414

Di e l e c t r i c Re s p o n s e Di a g n o s e s
Fo r Tr a n s fo r m e r Wi n d i n g s

Wo r k i n g Gr o u p
D1 .0 1 (T F 1 4 )

April 2010
 WG D1.01 (TF 14)

Dielectric Response Diagnoses

for Transformer Windings

Members
S.M. Gubanski, Chair (SE), J. Blennow (SE), B. Holmgren (SE),
M. Koch (AT), A. Kuechler (DE), R. Kutzner (DE), J. Lapworth (GB),
D. Linhjell (NO), S. Tenbohlen (DE), P. Werelius (SE)

Copyright © 2010
“Ownership of a CIGRE publication, whether in paper form or on electronic support only infers
right of use for personal purposes. Are prohibited, except if explicitly agreed by CIGRE, total or
partial reproduction of the publication for use other than personal and transfer to a third party;
hence circulation on any intranet or other company network is forbidden”.

Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it
accept any responsibility, as to the accuracy or exhaustiveness of the information. All implied
warranties and conditions are excluded to the maximum extent permitted by law”.

ISBN: 978- 2- 85873-101-5

 2

Content
DIELECTRIC RESPONSE DIAGNOSES FOR TRANSFORMER WINDINGS ........................ 1
ABSTRACT ............................................................................................................................... 9
1 INTRODUCTION .............................................................................................................. 10
1.1 BASICS ....................................................................................................................... 10
1.2 OBJECTIVE .................................................................................................................. 11
1.3 SCOPE ........................................................................................................................ 11
1.4 DIELECTRIC RESPONSE ................................................................................................ 12
1.5 MEASURING METHODS ................................................................................................. 13
2 MODELLING DIELECTRIC RESPONSE OF POWER TRANSFORMER ...................... 15
2.1 X-Y MODEL ................................................................................................................. 15
2.2 INFLUENCE OF PRESSBOARD TYPE, OIL QUALITY ON THE DIELECTRIC RESPONSE OF
LABORATORY PRESSBOARD MODELS ...................................................................................... 18
2.2.1 Influence of pressboard ..................................................................................... 18
2.2.2 Influence of oil conductivity ................................................................................ 20
2.2.3 Influence of acids ............................................................................................... 21
2.2.4 Practical aspects and error estimates................................................................ 24
3 MEASURING SYSTEM CONFIGURATIONS .................................................................. 28
3.1 UST CONFIGURATION .................................................................................................. 28
3.2 GST CONFIGURATION .................................................................................................. 30
3.3 CONTRIBUTION OF DIFFERENT PARTS OF THE INSULATION TO THE RESPONSE ................. 31
3.3.1 Ungrounded measurement CHL ........................................................................ 32
3.3.2 Measurements to ground CH and CL ................................................................ 33
3.3.3 Influence of winding arrangement...................................................................... 33
4 EXECUTION OF MEASUREMENTS ............................................................................... 35
4.1 GENERAL INSTRUCTIONS ............................................................................................. 35
4.1.1 Suggested check-list of transformer data to be registered for tested transformer
from its nameplate and other sources (instruction manual, construction drawings, etc.) 37
4.1.2 Supporting information about the transformer state .......................................... 37
4.1.3 Suggested check-list for execution of dielectric response measurements on
power transformers .......................................................................................................... 38
4.2 SUPPLEMENTARY RECOMMENDATIONS AND COMMENTS FOR FDS MEASUREMENTS......... 39
4.3 SUPPLEMENTARY RECOMMENDATIONS AND COMMENTS FOR PDC MEASUREMENTS ........ 40
5 CASE STUDIES ............................................................................................................... 41
5.1 KARL FISCHER TITRATION AS A SCALE FOR DIELECTRIC RESPONSE METHODS ................. 41
5.1.1 The KFT technique ............................................................................................ 41
5.1.2 Reliability and comparability of KFT analyses - Round Robin Test ................... 41
5.1.3 Consequences for interpretation of dielectric response measurements ........... 43
5.2 DIELECTRIC RESPONSE DIAGNOSTICS ........................................................................... 44
5.2.1 Comparison of moisture determination by means of dielectric response
measurements and KFT ................................................................................................... 44
5.2.2 Moisture determination in a heavily aged transformer........... ..........................45
5.2.3 Oil exchange and drying of power transformers ................................................ 48
5.3 INFLUENCE OF EXTERNAL ENVIRONMENT....................................................................... 51
5.3.1 Presence of different leakage paths .................................................................. 51

6 REFERENCES ................................................................................................................. 56
 3

Executive summary
The dryness of the oil-paper insulation systems in power transformers is a key factor
in both their short and long term reliability since moisture has deleterious effects on
dielectric integrity and insulation ageing rates. Traditionally, moisture in oil
measurements have been used to estimate the dryness of transformers, but this has
not been a particularly reliable approach because of possible sampling and analysis
errors and because moisture distributes unequally between cellulose and oil, the
greater part residing within the solid insulation, with the water content in the oil being
very dependent on temperature and also oil condition, and therefore not simply
correlated to solid insulation dryness. The moisture content in the solid cellulose
insulation is therefore the key parameter. Unfortunately it is impractical to determine
this directly by taking paper samples. Therefore, recent attention has been directed
to methods of determining moisture content by measuring the effects of the moisture
on electrical properties of the insulation. Rather than traditional measurements of
power frequency dissipation factor, the variation of various dielectric parameters in
the time and frequency domains have been studied in an attempt to isolate the
effects of moisture.

A previous CIGRE Working Group, 15.01.09 [1], evaluated the three foremost of
these so-called ‘dielectric response’ techniques:
• Return voltage measurements, sometimes also called recovery voltage –
(RVM)
• Polarisation and depolarisation current variation in time domain - (PDC)
• Capacitance and dissipation factor variation with frequency – (FDS)

It was concluded that all three techniques reflect the same fundamental polarisation
and conduction phenomena in transformer insulation, the special feature of which is
a combination of oil gaps and solid insulation, and as a consequence are influenced
strongly by both solid insulation moisture content and oil condition, and also, but less
strongly, by the geometry of the solid and liquid insulations. Mathematical modelling
was seen as the key to determining how measured responses are affected by oil
conductivity and moisture content. By contrast, the simple relationship claimed
between RVM dominant time constant and solid insulation moisture content was
shown to be flawed. Although these dielectric response techniques were considered
promising, it was concluded that solid insulation moisture contents determined from
measured responses still needed to be verified by comparison with basic chemical
measurements, and also that the influences of different types of pressboard/paper
and ageing products still had be to determined.

The present Working Group has investigated the influences of different types of solid
insulation and ageing products, provided detailed guidance on the practicalities of
making dielectric response measurements, and collected case examples illustrating
the value of such measurements. The work has focussed almost exclusively on the
two techniques which allow the clearest discrimination between the effects of oil
condition and solid insulation moisture content: FDS and PDC. It is concluded that
different types of pressboard and ageing products, in particular low molecular weight
acids, can have a significant effect on measured dielectric responses. As a result, as
shown by a case example, the solid insulation moisture content in aged insulations
can be overestimated and improvements to interpretation schemes are desired. As
regards verifying dielectric response moisture determinations by comparison with
basic chemical measurements, it is concluded that there are too many uncertainties
associated with the application of equilibrium diagrams based on moisture estimation
 4

in oil obtained by Karl Fischer titration for this to be a useful reference, but relative
saturation measurements are seen as a promising alternative.

Measuring configurations in time and frequency domains

When estimating the moisture content in the solid part of traditionally designed
transformers, one usually focuses on the main insulation, i.e. the insulation between
the coaxially arranged high voltage (HV) and low voltage (LV) windings. The
capacitance of this insulation is commonly denoted as CHL. Other capacitances of
interest can, for example, be winding capacitances to ground (tank, core, etc.), i.e.
CL or CH.
Dielectric response measurements, independently if in time or frequency domain, can
be performed within different configurations, UST (Ungrounded Specimen Test)
where the ground act as natural guard or GST (Grounded Specimen Test) with or
without guard. Examples of instrument connections for these configurations are
shown in Figures S1 and S2.

Figure S1: UST connection of dielectric response measuring system to a

transformer for CHL configuration.

In case of auto-transformers having usually a tertiary winding, provided it is not

internally earthed, the measurements of main to tertiary insulation, CHT, are possible
and preferred, as illustrated in Figure S3.

Time domain measurements can directly be related to the polarisation and

conduction phenomena in the investigated insulation system. Through measuring
polarisation current i p (t ) and depolarisation current id (t ) (PDC), information about
its dielectric response function f (t ) is obtained, provided it is charged for a
sufficiently long time. A rule of thumb says that the test object should be charged for
at least 5 to 10 times as long as it is depolarised. It also is possible to estimate the dc
conductivity, σ dc , of the system from the difference between the polarisation and
depolarisation currents. In an oil-barrier system most interesting information is
obtained for longer times up to a few thousands seconds. The barriers are charged
via the oil ducts immediately after voltage step application. Therefore the initial
 5

current values and the main time constant of the current curves are directly related to
the conductivity of oil in the main ducts. The polarisation and depolarisation current
and the difference between polarisation and depolarisation curves at longer times are
directly related to the conductivity of the paper/pressboard part of the insulation
system and its water content as indicated in Figure S4(a).

Figure S2: GST connection of dielectric response measuring system to a

transformer for CH configuration.

Figure S3: UST connection of dielectric response measuring system to an

autotransformer with tertiary winding.

An equivalent method in the frequency domain is to measure the responses from

sinusoidal excitations at different frequencies (FDS), say from 0.1 mHz to 1 kHz. On
 6


this basis the complex relative permittivity ε (ω ) at the frequency of the applied field,
assuming a capacitive test object, can be found. It is important to notice that the
imaginary part of the complex relative permittivity, ε ′′(ω ) , (loss part) contains both the
resistive (dc conduction) losses and the dielectric (polarisation) losses. Another way
of presenting the measured information of a FDS is to use the dissipation
factor tgδ(ω) = ε′′(ω) ε′(ω) . As shown in Figure S4(b), at the very low (<10-2 Hz)
frequency range the response is mainly influenced by the properties of the
pressboard. In this range the imaginary part, ε ′′(ω ) of the complex relative
permittivity, has usually a slope similar to that of the real part ε ′(ω ) , which represent
the polarization behavior of the solid part of transformer insulation, modified by water
content. The same is true for the higher frequency range (>10 Hz). The central part
of the response is, on the other hand, influenced by the properties of the oil, mainly
by its conductivity. Although insulation geometry is indicated to have a significant
influence on responses in particular parts of the responses, it still has an effect over
the entire frequency and time ranges.
The characteristic boundaries in PDC and FDS spectra differ for various moisture
contents, conducting ageing by-products and temperatures.

