insulation, moisture, oil conductivity
v
Acknowledgement
I would like to express deep gratitude to my supervisor and examiner Prof.
Stanislaw Gubanski for his guidance, encouragement and invaluable
comments on this report. In addition the guidance of Dr. Manjula Fernando,
during my stay in Sri Lanka, is highly appreciated.
I specially thank Dr. Janake Ekanayake for encouraging me to continue my
studies in high voltage engineering.
All the field measurements were performed in the Ceylon Electricity Board
premises. So I greatly appreciate the support of CEB technical staff in
Mahaweli and Laxapana complexes. Special thanks go to Eng. Kamal
Mularachchi and Eng. Anura Herath for their invaluable support and
encouragement.
Prof. Anders Cedergren in University of Ume helped me to learn Karl
Fisher titration measurements. I really appreciate your help. Prof. Klause
Frhlich and PhD. student Wolfgang Hribernik gave me an opportunity to
make measurements on pressboard samples in the high voltage laboratory of
ETH, Zurich. My special thanks go to them.
I sincerely thank all my department colleagues for your support and keeping
a friendly working environment.
My Sri Lankan friends Vishaka, Dhammika, Siri and Chandi provided me a
good Sri Lankan environment. Thank you all for everything.
I thank my colleagues Disala and Lilantha for the good friendship and
support.
I owe my deepest gratitude to my parents, sisters, brother and brother in law
for their love and support for my studies. At last but not least many thanks go
to Charani for her understanding and encouragement, especially when I was
writing my thesis.
I appreciate the finance support for this research through Sida SAREC.
vii
Table of contents
ABSTRACT ............................................................................................................ III
ACKNOWLEDGEMENT ....................................................................................... V
TABLE OF CONTENTS ...................................................................................... VII
1. INTRODUCTION............................................................................................. 1
2. DIELECTRIC RESPONSE OF INSULATION............................................. 5
2.1 TIME DOMAIN RESPONSE .............................................................................. 5
2.2 FREQUENCY DOMAIN RESPONSE................................................................... 8
2.3 PRINCIPLES OF DIELECTRIC RESPONSE MEASUREMENTS ............................. 10
2.3.1 Frequency domain measurements...................................................... 10
2.3.2 Time domain measurements............................................................... 11
2.4 TEMPERATURE DEPENDENCY OF DIELECTRIC RESPONSE............................. 14
2.5 RESPONSE OF OIL-PAPER INSULATION SYSTEMS ......................................... 15
3. COMPONENTS AND PROPERTIES OF TRANSFORMER
INSULATION SYSTEM ........................................................................................ 17
3.1 PAPER AND PRESSBOARD............................................................................ 17
3.2 MINERAL OIL.............................................................................................. 18
3.3 OIL IMPREGNATED PAPER INSULATION....................................................... 19
3.4 MOISTURE IN TRANSFORMER INSULATION.................................................. 20
3.4.1 Moisture in paper .............................................................................. 20
3.4.2 Moisture in oil.................................................................................... 20
3.4.3 Moisture in oil-paper insulation........................................................ 21
3.4.4 Source of excessive moisture in transformer insulation .................... 22
3.5 DEGRADATION OF OIL-PAPER INSULATION SYSTEMS .................................. 23
4. CONDITION ASSESSMENT OF POWER TRANSFORMER
INSULATION.......................................................................................................... 25
4.1 CHEMICAL AND PHYSICAL ANALYSES......................................................... 25
4.2 ELECTRICAL TESTS..................................................................................... 26
4.2.1 Traditional methods........................................................................... 27
4.2.2 Dielectric response measurements .................................................... 31
5. MODELLING DIELECTRIC RESPONSE.................................................. 35
5.1 DIFFERENT MODELLING TECHNIQUES......................................................... 35
viii
5.1.1 Debye model with single and distributed time constants....................36
5.1.2 General response function .................................................................37
5.1.3 X-Y model ...........................................................................................38
5.2 EFFECT OF DIFFERENT PARAMETERS OF X-Y MODEL ON THE FINAL FDS
RESPONSE................................................................................................................45
5.2.1 Influence of oil conductivity ............................................................45
5.2.2 Influence of spacers............................................................................47
5.2.3 Variation of permittivity at 1 kHz (
1kHz
)...........................................53
5.2.4 Conclusion of the analyses.................................................................55
5.3 MODELLING USING X MODEL......................................................................55
5.4 MODELLING USING DISTRIBUTED RELAXATION TIMES ................................57
6. TRANSFORMERS AT THE CEYLON ELECTRICITY BOARD............59
6.1 INTRODUCTION ...........................................................................................59
6.2 GENERATOR TRANSFORMERS UNDER STUDY...............................................59
6.3 MAINTENANCE OF GENERATOR TRANSFORMERS AT CEB...........................61
6.3.1 Maintenance tasks carried out ...........................................................61
6.3.2 Information recording........................................................................63
6.3.3 Generator transformer replacements .................................................64
6.4 DISTRIBUTION TRANSFORMERS UNDER STUDY............................................64
7. MEASUREMENTS..........................................................................................67
7.1 INSTRUMENTATION FOR DIELECTRIC RESPONSE MEASUREMENTS ...............67
7.2 FIELD MEASUREMENTS ...............................................................................69
7.3 LABORATORY MEASUREMENTS...................................................................70
7.3.1 Laboratory transformer......................................................................70
7.3.2 Oil test cell .........................................................................................71
7.3.3 Pressboard samples............................................................................72
7.3.4 Karl Fischer titration measurements..................................................73
8. RESULTS AND DISCUSSION.......................................................................75
8.1 PRESSBOARD SAMPLES................................................................................75
8.2 DISTRIBUTION TRANSFORMER IN LABORATORY..........................................77
8.2.1 Modelling using X-Y and X model......................................................77
8.2.2 Comparison of time domain and frequency domain spectroscopy
measurements.....................................................................................................80
8.3 POWER TRANSFORMERS IN THE FIELD.........................................................82
8.3.1 Single-phase power transformers.......................................................82
8.3.2 Three-phase power transformer.........................................................92
8.4 DISTRIBUTION TRANSFORMERS IN THE FIELD..............................................93
8.5 LIMITATION ON FDS MEASUREMENTS........................................................97
9. CONCLUSION.................................................................................................99
ix
10. FUTURE WORK....................................................................................... 101
11. REFERENCES .......................................................................................... 103
Chapter 1: Introduction
1
1. Introduction
Demand for a reliable electricity supply has significantly increased during the
last few decades. Therefore, fault free operation of power systems has
become very important. However, due to the high cost of power system
components, especially transformers, it is not economical to replace them in
order to increase the reliability, by considering their age. A relatively large
number of transformers that are still working in different power systems
around the world remain in fairly good condition although they have been
used longer than their designed lifetime. Therefore, correct condition
assessment of power transformers is needed before making any conclusions
about replacements or refurbishment.
Degradation of power transformer insulation, which mainly consists of oil
and paper, is the main factor for power transformer failures. Chemical
analyses and electrical measurements are used for monitoring the condition
of transformer insulation. Among these, chemical analyses provide direct
information on parameters, such as water content, degree of polymerisation
of paper, sludge content in the oil, acidity of the oil and quantity of different
gasses dissolved in the oil. However, most chemical analyses must be
performed under laboratory conditions and for some chemical analyses
(eg. Chromatography tests) paper samples are needed. On the other hand,
electrical measurements are simpler and it is possible to perform them
on-site. Because of this simplicity, electrical tests are currently preferred for
condition monitoring of transformer insulation instead of chemical tests
although they do not provide direct information about the above mentioned
parameters.
Traditional electrical tests, such as insulation resistance (IR), polarisation
index (PI) and loss factor (tan) provide very little information about
transformer insulation since they are limited to a single value measurement.
To overcome this disadvantage, dielectric response measurements, namely
return voltage measurements (RVM), polarisation and depolarisation current
measurements (PDC) and frequency domain spectroscopy measurements
(FDS), have been introduced for condition monitoring of transformer
insulation, especially for the evaluation of water content in transformer
pressboard. In the early stages, RVM was introduced because voltage
measurements were simpler than measurements of low currents. The other
two methods, requiring current measurements, were introduced recently due
Chapter1: Introduction
2
to improvements provided by the use of sophisticated electronic devices. Not
only measuring techniques but also the interpretation of results have also
been developed. However, for most of the present interpretation techniques, a
prior knowledge of the geometrical arrangement of the insulation is required.
Most of the power utilities lack this information about transformer
construction. Therefore, practical difficulties arise when power utilities utilise
these techniques. Hence, there are still needs for improvement in the
interpretation of the results of all these techniques and, therefore, further
investigations are essential [1]. Comparisons with chemical analyses are
necessary for calibrating the correlation between the dielectric response data
and the moisture content in the insulation.
Work presented in this thesis has been carried out to find out the potential for
using frequency domain dielectric spectroscopy measurements and
conductivity measurements on oil for assessing the insulation condition, i.e.
the moisture content in the paper insulation in power transformers without
using information about geometrical capacitance. Transformers used for this
study belong to the Ceylon Electricity Board (CEB) in Sri Lanka. A number
of power and distribution transformers of different ages were selected for the
measurements. In parallel to dielectric measurements on transformers,
moisture content in the oil and the frequency dependent conductivity of the
oil was also measured by means of Karl Fischer titration (KFT) and dielectric
spectroscopy measurements of oil samples, respectively. In addition similar
measurements were performed on well-defined oil-paper samples to get a
clear understanding of the influence of different parameters on the dielectric
response of the oil-paper insulation and, later, results of these measurements
were utilised for estimating moisture contents in transformers selected for
this study.
A short description of each of the chapters in this report is given below.
Chapter 2 This chapter describes the theory behind different dielectric
response measurement techniques on insulation systems. The influence of
temperature on dielectric response of oil-paper insulation systems is
discussed as well.
Chapter 3 In this chapter, properties of oil impregnated paper insulation
systems are discussed. The sources and effects of excessive moisture in
transformer insulation are also reviewed.
Chapter 4 Here, different methods utilised for condition assessments of
transformer insulation are presented.
Chapter 1: Introduction
3
Chapter 5 This chapter reviews techniques developed for modelling
dielectric response of insulation systems. The influence of different parts of
transformer insulation on total response is also discussed. Based on this
analysis, a simple model for transformer insulation is presented.
Chapter 6 A description of transformers selected for field measurements is
given in this chapter. Present maintenance tasks carried out by CEB are also
reviewed.
Chapter 7 Field and laboratory measurements conducted are briefly
described in this chapter.
Chapter 8 Results of the measurements are discussed.
Chapter 9 Conclusions.
Chapter 10 Future work proposed.
Publications
C. Ekanayake, W. Hribernik, S.M. Gubanski, M.A.R.M. Fernando
Calibrating the results of dielectric measurements on field aged power
transformers using oil analyses and similar measurements on well-
defined pressboard samples, Accepted to publish in Nordic Insulation
Symposium 2003 (NORD-IS 03), Tampere, Finland.
C. Ekanayake, M.A.R.M. Fernando, K.Mularachchi, S.M. Gubanski
Diagnostic of power transformer insulation using frequency domain
spectroscopy Ninth annual conference of IEE Sri Lanka, Colombo,
September 2002.
C. Ekanayake, S.M. Gubanski Correlation between results of dielectric
measurements and oil analyses for a power transformer Proc. of
Electrical Insulation Conference and Electrical Manufacturing & Coil
Winding Conference 2002, Cincinnati, USA, October 15-17, 2002.
C. Ekanayake, S.M. Gubanski, K.Mularachchi, M.A.R.M. Fernando,
Diagnostic of power transformers in Sri Lanka: Application of dielectric
spectroscopy in frequency domain Proc. of Electrical Insulation
Conference and Electrical Manufacturing & Coil Winding Conference
2001, pp. 593 - 596, Cincinnati, USA, October 16-18, 2001.
Chapter 2: Dielectric response of insulation
5
2. Dielectric response of insulation
2.1 Time domain response
Maxwell equations, which describe electromagnetic phenomena, are the basis
of mathematical modelling of the electromagnetic behaviour of insulation.
0
+

B
t
B
E
t
D
j H
D
(2.1)
Here, D dielectric displacement, free charge density, H magnetic field,
j ohmic current density, E electric field and B magnetic flux density,
respectively. However, dielectric materials studied here are assumed to have
zero magnetisation. Therefore, in further considerations magnetic properties
will not be considered.
In addition to Maxwell equations, D and E are interrelated by the properties
of insulating material.
P E D +
0
(2.2)
Where
0
is the permittivity of free space and P is the polarisation vector,
which is material dependent. The insulating materials considered here are
assumed isotropic, homogeneous and linear, which allows for applying the
above equations without considering microscopical electromagnetic
behaviour of the materials.
Dielectric displacement D is often linear with the applied electric field E.
Therefore, D and E can be interrelated using a proportional constant called
relative permittivity
r
.
E D
r 0
(2.3)
Moreover, by combining (2.2) and (2.3) P and E can be interrelated as,
Chapter2: Dielectric response of insulation
6
( ) E E P
r 0 0
1 (2.4)
where, is known as the dielectric susceptibility of the material.
Also by using (2.1) and (2.2),
t
P
t
E
E J
+
0
(2.5)
Here, J is the total current density due to the applied field and is the volume
conductivity of the material. Equation (2.5) shows the contribution of
polarisation to the total current through the insulation.
Polarisation in materials can be due to several mechanisms known as
electronic, ionic, atomic, dipolar, interfacial polarisation and hopping. Of
these, the first three mechanisms are much faster than the others. The time
these mechanisms take is under 10
-13
seconds. In contrast, interfacial
polarisation and hopping are slower, which take about 10
-3
- 10
4
seconds
depending on the material. Therefore, when a constant electric field is
applied, total polarisation of the material varies, as shown in Figure 2.1.
Figure 2.1. Time dependence of polarisation when a constant electric field E
0
is applied at
t=t
0.
In Figure 2.1, P
represents polarisation due to fast mechanisms and P

S
is the
total saturated polarisation of the material (after an infinite time).
t
0 t
P=P
P=P
S
E=E
0
E=0
P, E
7
Therefore, polarisation at any given time after t=t
0
can be written as,
( ) ( ) ( )
0 0
t t for t t g P P P t P
s
+

(2.6)
where g(t) is a monotonically increasing function that satisfies the following
conditions,
( )
( ) ( )
0
0
0 & 0
1
0
t t for t g t g
and
t
t t
if
if
t g

'
`
By substituting (2.4) to each polarisation term in (2.6), total polarisation due
to a constant electric field E
0
can be expressed as,
( ) ( ) ( ) ( ) [ ]
0 0 0
1 E t t g t P
S
+

(2.7)
Where,
0
and
are static and high frequency relative permittivities of the

material, respectively.
Equation (2.7) can be extended to find the polarisation of a linear material
due to any arbitrary electric field since any arbitrary function can be
represented by sum of infinite number of step functions. By applying the
superposition principle and using convolution integral polarisation at time t
due to any arbitrary electric field, E(t) can be given as,
( ) ( ) ( ) ( ) d E t f t E t P
t
+ ) ( 1
0 0
(2.8)
Where f(t) is a monotonically decreasing function known as the dielectric
response function. The first part of (2.8) corresponds to fast polarisation
processes in the material.
By combining (2.5) and (2.8), the total current density J(t) due to a constant
electric field can be written as follows,
8
( ) ( )
( )
( ) ( ) ( ) ( )
( ) ( ) ( )
( ) t E t f t t E
t
d E t f t E
t
t E
t E t J
t
]
]
]
]
+ +
'
'
+
+

3
2
0
1
0 0
0
1
. .

(2.9)
As shown in (2.9), total current density comprises three components:
1 Current density due to conductivity of the material.
2 Instantaneous current density due to fast polarisation processes.
3 Current density due to slow polarisation processes.
Moreover, one can see that in the time domain, the behaviour of the dielectric
material is characterised by conductivity , high frequency dielectric
permittivity
and the dielectric response function f(t).

2.2 Frequency domain response
When considering time varying electromagnetic fields, it is possible to
describe them by using a single frequency sinusoidal function. Then time
varying field E(t) can be written as,
( )
t j
e E t E

0
(2.10)
The real part of this function represents the physical electric field.
Making the same assumptions, which we made to derive the time domain
response and by substituting (2.10) in (2.2) and (2.8), dielectric displacement
D(t) can be expressed in the following form,
( ) ( )

d e E t f e E t D
j
t
m
t j
m

+
0 0
(2.11)
9
by substituting t
0
= t-,
( ) ( )
t j
m
A
t j
e E dt e t f t D

0
0
'
'

.
(2.12)
Term A of (2.12) is equivalent to the Fourier transform of f(t), which is
defined as frequency dependent electric susceptibility ( ) . Where,
( ) ( ) ( ) ( ) dt e t f j
t j
o
(2.13)
Here, () and () are real and imaginary components of the complex
susceptibility, respectively. Since both of them are derived from the same
function f(t), they are interrelated by the so called Kramers-Kronig (K-K)
transformation [2].
( )
( )
( )
( )
dx
x
x
a
dx
x
x x
a
a
a

0
2 2
0
2 2
lim
2
lim
2

(2.14)
In the frequency domain, the current density ) in a dielectric material due
to an external electric field of () can be written as,
( ) ( ) ( ) ( ) { [ ] ( )
( )
( ) ( )
( ) ( ) { ( )

E j j
E j j
E j j E J
B
A

0
2
1
0
0
0 0

]
]
]
]
]
]
]
'
'
+ +
+ +
.