1000
moisture of
Current (nA)

high cellulose
100 and aging
low
10
conductivity

Ipol
insulation
geometry

1 Idep high
low
oil

0.1
1 10 100 1000 10000
Time (s)

(a)
10 100
moisture of
Dissipation factor

cellulose, ε'
ε', ε ''

aging 10
1
high
high ε" ε'
low 1
0.1 low
0.1
moisture of
and aging

insulation
geometry
cellulose

0.01 high
0.01
oil ε"
conductivity low
0.001 0.001
0.0001 0.001 0.01 0.1 1 10 100 1000 0.0001 0.001 0.01 0.1 1 10 100 1000
Frequency (Hz) Frequency (Hz)

(b)
Figure S4: Main features of PDC (a) and FDS (b) responses in oil-
paper/pressboard barrier systems.
 7

Execution of measurements
A successful dielectric response measurement on a power transformer requires
adequate planning and coordination with its owner for finding information on winding
geometry and on means to disconnect the unit from network. The objectives of the
measurements, and thereby the resulting connection of the instrumentation to the
transformer, should preferably be decided forehand.
Good understanding of the factors that may influence the measurements under field-
conditions, and therefore also the interpretation of dielectric response results, is of
crucial importance for the diagnosis reliability. Constant and preferably not too low
temperature is advantageous for the interpretation quality – at higher temperatures
the time necessary for the measurements can be shortened. It is recommended to
make guarded CHL measurements between main transformer windings. In the case
of CH and CL measurements, it is also important to make sure that the transformer
bushings are dry and clean.
Before starting the measurements the transformer should preferably be completely
disconnected from the station power connections, i.e. all connections to the bushings
should be dismantled, and the station connections should be properly grounded. The
bushing terminals of the individual windings should, if possible, be shortened
together by conductors kept away of earthed structures. For CH and CL
measurements internal leakages and creeps from bushing surfaces remain as the
disturbing factors and therefore should be considered in the analyses of the results.
If any of the windings has a neutral point connected either directly to ground or
through impedance to ground, also this connection should be removed. The neutral
bushing should preferably be interconnected with the phase bushings during the
measurements. Sometimes a surge arrester is connected to the neutral point, and
preferably, this connection should also be isolated.
The measurements are often affected by interference in the substation, such as
parasitic leakage currents and induced electromagnetic disturbances. It is therefore
important to secure proper grounding connections of the transformer tank during the
measurements to minimize the ground interference. Proper placement of connecting
leads can also minimize the influence of capacitive coupled or radiated noise. Use of
shielded connecting cables is recommended in noise environments.
The temperature of the insulation system in transformer has a great influence on the
results of dielectric response measurements. Transformer temperature often varies
and it is not equal inside the tank – higher at the top than at the bottom. If the
measurements are made just after taking the transformer out of operation, its
temperature will be slowly decreasing. It is recommended to register the temperature
of oil just before starting the dielectric response measurements. The most accurate
way to determine the oil temperature is to take top and bottom oil samples and
measure the temperatures directly on-site in the sampled oil. After opening the tap,
cold oil flows out first, thus waiting for sufficient time is recommended in order to get
a representative sample. If oil sampling involves too much effort, indications from the
built-in temperature gauges may be used, though it should be bear in mind their
readings depend on location of the temperature probe. Alternatively, the average
winding temperature can be calculated from a comparison of the actual winding
resistance to a measurement at known temperature (e.g. factory acceptance test,
FAT).
It is also recommended to register the ambient temperature, relative humidity and
weather conditions in the station. In addition, making photographic documentation on
the transformer and the measuring setup as well as recording the transformer
nameplate information is advisable.
 8

Modelling of dielectric response

In core type transformers, the main insulation between windings consists of
cylindrical pressboard barriers in series with oil ducts and sticks. Interpretation of the
results of dielectric response measurements aiming at evaluating the amount of
moisture contained in the solid part of transformer insulation can be obtained by
means of so called X-Y model, which represents volume fractions of the insulation
system components.
Figure S5 shows the basic structure of a main insulation duct in a core type power
transformer and its X-Y representation, where X and Y often vary between 0.2 - 0.5
and 0.15 - 0.25, respectively. For proper modelling, the configuration of the
measured insulation system of the transformer should be taken into account. For
core type transformers the geometrical capacitance between windings can be
estimated, based on construction drawings, as the capacitance of concentric
cylinders. For shell type transformers the calculation is more complex, but still can be
done assuming the construction details are known. Further, information on dielectric
response of pressboard at different humidity contents and of oil is also necessary.
The total response of X-Y model at particular temperature, given moisture content in
pressboard and the conductivity of oil is in detail described in [1].

HV-winding
Barrier Y
Stick
Oil Oil Sticks
LV-winding Core
Barriers X
.
(a) (b)
Figure S5: Basic structure of the insulation duct seen from above (a) and its X-
Y representation (b).

Error analysis
Since the analysis and interpretation of dielectric response data is based on a
comparison between measured dielectric properties and a library of model data
obtained in laboratory on pressboard and paper samples, the quality of the latter
determines the accuracy of moisture analyses.
It is shown that different factors influence the quality of result interpretation. The type
of pressboard material (density) and origin as well as oil conductivity and acid
content in cellulose have an impact on the response. The resulting errors have been
estimated and are illustrated in Table A below. The errors found are “absolute”, i.e.
independent of the actual moisture level in the insulation. For example, for a dry
transformer the relative errors will end up being high. Both over and under estimates
of moisture are possible. It is important to be aware of these errors and take them
into account in the assessment of the transformer insulation quality. Certainly the
influence of the different errors will be judged differently from situation to situation
dependent on transformer vintage, type, service conditions, service record,
measurement conditions etc.
 9

Table A: Error sources and their maximum magnitudes in estimation of

moisture in cellulose (given in percent water by weight of cellulose)
Error source Approx. error Comment
Type of pressboard 0.7 % Assuming all insulation
HD => LD or v.v. changed.
Assuming HD board when
actual board is LD causes
underestimation of moisture.
Type of HD pressboard 0.5 %
Low molecular weight 0.08 -0.15 % per NV = 1 1 mgKOH/g pressboard
(hydrophilic) acids mgKOH/g in pressboard corresponds to about 0.01
For heavily acidic field mgKOH/g oil for this type of
aged transformers the acids. Corresponding total NV
error range is believed to of oil cannot be specified as
be up to 1% the percentage of low
molecular weight acids vary
very much between
transformers. The presence of
acids causes overestimation
of moisture.
Incorrect insulation Up to 0.6%
geometry parameters X
and Y
Temperature 0.07 %/K Calculated at a temperature
of 310 K, but does not seem
to be significantly different
over a temperature range of
273 – 353 K. Real temp
higher than assumed temp
causes moisture
overestimation.

[1] S.M. Gubanski, P. Boss, G. Csepes, V.D. Houhanessian, J. Filippini, P. Guuinic, U.

Gäfvert, V. Karius, J. Lapworth, G. Urbani, P. Werelius and W.S. Zaengl - "Dielectric
Response Methods for Diagnostics of Power Transformers”, CIGRE Technical Brochure, No.
254, Paris 2004.
 10

1 INTRODUCTION
1.1 Basics
The insulation system of power transformers consists of oil and cellulose.
Both materials generally change their dielectric properties during the life of the
transformer and among factors contributing mostly to the degradation of
transformer insulation moisture plays an important role. Presence of water in
solid part of the insulation, even in small concentrations, increases its aging
rate, lowers the admissible hot spot temperature and increases the risk of
bubble formation. In addition, moisture reduces the dielectric strength of
transformer oil as well as the inception level of partial discharge activity.
Methods for evaluating moisture content and ageing in pressboard and paper
are thus of great interest to transformer owners. Over the recent years,
attempts have been undertaken to replace the traditional way for determining
moisture content in pressboard and paper in insulation of power transformers,
based on chemical analyses of oil according to IEC 60422, by new methods
utilising dielectric response measurements. One of the main reasons driving
ahead the development of the new techniques is related to the inferior
accuracy of the former methodology, especially at lower temperatures. In
addition, the electrical measurements are simpler and it is possible to perform
them on-site.
CIGRE Task Force D1.01.09 – Dielectric Response Methods for Diagnostics
of Power Transformers - presented in 2004 [ 1] conclusions regarding the
state of the knowledge on the applicability of the dielectric response
techniques. The techniques compared were:
• Dielectric spectroscopy in time domain, i.e. measurements of
polarisation and depolarisation currents (PDC) and recovery (or
return) voltage (RVM),
• Dielectric frequency domain spectroscopy (FDS), i.e. measurements
of electric capacitance C and dissipation factor tgδ in dependency of
frequency.
This work confirmed that the dielectric response measurements can provide
valuable information on the state of oil-paper insulation in power transformers,
in particular the moisture content. All the dielectric response methods
compared (RVM, PDC and FDS) reflect the same fundamental polarisation
and conduction phenomena in transformer insulation, the special feature of a
combination of oil gaps and solid insulation. The oil-paper insulation system
as a composite of two different dielectric media, where an insulating liquid
with ionic conduction is mixed with a less conducting impregnated solid
(pressboard or paper), has its own dielectric response which not only reflects
properties of each material but also the way they are combined. Therefore the
influence of oil gaps, the condition of the oil, specifically its conductivity, has a
significant impact on dielectric response, and this must be taken into account
when attempting to estimate moisture contents in the solid insulation from the
results of all three methods.
 11

To further validate the indications of the dielectric response techniques the

work of CIGRE has continued. Related research activities were conducted at
various industrial and academic centres. Among them, an extensive
investigation program was carried out within a European research project
REDIATOOL – “Reliable Diagnostics of HV Transformer Insulation for Safety
Assurance of Power Transmission System” [ 2]. Particular attention has
concentrated on verifying the estimates of water content determined by the
dielectric response techniques in comparison with basic chemical
measurements. The influences of different types of pressboard and paper and
ageing products on dielectric response have also been addressed and
demonstrated for improving interpretation scheme of the dielectric response
measurements [ 3, 4, 5].
In this report the continuation of the work performed by CIGRE Task Force
D1.01.14 – Dielectric Response Diagnoses for Transformer Windings – is
summarized by presenting hands-on guidelines for performing meaningful
dielectric response measurements on power transformers. The work
described concentrates on diagnosing the moisture content in transformer
insulation system. The attempts to differentiate between the influence of
moisture and aging products in pressboard and paper on the dielectric
response still remains under consideration and the issue is not further
discussed here. A number of study cases are also reported in order to
exemplify different practical situations that should be considered when
performing diagnostic measurements of dielectric response on power
transformers.

1.2 Objective
The following objectives were considered when preparing this report:
• reporting on the present knowledge how the dielectric responses of
the materials in transformer insulation system depend on type,
manufacturer, moisture, contamination, aging, temperature, etc.,
• advising how practical measurements should be arranged to provide
best possible data for conclusive diagnosis,
• illustrating, by description of different study cases, difficulties and
possibilities arising from the applications of dielectric response
measurements.
• providing guidelines on correct interpretation of results of dielectric
response measurements.

1.3 Scope
The work reported in this document has limited its interest to investigations of
the influences of different types of solid insulation and ageing products on
interpretation of dielectric response measurements in terms of moisture
content in solid parts of transformer insulation. Detailed guidance on the
practicalities of making dielectric response measurements are provided and
collected case examples illustrate the value of such measurements. Although
 12

some companies claim they are able to detect by means of dielectric

response measurements other types of defects in transformers, such for
example bad shield contacts, the possibilities are not presented in this report
since relevant documentation has not yet been available and openly
evaluated.
Two dielectric response techniques (PDC and FDS) for which well developed
interpretation schemes based on mathematical models are available and
which take into account the separate influences of insulation system
geometry, oil conductivity and moisture.
Most of the work reported has been based on measurements between
windings, for which a guarded measurement setup is applicable, but
considerations on measurements of winding insulation to ground are included.
The results presented in this report were performed by means of
commercially available instruments, but this does not exclude the possibility of
using other dielectric response measuring systems, provided these are
suitable for application in a field environment and designed to follow
conditions formulated below.

1.4 Dielectric response

The purpose of this chapter is to formulate the basic relations for the dielectric
response characterization of electrical insulation. For more details on this
subject the readers are advised to reference literature [1, 6, 7, 8].
Under assumption that the considered insulation material behaves linearly
and is homogeneous and isotropic, the information obtained by performing
dielectric response measurements in either the time or the frequency domains
is the same. The results of measurements can therefore be transformed from
time to frequency domain and vice versa.
In time domain, the conductivity σ dc , the instantaneous (high frequency)
component of the relative permittivity, ε ∞ , and the dielectric response function,
f (t ) characterise the behaviour of a dielectric material [6]. The response
function, f (t ) is the response to a delta pulse. From this, the response - or
polarization, P can be calculated for any excitation shape, E (t ) through the
∞
convolution, P =
∫0 f (t )E (t − τ )dτ . It is also worth of notifying that current
measurements in time domain can directly lead to quantification or at least
estimation of f (t ) .

The Fourier transform of the response function f (t ) allows for the

transformation to frequency domain. It yields the complex susceptibility,

χ (ω ) = χ ′(ω ) − iχ ′′(ω ) . The relation between the susceptibility and the
permittivity is ε ′(ω ) = ε ∞ + χ ′(ω ) , ε ′′(ω ) = σ dc + χ ′′(ω ) and the dissipation
ε 0ω
factor is defined as tgδ = ε′′(ω) ε′(ω) . Therefore, in the frequency domain, the
dc conductivity, σ dc , the high frequency component of the relative permittivity,
 13


ε ∞ , and the complex dielectric susceptibility, χ (ω ) , characterise the dielectric
material, and as for the time domain, it is possible to find these parameters by
the measurements.
The oil-paper insulation system is a composite of two different dielectric
media, where an insulating liquid with ionic conduction is mixed with a lower
conducting impregnated solid (pressboard or paper). It is important to realise
that each has its own dielectric response and when putting them together the
total response will not only reflect properties of each material but also the way
they are combined. When these two media are put into contact (forming
interfaces), charge accumulation occurs at the interfaces due to the
differences between their electrical properties. This kind of polarisation is
called Maxwell-Wagner or interfacial polarisation [see for example 9, 10].

1.5 Measuring methods

Time domain measurements can directly be related to the polarisation and
conduction phenomena. Trough measuring polarisation current i p (t ) and
depolarisation current id (t ) (PDC) [7, 8], information about the dielectric
response function f (t ) is obtained, provided the test object is charged for a
sufficiently long time. To determine the response function f (t ) , it is easier to
use the depolarisation current id (t ) where there is no dc current present.
Many solid dielectric materials have dielectric response functions that
decrease slowly with time. A rule of thumb says that the test object should be
charged for at least 5 to 10 times as long as it is depolarised in order to get a
depolarisation current, which is proportional to the dielectric response
function. It also is possible to estimate the dc conductivity, σ dc , of the test
object from the difference between the polarisation and depolarisation
currents.
In an oil-barrier system most interesting information is obtained for longer
times up to a few thousands seconds. The barriers are charged via the oil
ducts immediately after voltage step application. Therefore the initial current
values and the main time constant of the current curves are directly related to
the conductivity of oil in the main ducts. The polarisation and depolarisation
current and the difference between polarisation and depolarisation curves at
longer times are directly related to the conductivity of the paper/pressboard
part of the insulation system and its water content. These features are
indicated in Figure 1(a).
 14

1000
moisture of

Current (nA)
high cellulose
100 and aging
low
10

conductivity
Ipol

insulation
geometry
1 Idep high
oil low
0.1
1 10 100 1000 10000
Time (s)

(a)

10 100
moisture of
Dissipation factor

cellulose, ε'

ε', ε ''
aging 10
1
high
high ε" ε'
low 1
0.1 low
0.1
moisture of
and aging

insulation
geometry
cellulose

0.01 high
0.01
oil ε"
conductivity low
0.001 0.001
0.0001 0.001 0.01 0.1 1 10 100 1000 0.0001 0.001 0.01 0.1 1 10 100 1000
Frequency (Hz) Frequency (Hz)

(b)
Figure 1: Main features of PDC (a) and FDS (b) responses in oil-
paper/pressboard barrier systems.

An equivalent method in the frequency domain is to measure the responses

from sinusoidal excitations at different frequencies (FDS), which is practically
an extension of the common C-Tan-Delta measurement at power frequency
and the measurements can be done from 0.1 mHz to 1 kHz. On this basis the

complex relative permittivity ε (ω ) at the frequency of the applied voltage,
assuming a capacitive test object, can be found. It is important to notice that
the imaginary part of the complex relative permittivity, ε ′′(ω ) , (loss part)
contains both the resistive (dc conduction) losses and the dielectric
(polarisation) losses. Another way of presenting the measured information of
a FDS is to use the dissipation factor tgδ(ω) = ε′′(ω) ε′(ω) .

As shown in Figure 1(b), at the very low (<10-2 Hz) frequency range the
response is mainly influenced by the properties of the pressboard. In this
range the imaginary part, ε ′′(ω ) of the complex relative permittivity, has
usually a slope similar to that of the real part ε ′(ω ) , which represent the
 15

polarization behavior of the solid part of transformer insulation, modified by

water content. The same is true for the higher frequency range (>10 Hz). The
central part of the response is, on the other hand, influenced by the properties
of the oil, mainly by its conductivity [1]. Although insulation geometry is
indicated to have a significant influence on responses in particular parts of the
responses, it still has an effect over the entire frequency and time ranges.
The characteristic boundaries in PDC and FDS spectra differ for various
moisture contents, conducting ageing by-products and temperatures.

2 MODELLING DIELECTRIC RESPONSE OF POWER TRANSFORMER

2.1 X-Y model
In core type transformers, the main insulation between windings consists of
cylindrical pressboard barriers in series with oil ducts and sticks. Interpretation
of the results of dielectric response measurements aiming at evaluating the
amount of moisture contained in the solid part of transformer insulation can be
obtained by means of so called X-Y model [1, 11], which represents volume
fractions of the insulation system components.

Figure 2 shows the basic structure of a main insulation duct in a core type
power transformer. The insulation consists of concentric cylinders made of oil
impregnated pressboard and axial sticks. The complex geometrical
arrangement shown in this figure can be simplified by separately combining
its components, i.e. barriers and spacers and oil ducts, into single material
blocks representing the components of the composite duct insulation, where:
total thickness of barriers
X =
width of the duct
total width of the spacers
Y=
periphery of the duct
In power transformers, X and Y often vary between 0.2 - 0.5 and 0.15 - 0.25,
respectively. The total response of X-Y model at particular temperature, given
moisture content in pressboard and the conductivity of oil is in detail
described in [1].

HV-winding
Barrier Y
Stick
Oil Oil Sticks
LV-winding Core
Barriers X
.
(a) (b)
Figure 2: Basic structure of the insulation duct seen from above (a) and
its X-Y representation (b).
 16

For proper modelling, the configuration of the measured insulation system of

the transformer should be taken into account. For core type transformers the
geometrical capacitance between windings can be estimated, based on
construction drawings, as the capacitance of concentric cylinders. For shell
type transformers the calculation is more complex, but still can be done
assuming the construction details are known. Further, information on
dielectric response of pressboard at different humidity contents and of oil is
also necessary.

Figure 3: Dielectric permittivity of pressboard containing different

amount of moisture (1%, 2%, 3% and 4%) at 50 ºC [3].
 17

polarisation polarisation polarisation polarisation

current (A) current (A) current (A) current (A)
100 n 100 n 100 n 100 n

w=4%
10 n 10 n 10 n 10 n
w=3%

1n 1n w=2% 1n 1n

w=1%
100 p 100 p 100 p 100 p

10 p 10 p 10 p 10 p

depolarisation depolarisation depolarisation depolarisation

current (A) current (A) current (A) current (A)
1p 1p 1p 1p
1 10 100 1000 1 10 100 1000 1 10 100 1000 1 10 100 1000
time (s) time (s) time (s) time (s)

Figure 4: Example of relaxation currents (polarisation and depolarisa-

tion currents) measured on samples with different moisture contents
[14].
The dielectric response of mineral oil is quite simple and can be characterised

by its essentially frequency independent permittivity, ε(ω) = const = 2.2 , and
volume conductivity, σ dc , whereas its dielectric response, f (t ) , can be
neglected [ 11]. Pressboard and paper are, on the other hand, characterised
by a strong dielectric response, f (t ) . This response is, in addition, dependent
on moisture content and ageing products. Results of measurements on
well-defined pressboard samples with different moisture content are used as
a database [3, 12, 13, 14]. Examples of the model responses in frequency and
time domains are respectively illustrated in Figure 3 and in Figure 4. The
temperature dependence of dielectric properties of pressboard/paper and oil
are characterised by the activation energies, typically 0.9 eV and 0.4 eV for
pressboard and oil, respectively [3]. If both components, i.e. oil and paper, are
sandwiched, it is possible to calculate the resultant behaviour for the
composite insulation if the geometrical arrangement is known.
For modelling FDS response from the information on the dielectric properties
of cellulose and oil, the frequency and temperature dependent permittivity and
dissipation factor of the composite insulation is calculated and compared with
results of measurements on real transformers.
For modelling PDC response similar procedure is followed. If the geometry of
the main insulation (i.e. the relative portions of oil ducts, barriers and spacers)
and a mean temperature are known, polarisation and depolarisation currents
can be calculated from the model and compared with measured currents.
By use of curve fitting procedure material parameters (oil conductivity,
moisture content in pressboard) and other variables (insulation geometry,
temperature) can be estimated.
 18

2.2 Influence of pressboard type, oil quality on the dielectric response of

laboratory pressboard models
Since the analysis and interpretation of dielectric response data is based on a
comparison between measured dielectric properties and a library of model
data obtained in laboratory on pressboard and paper samples, the quality of
the latter determine the accuracy of moisture analysis. This section presents
recently acquired information [4, 5, 15, 16, 17] on the influence of pressboard
material, oil conductivity and its acidic content on dielectric response spectra,
which in addition to the earlier gathered knowledge are complementary and
ought to be considered for improving common interpretation schemes. The
influence of the type of pressboard material is discussed in section 2.2.1, the
effect of the conductivity of the oil used for pressboard impregnation in section
2.2.2, and in section 2.2.3 it is demonstrated that different constituents
contributing to the conductivity of oil may have very different influence on the
response of pressboard.
2.2.1 Influence of pressboard
Two high density (HD1 and HD2) and two low density (LD1 and LD2)
pressboard types made by three manufacturers were compared through
measuring dielectric properties of pressboard discs in frequency range of
0.1 mHz to 1 kHz in a shielded measurement cell. The moisture content of the
four samples, determined by means of Karl Fischer titration, was 2.5 %; all
the samples were impregnated with new mineral oil. A coulometric titration
cell was used to determine the water content by heating the samples in an
external oven at 160°C [21].
Figure 5 shows the complex permittivity of the four pressboard samples. The
high density materials (1.1-1.3 g/cm³) exhibit higher permittivity of ε’ ~ 4.6
compared to the low density materials (0.9 g/cm³) having ε’ ~ 3.0. The high
density materials contain more cellulose fibres per volume, thus are more
polarisable. The losses, the imaginary part ε’’ of permittivity, are higher for the
high density materials as well.
 19

100
HD 2
HD 1
LD 2
LD 1
ε' 10

1
0.0001 0.001 0.01 0.1 1 10 100 1000
Frequency (Hz)
10
HD 2
HD 1
1 LD 2
LD 1
ε''

0.1

0.01
0.0001 0.001 0.01 0.1 1 10 100 1000
Frequency (Hz)
Figure 5: Dielectric permittivity of four pressboard types of different
origin having 2.5% moisture content measured at 23°C [5].

Moisture contents for the four materials LD1, LD2, HD1 and HD2 were
determined by means of analysis software [3] and were respectively 1.3; 1.5;
1.5 and 2.0 %. The difference between these results and the estimation of
water content of 2.5 % determined by Karl Fischer titration is probably caused
by use of different titration procedures when scaling the database of the
software.
When analysing results of measurements on a real transformer, the origin of
the pressboard is usually unknown. Therefore the observed variation of
pressboard dielectric properties may, to a certain extent, influence the
accuracy of moisture content estimates. Similar conclusion was drawn in [ 18],
where two types of high density pressboard from various manufacturers were
compared – the difference between moisture contents derived from titration
analyses and results of moisture determination by the software was 0.4-
0.5 %.
High density pressboard is typically used as a component of the main
insulation between low and high voltage windings and low density pressboard
in end insulations. Therefore, if possible, one should use the respective data
when modeling and interpreting measurement results.
 20

2.2.2 Influence of oil quality

New pressboard samples having 0.7 and 2.6% moisture content were
impregnated with different types of oils (one new and thee service aged, the
latter sampled from different transformers). Table 1 shows parameters of the
oils (neutralization number, conductivity and resistivity) as measured before
impregnating the pressboard samples. Typically the oil conductivity at room
temperature ranges from 0.05 pS/m for new oils to about 20 pS/m for aged
oils, in extreme cases it may raise up even to 1000 pS/m.

Table 1: Neutralization number, conductivity and resistivity of the four

oils used to impregnate the pressboard samples

NN Conductivity (pS/m) Resitivity (GOhm·m)

(mg KOH / g oil)
20°C 90°C 20°C 90°C
Oil 1 0.054 0.05 2.3 20'000 430
Oil 2 0.29 3.5 66,5 290 15
Oil 3 0.3 21 320 48 3.1
Oil 4 0.32 148 1945 6.8 0.51

Figure 6 depicts how the conductivity of impregnating oil influences the

dissipation factor of the new pressboard samples. Although the moisture
content of all samples was identical (2.6 %), the losses increase with the
conductivity of the oil used for impregnation. Analyses of these data by means
of modelling software [3] resulted in different estimates of the moisture
content, namely 2.1; 2.7; 3.0 and 3.8% for the respective oil conductivities,
meaning that a presence of conductive aging by-products in the oil may yield
an overestimation of moisture content in pressboard. An approach to
compensate for the influence of conductive aging products in oil and paper
insulation is illustrated in [17].
Figure 7 shows the behaviour of dielectric response for pressboard containing
only 0.7 % of moisture. In such a case the influence of conductive aging by-
products dissolved in oil becomes weaker.
 21

100

Dissipation factor
0.05 pS/m
3.5 pS/m
10
21 pS/m
1 148 pS/m

0.1

0.01

0.001
0.0001 0.001 0.01 0.1 1 10 100 1000
Frequency / Hz
Figure 6: Dissipation factor of new pressboard having 2.6% moisture
content impregnated with oils of various conductivities: 0.05; 3.5; 21
and 148 pS/m [5].
10
Dissipation factor

0.05 pS/m
3.5 pS/m
1 21 pS/m
148 pS/m
0.1

0.01

0.001
0.0001 0.001 0.01 0.1 1 10 100 1000
Frequency / Hz
Figure 7: Dissipation factor of new pressboard having 0.7 % moisture
content impregnated with oils of various conductivities: 0.05; 3.5; 21
and 148 pS/m [5].
The influence of oil quality on the conductivity of model samples of paper and
pressboard, as measured in time domain, was also reported in [15, 16]. The
results show that structure and kind of the solid material, its water content and
the conductivity of the impregnating oil have an influence on the dielectric
properties.