.
(2.15)
Where and are real and imaginary components of complex permittivity,
respectively.
10
Sections A and B of (2.15) represent the capacitive and resistive components
of the total current, respectively. Resistive current, which is in-phase with the
applied electric field, is coupled with the losses in the material. Term 1 of the
resistive current, where the conductivity term is located, is associated with
the conduction or ohmic losses due to free charge movement in the material.
Term 2 of the resistive current corresponds to the dielectric losses in the
media, which occur due to the inertia of bound charges when accelerated by
the driving field.
In the frequency domain conductivity , high frequency permittivity
and
electric susceptibility ) ( characterised the dielectric behaviour of
material.
For deriving equations in both domains, it was assumed that the insulating
material is isotropic, homogeneous and linear. Therefore, equal information
can be collected using measurements in either time or frequency domain. One
can transform the information from one domain to another by Fourier
transformation of f(t) or inverse Fourier transformation of ) ( .
2.3 Principles of dielectric response measurements
As described in the previous sections, response measurements can be
performed either in the time domain or the frequency domain. In the time
domain, two measuring techniques known as polarisation and depolarisation
current measurements and recovery voltage measurements are utilised. Both
of these techniques provide information on the conductivity and the
response function f(t). In the frequency domain, complex capacitance and the
loss factor are measured as a function of frequency. Information on the
complex permittivity ) ( and the conductivity are provided.
2.3.1 Frequency domain measurements
In this technique, slow polarisation processes in the insulation are studied by
measuring current due to a sinusoidal excitation. Since single frequency
component is considered at a time, resultant current can be written as follows
11
( ) ( ) ( ) ( )
( ) ( ) { ( )
( ) ( )

U C j
U C j C j
U j C j I

0
0

]
]
]
'
'
+ +

(2.16)
where C
0
is the geometrical capacitance and () is the applied voltage.
C() and C( ) are real and imaginary components of the complex
capacitance ). This shows possibility of calculating complex permittivity
by measuring magnitude and phase angle of the response current when
geometrical capacitance is known, but it is not sufficient to distinct ohmic
and dielectric losses in the insulation. However, at very low frequencies the
resistive part is often dominating. When this happens () has a slope of 1
in the log frequency scale and () does not vary with frequency. But when
hopping mechanisms are present, it is hard to find this type of behaviour. The
other possible way of separating these two losses is by calculating dielectric
losses using the K-K transformation of (). Results of this technique
depend on the size of the measured frequency window.
When the geometrical capacitance is unknown, the frequency dependent loss
factor tg() can be used to present the measured results.
( )
( )
( )

tg (2.17)
However, in such a case information on important dielectric parameters
( ) ( , and
) cannot be obtained since the loss factor is a ratio of real

and imaginary components of complex permittivity.
2.3.2 Time domain measurements
Polarisation depolarisation current measurement
Polarisation and depolarisation current measurements can be used to analyse
the slow polarisation processes in insulation materials. When a fixed dc
voltage (U
0
) is applied across a totally discharged material having the
geometrical capacitance of C
0
, the resultant current can be given as,
12
( ) ( ) ( )
0 0 0
0
0 t t for U C t f t t I
pol
< <
'
'
+ +

(2.18)
Here, the contribution of the delta function (t) is only at t=0. Therefore, the
polarisation current consists of two major components, related to the
conductivity and the response function f(t).
After a certain time t
0
, the applied voltage is removed and at the same time
the material is short-circuited. The resultant current due to reorientation of
polarised species can be expressed as,
( ) ( ) ( ) ( ) { < < + +

t for U C t t t f t f t I
depol
0
0 0 0
(2.19)
Measured polarisation and depolarisation currents have opposite signs, but
for convenience, when they are plotted, only the magnitudes are considered.
As shown in (2.19) depolarisation current does not have a conductivity term.
Therefore I
depol
can be used to calculate the response function by disregarding
the contribution of the delta function.
( )
( )
( )
0
0 0
t t f
U C
t I
t f
depol
+ + (2.20)
When the test object is charged for a sufficiently long time (t
0
), i.e. at least
5 - 10 times as long as the measurement time of depolarisation current,
( ) ( ) 0
0
> + >> t for t t f t f (2.21)
Therefore one can assume,
( )
( )
0 0
U C
t I
t f
depol
(2.22)
The response function of many solid dielectrics decreases slowly with time
resulting long measuring times to fulfil the requirement given in (2.21). In
such cases, (2.20) can be used to compute the response function with suitable
numerical calculations.
Measured polarisation and depolarisation currents can be used to estimate the
dc conductivity of the measured object from,
13
( ) ( ) { ( ) 0
0 0
0 0
0
> + + t for t t f t I t I
U C
depol pol

(2.23)
In solid materials, it is difficult to distinguish the influence of conductivity
and the dielectric polarisation on I
pol
if the charging period is not sufficient. It
is recommended to charge the object until the effect of the dielectric response
function has disappeared or f(t+t
0
) << /
0
, to obtain true conductivity.
Recovery voltage measurement (RVM)
Figure 2.2. Voltage and current variation during RVM measurement.
In this method, time dependent voltage measurements are utilised instead of
current measurements. First, a step voltage U
0
is applied across fully
discharged insulation for a period of t
c
. During this stage, the polarisation
current I
pol
flows through the insulation. The polarisation of particles having a
relaxation time less than t
c
, and free charge movement cause this current.
After the time t
c
, the material is short-circuited for a period of t
d
. Because of
the reorientation of the polarised particles, the depolarisation current I
depol
flows. During the period t
d
, species, which have a relaxation time less than t
d,
are totally relaxed. After t
d
, the short-circuited connection is opened and
voltage across the insulation is measured. Relaxation of remaining polarised
particles develops a voltage across the insulation. This voltage re-polarises
2
t t
R
dt
dU

charging
short circuit
open circuit
U=U
0
t
I
pol
I
depol
U=U
R
(t)
t
c
t
d
t
m
U=U
max
t=t
2
14
some of the species in the insulation, which yields some difficulties when
interpreting RVM data because the relaxation of polarisation processes in the
insulation is integrated over the dispersive material. During RVM
measurement, current through the insulation is zero. Therefore, the following
relation can be used to interpret the RVM result,
( )
( )
( ) ( ) {
( ) ( )
( ) 0
0
2
2 0
1 0 0 0
2

< <
+ + +
t t U
t t for d U t f
dt
d
t t f t f U
dt
t dU
t U
R
t
t
R
R
R

(2.24)
Resolving the conductivity , high frequency relative permittivity
and the
dielectric response function f(t) from the measured recovery voltage are
difficult compared to the previous methods. The formal way of doing this is
by assuming an analytical form of f(t) and then minimising the current given
in (2.24). This methods works well with systems that have a known shape of
f(t), which can be expressed using a simple parameterised analytical function.
2.4 Temperature dependency of dielectric response
The dielectric response of insulation is not only a function of frequency or
time but is also a function of temperature. However, for most of the
materials, the spectral shape of the response does not change with
temperature, at least over a temperature range during which the structure of
the material does not alter significantly [2]. This allows for normalising time
or frequency dependent spectra for different temperatures by shifting the
corresponding spectra until they coincide into a single curve, which is called
a master curve [2-4]. The master curve contains more information than a
single temperature measurement since it covers a wider range of frequency or
time span compared to a single measurement. For some dielectric materials, a
shift in the spectral function due to a change in absolute temperature from T
1
to T
2
can be expressed with an Arrhenius factor as,
,
`
.
|
,
`
.
|

1 2
2 1
1 1
exp ) , (
T T K
E
T T S (2.25)
15
Here, E is the activation energy and K is Boltzmans constant. Furthermore,
for most Arrhenius activated systems, this shift is along the frequency axis
and when the frequency is on a log scale it is a constant shift, independent of
frequency.
( ) ( )

,
`
.
|

2 1
2 1
1 1
log log
T T K
E
shift (2.26)
Where,
1
and
2
are any two frequencies corresponding to the same
magnitude of the spectral functions at T
1
and T
2
, respectively.
The master curve technique is applicable for both real and imaginary
components of the frequency dependent complex capacitance, since both of
them contain similar information. However, first the contribution of the high
frequency permittivity
and dc loss /
0
should be subtracted.
In the time domain, a corresponding shift of dielectric response function, to
obtain the master curve, is along a line with a -1 slope when it is plotted on a
log-log scale.
The temperature dependence of the dc conductivity can also be
characterised with an Arrhenius factor with the activation energy E
dc
and the
pre-exponential constant
0
as follows,
,
`
.
|

KT
E
dc
exp
0
(2.27)
where T is the absolute temperature.
2.5 Response of oil-paper insulation systems
The dielectric response of an oil-paper insulation system is affected by the
way the components are combined. Therefore, the dielectric response of such
insulation systems reflects the properties of each material, as well as the
geometrical arrangement of insulation materials. When the combined
insulation is subjected to an electrical stress, charge accumulation occurs at
the interface due to the difference in conductivity of the two materials. This
phenomenon is known as interfacial polarisation or the Maxwell-Wagner
effect [2].
16
The dielectric dispersion of mineral oil can be neglected in the frequency
range (< 1000 Hz) of interest. Therefore, the dielectric response of oil can
simply be described by a constant permittivity value (
r
= 2.2) and the dc
conductivity
,
which is dependent on the presence of ionic contaminants in
oil. The temperature dependence can be characterised by selecting its own
activation energy. On the other hand, the response of the pressboard and
paper is characterised by their dielectric response function, which is strongly
dependent on the presence of moisture and other ageing products in the
insulation.
Nevertheless, it has been shown that both oil and pressboard dominate in the
total response at different frequencies [5], when the oil channels are in series
with pressboard. Dominant influences of pressboard and oil at different
frequencies on the total response is shown in Figure 2.3.
10
-4
10
-2
10
0
10
2
10
4
10
-4
10
-2
10
0
10
2
frequency (Hz)
t
a
n

pressboard
dominates
oil dominates
pressboard
and geometry
dominates
Figure 2.3. Example of variation of loss factor of oil-paper insulation with frequency and
dominant influences.
When the geometry of insulation and the dielectric response of each material,
i.e. oil and paper, are known, it is possible to calculate the dielectric response
of the complex structure. This is the basis of some of the modelling
techniques described in Chapter 5.
Chapter 3: Components and properties of transformer insulation system
17
3. Components and properties of transformer
insulation system
This overview is focused on the insulation systems of oil impregnated power
transformers, where Kraft paper is used as the conductor insulation whereas
pressboard and ducts filled with mineral oil are mainly used as the main
insulation.
3.1 Paper and pressboard
Kraft paper is a key component of transformer insulation. It is a low cost base
material obtained from wood pulp with outstanding mechanical and electrical
properties. Paper is composed of 90 % cellulose, 6 7 % hemi-cellulose and
3 4 % lignin [6, 7].
Long cellulose fibres provide high mechanical strength to paper. However to
obtain higher density, which gives higher electric impulse strength and higher
dielectric permittivity, the length of fibres should be shorter [8]. Therefore, to
achieve good electrical properties a compromise is needed at the expense of
weakening mechanical strength. Although the density of cellulose fibres is
about 1.55 g/cm
3
, the maximum density of paper is about 1.15 g/cm
3
. This
reduction in density is due to the porous nature of paper. Because of its
fibrous structure, paper can bear higher mechanical stresses in the
longitudinal direction than in the transverse direction. The longitudinal
tensile strength varies between 40 150 N/mm
2
, whereas transverse tensile
strength lies between 25 85 N/mm
2
[9].
The important dielectric properties of paper, which stand for its insulation
quality, are dielectric permittivity , loss factor tan and conductivity . The
permittivity of dry paper varies from 1.5 to 3.5 and the loss factor varies
between 0.003 and 0.004. Dry paper has a very high volume resistivity, i.e.
values between 10
15
and 10
17
cm can be obtained [10]. However, paper has
to be protected from direct contact with moisture to maintain its good
dielectric properties, due to the high affinity of paper to water.
There are areas in transformers where the electrical and mechanical stresses
are high and these cannot be withstood by single layers of paper. Therefore,
pressboard is used instead. Pressboard is produced by wet pressing several
paper layers without any bonding material. The density of pressboard can
Chapter3: Components and properties of transformer insulation system
18
reach up to 1.3 g/cm
3
due to its low porosity, compared with that of paper.
Also the permittivity of pressboard is higher than the permittivity of paper,
i.e. around 4.5 [11].
3.2 Mineral oil
In the majority of power transformers, mineral oil is used as the insulating
liquid due to its wide availability and its low cost compared to other
equivalent products, like silicon oil and organic esters. The main purpose of
using mineral oil is to impregnate the paper insulation to avoid direct contact
between paper, air and moisture. The oil also acts as a heat transfer medium,
which controls unnecessary rises in the temperature of the apparatus due to
losses in conductors, core and dielectric materials [11, 12].
Transformer mineral oil is refined from hydrocarbons collected during the
distillation of crude petroleum. Characteristics of the refined oil depend on
the relative proportions of paraffinic, napthenic and aromatic hydrocarbons.
Of these the aromatic component is chemically unstable compared with the
two other, due to its unsaturated chemical structure. Therefore, aromatic
components in mineral oil have to be minimised in the final product using
appropriate refining methods. Depending on the refining method, the
proportions of paraffin, naphthene and aromatics can vary between 40
60 %, 30 50 % and 5 20 %, respectively [12, 13]. All these hydrocarbons
have little or no polarity. Apart from the above named constituents one can
find trace quantities of polar compounds (sulphur, oxygen and nitrogen) and
ionic species (organic salts) present in mineral oil. These constituents have an
immense influence on chemical and electrical properties of the mineral oil
[14].
To obtain better performance as a cooling medium and as electrical
insulation, transformer oil must have high dielectric strength, low viscosity,
high heat capacity and a low expansion coefficient. It should also be free
from moisture, gasses, chemical impurities and mechanical contaminants to
avoid unnecessary electrical discharges. Therefore, different standardised
measurement techniques that measure the above listed parameters, are used
to identify the suitability of the oil for use as transformer oil. Table 3.1 shows
the accepted limits of electrical properties for new mineral-based transformer
oil [13].
19
Table 3.1. Electrical properties of new transformer oil.
Property Standard Value
Breakdown strength
(~20 C)
IEC 156 ~30 kV/mm
Dissipation factor
(90 C)
IEC 247 <0.1 %
Relative permittivity
(50 Hz)
IEC 247 2.2
Volume resistivity
(~20 C)
IEC 247 10010
12
cm
Apart from the above mentioned electrical properties, one of the other
important parameters is water solubility in oil, since free water can create an
easy path for electrical discharges. This is discussed in detail in Section 3.4.
3.3 Oil impregnated paper insulation
Paper itself shows modest dielectric performance due to its porous structure.
When air is present, the dielectric strength of paper is predominantly
determined by the gaseous ionisation within the air space. Heat, water and
oxygen accelerate the degradation of paper insulation [7]. To control the
influence of the above mentioned critical factors, it is necessary to
impregnate paper insulation.
The impregnation of paper insulation is a sophisticated process. However,
present technology does not allow for achieving 100 % perfect impregnation
[8]. Imperfect impregnation may cause damage to the insulation during
service. This may happen due to the partial discharges through the
gas-trapped voids between paper layers or paper fibres.
Most of the electrical properties of impregnated paper, except for the
breakdown voltage, are closely dependent on the electrical properties of dry
paper, since the proportionality of oil is small compared with paper and these
two materials are chemically inert with each other.
20
3.4 Moisture in transformer insulation
Moisture in transformer insulation is one of the key factors that determines
the condition of the insulation. Therefore, in the following section moisture
distribution and its influence on transformer insulation are discussed.
3.4.1 Moisture in paper
Water in paper can be found in four states: as water molecules adsorbed to
surfaces, as vapour, as free water in capillaries and as imbibed free water [15,
16]. Moisture migration and adsorption in paper is due to the diffusion of
water molecules through the porous structure of paper and by attraction of
water molecules to active sites or surface polar groups. Diffusion is quite a
slow process that determines the time needed for an equilibrium moisture
distribution. The amount of water in oil impregnated paper mainly depends
on the temperature and the water vapour pressure (or the relative humidity) in
the state of the equilibrium. Water content in paper increases with increasing
water vapour pressure while it decreases with increasing temperature.
However, there is a significant difference in the sorption of water at a given
temperature and a relative humidity in pressboard and kraft paper, i.e. the
moisture content in pressboard is bit higher than the corresponding moisture
content in paper [16, 17].
To maintain the desired electrical and mechanical properties of paper, it is
necessary to minimise the moisture content in it. During the manufacture of
transformers, paper insulation is dried out, which brings down the moisture
content to a value between 0.5 1 %. During the operation, the moisture
content in paper has to be maintained below 2.5 % [17, 18]. A 0.5 1 %
increment of moisture decreases the lifetime of insulation by roughly a factor
of 2. Moreover, high moisture content, up to a certain limit, increases the
paper conductivity, which leads to high dielectric losses [19].
3.4.2 Moisture in oil
In mineral oil, water can exist in three states [16, 20]. Most of the water
resides in the dissolved state. Water also exists in the oil as tightly bound to
oil molecules, especially in deteriorated oil. In addition to these two states,
water in oil can be found as free drops when moisture in oil exceeds the
water saturation limit of the oil. Among the above mentioned three states, the
most critical state is the free water state since it provides easy breakdown
21
paths for electrical discharges. Water content in oil is directly proportional to
the relative saturation, up to the saturation level [21]. The saturation
solubility of water in oil, W
s
can be expressed in form of Arrhenius ,
,
`
.
|

T
B
A W
s
exp (3.1)
where the saturation solubility of water in oil, W
s
is in parts per million (ppm)
and T is the absolute temperature. Coefficients A and B are basically the same
for most unaged transformer oils, but may be slightly different for some
products due to differences in aromatic contents. Measurements of differently
aged transformer oils indicated that there were no significant differences in
moisture solubility among differently aged oils [20]. Table 3.2 shows the
water saturation solubility in mineral oil at 20 C and 100 C presented by
different authors [13, 22, 23].
Table 3.2. Water saturation solubility in oil presented by different authors.
Water saturation solubility in oil (ppm) Temp. (C)
Oommen [22] Griffin [23] Fofana [13]
20 53 56 45
100 880 777 650
As shown in Table 3.2, the water saturation limit in oil at 20 C is more than
10 times lower than that at 100 C. Therefore, water droplets can be formed
during a sudden cooling of a transformer, since the rate of diffusion of
moisture in paper is slow compared with the rate of moisture release from the
oil.
Moisture in oil has an immense effect on the dielectric strength of the oil,
when the distance between electrodes is shorter than 1 cm. On the other hand,
in saturated hydrocarbons, the effect of free water on conductivity is less than
10
-12
S/m [24, 25].
3.4.3 Moisture in oil-paper insulation
Paper has a higher affinity to water than oil. Therefore, in oil-paper insulation
systems, the moisture resides mainly within the paper part of the insulation.
22
Moisture distribution in oil-paper insulation is highly temperature dependent.
At high temperatures, the affinity of paper to water decreases and, at the same
time, the water saturation level of oil increases.
Under stable conditions, that is to say under constant temperature and vapour
pressure, moisture distribution between oil and paper is in an equilibrium
state. Several sets of moisture equilibrium curves for oil-paper systems can
be found in the literature [16, 22, 23, 26]. These curves have been derived
using different measurement techniques, data sources and generating
methods, which have resulted in differences between the curves. By using
these curves, it is possible to estimate the moisture content in paper by
measuring the moisture content in oil at a known temperature. In IEEE std
62 - 1995 for estimating the moisture content in pressboard, a correlation
multiplier is introduced instead of equilibrium curves [18]. In all these
techniques, the estimated moisture content can be biased since it is practically
hard to reach the equilibrium state in operating transformers. This problem
can be overcome, to some extent by locating moisture sensors in the suitable
positions of pressboard structure.
Bubble formation is one of the critical effects of moisture in oil-paper
insulation, and may occur when the moisture concentration in paper exceeds
3.5 % during actual overloading conditions of a transformer [27].
3.4.4 Source of excessive moisture in transformer insulation
Moisture in power transformer insulation continuously increases under
service conditions. There are three main sources that can produce excessive
water in power transformer insulation.
Direct moisture ingress from the atmosphere is the main source of water in
transformer insulation. This may take place when the insulation is directly
exposed to air during installation and repair. The other two main ways of
moisture entering from the atmosphere to transformer insulation are the
viscous flow of moisture through poor seals and water migration through
open breathers (about 0.1-0.2 % per annum) [17, 27].
Residual moisture trapped in thick structural components, such as wood and
plastic resin impregnated materials are also the sources of excessive water,
since residual moisture can remain in these structures due to their longer
drying time compared with pressboard. Heat generated during the service
drives this excessive moisture to other thin insulation structures [17].
23
The other main source of excessive water in transformer insulation is the
decomposition of paper. Due to thermal stresses acting on paper insulation,
high molecular weight cellulose chains in paper undergo scission reaction in
which water and furanic compounds are formed as by-products. The amount
of water produced from the decomposition of paper varies with the actual
condition of the insulation [7, 28].
3.5 Degradation of oil-paper insulation systems
Current designs of power transformers are optimal in size. Therefore, the
influence of the condition of transformer insulation on the correct operation
of the unit is critical. Changes in electrical, mechanical and chemical
properties of the insulation due to degradation phenomena during service can
lead to severe damage. The rate of degradation of oil-paper insulation
depends on the existing thermal, oxidative, hydrolytic, electrical and
mechanical conditions within the transformer [28].
High molecular weight cellulose chains undergo scission reactions during
service. These reactions decompose the cellulose chains to low molecular
weight chains, which yield the deterioration of the mechanical and electrical
properties of paper [7].
Heat, water and oxygen accelerate the decomposition of cellulose; all of these
are present in operating transformers. Degradation is a chemical reaction, it
obeys the Arrhenius law of reaction kinetics. However, above 140 C,
degradation is faster due to a change in the degradation mechanism [7, 29].
According to several studies, the rate of paper degradation is proportional to
water content in the insulation. The presence of oxygen increases the
degradation rate of paper by a factor of 2.5 [7, 29, 30].
Not only the paper but also the properties of oil are affected by the presence
of oxygen. Oxidation of napthenic and parafinic hydrocarbons increases the
acid number, whereas oxidation of aromatics remarkably increases the loss
factor [31, 32]. The rate of oxidation of oil highly depends on temperature,
the presence of light and catalysts. At elevated temperatures and with the
presence of active metals, such as Cu, Pb and their alloys, rapid oxidation of
oil can be experienced [31].
Ionisation is the other important process that increases the ageing of oil-paper
insulation. During initial processing of oil-paper insulation, it is not possible
to remove whole trapped gases within the insulation. However, during
24
service, these local areas, where air bubbles are trapped, may be stressed by
high electric fields. The field strength in the bubble can be higher than the
breakdown strength of air and can lead to discharges within the insulation.
Oil molecules absorb energy from these discharges and decompose into
hydrocarbons and hydrogen. If oil is saturated with gases, this hydrogen may
create more bubbles, which develop more discharging paths within the
insulation. Therefore, the breakdown strength of insulation decreases with
time. Also discharges initiated in cavities of paper insulation produce
conductive carbon particles [12, 32].
Chapter 4: Condition assessment of power transformer insulation
25
4. Condition assessment of power transformer
insulation
The condition assessment of power transformer insulation is essential since
the degradation of transformer insulation is unavoidable. Mainly chemical
analyses and electrical measurements are used for the condition assessment
of transformer insulation.
4.1 Chemical and physical analyses
Chemical analyses provide direct information about the actual condition of
the insulation. Either a paper or oil sample taken from the interested oil-paper
insulation is used for these analyses.
To assess the quality of the paper insulation within transformers, one makes
use of a factor called the degree of polymerisation DP, which represents the
average number of glucose rings in the cellulose polymer. DP is determined
by measuring the intrinsic viscosity of a paper solution in an appropriate
solvent. Original kraft paper has a DP of about 1100. After factory
processing, the DP drops down to a value about 750 and then decreases with
time. A limit of DP equal to 250 is often used as corresponding to the end of
the lifetime of paper insulation [33]. Apart from DP measurements, ion
chromatography tests are also used to characterise the thermal degradation of
paper insulation by determining various sugars (monosaccharides,
polysaccharides and anhydrous sugars) as described in[6]. However, all these
tests require a paper sample, which means that the transformer must be taken
out of service, and that steel covers have to be opened. Consequently, taking
paper samples might be deleterious to the transformer.
Oil analyses are regularly used for the condition assessment of transformer
insulation, since obtaining a sample of transformer oil is much easier.
Table 4.1 shows different physcio-chemical analyses of transformer oil and
the parameters determined by these tests.
Among the tests shown in Table 4.1, the results of Karl Fischer titration
(KFT), dissolved gas analyses and high performance liquid chromatography
can also be used to predict the condition of paper insulation.
26
Table 4.1. Different physcio-chemical analyses on transformer oil.
Test name Parameters / properties
Karl Fischer titration Water content in the oil
Neutralisation number Acidity of the oil
Dissolved gas analyses Amount of different gasses i.e. H
2
,
CH
4
, C
2
H
6
, C
2
H
4
, C
2
H
2
, CO, O
2
High performance
liquid chromatography
Furanic compounds in mineral oil
Colour Deterioration and contaminants
Interfacial tension Interfacial tension of oil against
water
For finding the amount of moisture content in transformer oil, and thereafter
in the paper, coulometric KFT is widely used. This technique provides a
better resolution compared with volumetric titration when measuring low
moisture contents [34]. The water content in paper is predicted using the
moisture equilibrium curves between oil and paper [16]. It is important to
keep the percentage of moisture saturation of oil and the moisture in paper
under 30 % and 2.5 %, respectively. This may help to increase the lifetime of
a transformer [18].
Most of the chemical analyses must be performed under laboratory
conditions, which is a major draw back of these techniques.
4.2 Electrical tests
Electrical measurements, used for the condition assessment of transformer
insulation are simpler compared to the chemical analyses and it is possible to
perform them on-site. Because of this simplicity, electrical tests are currently
preferred for monitoring the condition of transformer insulation rather than
chemical tests. However, electrical tests do not provide direct information
about the constituents of the insulation system. Therefore, results of the
electrical measurement must be calibrated.
Different electrical tests and their merits and demerits are discussed in the
following section.
27
4.2.1 Traditional methods
Insulation resistance, the polarisation index and loss factor are the widely
used traditional on-site electrical tests for assessing the condition of
transformer insulation. Partial discharge measurement is also a traditional
measuring technique for on-site and on-line insulation diagnostics, which has
been continuously improving.
Insulation resistance (IR)
Insulation resistance measurement is one of the conventional methods used
for determining the dryness of the transformer insulation. IR is measured by
applying a fixed dc voltage, usually 0.25-5 kV, across the insulation. The
resultant current, which is a combination of capacitive current, absorption
current and conduction current, monotonically decreases. Therefore, it is hard
to measure true dc resistance, especially on new transformer insulation where
the true dc level is reached only after a few hours. Usually the resistance is
measured after 1 minute from the moment the dc voltage has been applied.
The temperature of the insulation influences the IR. It has been observed that
an increase in temperature by 10 C will approximately decrease the IR by
half. Therefore, it is important to note the temperature of the insulation when
performing the IR measurement [18].
In this technique, no absolute values are defined as acceptance limits. Instead,
the IR values have to be compared with values from previous measurements
of the same transformer or from measurements of the same type of
transformers in order to evaluate the actual condition of the insulation. IR
measurements give a good indication of whether or not the insulation is wet
and contaminated, but it is rather difficult to identify partially wet insulation.
It is recommended to use a guard electrode in IR measurements to avoid the
influence of unwanted leakages, e.g. leakage current through bushing
insulation. If the measuring system does not provide a guard electrode then
the surfaces of bushings must be cleaned well before the measurement.
Results influenced by leakages can lead to wrong conclusions, especially
when the measured IR is low. Another reason for not totally relying on this
method is that the IR of a poor insulation might appear higher than that of a
good insulation, if it is not measured long enough and the total current varies
as shown in Figure 4.1. Therefore, when the IR is low, it is recommended
that other diagnostic tests be performed, too [4, 18, 35].
28
However, among utility companies this method is still very popular because
of its simplicity and lower equipment cost compared with other
instrumentation.
Figure 4.1. Variation of current during IR measurement on two different transformers.
Polarisation index (PI)
PI is an extension of IR measurement. In this technique, insulation resistance
is measured at two different times (i.e. at 1 minute and 10 minute intervals)
after the voltage has been applied. By definition, PI is the ratio of insulation
resistance at 10 minutes to that at 1 minute. PI measured on power
transformer multi-layer insulation is strongly influenced by macroscopic
interfacial polarisation and the geometrical composition of oil and
pressboard. However, PI is less temperature dependent than IR, since it is a
ratio of two values at a certain temperature.
In the present interpretations of PI, it is indicated that PI is greater than unity
for a good insulation system. However, when the charging current of the
insulation varies, as shown in Figure 4.2, it is impossible to distinguish
between the PI values of good and bad insulation. Transformers TF1 and TF2
have very close PI values although the insulation of TF1 is better than that of
TF2. Therefore, the reliability of this technique is limited. It is also not
possible to use this method for predicting the moisture content or
conductivity of the insulation.
I
tot
TF
2
1 min
I
1
I
2
t
After 1 min.
I
2
> I
1
Therefore
IR
2
< IR
1
TF
1
29
Figure 4.2. Variation of current during IR/PI measurements on two different transformers.
Loss factor (dissipation factor, tan)
Loss factor measurement is a traditional electrical measurement used for
identifying power loses in high voltage insulation under an alternating
voltage.
The loss factor of insulation is defined as the ratio between resistive and
capacitive currents caused by an ac voltage applied across the insulation
(Figure 4.3). Subsequently, the total loss of insulating material is
characterised by the loss tangent.
Figure 4.3. Phasor representation of insulation current.
Usually the loss factor is measured at power frequency, that is to say at 50 Hz
or 60 Hz depending on the operational frequency of the system. Nevertheless,
in some measuring systems frequencies close to the power frequency are
V I
res
I
cap
I
tot
cap
res
I
I
tan
I
tot
TF
1
TF
2
1
t (min) 10
30
used as the measurement frequency for eliminating power frequency
disturbances while taking on-site measurements. For example in the
literature, we can find an empirical relation between loss factor at 80 Hz and
50 Hz for a specific type of current transformer [32]. But this method should
be developed further if it is to be applied to other transformers.
In conventional loss factor measurement, a Schering bridge coupled with a
high voltage ac source is used. Test voltage used in typical field test sets
varying broadly (10 V-12 kV). A voltage lower than the rated voltage of the
instrument is used for loss factor measurement.
Loss factor is sensitive to temperature and varies with frequency. Hence, for
loss factor to be a meaningful parameter, it is necessary to provide
information on the frequency of applied voltage and the temperature of
insulation during measurements. Apart from this, the loss factor depends on
the geometrical composition of the oil-paper insulation. Because of this, the
loss factor of different insulation systems cannot be compared.
When partial discharges occur there is a tip-up in the loss factor versus
voltages curve [36]. The Garton effect is another phenomenon, which on the
other hand, reduces the loss factor with increase in test voltage [37].
Partial discharge measurements (PD)
Major sources of partial discharge activity on power transformers are defects
caused by the mechanical deformation of windings, deterioration and ageing
of the components and defects of the insulating structure of tap changer.
Partial discharges within transformer insulation can be detected using either
electrical or acoustic methods. Electrical methods are more sensitive
compared with acoustic methods, but it is rather hard to obtain good
sensitivity with these methods due to other sources of electric noise,
especially corona discharges. Acoustic methods are simple and they detect
mainly arcing processes within the transformer. A good knowledge of the
internal construction of the transformer is needed to correlate the results of
PD measurements with the deterioration mechanism. One of the advantages
of this technique is that it can be used as an on-line diagnostic tool. This
allows for identifying the faults and defects of insulation at an early stage
[36, 38].
31
4.2.2 Dielectric response measurements
The major disadvantage of the first three techniques (IR, PI and loss factor)
described in the previous chapter is that they do not provide sufficient
information about the condition of insulation, which is necessary for a
reliable condition assessment. Dielectric response measurements provide
more information compared with the above methods. Though these
techniques were recently introduced for assessing the condition of
transformer insulation in the field they have been used for many years,
especially for laboratory measurements.
Polarisation and depolarisation currents (PDC)
Polarisation and depolarisation currents measured between low voltage (LV)
and high voltage (HV) windings of oil-paper insulated power transformers
have been used for assessing the condition of the insulation [39-42]. For
measuring the polarisation current, fixed dc voltage (over 500 V) is applied
across the separately short-circuited LV and HV terminals. Then the
depolarisation current is measured by short circuiting the HV and LV
terminals through an electrometer after dc voltage has been removed. A
guard terminal is also exploited for avoiding the influence of leakage
currents. Presently, computer controlled PDC measuring systems are
available for on-site measurements [43, 44].
The major advantage of this method is that more information can be gathered
compared with the traditional IR and PI measurements. In-between two
consecutive PDC measurements, the end terminals have to be short-circuited
for a sufficiently long time to reduce memory effects. A simple rule of thumb
on time domain measurements is that the short-circuit should last for at least
as long as the previous charging time before starting a new measurement.
Very sophisticated noise suppression methods are needed for PDC
measurements, which is another draw back to using this method as an on-site
diagnostic test.
Recovery voltage measurements (RVM)
As described in Chapter 2, RVM is the other time domain technique for
studying slow polarisation processes in insulating materials. When it is
utilised for identifying the condition of transformer insulation, most often
several measurements between HV and LV terminals are conducted for
different charging times while the ratio t
c
/t
d
is maintained at 2 (Figure 2.2)
32
[45, 46]. Typically, a voltage between 500-2000 V is applied as the charging
voltage.
The typical measurement set-up used for RVM is shown in Figure 4.4.
Figure 4.4. Schematic diagram of RVM.
In a study conducted by Hungarian scientists [45, 46] a variation of the
central (dominant) time constant used for estimating moisture content in
paper insulation. When t
c
/t
d
is 2, the central time constant is the charging time
t
c
where the maximum recovery voltage is highest. As mentioned in [45, 46]
the central time constant decreases significantly with the ageing of insulation.
Although this interpretation is very simple, the predicted moisture content in
paper insulation is often higher than the actual amount [1, 47]. In another
approach, secondary time constants, corresponding to subsidiary peaks of
U
max
versus t
c
curve, are used for estimating the moisture content in paper
[47]. Here, a plot of the initial slope of the recovery voltage curve against
the maximum recovery voltage U
max
, which is known as Guuinic
representation, is used to confirm the oil peak and for identifying the
presence of secondary time constants above the dominant time constant.
These subsidiary peaks are considered as reflections of polarisation
phenomena of solid insulation [47].
Dielectric spectroscopy in frequency domain (FDS)
In the FDS technique, a known sinusoidal voltage is applied across the
insulation. The measurement is repeated for several frequency sweeps.
Usually these measurements are carried out from high frequency to low
frequency for minimising memory effects. For increasing the reliability of
measurements, at least two voltage cycles are applied at each measurement
frequency. Therefore, a rule of thumb is that the total measuring time, which
is a critical parameter for on-site measurements, is equal to four times the
length of the period of lowest frequency [41].
Voltage
source
SW1
SW2
Test
object
Volt
meter
33
The major advantage of FDS measurements is that the inherent small
bandwidth of electronic components is relatively insensitive to interference
and therefore a high voltage power supply is not necessary. It is also possible
to avoid measurements at power frequency by selecting a suitable frequency
sweep. The other advantage is the possibility of utilising three terminal
configuration, which excludes the unnecessary leakage current [41].
Chapter 5: Modelling dielectric response
35
5. Modelling dielectric response
Although the dielectric response function reflects the condition of insulation,
it is rather difficult to identify the actual state of the insulation (e.g. the exact
amount of moisture in the insulation) only by observing the relevant curves,
especially when the insulation is a complex combination of different
materials. This problem can be solved, to some extent, by comparing the
dielectric response of insulation under study with the dielectric response of
well-defined similar types of insulation samples. The dielectric response of
insulating materials can be expressed as an analytical function of time or
frequency, which offers possibilities for fast and easy mathematical
manipulations in such comparisons. In addition, when the insulation is a
complex combination of two or more materials, modelling the geometrical
arrangement of insulation structure provides better comparisons between the
responses of the insulation and the models.
5.1 Different modelling techniques
Two different approaches to finding an analytical function, which represents
the dielectric response of insulation, have been used in [4, 35, 48-50].
In one of these approaches, called the functional approach, [4, 48-50] an
analytical expression is fitted to a time or frequency dependent dielectric
response by selecting suitable coefficients for the expression. A few of
widely used such expressions for modelling slow polarisation processes are
discussed below.
In the other approach, called the equivalent circuit approach, the behaviour of
insulation is modelled by equivalent RC circuits [35], which is also discussed
in the latter part of this section.
36
5.1.1 Debye model with single and distributed time constants
Time and frequency domains dielectric response functions of the classic
Debye model are given in (5.1 A) and (5.1 B), respectively.
( )