2.2.3 Influence of acids

Various carboxylic acids are produced in the ageing of cellulose and oil.
Cellulose ageing mainly produces lower molecular weight acids being water
soluble, while oil ageing produces higher molecular weight acids
predominantly oil soluble. When modelling FDS response, the effect of acids
should be taken into account for both oil and cellulose. The effect of acids in
the oil is primarily to increase its conductivity. The effect of acids in cellulose
 22

represents the addition of polar molecules which should have an effect upon
the dielectric response of both paper and pressboard. Experiments were
performed on high density pressboard to find this effect and its dependence
upon the humidity level [4].
The pressboard had apparent density (i.e. before drying) of 1.2 g/cm3 and
thickness 1 mm, and moisture contents 0.2 % and 3.9 %. The samples were
impregnated with dried and degassed naphtenic transformer oil containing
either none added acids, a mix of short-chained carboxylic acids (modelling
cellulose ageing), or a mix of long-chained carboxylic acids (modelling oil
ageing). Each acid mix consisted of two acids contributing equally to the
overall neutralisation value of 0.4 mgKOH/g. The short-chained acids were
formic and acetic acid (1 and 2 carbon atoms) and the long-chained acids
were naphtenic acid (an ill-defined mix with an average number of carbon
atoms of 15) and stearic acid with 18 carbon atoms. As diffusion through
pressboard can take long time, the pressboard was stored in the oil for at
least 5 months before experiments started. The results are illustrated in
Figure 8 and show “master curves” [1, 6] for use at 20 °C, obtained for a very
wide frequency range by combining records from tests at several
temperatures.
Humidity alone causes a shift of the response curves towards higher
frequency, and for the loss ε” there is also a somewhat upward shift, as can
be observed in Figure 3, which also shows a humidity-dependent “hump” to
the left of the minimum, seen in many but not all studies on FDS of
pressboard.
Like moisture, also short-chained acids shift the response curves towards
higher frequencies, and for ε” there is a somewhat upward shift as well. The
effect of short-chained acids is fairly but not entirely equal to the effect of
moisture. The acids cause a change in the shape of ε” near the minimum.
This change is different from the moisture-dependent “hump” of Figure 3. The
shape change pushes the minimum further towards higher frequencies than is
the case at the lowest frequencies of the low-frequency branch. But overall, in
practical transformer diagnosis, it will probably be difficult to see a difference
between the effect of short-chained acids and moisture. Long-chained acids
possibly show a little contribution in pressboard of low humidity, but none
whatsoever when the paper or pressboard is moist.
So why is there such a big difference between long-chained and short-
chained acids? The answer is that short-chained acids, which are also water
soluble, diffuse into the cellulose with very little amount remaining in the oil
[ 19], while long-chained acids, which are fat-soluble, stay in the oil and hardly
enter the cellulose. In the dielectric response experiment, this was also
reflected in the conductivity and the remaining neutralisation value of the oil
after impregnation. Both the oil without acids and the oil originally containing
short-chained acids had a conductivity at 20 °C of 0.04 – 0.09 pS/m
depending on humidity, and the neutralisation value was reduced from the
original 0.4 to about 0.02 mgKOH/g, both indicating that hardly any acid
remained in the oil. Note that this also means that if an oil-analysis shows
significant amounts of short-chained carboxylic acids in the oil, the solid
 23

Figure 8: Dielectric response of pressboard with varying moisture and

acid content [4].

insulation will be full of them. With long-chained acids, the neutralisation value
ended up at 0.37, and the conductivity of the oil was 0.5 – 10 pS/m, i.e. a
factor 10 higher than for the oil which originally contained short-chained acids.
 24

In summary, in a transformer, the majority of the short-chained acids formed

by cellulose ageing remain inside the pressboard. They contribute a little to
the conductivity of the oil, whereas their contribution to the response of the
solid insulation is similar to the contribution from moisture. In contrary, long-
chained acids stay in the oil and contribute heavily to its conductivity, but
hardly enter the pressboard or paper and therefore contribute very little if
anything at all to the response of the solid insulation.

2.2.4 Practical aspects and error estimates

Previous sections have shown that certain parameters will have an impact on
the measured dielectric response and consequently affect the estimated
moisture content of the cellulose insulation. Practical aspects of these
parameters and error estimates are presented in this section.
The analyses presented in this section are based on FDS, but as one (at least
in principle) can calculate PDC from FDS and vice versa, the error estimates
should be valid both for FDS and PDC measurements.
Assume that an FDS measurement has been done on a transformer in order
to establish the moisture content of the solid cellulose dielectric (which is
usually mainly pressboard). Also assume that measurements on the oil have
determined its dielectric properties very well. Using modelling of the
transformer, it should then be possible to calculate the overall response of the
solid cellulose dielectric from the measured response of the transformer.
Assume that this has been done in a way which gives the entirely correct
response of the solid dielectric. This response should in principle reveal the
(average) moisture content of the solid insulation when using the known
influence of moisture on the shape and frequency shift of the response. But
as seen in sections 2.2.1 to 2.2.3, there are other influences that are hard to
distinguish from the influence of moisture.
In chapter 2.2.1 it was shown that the type of pressboard material (density)
has an impact on the response, but also that there were differences between
different materials of fairly similar density. Thus, if the pressboard material
used in the transformer is different from the material used in modelling, the
difference will cause an error in the moisture estimate. In particular, if
assuming the solid insulation is HD pressboard when a significant part of it is
LD, the moisture content will be underestimated. Changing from HD to LD
pressboard shifts the responses towards lower frequencies and lower values,
which is exactly the opposite of what moisture does. However, in power
transformers barriers and spacers in the main ducts are typically made of HD
pressboard while formable parts such as end insulation (angle rings) are
made of LD pressboard.
As seen in 2.2.2 and 2.2.3 the presence of low molecular weight acids and
possibly some of the other ageing products causing increased oil conductivity
will shift the curves towards higher frequencies and higher values, which is
the same as water does although some minor details differ. An overestimation
of the moisture content will therefore be the result. However, as both moisture
and acids are unwanted agents accelerating degradation, especially in
combination, the considerable curve shift should cause the transformer
 25

operator to consider refurbishment of the transformer. Still, detailed

knowledge of the reason for a shift would be preferable in order to select the
most appropriate refurbishment method.
The insulation geometry parameters X and Y affect the shape of the modelled
response that to a certain degree reassembles the impact of moisture.
Therefore, use of incorrect X and Y parameters will influence the estimated
moisture content. Too large an amount of cellulose in the model will
underestimate the moisture content and vice versa.
Another factor which can cause erroneous moisture estimate is temperature.
The response shifts upwards in frequency with an increase in temperature.
The insulation activation energy is a measure of this shift. If the pressboard
temperature is higher than assumed in the moisture estimation analysis, an
overestimation of the moisture content will be the result.
Following is an attempt to put numbers on these errors, still assuming that a
“perfect” response curve for the solid dielectric alone has been obtained from
the overall response of a transformer. This must only be considered as a
rough estimate of the errors, as the range of materials and the range and
concentration of possible ageing products tested so far is very limited. Most of
the error analyses shown below, and also most practical moisture analyses,
are based on what is probably the safest marker for moisture content in the
solid cellulose: the frequency shift of the low frequency branch of the obtained
solid cellulose response, i.e. the branch mainly dominated by DC conductivity.
Since the “hump” seen in this part of the response in some but not all
experiments may be an experiment-dependent artefact, the following
analyses are based on the response at frequencies lower than where the
“hump” occurs. In one case, analysis software has also been used.
Error estimates based on the conductivity of the surrounding oil have not
been included, because the precise influence on the cellulose depends on to
what degree the constituents of the particular mix causing that conductivity
are absorbed by the cellulose. Estimates based on the presence of the
particular constituent low molecular weight carboxylic acids are included,
however, since it is largely absorbed by the cellulose, as opposed to for
instance high molecular weight acids.
The impact of LD or HD pressboard, in this case all with a moisture content of
2.5 %, is shown in Figure 5. The HD response is shifted up in frequency
approximately a factor 5 compared to LD pressboard. In one experiment it is
estimated that the frequency shift depends upon moisture as e BM where M is
the moisture content of the pressboard in per cent weight and B is 1.6 %-1,
which is within the 1 – 2 %-1 range suggested by Figure 3. In this analysis, the
response change caused by the difference in pressboard types corresponds
roughly to a moisture difference of 1 %. Analysis software, both one
commercial and one made for use at a laboratory, indicate 0.7 %. This is if the
entire solid insulation is made from the “wrong” type of pressboard, and
scales down according to the actual percentage if only part of the insulation is
of the “wrong” type.
 26

The difference between the two HD pressboards studied corresponds to an

error in the moisture estimation of 0.3 % according to the exponential model
above and 0.5 % according to the analysis software.
A similar very crude analysis for low molecular weight acids indicates that the
presence of such acids in the pressboard to a neutralisation value of 1
mgKOH/g corresponds to 0.08 - 0.15 % moisture (depending on actual
moisture). Note that the specified neutralisation value is in the pressboard, not
in the oil. The corresponding equilibrium neutralisation value of the oil due to
the low molecular weight acids alone is about 0.01 mgKOH/g, but note that
this is valid for the particular acid mix used in the experiments, only formic
and acetic acid which have the lowest molecular weight of any carboxylic
acids. The equilibrium shifts more towards the oil the higher the molecular
weight of the acids are. An example from real life was a service aged oil
having an overall neutralisation value of 0.3 mgKOH/g, of which about 25 %
could be due to hydrophilic carboxylic acids [ 20], which is very much the same
as low molecular weight acids. That would correspond to a neutralisation
value of the pressboard of somewhere in the range 4 – 7, corresponding to an
error in moisture estimation of 0.6 – 1 % by weight, which is significant. The
few service aged oils analysed for the relative content of low molecular weight
acids have shown that 1 – 27 % of the acids were of low molecular type, with
12 – 16 % being the most typical range. The latter would have corresponded
to an error in the moisture estimate of roughly 0.5 %.
A sensitivity analysis of the insulation geometry parameters impact on
estimated moisture content was performed on a dry and wet transformer. For
these transformers the moisture content in the cellulose and geometrical
parameters, X and Y were known. In the modelling the geometrical
parameters were systematically varied and the error in the estimated moisture
content recorded. With a deviation in the geometrical parameters up to ±0.1 a
reasonable fit of the model to measured data could be achieved. For larger
deviations the fit became poor. Within the ±0.1 range of the geometrical
parameters the maximum error in estimated moisture content was up to 0.6%.
The temperature dependence of the positioning of the entire response curve
along the frequency axis is proportional to e − Ea kT where the activation energy
Ea is about 1 eV, k is Boltzmann’s constant and T is the absolute temperature.
For T ≈ 310 K (or 37 °C) this gives a shift of a factor about 1.12 in frequency
for each degree K of temperature difference. This corresponds to an error of
0.07 % moisture. Thus an error of a few K in the estimate of the temperature
does not mean much, but 10 K gives 0.7 % which is a significant error.
Pressboard reference responses for the determination of moisture are most
likely to be available at a given temperature, typically 20 °C. In addition there
should be information about the shift with temperature, where Ea is central.
This information is supposed to be used for compensating for a transformer
measurement temperature being different from the reference temperature.
Uncertainty about Ea value will give an uncertainty in the final moisture
determination. While one experiment has found Ea to be 1.02 ± 0.3 eV for
both paper and pressboard, independently of the moisture and acid content, a
wider range has been found in other experiments. For instance, one such
experiment found the range 0.90 – 1.15 eV with some moisture dependence,
 27

where the higher values typically belonged to the moistest samples [14]. For
each 0.1 eV error in Ea, the error in estimated moisture is 0.008 %/K,
calculated around room temperature. Thus as an example, with reference
temperature 293 K and measurement temperature 333 K, the error in the
moisture estimate will be 0.3 % by weight for 0.1 eV error in Ea. This is
moderate compared to the other possible error sources.
The results are summarized in Table 2.