t
e t f

(5.1 A)
( )

j +
1
(5.1 B)
where, is known as the dielectric relaxation time and is the dielectric
strength.
Debye model has been derived under the assumption that dipoles in the
relaxing medium do not interact with each other. Therefore, simple Debye
behaviour is not applicable to most materials other than polar liquid [2]. Time
and frequency dependent Debye responses are shown in Figure 5.1 Figures
(A) and (B), respectively.
(A) (B)
Figure 5.1. Time and frequency dependency of different dielectric response functions.
t==1/
p log(t)
l
o
g

(
f
(
t
)
)
Debye
General
response
-n
-1-m
p log()
l
o
g

(
)
)
m

n-1
Debye
General
response
37
log ()
l
o
g

(
)
)
total
response
single Debye
loss peaks
log (t)
l
o
g

(
f
(
t
)
)
total
response
single Debye
responses
Dielectric responses, which depart from the ideal Debye behaviour, can
usually be interpreted by the distribution of relaxation times. Here, the total
response is modelled by the addition of single Debye processes as,
( )
( )
0
1

d
j
g
(5.2)
where function g() defines the distribution of relaxation times, which is
considered as a discrete function in most of the interpretations. Graphical
representations of this operation in frequency (loss) and time domains are
shown in Figure 5.2.
Figure 5.2. Schematic representation of non-Debye dielectric response by distribution of
relaxation times.
One could also explain the physical significance of distributed relaxation
times in solid materials as the presence of inhomogeneities in materials,
which can create different surrounding environments for different dipoles [2].
5.1.2 General response function
The general response function can be used to model a dielectric response
function, which shows a transition of two processes at t = 1/
p
, as shown in
Figure 5.1. In time domain general response function can be written as,
( )
( ) ( )
1 , 0
1
< <
+
+
m n
t t
A
t f
m
p
n
p

(5.3)
38
A logarithmic representation of this expression shows two straight lines with
different slopes. When t >> 1/
p
, the slope of the line is (1+m) and for
short times (t << 1/
p
) this slope is -n. An approximate frequency dependent
loss corresponding to this function can be written as [2],
( )
n
p
m
p

,
`
.
|
+
,
`
.
|

1
1
(5.4)
The real component of the frequency dependent general response function
varies as shown in Figure 5.1 (B). Values of n can be used to classify the
material as a charge carrier system (0 < n < 1) or dipolar system (1 < n < 2).
The dielectric response of pressboard and Kraft paper is well-defined by
general response [2].
However, the general response function cannot be analytically Fourier
transformed. Therefore sometimes another function, which has the same
asymptotic behaviour as the general response function, is used for
modelling. This function has an analytic Fourier transformation and its time
and frequency dependencies can be written as,
( )
'
'
,
`
.
|
,
`
.
|
+
,
`
.
|

n t m t
t
e
t
e A t f

1 (5.5 A)
( )
( ) ( ) ( )
( )
'
'
,
`
.
|
+

,
`
.
|
+

n
n
n
n
m
m
i
n
i
n
i
m
A
1 1 1
1
1
1
1
1
(5.5 B)
However, this expression does not have any physical significance [50].
5.1.3 X-Y model
The functional approach is further developed in modelling the power
transformer insulation system by considering its geometrical arrangement
[48]. Figure 5.3 shows the typical winding and insulation arrangement of a
power transformer. As shown in the figure, the LV winding is usually
0 < m < 1
0 < n < 2
39
surrounded by the HV winding and these windings are separated by the main
insulation duct that consists of pressboard layers and oil channels. Therefore,
response measurements between the HV and LV windings are affected by
this composite insulation system.
In core type transformers, the main insulation consists of cylindrical
pressboard barriers in series with oil ducts and spacers as shown in
Figure 5.4. The complex geometrical arrangement shown in this figure can be
simplified by combining all oil ducts, barriers and spacers separately, which
simplifies the modelling. Then the main insulation is simplified to the so
called X-Y model, as shown in Figure 5.5 [41, 48].
Figure 5.3. Typical winding configuration of a power transformer.
Yoke
Yoke
C
o
r
e
L
V

w
i
n
d
i
n
g
H
V

w
i
n
d
i
n
g
40
Figure 5.4. Cross section of main insulation of a core type transformer.
Figure 5.5. Simplified insulation structure of a core type power transformer.
Here,
duct the of periphery
spacer the of width total
duct the of width
barriers total of thickness

Y
X
In real power transformers, X and Y often vary between 0.2 - 0.5 and
0.15 - 0.25, respectively [48].
oil
spacers
barriers
Y 1-Y
X
1-X
LV
HV
barriers
spacers
41
The dielectric response across the X-Y system can be calculated when the
individual dielectric response of oil, spacers and barriers are known. The
response of the oil is described using the constant values of dc conductivity
and permittivity
r
( = 2.2). Since both barriers and spacers are formed by
pressboard, they can be treated as a single material. The dielectric response of
pressboard has significant dispersion in the time and frequency range of
interest. Hence, the response of pressboard is described by its dielectric
response function or susceptibility in time and frequency domains,
respectively. Dielectric measurements of well-defined pressboard samples are
used for forming a database, where information of relevant response
functions with different moisture content is stored [1, 41, 48]. These
responses can be well described by the general response function [2, 48]. The
temperature dependence of oil and paper is treated as described in
Chapter 2.4. This is characterised by the activation energies of 0.9 eV and
0.7 eV for pressboard and oil, respectively [48].
In addition to above information, when X and Y for a particular transformer
are known, (5.6) can be used for calculating the corresponding frequency
dependent dielectric response at a known temperature for a given moisture
content in pressboard and the conductivity of oil.
( )
barrier oil barrier spacer
duct
X X
Y
X X
Y
T

1
1

1
,
+
+
+
(5.6)
Where,
( )

0
,
) (
,
T
j
T
oil r oil
pressboard barrier spacer

(5.7)
Furthermore, the following set of equations can be used for calculating the
polarisation current I
pol
of the same insulation system in the time domain.
b oil s pol
I I I I + (5.8)
so b
U U U + (5.9)
Here I
s
, I
oil
and I
b
are the polarisation current through spacers, oil and
barriers, respectively. U
b
and U
so
are the voltages across the barrier and the
42
spacer components, respectively. The respective currents can be calculated by
applying (2.18) to each material separately.
When charging time t
c
and discharging time t
d
are specified, (2.24) combined
with (5.9) can also be used for deriving the return voltage spectrum
corresponding to the X-Y arrangement of the considered transformer.
The measurement techniques described in Chapter 4.2.2 can be utilised for
performing FDS, PDC and RVM measurements. Then the results obtained
from these measurements are compared with the relevant derived quantities
using the X-Y model. Error between derived and measured quantities is
minimised by changing the amount of moisture content in the pressboard and
the conductivity of the oil until a best fit is obtained. This allows for
predicting the moisture content in paper and the conductivity of oil in the
investigated transformer.
In the second modelling approach, each material is represented by an
equivalent RC circuit, instead of a single analytical function. An example of
such a circuit representing a single material is shown in Figure 5.6.
Figure 5.6. Equivalent RC circuit representing the dielectric response of an insulation.
Here, the dielectric dispersion of this material is characterised by n parallel
connected RC elements. When the geometrical capacitance C
0
is known, for a
long charging time t
c
the response function f(t) can be written as,
( ) ( )
i
n
i
i
depol
t A
U C
I
t f / exp
1 0

(5.10)
C
R
R
1
R
2
R
3
R
i
C
1
C
2
C
3
C
i
R
n
C
n
U
43
where,
i i i
i
c
i
i
C R
t
R C
A
]
]
]
,
`
.
|

exp 1
1
0
Therefore, using measurements of well-defined insulating samples, it is
possible to calculate the equivalent time constants
i
, using a suitable
sequential algorithm [35].
The application of this method for the X-Y model is shown in Figure 5.7
[35]. In the figure, both barriers and spacers are formed by pressboard.
Therefore, when barriers and spacers are connected in series, the equivalent
can be assumed as a single unit (index sb - stands for serial connection of
spacers and barriers).
Figure 5.7. Equivalent RC circuit of the X-Y model.
R
oil
C
b
R
b
C
0il
R
sbn
C
sbn
R
b1
R
b2
R
bn
C
b1 C
b2
C
bn
I
pol
I
pol
B
A
I
oil
I
b
U
so
U
b
I
sb
R
sb C
sb
R
sb1
R
sb2
C
sb1 C
sb2
R
sbi
C
sbi
R
bi
C
bi
44
The capacitance and resistance of oil can be directly calculated when
permittivity
r
and dc conductivity are known, see (5.11) and (5.12).
X
Y
C C
r oil
1
1
0 0
(5.11)
,
`
.
|
Y
X
C
R
oil
1
1
0
0
(5.12)
Therefore, for a particular transformer, the geometrical arrangement of which
is known (C
o
is known), the polarisation current I
pol
of the main insulation
can be calculated using the following set of equations {(5.13) - (5.17)}.
dt
U d
C
R
U
I
so
oil
oil
so
oil
+ (5.13)
+ +
n
i
C R
t
bi
b b
b
b
b
b
bi bi
e
R
U
dt
U d
C
R
U
I
1
(5.14)
+ +
n
i
C R
t
sbi
sb
sb
sb
sbi sbi
e
R
U
dt
U d
C
R
U
I
1
(5.15)
so b b oil
U U U and I I + (5.16)
sb b pol
I I I + (5.17)
The calculated polarisation current is fitted with measured current by
changing conductivity and the RC parameters of distributed RC circuits as
mentioned in [35]. The same parameters can be used to calculate the
frequency dependent dielectric spectroscopy and the return voltage spectrum,
as described earlier in this chapter.
One of the major assumptions made in X-Y modelling is that the dielectric
response is linear. However, in some cases this assumption may not be true,
especially when insulating oil is new. Therefore, it is advisable to keep the
measuring voltage at the minimum possible level [48].
45
5.2 Effect of different parameters of X-Y model on the final
FDS response
The X-Y model is widely used for the diagnostics of transformer insulation
due to its simplicity and the direct relation between the model and the
construction of transformer insulation [1]. However, the main disadvantage
of this approach is the necessity for knowing the geometry of transformer
insulation, which is rather hard to find for utility engineers. Therefore,
additional simplification of X-Y model would be useful to enable engineers
to study the condition of insulation with little knowledge of its geometrical
construction. For this purpose, it is essential to study the influence of
different model parameters in the X-Y model on the final response of the
system. Such analyses are presented in the following section. These analyses
are based on the measurements made of the pressboard samples described in
Chapter 7.
5.2.1 Influence of oil conductivity
The conductivity of oil is highly temperature dependent. Therefore, it is
necessary to observe the influence of conductivity values in a wide range
since, in our study, field measurements were performed in a temperature
range varying between 25 C and 65 C. According to IEC 6422, the
conductivity of new mineral oil should be less than 16.5 pS/m at 90 C [51].
By assuming an activation energy of 0.7 eV [48] (2.27) gives the
corresponding conductivity at 25 C as 0.1 pS/m. Hence, 0.05 pS/m at 25 C
(or 1.25 pS/m at 65 C) is taken as the lower limit of the conductivity range
selected. The upper limit of the range is selected as 10
4
pS/m at 65 C (or
410
2
pS/m at 25 C), which is more than 10 times higher than the acceptance
limit of the conductivity of used oil.
Figure 5.8 shows the influence of oil conductivity on FDS response at 25 C ,
when the amount of moisture content in paper and the amount of pressboard
in the insulation system are kept at a minimum level. In the figure, the
variations of permittivity and loss clearly show the Maxwell-Wagner
behaviour of a parallel system, which is mainly influenced by the conduction
in the oil and, at the same time, the dispersion of the spacer component is
negligible. However, one can observe that when oil conductivity is less than
410
2
pS/m, it has little influence on the permittivity of the system at
frequencies above 100 Hz.
46
Figure 5.9 shows the influence of oil conductivity at 65 C, when the amount
of moisture content in paper and the amount of pressboard in the insulation
system are kept at a maximum level. Due to the significant influence of low
frequency dispersion in the paper, the influence of oil conductivity on the
total permittivity of the system is weaker in comparison with the previous
case. In addition, oil conductivity, which lies within the same range, has little
influence on the permittivity of the system at frequencies above 100 Hz.
10
-4
10
-2
10
0
10
2
10
4
10
0
10
1
10
2
Frequency (Hz)

'
0.05 pS/m
4.5 pS/m
400 pS/m
10
-4
10
-2
10
0
10
2
10
4
10
-4
10
-2
10
0
10
2
Frequency (Hz)

'
'
0.05 pS/m
4.5 pS/m
400 pS/m
Figure 5.8. Derived real and imaginary
components of complex permittivity for
the X-Y model for different values of oil
conductivity at 25 C, when X=0.2,
Y=0.15 and moisture content=0.2%.
10
-5
10
-3
10
-1
10
1
10
3
10
5
10
0
10
1
10
2

'
10
-5
10
-3
10
-1
10
1
10
3
10
5
10
-2
10
0
10
2

'
'
1.25 pS/m
112 pS/m
10000 pS/m
1.25 pS/m
112 pS/m
10000 pS/m
Frequency (Hz)
Frequency (Hz)
Figure 5.9. Derived real and imaginary
components of complex permittivity for
the X-Y model for different values of oil
conductivity at 65 C, when X=0.5,
Y=0.25 and moisture content=5%.
47
The total permittivity at 1 kHz under different conditions is given in
Table 5.1. Case 1 and Case 2 in the table correspond to the two cases
described in Figure 5.8 and Figure 5.9, respectively. This reveals that in both
cases the influence of oil conductivity, which lies within the specified range,
has little influence on the at 1 kHz . The significant difference of at
1 kHz between the two cases considered is mainly caused by the variation of
other parameters.
Table 5.1. at 1 kHz under different conditions.
Case No.
Conductivity
(pS/m)
at
1 kHz
0.05 2.66
4.5 2.66 Case 1
400 2.66
1.25 3.67
112 3.67 Case 2
10000 3.68
5.2.2 Influence of spacers
According to previous studies, in a typical power transformer insulation, the
peripheral area covered by spacers can be usually vary from 15 % - 25 %
[48]. Based upon this information, the maximum influence of spacers on the
final dielectric response can be encountered when Y equals 0.25. Hence, the
influence of spacers is examined by comparing the dielectric response of the
X-Y system with Y equals 0.25, with the same response with Y equals 0. The
influence of spacers is studied under different conditions since the total
response is dependent on the amount of barriers, moisture content, oil
conductivity and temperature (Figure 5.10 Figure 5.13). These figures also
show the residual curves representing the difference between two response
curves corresponding to Y = 0.25 and Y = 0.
Table 5.2 describes the way the spacers influence the total dielectric response
under different conditions, based on the results shown in
Figure 5.10 - Figure 5.13.
48
10
-4
10
-2
10
0
10
2
10
4
10
-2
10
0
10
2
10
4
Frequency (Hz)