Table 2: Error sources and their maximum magnitudes in estimation of

moisture in cellulose (given in percent water by weight of cellulose)
Error source Approx. error Comment
Type of pressboard 0.7 % Assuming all insulation
HD => LD or v.v. changed.
Assuming HD board when
actual board is LD causes
underestimation of
moisture.
Type of HD pressboard 0.5 %
Low molecular weight 0.08 -0.15 % 1 mgKOH/g pressboard
(hydrophilic) acids per NV = 1 corresponds to about 0.01
mgKOH/g in mgKOH/g oil for this type
pressboard of acids. Corresponding
For heavily total NV of oil cannot be
acidic field aged specified as the
transformers the percentage of low
error range is molecular weight acids
believed to be vary very much between
up to 1% transformers. The
presence of acids causes
overestimation of moisture.
Incorrect insulation Up to 0.6%
geometry parameters X and
Y
Temperature 0.07 %/K Calculated at a
temperature of 310 K, but
does not seem to be
significantly different over
a temperature range of
273 – 353 K. Real temp
higher than assumed temp
causes moisture
overestimation.

The errors found above are “absolute”, i.e. independent of the actual moisture
level in the insulation. For a dry transformer the relative errors will end up
being extremely high. Also, for a fairly dry transformer, the errors may in
principle even shift the response curves further to the left of the curves for dry
HD pressboard, yielding an essential underestimation of moisture level.
Both over and under estimates of moisture is possible as seen above. It is
important to be aware of these errors and take them into account in the
 28

assessment of the transformer. Certainly the influence of the different errors

will be judged differently from situation to situation dependent on transformer
vintage, type, service conditions, service record, measurement conditions etc.
In the experiments described above (section 2.2.3) the oil conductivity was
increased by adding various carboxylic acids. In contrast to this, the increased
oil conductivity, as presented in section 2.2.2, was caused by presence of
aging by-products developed in service-aged power transformers. These by-
products are not necessarily of acidic nature. There seem to be service-aged
oils where only a minor part of the oil conductivity can be attributed to acids,
but only influence of acids have so far been tested for their effect on the
dielectric response.

3 MEASURING SYSTEM CONFIGURATIONS

When estimating the moisture content in the solid part of traditionally
designed transformers, one usually focuses on the main insulation, i.e. the
insulation between the coaxially arranged high voltage (HV) and low voltage
(LV) windings. The capacitance of this insulation is commonly denoted as
CHL. Other capacitances of interest can, for example, be winding
capacitances to ground (tank, core, etc.), i.e. CL or CH.
Dielectric response measurements can be performed within different
configurations, UST (Ungrounded Specimen Test) where the ground act as
natural guard or GST (Grounded Specimen Test) with or without guard. Basic
circuit connections for the test configurations are presented below.

3.1 UST configuration

Ungrounded Specimen Test (UST) measurement is performed between two
object terminals that are not connected directly to ground. Examples are the
main insulation between high and low voltage windings in power transformers
CHL or bushing insulation between the high voltage electrode and capacitive
tap.
There may also be one or more other terminals available, both grounded and
ungrounded, as for example tertiary winding of the transformer and
transformer tank. In case of one such terminal there are three capacitances
present. These are, in general, the capacitance of the insulation part to be
measured, Cx and two stray capacitances, Cs1 and Cs2 between the
terminals of Cx, as schematically indicated in Figure 9 below. Without guard
both stray capacitances are measured together with Cx. By connecting the
stray capacitances to ground, which acts as a guard, their influence is
eliminated.
 29

Figure 9: UST measurement configurations.

A connection diagram for a UST measurement on a power transformer is
shown in Figure 10 and indicates how the measurement leads should be
attached. The capacitance to be measured is the main inter-winding insulation
CHL, where the voltage electrode (Hi) is connected to high voltage bushings
and the measuring electrode (Lo) to low voltage bushings. The ground
electrode can, in this case, be connected to the third available terminal -
grounded transformer tank, thus eliminating the influence of capacitances
between windings and tank.

Figure 10: UST connection of dielectric response measuring system to a

transformer for CHL configuration.
 30

3.2 GST configuration

Grounded Specimen Test (GST) measurement is performed between two of
the object terminals, of which one is directly connected to ground. Examples
are the insulation between high or low voltage windings and tank or core, CH
or CL, and bushing insulation between capacitive tap and flange. There may
be one or more other terminals available, both grounded and ungrounded, as,
for example, low-voltage and/or tertiary winding of the transformer and
transformer tank. Similarly, as for UST configuration, the presence of one
such terminal results in additional stray capacitances, Cs1 and Cs2, as shown
in Figure 11 below. Without guarding these capacitances are measured
together with Cx (Figure 11 upper graph), whereas by guarding, their influence
is eliminated (Figure 11 lower graph).

Figure 11: GST measurement configurations.

A connection diagram for a GST measurement on a power transformer is

shown in Figure 12.
 31

Figure 12: GST connection of dielectric response measuring system to a

transformer for CH configuration.

3.3 Contribution of different parts of the insulation to the response

In a transformer different insulation segments form capacitive couplings
between each other as well as to the transformer tank and core. The
measured dielectric response therefore depends on which terminal the
measurement leads are attached to, on the test equipment configuration
(UST, GST) and on the winding arrangement in the specific transformer. For
that reason it is important to connect the measuring instrument in a proper
way and be aware which segments of the transformer insulation that is
included in the measurement.

Figure 13 schematically shows the windings of phase A of a two winding

transformer. In the figure the capacitive coupling between insulation segments
and tank/core are symbolically depicted forming the three main capacitances:
• between HV winding and LV winding (CHL)
• between HV winding to tank and to core (CH)
• between LV winding to tank and to core (CL)
The other two phases have similar couplings.
 32

a A

L H

CH-tank CL-core CHL

CH-core
CL-tank

Core

Figure 13: Schematic drawing of phase A windings in a transformer. The

capacitive couplings between insulation segments and tank/core are
symbolically represented by respective capacitances.

3.3.1 Ungrounded measurement CHL

If possible one should measure one capacitance segment at a time with help
of guarded measurement configuration in order to assess that specific part of
the insulation system. An example of such a measurement is the CHL
measurement.
In this configuration voltage is applied to the high voltage winding and current
is measured at the low voltage winding (UST measurement configuration). All
other terminals are grounded. This measurement is similar for all types of
transformers except for auto transformers where the series and common
windings are galvanic-connected. The insulation assessed is the insulation
between high and low voltage winding, CHL in Figure 13 This is the preferred
configuration for moisture estimation, for which the interpretation schemes
based on the modelling of the response data [1] work most reliably.
Most auto-transformers have a tertiary winding, so, provided this is not
internally earthed, measurements of main to tertiary insulation, CHT, are
possible and preferred, as illustrated in Figure 14.
 33

Figure 14: UST connection of dielectric response measuring system to

an autotransformer with tertiary winding.
3.3.2 Measurements to ground CH and CL
In these configurations the voltage is applied to one of the windings and all
other winding terminals are connected to guard. All currents flowing to the
tank (ground) are measured. These are GST guard measurements and
include additionally contributions from bushings and creep to ground.
In the CH measurement voltage is applied to the high voltage winding and
current is measured at all parts connected to ground (GST guard
measurement). The insulation measured is primarily the insulation between
the high voltage winding and the tank, but it also includes the end insulation
between high voltage winding and core, see
Figure 13. This configuration is less suitable for moisture assessment using
the modelling of the response data [1] but is however important for diagnosing
insulation to ground.
In a CL measurement voltage is applied to the low voltage winding and
current is measured at all parts connected to ground (GST guard
measurement). The insulation measured is primarily the insulation between
the low voltage winding and the core but it also includes the end insulation
between low voltage winding and winding leads and tank, see Figure 13.
When measurements between high and low voltage windings are for any
reason difficult or impossible, an acceptable alternative is created through
measurements between the innermost winding and the core.

3.3.3 Influence of winding arrangement

The winding arrangement is also important when performing diagnostic
measurements. In some transformers the high voltage winding is split, as
 34

schematically illustrated in Figure 15. The two parts of high voltage winding
HV surround the low voltage winding LV. For such an arrangement the
measurement between HV winding and LV winding is applicable but one
should be aware that in this case two insulation segments are assessed in
parallel, i.e. CH1L and CH2L, see Figure 15.
The CL measurement will in this particular case be different from the
previously discussed CL measurement. The surrounding high voltage
windings, which are connected to guard, will shield away most of the current
from the low voltage winding to tank and to core. The response is dominated
by the end insulation of the low voltage winding, the bushings, and creep
currents to ground. It is therefore advised to check the capacitance value of
the configuration measured. For power transformers, values below 1 nF
usually indicate that the loss is dominated by end insulation and higher
dissipation factors can be accepted.
The CH measurement will also, for this winding arrangement, assess several
insulation segments. In this case the contribution comes from the insulation
between inner high voltage winding and core/tank in parallel with the
insulation between outer high voltage winding and core/tank, see Figure 15.

a A

HV L HV

CH1-tank CH2-core CH2L CH1L

CL-core CH1-core
CL-tank CH2-tank

Core

Figure 15: Schematic drawing of phase A windings for a transformer in

which the high voltage winding is split into two parts. The capacitive
coupling between insulation segments and tank/core are symbolically
represented by capacitances. For this winding arrangement the low
voltage winding is surrounded by the high voltage main and regulation
windings.
 35

There may exist in practice transformers with more complicated winding

configurations than the case described above. The general principle is still
that guarded measurements between windings are preferred.

4 EXECUTION OF MEASUREMENTS
4.1 General instructions
A successful dielectric response measurement on a power transformer
requires adequate planning and coordination with its owner for finding
information on winding geometry and on means to disconnect the unit from
network. The objectives of the measurements, and thereby the resulting
connection of the instrumentation to the transformer, should preferably be
decided forehand.
Good understanding of the factors that may influence the measurements
under field-conditions, and therefore also the interpretation of dielectric
response results, is of crucial importance for the diagnosis reliability. Constant
and preferably not too low temperature is advantageous for the interpretation
quality – at higher temperatures the time necessary for the measurements
can be shortened. It is recommended to make guarded CHL measurements
between main transformer windings. In the case of CH and CL
measurements, it is also important to make sure that the transformer
bushings are dry and clean.

Before starting the measurements the transformer should preferably be

completely disconnected from the station power connections, i.e. all
connections to the bushings should be dismantled, and the station
connections should be properly grounded. The bushing terminals of the
individual windings should, if possible, be shortened together by conductors
kept away of earthed structures. For CH and CL measurements internal
leakages and creeps from bushing surfaces remain as the disturbing factors
and therefore should be considered in the analyses of the results.

If any of the windings has a neutral point connected either directly to ground
or through impedance to ground, also this connection should be removed.
The neutral bushing should preferably be interconnected with the phase
bushings during the measurements. Sometimes a surge arrester is connected
to the neutral point, and preferably, this connection should also be isolated.
In exceptional cases CHL measurements can be performed with only
disconnectors opened. In such a case all the equipment representing direct
galvanic connections to ground, such as voltage instrument transformers and
neutral point impedances, must be disconnected. In such cases, special
safety precautions must also be observed, so that the measurement
procedures do not violate national regulations and company instructions.
Description of such procedures is not included in this guide.

The measurements are often affected by interference in the substation, such

as parasitic leakage currents and induced electromagnetic disturbances. It is
therefore important to secure proper grounding connections of the transformer
tank during the measurements to minimize the ground interference. Proper
 36

placement of connecting leads can also minimize the influence of capacitive

coupled or radiated noise. Use of shielded connecting cables is
recommended in noise environments.

The temperature of the insulation system in transformer has a great influence

on the results of dielectric response measurements. Transformer temperature
often varies and it is not equal inside the tank – higher at the top than at the
bottom. If the measurements are made just after taking the transformer out of
operation, its temperature will be slowly decreasing. It is recommended to
register the temperature of oil just before starting the dielectric response
measurements. The most accurate way to determine the oil temperature is to
take top and bottom oil samples and measure the temperatures directly on-
site in the sampled oil. After opening the tap, cold oil flows out first, thus
waiting for sufficient time is recommended in order to get a representative
sample. If oil sampling involves too much effort, indications from the built-in
temperature gauges may be used, though it should be bear in mind their
readings depend on location of the temperature probe. Alternatively, the
average winding temperature can be calculated from a comparison of the
actual winding resistance to a measurement at ambient temperature.