'
Y=0.25
Y=0
Resi dual
10
-4
10
-2
10
0
10
2
10
4
10
-4
10
-2
10
0
10
2
10
4
Frequency (Hz)

'
'
Y=0.25
Y=0
Resi dual
X = 0.2
X = 0.5
X = 0.2
X = 0.5
X = 0.2 X = 0.5
X = 0.2
X = 0.5
Figure 5.10. Influence of spacers with different amounts of barriers.
49
10
-2
10
0
10
2
10
4
10
0
10
2
10
4
Frequency (Hz)

'
Y=0.25
Y=0
Resi dual
10
-2
10
0
10
2
10
4
10
-2
10
0
10
2
10
4
Frequency (Hz)

'
'
Y=0.25
Y=0
Resi dual
mc = 0.2 %
mc = 5 %
mc = 0.2 %
mc = 5 %
mc = 5 %
mc = 0.2 %
mc = 0.2 %
Figure 5.11. Influence of spacers with different amounts of moisture content.
50
10
-2
10
0
10
2
10
4
10
-2
10
0
10
2
10
4
Frequency (Hz)

'
Y=0.25
Y=0
Resi dual
10
-2
10
0
10
2
10
4
10
-2
10
0
10
2
10
4
Frequency (Hz)

'
'
Y=0.25
Y=0
Resi dual
= 1000 pS/m
= 10 pS/m
= 10 pS/m
= 1000 pS/m
= 1000 pS/m
= 1000 pS/m
= 10 pS/m
= 10 pS/m
Figure 5.12. Influence of spacers at different oil conductivity values.
51
10
-2
10
0
10
2
10
4
10
1
Frequency (Hz)

'
Y=0.25
Y=0
Resi dual
10
-2
10
0
10
2
10
4
10
-2
10
0
10
2
10
4
Frequency (Hz)

'
'
Y=0.25
Y=0
Resi dual
T = 25
o
C
T = 65
o
C
T = 25
o
C
T = 25
o
C
T = 25
o
C
T = 65
o
C
Frequency (Hz)
Figure 5.13. Influence of spacers at different temperatures.
52
Table 5.2. Influence of spacers under different conditions.
State define
parameter
Mean
relative
error (in %)
Permittivity () at
1 kHz
Fig.
No.
Name Value
Values of
other
parameters
mc-% ;
-pS/m ; T-C

Y=
0.25
Y=
0
Error
(%)
5.10 X
0.2
0.5
mc 5

(25 C )
10
T 25
14
9
20
19
3.1
3.6
2.5
3
19
17
5.11
mc
(%)
0.2
5
X 0.2

(25 C )
10
T 25
12
19
33
30
2.8
3.1
2.4
2.5
14
19
5.12

(25 C )
(pS/m)
10
10
3
X 0.2
mc 5
T 25
19
10
30
15
3.1
3.1
2.5
2.5
19
19
5.13
T
(C)
25
65
X 0.2
mc 5

(25 C )
10
19
20
30
35
3.1
3.2
2.5
2.5
19
22
As shown in all four figures (Figure 5.10 Figure 5.13) within the specified
ranges of all the parameters of interest, the affect of spacers on the shape of
the final dielectric response curve is not particularly significant. This effect
could be explained by the influence of the more conductive oil component,
which is in parallel to the less conductive spacer component. One can clearly
observe this in Figure 5.12, where dielectric responses with two different
values of oil conductivity are plotted. The figure shows that when the oil
conductivity increases residual substantially decreases.
In Table 5.1, the column corresponding to the relative error, shows a mean
error introduced by removing spacers from the model. The mean error
53
introduced to the real component of the permittivity is always less than 20 %.
However, in the imaginary component, this can reach up to 35 %.
According to the results shown in Table 5.1, the maximum influence of
spacers on the total dielectric response can be experienced at higher
temperatures with a minimum level of barriers, high moisture content and
low oil conductivity.
5.2.3 Variation of permittivity at 1 kHz (
1kHz
)
As described in the Section 5.2.2, oil conductivity has little influence on
permittivity over 100 Hz. Moreover, Figure 5.13 and Table 5.1 show that the
effect of temperature on permittivity at 1 kHz is also insignificant.
Furthermore, the last right hand column of Table 5.2 shows the estimated
error of permittivity at 1 kHz when the spacers are removed. One can see
that this error is always under 22 %.
The other two parameters, which affect at 1 kHz are moisture content in
paper and the amount of barriers in insulation. Figure 5.14 shows the
influence of these two parameters with-in the region of interest of each
parameter.
0 1 2 3 4 5
2.5
2.6
2.7
2.8
2.9
3
3.1
Moi sture content (%)

'

@

1

k
H
z
Mean = 2.45
Mean = 2.59
Mean = 2.76
Mean = 2.94
X = 0.2
X = 0.3
X = 0.4
X = 0.5
Figure 5.14. Variation of at 1 kHz with moisture content in paper and amount of barriers
when Y=0.
54
As shown in Figure 5.14, any variation of at 1 kHz with moisture content
is dependent on the amount of barriers. When X = 0.5, the dielectric response
of the barriers has the greatest influence on the total response. This gives a
maximum variation of at 1 kHz (
1 kHz
) with moisture content, the value
of which is 0.22. Table 5.3 shows a variation of at 1 kHz as a percentage of
lowest at 1 kHz for each X value. Therefore, we assume that at 1 kHz
does not vary with moisture content for a given X and is equal to the mean
value of at 1 kHz (Figure 5.14), then the introduced maximum percentage
error should be around 4 % (Table 5.4).
Table 5.3. Percentage variation of at 1 kHz at different X.
X
1 kHz
(
1 kHz
* 100)/min(
1 kHz
)
0.2 0.06 2.5
0.3 0.11 4.3
0.4 0.16 6.0
0.5 0.22 7.8
Table 5.4. Possible maximum percentage error if constant at 1 kHz (mean) value is
assumed at each X.
X Mean
1 kHz
Maximum error (%)
0.2 2.45 1.2
0.3 2.59 2.4
0.4 2.76 3.4
0.5 2.94 4.3
The variation of the mean at 1 kHz with X is shown in Figure 5.15. The
points obtained from the calculation are points on a quadratic curve described
by the following expression, in which
2 . 2 94 . 0
2
+ + x x y (5.18)
This provides an easy way for calculating the mean at 1 kHz for a given X.
55
0.2 0.25 0.3 0.35 0.4 0.45 0.5
2.4
2.5
2.6
2.7
2.8
2.9
3
3.1

y = 1*x
2
+ 0.94*x + 2.2
X
M
e
a
n

'

@

1

k
H
z
Mean from Fi gure 5.14
Fi tted quadrati c curve
Figure 5.15. Variation of mean
1 kHz
with amount of barriers.
5.2.4 Conclusion of the analyses
The above analyses reveal that the influence of spacers on the total dielectric
response of the X-Y system is relatively low and permittivity at 1 kHz is
mainly determined by the amount of barriers. Therefore, when there is a lack
of construction details of transformer insulation one can reduce the so called
X-Y model to a simple X model, while keeping the resultant errors within a
reasonable limit.
5.3 Modelling using X model
The dielectric response of the system shown in Figure 5.16 can be
characterised by the conductivity and permittivity of oil, the moisture content
in paper, the amount of barriers and temperature. Of these parameters, the
moisture content in paper and the amount of barriers are the only two
unknown, when separate measurements can be carried out to measure the
conductivity of oil.
56
Figure 5.16. Further simplified X model of the transformer insulation.
The dielectric response of the X model can be written as shown in (5.19).
( )
( )
X
T j
X
T
barrier
0
oil
model X
2 . 2
1
1
,

(5.19)
Equation (2.27) is used to calculate the conductivity at temperature T from
the measured conductivity at a known temperature. The dielectric response
measurements of pressboard samples at a known temperature are also
transformed into the temperature T by using (2.26).
In the modelling technique proposed, first a barrier percentage is assumed
and then (5.18) is used for calculating the corresponding at 1 kHz and this
value is utilised to calculate the geometrical capacitance of the transformer,
as shown in (5.20).
kHz
measured kHz
C
C
1
, 1
0

(5.20)
This provides a path for transforming the measured complex capacitance into
complex permittivity.
Subsequently, the percentage error between the complex permittivity of the
model and the transformer is minimised by changing the amount of moisture
barrier
oil
X
1-X
57
content in paper. This error minimisation is performed using an optimisation
routine developed in MATLAB software. The same procedure is followed for
different X values until the minimum percentage error is obtained.
5.4 Modelling using distributed relaxation times
Another modelling technique that can be utilised to model transformer
insulation without considering geometrical information, is presented below.
The complex insulation system is modelled using distributed relaxation
times. Each Debye relaxation can be represented by an equivalent series RC
circuit. Therefore, the insulation system is denoted as shown in Figure 5.6.
Subsequently, the complex capacitance ) of insulation can be written as
( )

'
'
+
+
+
]
]
]
]
'
'
+
+
+
]
]
]
+
n
i
n
i
i
i i
i
i
hf
n
i
n
i
i
i i
i
i
hf
n
i i
i
hf
C G
j
C
C
j C
j
j
C C
1 1
2 2 2 2
1 1
2 2
0
2 2
0
1 0
0
1 1
1 1
1

(5.21)
where,
C
hf
is the capacitance of insulation at highest measurable frequency
C
i
is the strength of relaxation process corresponding to relaxation time
i
G is the dc conductance of the insulation.
When measured complex capacitance data ( )
m
) are available, the
technique described in the following section can be utilised to calculate
unknown parameters in (5.21).
By considering real components of the measured and modelled complex
capacitances,
( )
hf real m
n
i
i
i
C C
C

+
1 1
2 2
(5.22)
58
The coefficients
i
and C
i
are determined by a sequential algorithm where
least square optimisation is implemented. In the first step of this algorithm, n
time constants, which are equally distributed in a logarithmic scale, are
chosen. Maximum and minimum
i
are selected as the reciprocal of the
minimum and maximum angular frequency involved in the corresponding
measurement. Then,
min
max
max
min
1
1
(5.23)
Subsequently, the constrained least square optimisation technique provided in
MATLAB 6 is used for calculating the strength of relaxations, C
i
. During
the optimisation the selected
i
values also change within a specified range
for obtaining a better fitting.
Thereafter, (5.24) is used for determining the dc conductance of insulation
using measured loss (
m imag
).

n
i
i
i i
imag m
C
C G
1
2 2
2
1

(5.24)
Chapter 6: Transformers at the Ceylon Electricity Board
59
6. Transformers at the Ceylon Electricity Board
6.1 Introduction
The Ceylon Electricity Board (CEB), a power utility of Sri Lanka, has a
monopoly on generation and transmission on the island. About 90 % of the
generation is owned by CEB and the balance is owned by independent power
producers. However, distribution is shared with the Lanka Electricity
Company Limited (LECO) and Local Authorities (LA). More than 15
medium scale generator stations (highest capacity is 210 MW), about 30 grid
substations (mainly 132/33 kV), more than 100 primary substations
(33/11 kV) and more than 10 000 low tension substations (33 or
11 kV/400 V) are in the CEB system, which annually delivers about
6500 GWh of energy. At the end of year 2000, the installed capacity was
2000 MW. Until the year 2000, hydro electricity was the major component of
the capacity, which is over 65 % of the total generation. Lakshapana and
Mahaweli are the two main complexes that generate 40 % and 51 % of the
total hydro electricity, respectively. Of these, Lakshapana is the oldest hydro
generation scheme and started operation in 1952 [52].
In all the above generator stations and grid substations with more than 150
power transformers of different rating and voltage levels, different
manufactures, and of broadly varying age are installed. Also more than 200
transformers of rated power less than 10 MVA are installed at the primary
substations. Numerous locally produced distribution transformers also
operate in the system.
6.2 Generator transformers under study
In this study, we have mainly focused on generator transformers in the
Laxapana and Mahaweli complexes. Table 6.1 provides information on the
year of manufacture and the rated power of all the transformers under
operation in these two schemes. According to the table, the age and rated
power of these transformers is between 2 - 50 years and 2 50 MVA,
respectively. During the study period, it was not possible to perform
electrical measurements on all these transformers. By utilising possibilities
arising from the CEB maintenance schedule, the transformers shown in
Table 6.2 were selected for electrical measurements.
60
Transformers from T 1 to T 11 are located in 3 generator stations in the
Laxapana complex. Others are installed in the Mahaweli complex. Annual
plant factors shown in the Table 6.2 are for the year 1999. However, these
values do not vary broadly since the annual rainfall, which is almost constant
in the area, is the main decisive factor if no major maintenance is involved.
Table 6.1. Generator transformers in Mahaweli and Laxapana complexes.
Year of
manufacture
Power
(MVA)
Phase No. of
Transformers
1963 11 Single 7
1965 18 Single 7
1974 24 Single 7
1985 26/38 Three 1
1987 28.5/38 Three 1
1989 13 Single 3
Laxapana
complex
~1950 5 Single 6
1975 27 Three 1
1976 50 Three 1
1982 30 Single 9
1984 32 Single 10
1984 34.5 Three 2
1985 27 Single 7
1985 2 Three 2
1990 27 Three 1
Mahaweli
complex
2000 30 Single 1
61
6.3 Maintenance of generator transformers at CEB
6.3.1 Maintenance tasks carried out
Routine tasks
For cheap, reliable, safe and efficient operation of power transformers correct
routine maintenance is desired. This allows for precautions to be taken before
a sudden failure of the unit occurs.
In general, manufacturers specify tasks and the performance frequency of
each task needed to maintain the quality of the transformer. Some of these
tasks depend on the design of the transformer. Therefore, routine
maintenance tasks carried out on CEB power transformers, vary from station
to station. For example, Table 6.3 shows the routine maintenance schedule of
transformer T 12 and other identical generator transformers in the same
station.
Table 6.2. Transformers under study.
ID No. Year of
manufac.
Manufacturer Power
(MVA)
Voltage
(kV)
Phase Plant
factor
(%)
T 1 7 1965 Canadian GE 18 12/132 Single 70
T 8
10
1963 Le materiel
electrique
France
11 11/132 Single 25
T 11 1974 Alsthom 24 12/132 Single 45
T 12 1984 GEC 32.5 12.5/220 Single 45
T 13 1990 Hyosung,
Korea
27 12.5/132 Three 50
As shown in Table 6.3, the two major insulation tests carried out on the
transformer insulation system are insulation resistance (IR) measurements
and the dielectric strength of warm oil. The other important test is chemical
62
analysis of the insulating oil, which should be performed annually. However,
in CEB this analysis is not performed as an annual test due to a lack of
internal resources. Instead, the analysis is performed when the other
measurements indicate inferior quality of the insulation.
On-site insulation improvements
CEB has its own facilities for on-site purification of transformer oil. Usually
this is done when the results of the above mentioned measurements (IR,
chemical analyses) fall outside acceptable range. The purification is mainly
used for removing excess moisture from transformer insulation. In addition,
particles of dirt and dissolved gasses are also removed to some extent.
However, with the available purification plants, at least 10 days of
continuous purification is needed for a significant improvement of
transformer insulation. The purification time depends mainly on the quality
of transformer insulation before purification and the individual specifications
of the purification plants.
Table 6.3. Routine maintenance schedule of transformer T-12.
Frequency of the
task (in months)
Jobs to be carried out
1
Check the drycol operation counter and whether water
is being ejected from the drycol; inspect the
transformer carefully to see whether there is any oil
leakage or abnormal vibration conditions; check the oil
level of conservator and bushing; check the general
cleaning of the transformer.
3
Check the condition of silica gel at the breather of
storage oil tank.
12
Measure insulation resistance (IR) between windings
and from each winding to the tank; measure the oil
dielectric strength while oil is warm; operate the tap
changer in full range (no load); check the operation of
relays; grease the bearings; measure IR values of
fan/pump motors; measure motor starting and running
current; measure power factor of the bushing; check
the condition of the paint; check the condition of motor
bearings.
63
6.3.2 Information recording
Keeping stringent records of the collected data and other information is
essential for identifying the condition of transformers under operation.
However, the recording system used by CEB is not up to contemporary
standards. Therefore, in this study, it was hard to find information on the
measurement history of the all transformers of interest. For example,
Table 6.4 shows the past measurement history of transformer T 13.
According to the data shown in Table 6.4, the insulation condition of this
transformer was drastically reducing during the period 1991 - 2002.
However, other factors that influence these measurements, such as condition
of the bushings, atmospheric conditions during the measurements, the
instruments used for the measurements are not clearly mentioned. Therefore,
it is hard to draw firm conclusions using the results of such measurements.
Table 6.4. Measurement history of transformer T-13 (LT low voltage winding; HT high
voltage winding; E tank).
IR (G) PI
Date
LT E HT E HT LT LT E HT E HT LT
1991.11.22 5 15 --- 3 --- ---
1992.07.31 4 5 7 1.8 2 1.4
1993.11.25 1.7 2.1 7 1.5 1.7 ---
1995.05.30 1.5 0.9 4 2.3 1.4 1.9
1995.11.24 2.5 0.9 5 1.2 1.3 1.2
1996.12.03 1.8 0.9 5 2.2 1.5 1.8
1997.12.08 1.3 0.7 3 2.1 1.4 2.3
1999.01.13 3.8 0.6 1.4 2 1.5 1.8
2002.09.16 0.8 0.4 0.9 1.5 1.3 1.8
--- IR is higher than the measurable range of the instrument.
64
6.3.3 Generator transformer replacements
There are only four cases of replacing generator transformers reported in the
Laxapana and Mahaweli complexes, as follows below:
A 27 MVA 12.5/132 kV generator transformer was burnt down in 1989
due to a guerrilla attack. Later transformer T 13 replaced it.
A 30 MVA 13.8/132/220 kV three winding single-phase transformer was
burnt down due to an insulation failure after being in operation for 16
years. During investigations performed after the failure it was found that
whole windings were coated with thick carbon layer. Therefore one
possible reason for this breakdown could be the contaminants produced
on load tap changer (OLTC) since in this particular design OLTC and the
windings were immersed in the same oil bath. After the investigations it
was decided to replace it with a new transformer since refurbishment of
the damaged transformer would be too expensive.
13 MVA 11/132 kV three single-phase generator transformers were
replaced by new ones in 1989. Major reasons for replacement were the
age (about 50 years) of the transformers and high number of faults noted.
Six other 5 MVA single-phase generator transformers in the same power
station will be replaced during an on-going rehabilitation project. Similar
to the previous case, the major reasons for this are the age (operation
started in 1958) of the transformers and high number of faults noted.
6.4 Distribution transformers under study
The CEB distribution system consists of two main categories called medium
voltage (33/11 kV) distribution and low voltage (415 V) distribution systems.
Therefore, most of the distribution transformers operate at 33 kV, 11 kV and
415 V. Nevertheless, due to the rapid increase in demand for electricity and
for obtaining a more reliable supply, most of the 11 kV system is currently
being replaced by a 33 kV system.
The majority of the distribution transformers are 33/0.4 kV (less than
500 kVA) sealed units. There are no specific routine maintenance procedures
followed by CEB for these transformers. Once a major fault in a distribution
transformer is identified, the transformer is replaced with a new one and the
65
old one is sent for possible factory maintenance. Table 6.5 shows details of
some of the distribution transformers selected for this study.
Table 6.5. Some of the distribution transformers under study.
ID No. Year of
manufac.
Manufacturer Power
(kVA)
Voltage
(kV)
Vector
DT 1 1974 Mitsubishi 500 33/0.4 Dyn11
DT 2 1981 BBC 200 33/0.4 Znyn11
DT 3 1996 Lanka
Transformers
250 33/0.4 Dyn11
The effectiveness of the oil purification process carried out on transformer
DT 1 was studied by performing an oil analysis on oil samples and FDS
measurements on the transformer, before and after oil purification.
Of the above listed transformers, transformer DT 3 has never been energised.
The other two transformers are still in operation. Due to lack of records, it
was hard to find information about previous measurements or analyses of
these transformers.
Chapter 7: Measurements
67
7. Measurements
7.1 Instrumentation for dielectric response measurements
During the study presented in this thesis two measurement set-ups were
utilised for dielectric response measurements.
FDS measurements were performed using a commercially available FDS
measuring system IDA 200. Figure 7.1 presents a schematic diagram of this
instrument.
Figure 7.1. Schematic diagram of frequency domain dielectric measuring system IDA 200.
Non-grounded measurements are often used for transformer measurements.
The tank is connected to the guard electrode. Channels Ch1 and Ch2 in the
digital signal processing (DSP) board are used for measuring the magnitude
and phase of the applied voltage and the resultant current, respectively. Then
the complex capacitance of the test object is calculated using (2.16).
The main features of the IDA 200 are given in Table 7.1 [53].
V
A
I
U
Voltage
source
Computer with
DSP - board
Sample
Z
Electrometer
Measured voltage
Measured current
Control
voltage
Voltmeter
Ch1
Ch2
Guard
connection
68
Table 7.1. Features of IDA 200.
Parameter Value
Voltage sources 1 0-10 V
peak
2 0-200 V
peaak
Frequency 0.1 mHz 1kHz
Current 0-50 mA
peak
PDC measurements were performed under laboratory conditions. A Keithley
6571 electrometer was used for these measurements. Figure 7.2 shows the
schematic diagram of a typical PDC measurement set-up.
Figure 7.2. Schematic diagram of PDC measurements.
In the electrometer settings the integration time for A/D conversion of the
current measurements was set to 20 ms, which was good enough to reduce
the influence of noise, even at the lowest current levels measured. During the
measurements, the tank was isolated from the ground and connected to the
guard electrode. The settings of the electrometer were such that the low
potential terminal of the voltage source was internally connected to the low
end of the electrometer. Also when the voltage source was switched off, its
two terminals were internally short-circuited. These features allowed for
performing the polarisation and depolarisation current measurements without
using any additional switching arrangement.
Electrometer
Voltage
source
SW1
SW2
Test
object
Guard
connection
69
7.2 Field measurements
Field measurements were carried out in Sri Lanka on the transformers
belonging to CEB as specified in Table 6.2 and Table 6.5. The measurement
configuration used is shown in Figure 7.3.
Figure 7.3. Measurement configuration used in field measurements.
During the measurements, all the terminals of HV and LV windings were
short-circuited separately. The grounded tank was connected to the guard
terminal.
Field measurements were performed at different temperatures (between 30 C
and 60 C) depending on the operational situation. When the measurements
were performed at elevated temperatures, the temperature of the oil was
noted both at the beginning and at the end of each measurement. In most of
the cases, the temperature difference was less than 3 C. Therefore,
throughout this study the starting temperature was considered as the
measurement temperature.
Apart from the FDS measurements, whenever possible oil samples were also
taken from the transformers for further chemical and electrical analyses. Oil
sampling was done according to IEC 60475 [54].
HV
LV
To IDA 200
Voltage
terminal
Current
terminal
Guard
terminal
Tank
70
7.3 Laboratory measurements
7.3.1 Laboratory transformer
The transformer selected for study in a laboratory environment was a
distribution transformer (100 kVA, 20 kV/400 V), which was built in 1979
by ASEA. It consists of a foiled low voltage winding and an enamel coated
high voltage winding. In 2000, the transformer was renovated and the oil was
changed. After the renovation, the transformer was not used until the tests
described here were conducted. The insulation of distribution transformers
usually differs from what is typical for large power transformers. In the
transformer investigated, the insulation between the low and high voltage
windings consisted of pressboard barriers and glued masonite spacers.
In order to balance the water content in the oil and the paper, the transformer
was heated to above 70 C for 2 weeks. This was done by continuously
feeding the high voltage windings with a current of 2.1 A, which is about
70 % of the full load, while the low voltage windings were short-circuited.
To minimise the heat losses and to keep the temperature at the level required,
the transformer was thermally isolated. The electric resistance of the high
voltage winding was measured to calculate the temperature inside the
transformer.
Thereafter, oil samples were taken from the hot transformer (70 C) for KFT
analyses and for dielectric characterisation.
For checking the effect of thermal treatment on the insulation, FDS
measurements were repeated before and after the transformer was heated.
These measurements were performed at 20 C in frequency range
10
-4
Hz - 10
3
Hz and the voltage used was 200 V
peak
. Instrumentation was
connected to the transformer as described in Section 7.2.
PDC measurements of this transformer were performed using the set-up
described in Section 7.1. The PDC measurements were performed after the
transformer was cooled down to room temperature (20 C) after heat
treatment. A voltage of 500 V dc was applied across the main insulation for a
period of 7.510
4
s for measuring the polarisation current. Then the
depolarisation current was measured for a period of 10
4
s by short-circuiting
the two winding sets through the electrometer.
71
7.3.2 Oil test cell
A three terminal stainless steel cylindrical oil test cell was used for
performing FDS measurements on the transformer oil-samples. Results of
these measurements were used to calculate the conductivity of the oil. The
parameters of this cell are given in Table 7.2.
Table 7.2. Parameters of the test cell.
Parameter Value
Volume 45 ml
Electrodes distance 2 mm
Geometrical capacitance 70 pF
Figure 7.4 shows a schematic diagram of the cell used for FDS
measurements. Before the cell was filled with oil it was cleaned with hexane
and dried. During the measurements the cell was placed on an insulating
plate for isolating the voltage terminal from the ground.
Terminals of IDA 200 instrument were connected to three terminals of the
cell as indicated in Figure 7.4. To avoid non-linear effects, the applied
voltage was limited to 5 V
peak
.
Figure 7.4. Three terminal test cell.
3 2 1
Insulating plate
1- Measuring terminal
2- Guard terminal
3- Voltage terminal
72
Since there are no relaxation processes in oil within frequency window of
interest, the conductivity of oil can be calculated according to,
( )
( )

0
0
C
C
(7.1)
where;
() is the frequency dependant conductivity of oil,
C() is the imaginary part of complex capacitance at frequency ,
C
0
is the geometrical capacitance of the cell.
7.3.3 Pressboard samples
Impregnated pressboard samples containing different amounts of moisture
were taken for this analysis. These pressboard samples were prepared in the
central laboratory of the Weidmann company in 1995 and stored in separate
oil-filled cylindrical sealed ducts. All of these ducts were placed in the high
voltage laboratory of ETH Zrich, where all the measurements were carried
out in 2002. The same samples were also used for the measurements
described earlier in [35].
Figure 7.5. Schematic diagram of the test cell.
Additional
weight
Voltage
electrode
Guard
electrode Measuring
electrode
Pressboard
sample
Transformer
oil
73
FDS measurements were performed on the impregnated samples to form a
database on the correlation between the frequency dependence of the
complex permittivity of pressboard and moisture content. These
measurements were performed in a special test cell (Figure 7.5) using the
IDA 200 system. This cell has been described in detail in [35]. All the
measurements were performed at room temperature (between 20 C and
27 C).
The thickness and diameter of all the pressboard samples under study were
2 mm and 159 mm, respectively. The diameter of the measuring electrode
was 113 mm, which yielded a geometrical capacitance between the electrodes
of 44.4 pF. An additional weight of 2981 grams was placed on the top
electrode to apply equal pressure on the pressboard samples during the
measurements. FDS measurements were performed with 50 V
peak
ac voltage.
7.3.4 Karl Fischer titration measurements
The moisture content in pressboard samples was also measured using the
coulometric KFT technique. In this case, the indirect stripping oven
technique was utilised. In this method, a known weight (about 0.5 g) of
pressboard sample was placed in the oven, which was heated to 140 C.
Moisture released from the pressboard was led away to the titration vessel by
a dry N
2
gas flow. The procedure described in IEC 60814 was followed for
determining the moisture content in the pressboard [55].
For calibrating the moisture content estimated from FDS measurements, KFT
analyses were performed on oil samples. The direct coulometric KFT
technique was utilised and the methodology described in IEEE 62 - 1995 was
followed for interpreting the results [18].
Chapter 8: Results and discussion
75
8. Results and discussion
8.1 Pressboard samples
Complex permittivities derived from the measured complex capacitance of
the five different pressboard samples are presented in Figure 8.1. These
results are normalised at 27 C by assuming an activation energy of 0.9 eV.
10
-2
10
0
10
2
10
4
10
0
10
1
10
2
Frequency (Hz)

'
mc=0.16 %
mc=0.9 %
mc=1.4 %
mc=2.5 %
mc=4.5 %
10
-4
10
-2
10
0
10
2
10
4
10
-2
10
0
10
2
Frequency (Hz)

'
'
mc=0.16 %
mc=0.9 %
mc=1.4 %
mc=2.5 %
mc=4.5 %
Figure 8.1. Real and imaginary components of complex permittivity as a function of
frequency at 27 C for pressboard samples containing different amounts of moisture (mc).
76
As shown in the Figure 8.1, permittivity increases substantially at low
frequencies with increasing moisture content in the pressboard. Dielectric
losses also substantially increase with increasing moisture content. At the
same time, the variation of with moisture content at 1 kHz is relatively
small. This behaviour is one of the basic observations used in the
development of the X model, described in Chapter 5.2.
These data were used to form the database for modelling the dielectric
response according to the X model and for estimating the moisture content in
the insulation of the transformers measured in the field and under the
laboratory conditions. Linear variations of the logarithmic values of both the
permittivity and the loss between two consecutive moisture contents
were assumed for calculating responses corresponding to moisture contents
not included in the database.
The moisture content in the pressboard samples estimated by KFT analyses
are given in Table 8.1. MODS software, provided by the manufacturer of the
IDA 200 instrument, was also used for estimating the moisture content in
these samples. This software uses two different databases for modelling; one
called cellulose and another called pressboard. The former one is based
on dielectric measurements performed on paper samples, whereas the latter
one is based on measurements on pressboard samples. The estimated
moisture contents using both databases and their percentage difference with
the results of KFT analyses are also given in Table 8.1.
Table 8.1. Moisture content in pressboard samples.
MODS-Pressboard MODS-Cellulose
Sample
No.
mc
using
KFT
(%)
mc (%) pd (%) mc (%) pd (%)
1 0.16 0.16 0 0.002 -99
2 0.9 0.68 -24 0.2 -77
3 1.4 1.14 -19 0.35 -85
4 2.5 2.73 9 1.5 -40
5 4.5 4 -11 2.6 -19
mc moisture content ; pd percentage difference.
77
As seen in the table, the estimates based on the database cellulose were
always lower than the estimates based on the database pressboard. At the
same time, the difference between the estimated moisture contents from the
database pressboard and from our own KFT analyses are relatively low
(max 24 %).
Since insulation in power transformers is mainly comprised of pressboard
and oil, Pressboard database of MODS software as well as the database
presented in Section 8.1 were utilised for modelling the dielectric responses
of power transformer insulation presented in this thesis. However, the
construction of distribution transformers is often different. Therefore, for
modelling the results of measurements of distribution transformers, the
database cellulose was also used.
8.2 Distribution transformer in laboratory
8.2.1 Modelling using X-Y and X model
Results of FDS measurements obtained before and after the thermal treatment
of the transformer are shown in Figure 8.2. One may notice a small shift of
the curves to the lower frequency region after the treatment. This shift is
mainly seen at frequencies where the contribution from oil conductivity
dominates the total dielectric response of an oil paper insulation system. The
curves were modelled using both the X-Y model (MODS) and the X model.
The modelled parameters and the results of oil analyses are shown in
Table 8.2.
Table 8.2. Modelled and measured parameters of lab transformer.
pressboard
X-Y model
cellulose
X-Y model
X model
Oil
analyses
mc X Y mc X Y mc X mc
A 3.2 13 20 22 2.1 13 20 22 -- -- -- --
B 3.2 8.5 20 22 2.1 8.5 20 22 2.4 20 2 4
mc - moisture content in pressboard (%) ; - oil conductivity at 20 C (pS/m) ; A - before
treatment ; B - after treatment ; X and Y are in (%).
78
10
-4
10
-2
10
0
10
2
10
4
10
-10
10
-9
10
-8
10
-7
Frequency (Hz)
C

'

(
F
)
before treatment
after treatment
10
-4
10
-2
10
0
10
2
10
4
10
-12
10
-10
10
-8
10
-6
Frequency (Hz)
C

'
'

(
F
)
before treatment
after treatment
Figure 8.2. Comparisons of real and imaginary parts of the complex capacitance measured
on lab transformer at 20 C before and after thermal treatment.
Two values for oil conductivity were obtained from the X-Y modelling. They
were 13 and 8.5 pS/m, from the data corresponding to the measurements
before and after the thermal treatment, respectively. This explains the earlier
conclusion based on the frequency shift between the two dielectric responses
in Figure 8.2. The conductivity of the oil measured after the thermal
treatment was 4 pS/m. This value is similar, but somehow lower than the
values derived from the modelling.
Figure 8.3 shows the frequency dependent capacitance C(), loss C() and
of the oil, sampled from the lab transformer. The results presented were
79
measured at 20 C. As shown in the figure, slope of frequency dependent loss
is nearly 1, which shows the dominance of oil conductivity in the dielectric
response of oil. Also it can be observed that the capacitance C() is almost
constant, which indicates the insignificant dielectric dispersion of oil within
the measured frequency range. These two observations verify the validity of
modelling the dielectric response of oil with a parallel connected resistor and
a capacitor.
10
-3
10
-1
10
1
10
3
10
-13
10
-12
10
-11
10
-10
Frequency (Hz)
C

'
,
C
'
'

(
F
)

a
n
d

(
S
/
m
)
C'
C''
Figure 8.3. Capacitance (C), loss (C) and conductivity () as a function of frequency at
20 C for oil sampled from lab transformer.
The moisture contents estimated by means of MODS are different for both
databases used, but identical within each of the databases for Cases A and B.
The value obtained from the cellulose database is lower and similar to the
value obtained from oil analyses. The moisture content estimated from the X
model is also comparable with the one obtained from oil analyses. According
to IEEE std. 62-1995 [18] this value is close to the margin of the acceptance
limit, which is equal to 2.5 %.
As shown in Table 8.2, the barrier contents derived are also similar.
However, due to the lack of exact information on the construction of this
transformer it was not possible to compare them with the real barrier content.
80
8.2.2 Comparison of time domain and frequency domain
spectroscopy measurements
The FDS response of the lab transformer was modelled using the distribution
of relaxation times and later it was used for deriving the corresponding time
domain response.
The distribution of relaxation times was derived in two different ways. In the
first approach (Distribution 1), we assumed that there were no other
relaxation processes outside of the angular frequency window used for the
measurements (2.910
-03
rad/s to 6.310
3
rad/s). Therefore, all the derived time
constants were limited to the time span corresponding to this window. In the
second approach (Distribution 2), the measured data were extrapolated at the
low frequency end for two decades further by assuming that C() and C()
changed proportionally according to the fractional power law.
( ) ( )
n
C C