It is also recommended to register the ambient temperature, relative humidity

and weather conditions in the station. In addition, making photographic
documentation on the transformer and the measuring setup as well as
recording the transformer nameplate information is advisable.
The following specific recommendations are advised to be considered before
commencing the measurement procedure:
• Check functionality of the measuring instruments before going to the
measuring site - this can be done by performing check measurements on
a capacitance box.
• Upon arrival to the substation the measuring team should report to the
site supervisor for discussing the activity plan, for getting a briefing on
safety regulations and a summary on all earlier performed operations
(disconnections, etc.). Always follow local safety instructions.
• Check whether parallel work is or will be carried out on the transformer
that might restrict its availability.
• Inspect if all disconnecting/dismantling operations were performed in
accordance to earlier agreements. In case of discovered discrepancies,
obtain permission to have the complementary operations performed.
Typical example is remaining connections of the neutral point impedance.
• Request availability of uninterrupted 1−Φ power supply suitable for
supplying the measurement instruments.
• Obtain formal permission to start working on the transformer. Usually the
transformer is at this stage grounded for safety reasons.
• Check for availability of a reliable grounding point; if ground grid is out of
reach install ground rod.
• Plan methods and means to secure reliable connection of the measuring
instruments to transformer terminals,
 37

• If winding arrangement is not known, it should be identified by

capacitance measurements.
• Protect measuring instrumentation from possible undesired influences by
weather conditions.
• After completing the measurement, get assistance in re-establishing the
protective ground connections on to the transformer before attempting to
make any further disconnections/reconnections on the transformer top.
• Restore external transformer couplings to the initial state and report to the
supervisor that the measurements are finished and that the transformer is
restored to the state in which it was approached.

4.1.1 Suggested check-list of transformer data to be registered for tested

transformer from its nameplate and other sources (instruction manual,
construction drawings, etc.)
• Rated voltages, MVA, 1-Φ or 3−Φ, manufacturer and year and serial
number. Take a photo of the nameplate.
• Number of windings and winding arrangement, presence of tertiary or split
winding and if it has one point of delta internally earthed,
• Sealed type or free breathing type,
• Type(s) of cooling and the type most frequently used,
• Presence of oil duct between windings, if possible dimension of windings
and of main oil duct,
• On-load and/or de-energized tap changer on HV or LV side - record tap
changers positions,
• Core, frame and yoke internally connected to tank or brought out
externally through a bushing on tank cover. Are they directly grounded or
grounded via any resistor?

4.1.2 Supporting information about the transformer state

• Filled with oil or without oil, any vacuum/pressure applied,
• New, in-service, out-of-service or faulty,
• Rewound or modified in the past,
• Past records on DGA and furans for trouble checking or e.g. thermal
ageing,
• Past records of oil analyses (moisture, conductivity, breakdown strength,
dissipation factor, acidity, etc…),
• Records of oil and winding temperatures in service, e.g. from oil sampling
records,
• Details on fault or tripping events, if the transformer is tripped from
operation,
 38

• Dielectric response measurement is to be performed before or after

refurbishment; if after has it been done before?
• Request oil sampling from tank and tap changer before switching off from
service (test oil conductivity and moisture in oil by KF titration)

4.1.3 Suggested check-list for execution of dielectric response measurements

on power transformers
• The transformer must be disconnected from the net, which means that
any connection to the transformer bushings including neutral(s) must be
disconnected. There are procedures measuring UST-configurations, e.g.
CHL with e.g. limited length of cables still connected to the bushings.
These procedures will not be included in this guide, please contact the
instrument manufacturer for instructions.
• Short the high voltage windings including their neutral point if present and
available. Short other interconnected windings and their neutral, if
existing, together. Use copper wire or similar connectors.
• For accuracy in GST measurements, inspect severity of pollution on
bushing surfaces. Pollution in combination with rain and/or high air
humidity during the measurements may cause relatively large leakage
currents flow to the ground. This problem can however be reduced if the
bushings are cleaned or by placing protective conductor rings on the
external bushing surfaces and connecting them to the guard terminal of
the instrument, thus reducing the creep currents.
• The transformer tank must be properly grounded.
• If the transformer is equipped with a Load Tap Changer (LTC) put the tap
in a position away from neutral (some load tap changers have an over-
voltage protection that is not correctly short-circuited if the tap is in neutral
position). Record LTC position and DETC (De-Energized Tap Changer)
positions.
• Make sure to record serial number, estimate of insulation temperature
(top oil temperature, winding temperature), ambient temperature and
weather conditions including relative humidly and temperature. See also
earlier sections.
• Before connecting the main cable to mains outlet, connect the instrument
protective earth/ground (PE) cable to PE at or in connection to the
transformer tank. The PE cable is first connection made and last
removed.
• Connect the instrument to a mains outlet that fulfil local safety
requirements (usually PE of the transformer and the outlet must be
metallically connected); otherwise use an isolating transformer.
• Power on the instrument.
• The measurement used for moisture assessment is preferably a UST
measurement, e.g. CHL, with a reasonable capacitance usually above 1
nF. The capacitance depends on winding arrangement and possible
 39

electrostatic shields, and if the capacitance is too low, choose another

configuration, e.g. CHT (or if no UST configurations are possible, CL).
• Connect the instrument cables first to the instrument and thereafter to the
interconnected bushings, as planned forehand. Usually the voltage cable
should be connected to the winding that is believed to be exposed most to
electromagnetic disturbances, which usually is the HV winding. Thus, the
current cable should be connected to the less exposed winding, i.e. the
LV winding. No winding should be left on floating potential, i.e. windings
should either be connected to guard or to ground. It is essential to keep
detailed records of the connections made.
• Get permission and assistance for removing transformer protective
ground connections.
• Execute measurements.
• Possibly rearrange test leads and perform more measurements.
• After completing the measuring series disconnect the instrument cables
first from the interconnected bushings and thereafter from the instrument.
• Backup test data.
• Switch off test equipment and disconnect the instrument from mains
outlet.
• As the last operation, disconnect the PE-cable.

4.2 Supplementary recommendations and comments for FDS measurements

As regards FDS measurements, the following recommendations are to be
obeyed:
• Tests are normally performed with voltage levels up to 200 V peak, but it
is advisable to choose as low voltage as possible with respect to
background noise level. A wide frequency test should be performed with
at least one configuration, preferably UST; usually the range 1 mHz – 1
kHz is used but lowest frequency can be omitted for measurement at
elevated temperatures. In case of low temperature or dry transformers,
the lowest frequency should preferably be 0.1 mHz. Other configurations
are either omitted or measured in a shorter frequency range, e.g. 10 mHz
– 1 kHz. Avoid including the power frequency or it first harmonics (e.g. 50,
100, 150 Hz…) in the list of measuring frequencies.
• Initial control tests for checking the object capacitance may include
measuring it at highest frequencies, a few tests to ensure that the
instrument is properly connected and winding that should be floating is not
grounded or faulty (similar to depolarization current test).
• If the instrument allows for viewing the wave forms of both the applied
voltage and the current measured, they should be observed for contents
of harmonics. A high offset level or dominating power frequency
component may lead to (i) difficulties with carrying out the measurements,
(ii) prolonged measuring times or (iii) inaccuracies of the results.
 40

• Instrument output current is limited, often to about 50 mA peak, which,

depending on measuring voltage and load capacitance, sets the
maximum frequency of the measurements to fmax= 0.05/(2*pi*U*C). In
such case, it is usually advisable to perform the measurement at both
200V peak and at a lower voltage for comparing the results afterwards.
• As long as UST (winding to winding) measurements are possible to
perform, results of these should be used for interpretation in terms of
water content in solid part of transformer insulation system. UST
measurements minimize the influences from loading capacitances to
ground as well as from internal and external leakage currents to ground.
• GST measurements (winding to ground) are less useful for the estimation
of moisture content, since the measurements between winding and
ground involve more complicated configuration of insulation geometries
(i.e. bushings, windings to core, tap changer components, etc.). The
frequency range for GST measurement can therefore be reduced to 10
mHz to 1 kHz.

4.3 Supplementary recommendations and comments for PDC measurements

A three-step measurement procedure is advised in order to obtain best
accuracy:
• Initial reconnaissance measurement to assess remaining charge without
any voltage applied.
• Ground all bushing terminals after shutdown for discharging - it is required
that the test object is de-polarized prior to the measurements,
• Control measurement of short duration (e.g. 5 s) and with low charging
voltage (e.g. 100 V) to verify the amplitude of measured current. The next
step can be started after the depolarisation current decreases to
minimum. The depolarisation duration before the main measurement
(during control measurement) shall be long enough to achieve a current
significantly lower than the lowest current to be expected. A good
grounding system is essential in obtaining very low remaining currents.
This is sometimes difficult when the test is carried out in the area without
ground grid e.g. in workshop or storage. Then an installation of a ground
rod for testing is essential.
• Any fault in the winding under test increases the depolarization current
during initial measurement. The quite-constant pattern with the current
magnitude of several hundreds nA to µA range is the symptom. The test
shall finish in this first step. Further investigation can be done by means of
the initial measurement of different connection, e.g. insulation of winding
to ground.
• Main measurement with charging voltage of 100 – 500 V for the insulation
between windings (CHL) and duration of 104 s/ 104 s for
polarisation/depolarisation currents. Test duration may be reduced at
elevated temperatures.
 41

5 CASE STUDIES
A number of study cases are reported below in order to exemplify different
practical situations that should be considered when performing diagnostic
measurements of dielectric response on power transformers. Each of the
study cases is preceded by indicating the aim for which it was selected.

5.1 Karl Fischer Titration as a scale for dielectric response methods

Aim:
To illustrate difficulties arising when evaluating moisture content by
means of KFT analyses of transformer oil and further consequences for
scaling the results of dielectric response measurements.

To evaluate the moisture content in the solid insulation of transformers based

on results of dielectric response measurements, the titration according to Karl
Fischer (KFT) served in this report as a conventional and direct measurement
method allowing for benchmarking of the estimates. KFT measures water
content in oil and in cellulose. Though the method is considered as
representing the state of the art in the measurements of water mass in the
solid and liquid parts of impregnated insulation, there exist a number of
concerns regarding its accuracy. This chapter introduces the measurement
principle and points to the conditions necessary to be obeyed for obtaining
accurate results.

5.1.1 The KFT technique

Karl Fischer titration allows for determining trace amounts of water in a
sample using volumetric or coulometric titration. Its principle is to add a
reagent (titre iodine) to a solution containing an unknown mass of water until
all water reacts with the reagent. From the amount of reagent the mass of
water can be calculated. In volumetric titration the added volume of the
reagent iodine is measured and the water content calculated. The sensitivity
of volumetric titration is limited to some 10 µg of water so that it is hardly
applicable for dry transformer oils. In coulometric titration an electrode
generates the reagent iodine and the water mass is calculated using the
electrically generated iodine. The moisture can enter the electrolysis vessel
either by direct injection or through transfer from an external oven. The
detection limit of coulometric titration reaches a few µg of water.

5.1.2 Reliability and comparability of KFT analyses - Round Robin Test

Several factors may affect the results of KFT analyses:
- There is always ingress of moisture from the atmosphere during
sampling, transportation and sample preparation. This happens
particularly during paper sampling from open transformers.
 42

- Cellulose binds water with chemical bonds of different strengths. It is

uncertain whether the thermal energy supplied releases all the water.
Heating temperature and time certainly changes the released water.
- Laboratories differently treat constraints that are not covered by
standards, such for example as the type of solvent used for oil
extraction.
- Sometimes the direct injection and the heating methods yield different
results of moisture in oil analyses. This might result as an influence of
additives and aging by-products in oil.
To elucidate this influences and to evaluate the discrepancies that may result
from KFT analyses, a round robin test (RRT) was carried out among seven
laboratories from four European countries. It concentrated on analysing the
water content in paper relative to weight and the water content in oil relative to
weight in three oil and paper samples according to the respective laboratory’s
standard procedures. The obtained results revealed an unsatisfactory
comparability between the laboratories [ 21], as shown below.

20
Moisture in paper / %

Deviation from average / %

A
B
C
4 D 10
E 3,64 3,40
F 3,38 3,27
G 3,02 2,97
3 0
2,43
2,25 2,25 2,19
1,97 1,78
2 1,701,64 1,79 1,65 -10
1,45
1,24 1,36
1,021,08
1 -20

0 -30
Sample A Sample B Sample C A B C D E F G

Figure 16: Moisture content in paper in % relative to weight as measured

by the laboratories (left) and deviation of each laboratory from the
average (right).

The results of moisture in paper analyses are depicted in Figure 16. For
example, for sample A the moisture estimates varied between 1.0 and 2.0 %.
The laboratories used different oven temperatures during the titration of paper
samples. One systematic influence is therefore obvious - the heating
temperature of paper in the oven increases the released mass of water. This
is due to existence of chemical bonds of different strengths, which hold water
molecules in paper. On the other hand, the heating temperature must remain
below the threshold for decomposition of cellulose by pyrolysis, which if
exceeded will result in additional water production. Beside this, other
 43

differences could only be explained by stochastic influences. For sample A,

containing little water, the comparability was worst and hardly acceptable.