1
(8.1)
where n was set to be equal to 0.62. This allowed for increasing the time span
of the time constants derived by two decades. The distributions of dielectric
strengths and relaxation times obtained using the two approaches are shown
in Figure 8.4. Both distributions contain a weak local peak at about 100 s.
The depolarisation currents derived from these distributions are compared
with the measured depolarisation current, as shown in Figure 8.5. One may
notice that the current calculated from the Distribution 1 deviates from the
measured data at longer times (>100 s). On the other hand, the depolarisation
current calculated from the Distribution 2 agrees well with the measured data.
This indicates the necessity extrapolations in the low frequency data for
deriving correct parameters.
81
10
-4
10
-2
10
0
10
2
10
4
10
6
10
-13
10
-11
10
-9
10
-7
Rel axati on ti mes (s)

C
'

(
F
)
Di stri buti on 1
Di stri buti on 2
Figure 8.4. Dielectric strengths of relaxation times derived by the two approaches used.
10
-4
10
-2
10
0
10
2
10
4
10
6
10
-15
10
-13
10
-11
10
-9
ti me (s)
D
e
p
o
l
a
r
i
s
a
t
i
o
n

c
u
r
r
e
n
t

(
A
)
Calculated depol. current distri 1
Calculated depol. current distri 2
Measured depol. current
Figure 8.5. Calculated and measured depolarisation currents.
82
8.3 Power transformers in the field
Results presented in this section are based on measurements performed on
the field transformers owned by CEB in Sri Lanka. Information on these
transformers is provided in Chapter 6.
8.3.1 Single-phase power transformers
10
-2
10
0
10
2
10
4
10
-9
10
-8
10
-7
Frequency (Hz)
C

'

(
F
)
T4
T5
T6
T7
10
-2
10
0
10
2
10
4
10
-11
10
-9
10
-7
C

'
'

(
F
)
T4
T5
T6
T7
Frequency (Hz)
Figure 8.6. Real and imaginary components of the complex capacitance measured on power
transformers T4 T7 as a function of frequency at 30 C.
Data for complex capacitance measured on transformers T4 T7 are
presented in Figure 8.6. These single-phase power transformers are identical
83
in construction, in rating, and in age. They also have a similar number of
operating hours. All the results presented in the figure were obtained when
the corresponding generator was shut down for annual maintenance.
Therefore, the temperature of these transformers was stable, around 30 C,
when the measurements were performed. As shown in Figure 8.6, the
dispersion of capacitance C() at frequencies lower than 1 Hz is relatively
high. In addition, some deviation in C() between the transformers is also
noticeable. The measured capacitance of T7 at the lowest frequency is higher
than the corresponding values for the other three transformers. One may
assume that this difference is caused by a higher moisture content in T7. This
can be seen when comparing the results of the oil analyses. Losses C() of
these transformers have noticeable differences especially within the
frequency range where oil conductivity dominates the total response. Losses
C( ) of transformers T4 and T7 are slightly higher than those of the other
transformers. The effect is caused by the higher oil conductivity in theses two
transformers.
Table 8.3. Comparison of modelled FDS results and the results from oil analyses.
Oil analyses X-Y model parameters
X model
parameters
ID
No.
ST
mco mc
mc X Y C
0
mc X C
0
T1 43 19 35 2.1 20 4 20 25 0.9 3.9 30 1
T2 40 25 36 2.6 22 4 20 25 0.9 3.8 31 1
T3 38 48 56 4.5 54 3.9 21 25 0.9 3.6 32 1
T4 47 190 86 3.9 194 4 21 25 1 4.2 31 1
T5 48 70 82 3.7 77 4 21 25 0.9 4.1 30 1
T6 54 142 140 4.9 150 4 20 25 1 4.1 29 1
T7 35 150 150 10? 182 4.3 20 25 0.9 4.4 28 1
ST oil sampled temperature; - oil conductivity at 27 C; mco - moisture content in
oil (ppm); mc - moisture content in paper (%); C
0
Geometrical capacitance (nF).
Figure 8.7 shows the variation of capacitance and loss of transformers
T1 - T3, which are also identical in construction to transformers T4 - T7.
However, these transformers were measured at elevated temperatures (i.e. at
84
40 C and 45 C) since they were in operation just before the measurements
began.
The measured complex capacitances of all seven transformers were used to
model their dielectric response. The parameters derived from the MODS
software and the X model are listed in Table 8.3. Results of the oil analyses
of the corresponding transformers are also shown in the table.
10
-4
10
-2
10
0
10
2
10
4
10
-9
10
-8
10
-7
Frequency (Hz)
C

'

(
F
)
T1
T2
T3
10
-4
10
-2
10
0
10
2
10
4
10
-12
10
-10
10
-8
10
-6
Frequency (Hz)
C

'
'

(
F
)
T1
T2
T3
Figure 8.7. Real and imaginary components of complex capacitance of transformers T1, T2
and T3 at 40 C, 45 C and 40 C respectively.
The derived and measured oil conductivity values are compared after
recalculating their values at 27 C by assuming an activation energy of
85
0.7 eV. As shown in Table 8.3, oil conductivity values obtained from MODS
and the corresponding measured values are similar.
Moisture contents estimated from MODS are rather close to the
corresponding values estimated from the X model. However, these values are
different from the values estimated from the results of oil analyses, especially
in T1, T2 and T7. Estimating the moisture content in paper based on the
moisture content in oil is highly temperature dependent. Therefore, this
difference can be due to errors in the temperatures of transformer insulation
measured when oil samples were taken. The results of oil analyses shows a
10 % moisture content in transformer T7, which is an extremely high and
unexpected result for an operating transformer. These results may be due to
an imbalance in moisture distribution between oil and paper during the time
the oil sample was taken, or due to the presence of contaminants in the oil,
which react with iodine in the KFT solvent and yield an overestimate of
moisture content.
The derived parameter X and the geometrical capacitance obtained from the
X model are nearly 10 % higher than those obtained from the X-Y model.
This higher estimate can be due to the negligence of spacer content in the X
model. Due to the lack of construction details, it was not possible to check
the accuracy of these results.
The estimated moisture content showed that all the transformers tested
contained an excessive amount of moisture. Therefore, CEB has been
recommended to perform vacuum oil purification on all these units as soon as
possible.
The results of measurements made on three other identical single-phase
transformers (T8 T10) are presented in Figure 8.8, whereas the estimated
parameters for these transformers are listed in Table 8.4. As seen in
Figure 8.8, the C() and C() of these three transformers are similar.
Therefore, we can clearly state that the current conditions of the insulation in
these three transformers are similar. This statement is further supported by
the parameters derived from the oil analyses. Moreover, the parameters
obtained from the modelling are also fairly close.
Figures 8.8 8.10 illustrate the measured and the derived permittivities
and losses for transformers T8-T10. The derived quantities were obtained
from the X model. It can be noticed in all three figures, that the measured and
the derived losses do not match well in the frequencies above 100 Hz. This
difference can be caused by the negligence of the influence of spacers in the
86
X model. Furthermore, the modelled loss curve of transformer T9 is
moderately higher than the corresponding measured curve within the
frequency range where oil conductivity dominates the total response. This
difference is due to the high oil conductivity value assigned to the X model,
which was obtained from the measured oil conductivity and by assuming an
activation energy of 0.7 eV.
Table 8.4. Comparison of modelled FDS results and the results from oil analyses for
transformers T8 T10.
Oil analyses MODS parameters
X model
parameters
ID
No.
ST
mco mc mc X Y C
0
mc X C
0
T8 35 51 44 4.2 63 5.2 20 15 0.6 4.7 24 0.7
T9 47 52 108 4.9 57 5 20 15 0.6 4.7 24 0.7
T10 50 47 115 4.6 52 4.5 20 15 0.6 4.3 24 0.7
T12 41 3 42 2 3 2.4 20 15 0.8 1.9 25 0.9
ST oil sampled temperature; - oil conductivity at 27 C; mco - moisture content in
oil (ppm); mc - moisture content in paper (%); C
0
Geometrical capacitance (nF).
Figure 8.12 illustrates the modelled and measured results for transformer
T12, the age of which is less than 20 years. All the estimated parameters for
this transformer are also shown in Table 8.4. The estimated moisture content
and the oil conductivity of T12 are much lower than estimates for the
previously considered transformers. As described in the previous case, the
difference between the modelled and measured loss curves are mainly
caused by the oil conductivity assigned to the X model.
87
10
-1
10
1
10
-8
10
-7
Frequency (Hz)
C

'

(
F
)
T8
T9
T10
10
-1
10
1
10
-10
10
-8
10
-6
Frequency (Hz)
C

'
'

(
F
)
T8
T9
T10
Figure 8.8. Real and imaginary components of complex capacitance of transformers
T8 - T10.
88
10
-3
10
-1
10
1
10
3
10
0
10
1
10
2
mc=4.7 %
X=24 %
Frequency (Hz)

'
Measured
Model ed
10
-3
10
-1
10
1
10
3
10
-4
10
-2
10
0
10
2
Frequency (Hz)

'
'
Measured
Model ed
X=24 %
mc=4.7 %
Figure 8.9. Comparison of modelled and measured permittivity and loss of transformer T8.
89
10
-3
10
-1
10
1
10
3
10
0
10
1
10
2
mc=4.7 %
X=27 %
Frequency (Hz)

'
Measured
Model ed
10
-3
10
-1
10
1
10
3
10
-4
10
-2
10
0
10
2
Frequency (Hz)

'
'
Measured
Model ed
X=27 %
mc=4.7 %
Figure 8.10. Comparison of modelled and measured permittivity and loss in transformer T9.
90
10
-3
10
-1
10
1
10
3
10
0
10
1
10
2
mc=4.3 %
X=24 %
Frequency (Hz)

'
Measured
Model ed
10
-3
10
-1
10
1
10
3
10
-4
10
-2
10
0
10
2
Frequency (Hz)

'
'
Measured
Modeled
X=24 %
mc=4.3 %
Figure 8.11. Comparison of modelled and measured permittivity and loss of transformer
T10.
91
10
-3
10
-1
10
1
10
3
10
0
10
1
10
2
mc=1.9 %
X=25 %
Frequency (Hz)

'
Measured
Model ed
10
-3
10
-1
10
1
10
3
10
-4
10
-2
10
0
10
2
Frequency (Hz)

'
'
Measured
Model ed
X=25 %
mc=1.9 %
T12.
92
8.3.2 Three-phase power transformer
10
-4
10
-2
10
0
10
2
10
4
10
0
10
1
10
2
mc=3.5 %
X=43 %
Frequency (Hz)

'
Measured
Model ed
10
-4
10
-2
10
0
10
2
10
4
10
-4
10
-2
10
0
10
2
10
4
Frequency (Hz)

'
'
Measured
Model ed
X=43 %
mc=3.5 %
T13.
In this study, there were not many possibilities for performing measurements
on three-phase power transformers, since such transformers are rarely
installed in accessible generator stations. Opportunities for measuring grid
transformers were, at the time, excluded. T13 is the only three-phase power
transformer on which measurements could be carried out. Although this
transformer was installed very recently (1990), its insulation resistance (IR)
value has been drastically reduced throughout the past years (see Table 6.4).
Furthermore, the records of this transformer contain information that the oil
93
seals were not good enough and oil was leaking through the seals. Therefore,
one possible reason for reducing IR was moisture ingress from the
atmosphere through these poor oil seals. Therefore, one could expect fairly
high moisture content in its insulation.
Figure 8.13 illustrates the measured and modelled permittivity , and loss
of transformer T13. The X model yielded a moisture content of 3.5 % in the
insulation. The corresponding values derived from the X-Y model was 3.8 %.
However, the corresponding moisture content derived from the oil analyses
was 2.3 %, which was much lower than the modelled values. At the same
time, the oil conductivity estimated from MODS (11 pS/m) and the measured
oil conductivity (9 pS/m) were simillar. The estimated geometrical
capacitance derived from MODS was 1.5 nF, and similarly, the geometrical
capacitance derived from the X model was 1.6 nF.
These results confirm that the moisture content in this transformer was
elevated, although it has been in operation for less than ten years. Therefore,
immediate action must be taken to repair the damaged oil seals and, if
possible, to perform vacuum oil purification.
8.4 Distribution transformers in the field
In the first instance, FDS measurements were performed on the distribution
transformer DT1 before and after the purification procedure. The resulting
dielectric response curves are shown in Figure 8.14. One should notice the
shifts between corresponding curves, which were mainly caused by
temperature differences during the measurements. The main conclusion from
the results obtained is that the purification procedure did not improve the oil
quality. To check it, additional analyses of the oil samples, taken before and
after the purification, were done. Oil resistivity and other typical parameters
of the oil were measured. The temperature dependencies of the resistivities
are shown in Figure 8.15 and it is clear that the purification did not lead to
any improvement of this parameter. Also the other analyses performed at the
company ABB in Vsters, Sweden showed similar behaviour. The moisture
content decreased from 26 to 14 ppm, the neutralisation number changed
from 0.28 to 0.24 mg KOH/g-oil, the breakdown voltage from 76 to
74 kV/2.5 mm, the loss factor from 0.519 to 0.476, and colour from 7.0 to
6.5. These results confirmed the earlier conclusion based on FDS
measurements. Estimated parameters derived from MODS, from the X model
and from oil analyses are given in Table 8.5.
94
10
-3
10
-1
10
1
10
3
10
-9
10
-8
10
-7
Frequency (Hz)
C

'

(
F
)
C' at 25
o
C before puri fi cati on
C' at 35
o
C after puri fi cati on
10
-3
10
-1
10
1
10
3
10
-10
10
-9
10
-8
Frequency (Hz)
C

'
'

(
F
)
C'' at 25
o
C before puri fi cati on
C'' at 35
o
C after puri fi cati on
Figure 8.14. Real C() and imaginary C() components of the complex capacitance of
transformer DT1 before and after oil purification.
Estimated moisture contents derived from the pressboard database and the
X model were similar. On the other hand these values are much higher than
the corresponding estimates derived from the cellulose database and oil
analyses. This might be due to differences in the internal construction of
distribution transformers from that of power transformers, and the models
used were mainly based on the response of oil impregnated pressboard
insulation systems.
95
Table 8.5. Estimated parameters for transformer DT1.
Before purification After purification
Method
mc X Y C
0
mc X Y C
0
cellulose 3.9 150 20 20 0.4 2.9 87 20 20 0.4
M
O
D
S
pressboard 5.2 170 20 20 0.4 4.4 106 20 20 0.4
X model 5.6 -- 30 -- 0.4 4.2 -- 30 -- 0.4
Oil analyses 3.2 180 -- -- -- 1.8 140 -- -- --
mc - moisture content in paper (%); - oil conductivity at 27 C; C
0
Geometrical
capacitance (nF).
18
19
20
21
22
23
24
25
26
0,0026 0,0028 0,003 0,0032 0,0034 0,0036 0,0038
After purification
Before purification
y = 3,2699 + 5812,1x R= 0,99545
y = 2,6634 + 5919,4x R= 0,99849
l
n
(
R
e
s
i
s
t
i
v
i
t
y
)
1/T [1/K]
Figure 8.15. Temperature dependence of the oil resistivity in transformer DT1 before and
after purification. The slope corresponds to an activation energy of 0.5 eV.
FDS measurements performed on the transformers DT2 and DT3 revealed a
different insulation state. Since these two units were sealed transformers, oil
samples were not available for further analyses. Hence, modelling using the
96
X model, in which measured oil conductivity was one of the input
parameters, was also not carried out. Therefore, modelled parameters
presented in Table 8.6 are only from the X-Y model (MODS). FDS results,
presented for the purpose of comparison, as () and (), are shown in
Figure 8.16. Geometrical capacitance estimated by MODS was used to derive
these curves from the measured complex capacitance. Much higher losses
and a stronger dispersive character of the real () component of the
complex permittivity *() were found in the insulation of the used
transformer DT2. This indicates the necessity for taking preventive measures
to improve the state of insulation in this transformer.
Table 8.6. Modelled parameters for transformers DT2 andDT3.
pressboard
X-Y model
cellulose
X-Y model
ID
No.
mc X Y mc X Y
DT2 3.1 12 20 20 4 12 20 20
DT3 1.5 0.7 20 20 1.5 0.7 20 20
mc - moisture content in paper (%) - oil conductivity at 27 C.
97
10
-3
10
-1
10
1
10
3
10
0
10
1
Frequency (Hz)

'
DT2
DT3
10
-3
10
-1
10
1
10
3
10
-2
10
0
10
2
Frequency (Hz)

'
'
DT2
DT3
Figure 8.16. Dielectric response results for used (DT2) and spare (DT3) transformers.
8.5 Limitation on FDS measurements
It was revealed during this study that in some transformers difficulties appear
when attempting to perform FDS measurements across the main insulation.
The results presented in Figure 8.17 are from the FDS measurements made
on transformer T11.
As shown in the figure, the capacitance measured is extremely low (~40 pF),
for an oil-paper insulated power transformer rated 24 MVA. Moreover, the
measured loss C() was negative for frequencies higher than 10 Hz. The
98
results of FDS measurements made on other transformers of the same type
were similar. This type of behaviour can be expected when a grounded metal
plate is installed between the LV and HV windings. However, it was not
possible to find information on the internal construction of these
transformers, which would be necessary for further explanation of this
problem.
10
0
10
1
10
2
10
-11
10
-10
Frequency (Hz)
C

'

(
F
)
10
0
10
1
10
2
-3
-2
-1
0
1
x 10
-10
Frequency (Hz)
C

'
'

(
F
)
Figure 8.17. Measured capacitance C() and loss C() of transformer T12.
Chapter 9: Conclusion
99
9. Conclusion
The purpose of this study was the further investigation of possibilities
provided by the use of dielectric response measurements for the diagnostics
of oil-paper insulating systems. The work was focused on the application of
frequency domain spectroscopy (FDS) measurements. The majority of the
measurements were performed on field-installed power and distribution
transformers in Sri Lanka, all of them owned by the Ceylon Electricity Board
(CEB). Some of the measurements were performed in a laboratory
environment too. The latter were carried out on both transformers and
oil-impregnated pressboard samples.
Presently, when using results of FDS measurements for estimating the
moisture content in transformer insulation the so called X-Y model is
utilised. This model takes into account the complex nature of transformer
insulation, i.e. the presence of barriers and spacers, and for this reason there
is a need for knowing the geometrical design of the transformer in detail.
However, power utility companies, such as CEB, usually do not have access
to such data. Subsequently, searching for possibilities to interpret FDS data
with limited information on geometrical design of the insulation was
desirable. Furthermore, a sensitivity analysis of the X Y model revealed
that influences of moisture content, oil conductivity and spacer content on the
dielectric permittivity of the transformer insulation system at 1 kHz were
relatively low. In addition, it was recognised that the influence of spacer
content on the total dielectric response in the X-Y model was also low. These
two effects were observed when the barrier content was within the range of
20 - 50 %, spacer content was within the range of 15 - 25 %, oil conductivity
was between 10 - 400 pS/m at 25 C, and moisture content in pressboard was
varied from 0.2 to 5 %. In the large population of the power transformers, the
values of these parameters fall within the specified limits. Based on these
observations, a simplified model, called the X model, was introduced. In the
X model, the presence of spacers in the transformer insulation has been
neglected. On the other hand, an additional measurement on oil sample has
been introduced for estimating oil conductivity, which is then used in the
modelling.
The results presented show that moisture contents estimated by means of the
X model were fairly close to the values obtained from the X - Y model.
Chapter 9: Conclusion
100
However, both these estimates were usually a bit higher than the ones
obtained from oil analyses. It is therefore necessary to continue
measurements of the frequency response of pressboard under different
conditions. There is also a need for checking the models against a much
larger population of power transformers by comparing the modelling results
of FDS measurements with results of KFT analyses. Such comparisons
would be helpful in improving the quality of the modelling. One should also
compare the estimated pressboard contents from the X model with the actual
barrier contents when there is potential for obtaining data on the internal
construction of transformers. This might help to assess the validity of all the
assumptions made and the accuracy of derived quantities.
Most of the tested field transformers were identified as wet transformers.
Therefore, necessary precautions must be taken to avoid further ageing in
these transformers. Furthermore, CEB is advised to set up a proper and
well-organised database on maintenance and measurement history of all the
power transformers they use.
Chapter 10: Future Work
101
10. Future Work
Based on the conclusions presented in the previous section, we can point out
the following areas for future study which, when completed, will further
increase our understanding of the behaviour of oil paper insulation in power
transformers and will improve the quality of the interpretation of diagnostic
measurements.
Further frequency domain spectroscopy (FDS) measurements on
pressboard samples under broadly varying conditions, such as different
moisture contents, differently aged and impregnated with oils with
different qualities.
Further measurements of field-installed power transformers for
calibrating the results of dielectric response analyses against results of
detailed oil analyses.
Further Study of the effects of thermally driven unequal moisture
distribution in transformer insulation on the results of dielectric response
measurements.
Extending the frequency range used for the measurements, especially for
analysing the response of pressboard at higher frequencies.
Investigating influences of electrical noise generated by other high
voltage units operating near the transformer under study and the
influence of parasitic creepage on the results of FDS measurements.
Attempts must be undertaken to identify reasons for apparent difficulties
in performing FDS measurements on some types of transformers.
Chapter 11: References
103
11. References
[1] S. M. Gubanski, P. Boss, G. Csepes, V. D. Houhanessian, J. Filippini,
P. Guuinic, U. Gafvert, V. Karius, J. Lapworth, G. Urbani, P. Werelius, and
W. S. Zaengl, "Dielectric response methods for diagnostics of power
transformers", Electra, No. 202, pp. 23-34, June, 2002
[2] A. K. Jonscher, Dielectric relaxation in solids, 2nd ed, Chelsea
Dielectrics Press Limited, London, UK, 1996
[3] U. Gafvert, H. Kols, and J. Marinko, "Simple method for determining
the electrical conductivity of dielectric liquids", Nordic IS, Helsinki, Finland,
1986, pp. 23:1 - 23:5, 1986
[4] A. Helgeson, "Analysis of Dielectric Response Measurement
Methods and Dielectric Properties of Resin-Rich Insulation During
Processing", PhD., Kungle Tekniska Hgskolan, Stockholm, 2000
[5] R. Eriksson and S. M. Gubanski, "Condition assessment of HV
apparatus through measurement of dielectric response", Jubilee congress of
the University of Peradeniya, Peradeniya, Sri lanka, 2000
[6] M. C. Lessard, L. Van Nifterik, M. Masse, J. F. Penneau, and R.
Grob, "Thermal aging study of insulating papers used in power
transformers", Proceedings of Conference on Electrical Insulation and
Dielectric Phenomena - CEIDP '96, New York, NY, USA, pp. 854-9 vol.2,
1996
[7] A. M. Emsley, X. Xiao, R. J. Heywood, and M. Ali, "Degradation of
cellulosic insulation in power transformers. Part 3: effects of oxygen and
water on ageing in oil", IEE Proceedings-Science, Measurement and
Technology, Vol. 147, No. 3, pp. 115-19, 2000
[8] R. Neimanis, "On estimation of Moisture Content in Mass
Impregnated Distribution Cables", PhD. Thesis, KTH, Stockholm, Sweden,
2001
[9] J. B. Whitehead, Impregnated paper insulation, John Willey & Sons,
New York, 1935
[10] D. Kind, High Voltage Technology, Vieweg, 1985
104
[11] L. Centurioni and G. Coletti, ''Transformer Insulation'', Wileys
Encyclopedia of Electrical and Electronics Engineering online, 2000
[12] T. O. Rouse, "Mineral insulating oil in transformers", IEEE Electrical
Insulation Magazine, Vol. 14, No. 3, pp. 6-16, 1998
[13] I. Fofana, V. Wasserberg, H. Borsil, and E. Gockenbach,
"Retrofilling conditions of high voltage transformers", IEEE Electrical
Insulation Magazine, Vol. 17, 2001
[14] P. Griffin and J. D. Christie, "Effects of Water and Benzotriazole on
Electrostatic charge generation in Mineral Oil/Cellulose Systems", Static
Electrification in Power Transformers, June 1993, 1993
[15] C. A. Eckelman, ''Wood Moisture Calculation'', Purdue
University,Department of Forestry and Natural Resouces,
http://www.ces.purdue.edu/extmedia/FNR/FNR-156.html, 2002
[16] Y. Du, M. Zahn, B. C. Lesieutre, A. V. Mamishev, and S. R.
Lindgren, "Moisture equilibrium in transformer paper-oil systems", IEEE
Electrical Insulation Magazine, Vol. 15, No. 1, pp. 11-20, 1999
[17] V. Sokolov, P. Griffin, and B. Vanin, "Moisture equilibrium and
moisture migration within transformer insulation systems", CIGRE WG
12.18 '' Life management of transformers '', Draft 3
[18] "IEEE guide for diagnostic field testing of electric power apparatus -
part 1: oil filled power transformers, regulators, and reactors", IEEE std.62-
1995, 1995
[19] S. Itahashi, H. Sakurai, H. Mitsui, and M. Sone, "Analysis of state of
water in oil impregnated Kraft-paper and its effect on conduction
phenomena", Proceedings of 1993 IEEE 11th International Conference on
Conduction and Breakdown in Dielectric Liquids (ICDL '93), New York,
NY, USA, pp. 472-6, 1993
[20] Y. Du, A. V. Mamishev, B. C. Lesieutre, M.Zahn, and S. H. Kang,
"Moisture Solubility for Different Conditioned Transformer Oils", IEEE
transactions on Dielectrics and Electrical Insulation, Vol. 8, No. 5, pp. 805 -
811, October 2001
[21] R. B. Kaufman, C. H. Shimansky, and E. J. McFadien, "Gas and
moisture equilibrium in transformer oil", Transactions AIEE, Vol. 74, pp.
111, 1955
105
[22] T. V. Oommen, "Moisture equilibrium in paper-oil insulation
systems", Proceedings of the 16th Electrical/Electronics Insulation
Conference, New York, NY, USA, pp. 162-6, 1983
[23] P. Griffin, C. M. Bruce, and J. D. Christie, "Comparison of water
equilibrium in silicon and mineral oil transformers", Minutes of the Fifty-
Fifth Annual international conference of Doble clients, pp. sec. 10 - 9.1, 1988
[24] K. Farooq, "The effect of particulate and water contamination on the
dielectric strength of insulating oils", Conference Record of the 1996 IEEE
International Symposium on Electrical Insulation, pp. 728-732 vol.2, 1996
[25] S. Itahashi, H. Mitsui, T. Sato, and M. Sone, "State of Water in
Hydrocarbon Liquids and its Effect on Conductivity", IEEE transactions on
Dielectrics and Electrical Insulation, Vol. 2, No. 6, pp. 1117 - 1122,
December 1995
[26] J. Fabre and A. Pichon, "Deteriorating Processes and Product of
Paper in Oil. Application to Transformers", CIGRE, Paris, France, pp. Paper
137, 1960
[27] V. Jaakkola, O. Jarvinen, M.-L. Surakka, R. Andersson, and M.
Lahtinen, "The influence of moisture on the dielectric strength and aging of
oil-paper insulation", Nordiskt Symposium om Elektriska Isoleringar:
NORD-IS 86 (Nordic Symposium on Electrical Insulation: NORD-IS 86),
Trondheim, Norway, pp. 8/1-10, 1986
[28] M. Darveniza, D. J. T. Hill, T. T. Le, T. K. Saha, and B. Williams,
"Chemical degradation of cellulosic insulation paper for power transformers",
Proceedings of 1994 4th International Conference on Properties and
Applications of Dielectric Materials (ICPADM), New York, NY, USA, pp.
780-3 vol.2, 1994
[29] W. J. McNutt, "Insulation thermal life considerations for transformer
loading guides", Power Delivery, IEEE Transactions on, Vol. 7, No. 1, pp.
392-401, 1992
[30] H. P. Moser and V. Dahinden, "Transformerboard 2", H.Weidmann
AG, CH-8640 Rappersvil, Switzerland, 1988
[31] P. R. Krishnamoorthy, K. R. Krishnaswamy, S. Vijayakumari, and K.
Kumar, "Ageing of mineral oils-a diagnostic study", Properties and
Applications of Dielectric Materials, 1991., Proceedings of the 3rd
International Conference on, pp. 59-62 vol.1, 1991
106
[32] R. Neimanis, "Dielectric Diagnostics of Oil-Paper Insulated Current
Transformers", Technical report for Lic., Chalmers University of
Technology, Gteborgy, Sweden, 1997
[33] A. White, "The desired properties and their effect on the life history
of insulating papers used in a fluid-filled power transformer", IEE
Colloquium on 'Assessment of Degradation Within Transformer Insulation
Systems' (Digest No.184), London, UK, pp. 4/1-4, 1991
[34] R. Gilbert, J. Jalbert, and P. Tetreault, "Bias assessment of current
technologies used for the determination of low level of moisture in mineral
oil samples", Analytical chemistry, Vol. 73, No. 3, pp. 520-526, 2001
[35] V. D. Houhanessian, "Measurement and Analysis of Dielectric
Response in Oil-Paper Insulation Systems", PhD., Swiss Federal Institute of
Technology, Zurich, 1998
[36] H. Edin and U. Gafvert, "Harmonic content in the partial discharge
current measured with dielectric spectroscopy", Electrical Insulation and
Dielectric Phenomena, 1998. Annual Report. Conference on, pp. 394-398
vol. 2, 1998
[37] P. J. Harrop, Dielectrics, Butterworth, London, 1972
[38] E. Robles, "Field Diagnostic Testing of Power Generators and
Transformers using Modern Technique", HV testing,monitoring and
Dignostic Workshop, Alexandria, Virginia, 13 & 14 September 2000, pp.
12.1 - 12.6, 2000
[39] V. Der Houhanessian and W. Zaengl, "On-site diagnosis for power
transformers: relaxation currents reveal the condition of insulation of high-
voltage components and installations", Bulletin des Schweizerischen
Elektrotechnischen Vereins & des Verbandes Schweizerischer
Elektrizitaetswerke, Vol. 87, No. 23, pp. 19-28, 1996
[40] V. der Houhanessian and W. S. Zaengl, "On-site diagnosis of power
transformers by means of relaxation current measurements", Conference
Record of the 1998 IEEE International Symposium on Electrical Insulation,
New York, NY, USA, pp. 28-34 vol.1, 1998
[41] U. Gafvert, L. Adeen, M. Tapper, P. Ghasemi, and B. Jonsson,
"Dielectric spectroscopy in time and frequency domain applied to diagnostics
of power transformers", Proceedings of the 6th International Conference on
Properties and Applications of Dielectric Materials, 2000
107
[42] T. Leibfried and A. J. Kachler, "Insulation diagnostics on power
transformers using the polarisation and depolarisation current (PDC)
analysis", Electrical Insulation, 2002. Conference Record of the 2002 IEEE
International Symposium on, pp. 170-173, 2002
[43] J. J. Alff, V. Der Houhanessian, W. S. Zaengl, and A. J. Kachler, "A
novel, compact instrument for the measurement and evaluation of relaxation
currents conceived for on-site diagnosis of electric power apparatus",
Conference Record of the 2000 IEEE International Symposium on Electrical
Insulation, Piscataway, NJ, USA, pp. 161-7, 2000
[44] ''Electrical Insulation Diagnostic System PDC -Analyser-1MOD'',
ALFF Engineering, Switzerland, http://www.alff-engineering.ch/index.htm,
2002
[45] A. Bognar, G. Csepes, L. Kalocsai, and I. Kispal, "Spectrum of
polarization phenomena of long time-constant as a diagnostic method of oil-
paper insulating systems", Properties and Applications of Dielectric
Materials, 1991., Proceedings of the 3rd International Conference on, pp.
723-726 vol.2, 1991
[46] G. Csepes, I. Hamos, R. Brooks, and V. Karius, "Practical
foundations of the RVM (recovery voltage method for oil/paper insulation
diagnosis)", Electrical Insulation and Dielectric Phenomena, 1998. Annual
Report. Conference on, pp. 345-355 vol. 1, 1998
[47] J. Lapworth and R. J. Heywood, "The determination of the dryness of
transformer insulation : recent experience with polarisation tests",
International conference on power transformers, Bydgoszcz, poland,
2001.09.05, pp. 84-95, 2001
[48] U. Gafvert, G. Frimpong, and J. Fuhr, "Modelling of dielectric
measurements on power transformers", Proceedings of 37th Sessions (Large
High Voltage Electric Systems), 1998
[49] U. Gafvert, "Conditionm Assessment of Insulation Systems", Nordic
Insulation Symposium, Bergen, June 10 - 12, 1996, pp. 1 - 20, 1996
[50] A. Helgeson and U. Gafvert, "Dielectric response measurements in
time and frequency domain on high voltage insulation with different
response", Proceedings of 1998 International Symposium on Electrical
Insulating Materials, 1998
108
[51] "Supervision and maintenance guide for mineral insulating oils in
electrical equipment", Report ISSN IEC 6422, IEC, 1989
[52] "Statistical Digest 2001", Ceylon Electricity Board, Colombo, 2002
[53] "User's manual for insulation diagnostic system IDA 200",
Programma Electric AB, 2002
[54] "Methods of sampling liquid dielecrics", IEC, IEC 60475, 1974
[55] "Insulating liquids-oil impregnated paper & pressboard -
Determination of water by automatic coulometric Karl Fischer Titration",
IEC, IEC 60814, 1997
109

Lic Thesis

Загружено:

Сведения о документе

Исходное описание:

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Lic Thesis

Загружено:

Авторское право:

Доступные форматы

THESIS FOR THE DEGREE OF LICENTIATE OF ENGINEERING

Application of Dielectric Spectroscopy for

represents polarisation due to fast mechanisms and P

are static and high frequency relative permittivities of the

and the dielectric response function f(t).

) cannot be obtained since the loss factor is a ratio of real

Вам также может понравиться