60 80

Deviation from average / %

Moisture in oil / ppm
A 54,8 340
B
50 C 60
D 44,3
E

without sample C
F 39,8 40
40 40
G 35,3
32,8
30 20

19,8 0
20
16,215,2
11,2 12,1 12,2
10 6,7 9,5 8,9 7,5 -20
5,8
4,7 3,5 4,8
0 -40
Sample A Sample B Sample C A B C D E F G

Figure 17: Moisture content in oil in ppm relative to weight as measured

by the laboratories (left) and deviation of each laboratory from the
average (right)
On the other hand, the results of moisture in oil analyses, shown in Figure 17,
provided a reasonable conformance only for the wet oil sample C, although
unexpected discrepancies could also be observed. For the drier samples A
and B only a trend was recognizable; the results varied from 3.5 to 12.1 ppm
for sample A and from 5.8 to 19.8 ppm for sample B. Systematic differences
were obvious. It has to be mentioned that for the dry oils, the results also
varied within one single laboratory and a standard deviation of 20% is not
unusual.

5.1.3 Consequences for interpretation of dielectric response measurements

Water content determination by means of dielectric response methods is often
calibrated by comparing them with evaluations based on KFT. However, as
shown above, KFT results also suffer from a poor comparability between
different laboratories. The user must therefore be aware of this fact, and
understand that a deviation in the comparison does not necessarily point out
weaknesses of the dielectric methods. The sampling procedure and the
titration parameters used by specific chemical laboratory determine
comparability to dielectric response based results.
This also applies while comparing the moisture evaluation result of various
dielectric response methods between each other. The moisture analyses are
based on comparison to results obtained on model samples. If different
titration techniques were used to scale the models, the dielectric response
methods stringently will come to different results as well.
 44

In recent years new moisture sensors have become available for continuous
on-line monitoring. A technical brochure provided by the CIGRÈ WG A2.30
represents the practical application of these sensors especially in comparison
to the traditional approach of oil sampling, Karl Fischer titration and
subsequent application of an equilibrium diagram [ 22].

5.2 Dielectric response diagnostics

Aim:
To present possibilities and challenges arising under application and
interpretation of dielectric response measurements. Comparisons
between different moisture determination approaches (dielectric
response, KFT and moisture equilibrium) are demonstrated, including
illustration of the influence of aging of insulation components on
estimation of moisture content.

5.2.1 Comparison of moisture determination by means of dielectric response

measurements and KFT
The main aim of the work presented in this section aimed at calibrating the
estimated moisture content in paper obtained from FDS measurements
against results of KFT analyses of oil and paper samples taken from the
investigated transformer. Generally, the possibilities to obtain paper or
pressboard samples from power transformers that are temporarily taken out
of operation are very limited. Such studies can thus be performed on
defective units, undergoing repair or scrapping, where the solid insulation can
readily be accessed. However, the number of units available for this type of
investigations is anyhow fairly low. Therefore, an offered opportunity to
perform measurements on a defective transformer that had to be opened and
repaired was appreciated.
The transformer studied was a three-phase unit, manufactured in 1967, rated
50 kV/10 kV and 40 MVA. It had Y0y0 phase configuration and OFAF cooling.
Indications for a presence of fault in the tap-changer, previously also
experienced in other two identical units, initiated broader investigations, which
aim was to judge whether the unit could be repaired or should be scrapped.
Insulation condition was assessed and it was found that the solid insulation
was significantly aged with DP values around 350 and possibly even down to
300. Further, the oil was severely degraded, showing high level of acidity. It
was however concluded that the transformer could still be shipped for repair
and the oil should be replaced by new oil or refurbished one.
The unit was out of operation for a period of 4.5 months and the FDS
measurements and other analyses were performed at several occasions
during this period, i.e. before, during, and after the repair. KFT analyses,
performed on oil samples and on samples of paper and pressboard, allowed
for comparison of moisture content estimations obtained by different
techniques. Samples of the paper were taken when the transformer was
 45

opened in workshop, whereas the oil samples were taken at several

occasions throughout the disconnection period.
Figure 18 shows in detail the positions and when the samples of solid
insulation were taken together with the respective moisture content values
obtained from KFT analyses.
The analyses of results from dielectric spectroscopy measurements showed
that the estimated moisture contents varied and were in the range of
1.4-2.4%. These estimates were in good agreement with the results from KFT
analyses performed directly on paper and pressboard samples, 1.0-2.5%.
Also estimates based on oil samples gave similar results, about 2.4%, if the
in-service temperature was used as the reference. Since the time constant for
moisture migration between oil and paper in large power transformers is
extremely long, especially at lower temperatures, the in-service temperature
is the proper selection for the KFT analyses, even if the transformer has been
taken out of operation and was allowed to cool for a few days.
The general conclusion is that moisture level in solid part of the insulation in
the investigated transformer is at the level of about 2%, which indicate a fairly
dry condition. More details on the analyses of this and other study cases can
be found in [ 23, 24, 25].

5.2.2 Moisture determination in a heavily aged transformer

A heavily aged transformer was designated for scrapping. Such a case
provides a good opportunity for comparing different methods of moisture
assessment in paper samples. The transformer, built in 1950, and rated
104/23,4 kV, 30 MVA, had an ONAF cooling. Mineral oil type Shell Diala 6KX
from 1954 and having neutralization number of 0,49 mg KOH/g filled the
transformer. Paper and oil samples were taken out after measuring the
dielectric properties (polarization and depolarization currents in time domain
as well as complex capacitance in frequency domain). The insulation
temperature was 24°C during the measurements.
Figure 19(a) displays the polarisation and depolarisation currents measured
through the main transformer insulation CHL as a superposition of
conductivity and polarisation phenomena. As seen in the figure, the
polarisation current approached its saturated value quite early, at around
300 s, and almost no decrease occurred afterwards. Further decay of
polarisation was solely dominated by the conductive component of the
current. Since the final current value could be estimated at 300 s, the
measurement might have been stopped at that time, which indicated an
unusually high conductivity of insulation system in this transformer.
Figure 19(b) displays the dissipation factor as measured between HV- and
LV-winding. The losses in the insulation are very high even at power
frequency, e.g. 0.13 at 50 Hz. Conductivity of the oil influences the dielectric
response of this transformer only down to 0.1 Hz, below this frequency the
properties of cellulose dominate. Because of the highly conductive nature of
this insulation the measurements might have been stopped at 0.01 Hz since
sufficient amount of information for the analysis was gathered.
 46

Pressboard On-site
Paper taken on
Paper In workshop
opposite side
Wood After repair

Figure 18: Moisture content estimated by means of KFT in samples of

transformer solid insulation at different locations and sampling events.

1E-04 10
Dissipation Factor
Current (A)

1E-05 1

1E-06 0,1
Ipol
Idep
1E-07 0,01
1 10 100 1000 10000 0,0001 0,01 1 10 1000
Time (s) Frequency (Hz)

(a) (b)
Figure 19: Polarisation currents (a) and dissipation factor (b) measured
between HV- and LV-winding of the studied transformer.
 47

The oil conductivity of this transformer was both measured and estimated
based on the dielectric response model, yielding respectively σmeas = 0.9 nS/m
at 22°C and σmod = 1.3 nS/m at 24°C. Both the results agree well, as
observed for all comparisons of measured to modelled oil conductivities. The
oil conductivity likewise the neutralisation number value of 0.49 g KOH/kg oil
were here unusually high (normally, at room temperature, the oil conductivity
is within the range of 0.05-20 pS/m), which indicate a high content of
conductive aging products.
Figure 20compares the results of the analyses of the time domain
measurements performed. Karl Fischer titration of the paper samples yielded
2,6 % moisture by weight (KFT). Results of the modelling of the dielectric
response measurements by means of different software differ from each
other: Two algorithms (DA1, DA2) had no compensation for the influence of
conductive aging products and came to 3.8 and 4 % moisture by weight.
Another algorithm (DA3) with build-in compensation for conductive aging
products [17] indicates 2.9 % moisture relative to weight.
In the oil sample the moisture saturation was measured directly onsite and the
moisture content in ppm by Karl Fischer titration in a laboratory. When using
moisture sorption isotherm [22], the relative saturation reading led to 2,5 % of
moisture in cellulose (RS), which well agrees with the KFT analysis of the
paper samples and the dielectric response analysis with compensation for
conductive aging products. At the same time, direct application of equilibrium
curves, based on moisture content in oil in ppm [ 26], yielded much too high
content of moisture in paper - 6,0 %.
This study case indicates that when analysing the moisture content in a
strongly aged transformer the methods and algorithms not adopted for taking
into account the aging state and moisture adsorption capacity of transformer
insulation may overestimate the moisture content.
6,0
6
Moisture content (%)

5
4,0 3,8
4
2,9
3 2,6 2,5

0
KFT DA1 DA2 DA3 PPM RS
Paper Dielectric Oil sample
sample response
Figure 20: Moisture content in the solid insulation obtained from Karl
Fischer titration of paper samples (KFT), dielectric response analyses
(DA1, DA2, DA3) and from use of equilibrium diagrams for moisture by
weight in oil (PPM) and from the relative saturation of oil (RS).
 48

5.2.3 Oil exchange and drying of power transformers

The following study case is presented to exemplify the effects of oil exchange
and drying of transformer insulation on the results of dielectric response
measurements.
A large amount of sludge was found in the oil of a 300 MVA transformer after
23 years of uninterrupted operation. The owner of the transformer decided to
exchange the oil. Prior and after this operation, measurements of PDC, RVM
and tan δ (at 0.1 Hz) were performed [ 27] and the results are shown in Figure
21 and summarised in Table 3. Furthermore, a paper sample was shaved at a
lead of a tap winding and analysed. The degree of depolymerisation was
DP = 352, which indicated a normal thermal ageing of the paper.

Table 3: Results of diagnostic measurements carried on a 300 MVA

transformer prior and after oil exchange
Measurements Prior to oil After oil
change change
tan δ at 0.1 Hz, directly measured 0.894 0.187
tan δ at 0.1 Hz, from PDC analysis 0.901 0.191
conductivity σoil of the oil, from PDC analysis in 1/Ωm ≈ 4.5⋅10
-11
≈ 3.5⋅10
-12

Moisture in the solid insulation material, from PDC analysis 3% 2.7 %

Moisture in the solid insulation material, from RVM measurements 3.45 % 2.48 %
Depolymerisation degree 352
 49

(a)

(c)

(b)

(d)

Figure 21: Polarisation currents measured between HV and LV windings

in 300 MVA transformer before and after oil exchange (a) with calculated
tanδ spectrum in frequency domain (b); RVM polarisation spectra before
(c) and after (d) oil exchange compared with spectra calculated from
PDC results [27].

The PDC analysis showed clearly that the oil exchange reduced oil
conductivity by about one order of magnitude, which can directly be seen from
the current values at t = 1 s, Figure 21(a). This was also confirmed by
separate oil conductivity measurements. Furthermore it is concluded that the
water content in the barriers was not changed significantly. Respective
estimates of water contents, 3% before and 2.7% after the oil exchange, were
estimated by comparing PDC curves with calculated ones, as shown in Figure
22 (a) and (b). The estimates were obtained by taking into account the
insulation geometry. Pressboard conductivity, which is related to the water
content, could also be calculated based on the long term current values.
Figure 21 also shows frequency dependence of tan δ and RVM spectra,
which were calculated from the time domain data of PDC measurements. As
it can be seen from the figure, these agree well with the measured data.
Due to the low moisture absorption capability of oil, it is not possible to de-
crease significantly the moisture in the solid insulation by exchanging the oil
 50

of a transformer. Assuming a weight of the solid insulation of 10000 kg and a

moisture content of 3 %, we have a total water content of 300 kg. Assuming
further an oil weight of 50000 kg and water in oil content of 30 ppm, which is a
high value, we get a water mass of only 1.5 kg stored in the oil. Thus, the
water content in the transformer prior and after the oil exchange remains al-
most the same. This can clearly be seen from the PDC measurement and
from its frequency domain transformation, but not from the so called RVM
spectra [27]. The differences between the RVM spectra seen in Figure 21
(before (c) and after (d) the oil exchange) are obviously caused by the
difference in oil conductivity but not by different properties of the solid
insulation.
Because the operation of oil exchange did not significantly reduced wetness
of the solid insulation, it was decided to dry the transformer [14]. Results of
PDC measurements that accompanied this operation showed a significant
reduction of water content, which remained in agreement with the amount of
extracted water, Figure 22 (c) and (d),
Figure 22 also illustrates the evaluation procedure of PDC results. Based on
XY insulation model, PDC curves are simulated assuming material properties
at different water contents. The best fit with the measured curves indicates
the water content in the cellulose barriers.

-4 before the oil exchange (a) before drying (c)

10 -5
polarization currents/ A 10
A

10
-5
measurement -6 measurement
10

-6
3.5 % -7
10 10 3%
3% rM = 2.7 %
2.5 %
rM = 3.0 % 2.5 %
-7 -8
10 10

after the oil exchange (b) after drying (d)

-5
10 -5
10 A
A
-6 measurement
10 -6 measurement
10
3%
-7
10 2.5 % -7
rM = 2.7 %
10 rM = 1.7 % 2%
1.5 %
-8
10 -8
10
1 10 100 1000 s 1 10 100 s 1000 t

Figure 22: Polarization current measured on 300 MVA power transformer

before and after oil exchange (a and b) as well as before and after drying
(c and d). Moisture content rM, as marked in the graphs, is determined
by curve fitting.
 51

5.3 Influence of external environment

Aim:
Pointing to different factors that may influence results of dielectric
response measurements in field conditions

5.3.1 Presence of different leakage paths

Although the determination of moisture content in power transformer
insulation is nowadays rather straight forward, there may appear situations
where results of dielectric response measurements become influenced by
local conditions in substation. Some selected cases are therefore presented
below to secure a better insight to possibilities and limitations of the
techniques.
Leakage paths in CHL measurements
In CHL measurements, the voltage usually is applied to the transformer HV
side and the resulting current is measured on the LV side. If the guard cable
is connected to the grounded transformer tank and no other elements are
connected to the bushings, the whole current flows across the main insulation
between windings, following the desired current path illustrated in Figure 23.
The capacitance (C') and the losses (C'') of the insulation between the HV
and LV windings can be characterized well. If, on the other hand, the
measurements are done when only opening the disconnectors on both sides,
letting in this way some elements remain connected to transformer bushings
(cables, arrestors, string and post insulators, etc.), the current required to load
these element capacitances loads the HV terminal of the measuring
instrument. Although the capacitance of these elements can sometimes be
quite high, the additional current does not severely affect the measurement.
However, this situation can create a problem in case when the instrument
gets into its current limit and decreases subsequently the supply voltage or a
parasitic frequency dependent impedance is present, for example a voltage
transformer (VT) on the LV side. Another error source is when the voltage
drop in the Hi measuring cable becomes significant due to the high current
caused by the additional load and thereby introduces an error in the
measured impedance. This error is larger at higher frequencies since the load
current increases with frequency.
 52

Figure 23: Main and parasitic current paths in CHL measurement with
opened disconnectors.

Leakage paths in CH and CL measurements

In CH measurements the capacitance C’ and the losses C’’ of the insulation
between the HV winding and the tank are measured by applying voltage to
the transformer HV side and measuring the current returning to the ground.
As Figure 24 shows, the guard should be connected to LV winding in order to
avoid measuring the inter-winding capacitance (CHL) together with HV-to-
tank one (CH).

Figure 24: Main and parasitic current paths in CH measurement with

opened disconnectors.
 53

Even in case all the shunt elements are dismantled from the transformer,
bushing capacitance and internal creeps are measured in parallel with the
CH, as they cannot be guarded. This means that the addition of the bushing
capacitance and the internal creep in CH measurements can yield deviations
of the dielectric response measured.
The influence may also be strong if the measurements are performed with
some circuit elements connected to the bushings. In this case the
capacitances of these elements (insulators, surge arrestors, cables, etc.) are
measured in parallel with the CH. The cables connected to the transformer
HV side have sometimes a length of a few hundred meters and are insulated
with oil-impregnated paper. Thus their capacitance is often larger than and
having similar properties to the transformer internal insulation. In such a case
the CH measurements are more representing the behavior of the cable
insulation than that of the transformer insulation.
In CL measurements the insulation between the LV winding and the core is
mainly measured. A similar situation to that analyzed in CH measurements
appears. In this case also the effect of the instrument VT should be
considered, if applicable, in addition to the effect of LV cables. More detail
explanations and analyses on the above described cases can be found in
[24].
Leakage paths along transformer bushings
Appearance of parasitic current paths along the surfaces of transformer
bushings can further influence dielectric response measurements [24]. The
risk of creep from bushings increases when their surfaces are contaminated,
and especially in conditions of high humidity or rain.
As indicated above, the bushing creep has not much effect on CHL (UST –
with guard) measurements, as the surface leakage currents are flowing to the
tank, which is guarded. Major effects can, on the other hand, be experienced
during CH or CL (GST – with guard) measurements. The latter effect is
illustrated below as observed during FDS measurements in field conditions.
Surfaces of bushings in the tested transformer were covered with old, dried
and polluted silicone grease. CH measurements were performed during a
rainy day and then compared with results of measurements performed on a
sunny day. In Figure 25 the results of measurements (C´ and C´´), performed
on a sunny day are compared with measurements made while it rained (C’
rainy and C’’ rainy). Large discrepancies were found between the results,
caused by creep current flow along the bushings surfaces under the rain. The
interpretation of the two measurements under dry and wet conditions yielded
2.7% and 5% moisture content in paper, respectively. Thus, the presence of
the leakage current resulted in a strong over-estimation of the moisture
content in the tested transformer.
 54

1.00E-06
C'
C''
C' rain
C'' rain
1.00E-07 C' rain guard ring
C'' rain guard ring

1.00E-08

C' and C'' (F)

1.00E-09

1.00E-10

1.00E-11

1.00E-12
0.001 0.01 0.1 1 10 100 1000
Frequency (Hz)

Figure 25: Effect of bushing surface creep during CH measurements in

field conditions.

To avoid the effect of surface creep, soft copper straps were fixed around the
bushing’s porcelain body and were connected to guard terminal of the
measuring device together with the LV winding, as illustrated in Figure 26. In
this way, even though the measurements were performed during rain, the
influence of the currents flowing along the bushing surfaces was eliminated.
This can be seen in Figure 25, where the results marked as C’ rain guard ring
and C’’ rain guard ring, are very similar to the ones obtained under dry
conditions, thus strongly damping the creep effect.
 55

Guard ring
around bushing porcelain

Figure 26: Guard applied around bushing for avoiding influence of

surface leakage currents on results of GST - with guard measurements
under strongly polluted conditions.
Electromagnetic disturbances
When disconnecting a transformer for performing dielectric measurements,
other circuits in the substation usually remain energized, representing a
potential source of electromagnetic disturbances. A transformer under test is
most prone to pick-up the disturbances through the air if the measurements
are performed with only opened disconnectors and long conducting lines,
acting as antennas, remain connected to the bushings. Disturbances can also
be conductively coupled through galvanic paths to the substation’s ground
system, for example via connected voltage instrument transformers.
Measurements of voltages induced at tested transformer terminals in one
such a substation were at a level of a few hundred volts when lines with a
length of some tenths of meters remained connected. Under these conditions
it was not possible to carry out any dielectric measurements at all.
The influence of electromagnetic disturbances, their limits and the measures
to be adopted for reducing their influence on the dielectric measurements are
still under investigations.
 56

6 REFERENCES

1 S.M. Gubanski, P. Boss, G. Csepes, V.D. Houhanessian, J. Filippini, P. Guuinic, U.

Gäfvert, V. Karius, J. Lapworth, G. Urbani, P. Werelius and W.S. Zaengl - "Dielectric
Response Methods for Diagnostics of Power Transformers”, Electra, No. 202, 2002,
pp 23-34; also in CIGRE Technical Brochure, No. 254, Paris 2004.
2 S.M. Gubanski, J. Blennow, L. Karlsson, K. Feser, S. Tenbohlen, C. Neumann, H.
Moscicka-Grzesiak, A.Filipowski, L. Tatarski - " Reliable Diagnostics of HV
Transformer Insulation for Safety Assurance of Power Transmission System
REDIATOOL – a European Research Project”, CIGRE Session 2006, Paris, France,
paper no. D1-207.
3 C. Ekanayake - ”Diagnosis of Moisture in Transformer Insulation – Application of
Frequency Domain Spectroscopy”, Dept. of Materials and Manufacturing Technology,
Chalmers University of Technology, Gothenburg, Sweden, PhD thesis No. 2409,
2006.
4 D. Linhjell, L. Lundgaard, U. Gäfvert - ”Dielectric Response of Mineral Oil
Impregnated Cellulose and the Impact of Ageing”, IEEE Trans. Diel. and El. Ins., Vol.
14, No. 1, 2007, pp. 156-169.
5 M. Koch - ”Reliable Moisture Determination in Power Transformers”, PhD thesis,
Institute of Energy Transmission and High Voltage Engineering, University of
Stuttgart, Sierke Verlag Göttingen, Germany, 2008.
6 A.K. Jonscher - “Dielectric Relaxation in Solids, Chelsea Dielectric Press, London,
1983, pp. 36-52.
7 W. S. Zaengl - “Dielectric spectroscopy in time and frequency domain for power
equipment – theoretical considerations”; IEEE Electrical Insulation Magazine, Vol. 19,
No. 5, 2003, pp. 5-19.
8 W. S. Zaengl - “Applications of dielectric spectroscopy in time and frequency domain
for HV power equipment”; IEEE Electrical Insulation Magazine, Vol. 19, No. 6, 2003,
pp. 9-22.
9 J.C. Maxwell - ”A Treatise on Electricity and Magnetism”, vol. 1, Clarendon Press,
Oxford, 3rd edition, reprint by Dover, 1981, pp. 450-464.
10 K.W. Wagner - ”Erklärung der dielektrischen Nachwirkungsvorgänge auf Grund
Maxwellscher Vorstellungen”, Archiv für ElektrotechnikV II, No. 9, 1914, pp. 371-387.
11 U. Gäfvert, G. Frimpong, J. Fuhr - "Modelling of Dielectric Measurements on Power
Transformers", CIGRE Session 1998, Paris, France, paper no. 15-103.
12 C. Ekanayake, S.M. Gubanski, A. Graczkowski, K. Walczak, B. Garcia, S.M.
Gubanski - "Frequency Response of Oil Impregnated Pressboard and Paper
Samples for Estimating Moisture Transformer Insulation", IEEE Trans. on Power
Delivery, Vol. 21, No. 3, 2006, pp. 1309-1317.
13 V. Der Houhanessian – “Measurement and Analysis of Dielectric Response in Oil-
Paper Insulation Systems”, ETH, Zurich, Switzerland, Zurich, PhD thesis No. 12832,
1998.
14 B. Breitenbauch, A. Küchler, T. Leibfried, W.S. Zaengl - “Insulation diagnosis by
polarization and depolarization current measurements” 13th Int. Symp. on High
Voltage Engineering, ISH03; Delft, The Netherlands, 2003.
15 A. Küchler, F. Hüllmandel, K. Böhm, M. Liebschner, Ch. Krause, B. Heinrich -
“Parameters Determining the Dielectric Properties of Oil Impregnated Pressboard and
Presspaper in AC and DC Power Transformer Applications”, 15th Int. Symp. on High
Voltage Engineering, ISH07; Ljulbjana, Slovenia, 2007,p. 356.
16 A. Küchler, F. Hüllmandel, K. Böhm, C. Neumann, N. Koch, K. Loppach, C. Krause,
J.-J. Alff - “Condition Assessment of Aged Transformer Bushing Insulations”, CIGRE
Session 2006, Paris, France, paper no. A-102.
17 M. Koch, M. Krüger - “A Fast and Reliable Dielectric Diagnostic Method to Determine
Moisture in Power Transformers”, Intern. Conf. on Condition Monitoring and
Diagnostics, CMD 2008, Beijing, China, April 21-24, 2008, paper no. 143.
18 J. Gielniak, K. Walczak - “Influence of Oil Parameters and Pressboard Mass Density
on Dielectric Response of Oil-Paper Insulating System", 2nd Conference on
 57

Advances in Processing, Testing and Application of Dielectric Materials, APTADM07,

Wroclaw, Poland, 2007, p 118-121.
19 S. Ingebrigtsen, M.Dahlund, D.Linhjell, L.E.Lundgaard - ”Solubility of Carboxylic Acids
in Paper(Kraft) – Oil Insulation Systems”, IEEE Conference on Electrical Insulation
and Dielectric Phenomena,CEIDP2004, Boulder, Colorado, USA, 2004, pp. 253-257.
20 D. Linhjell, W. Hansen, S. Ingebrigtsen and L. Lundgaard - ”Moisture absorption in
service aged transformer oils”, Nordic Insulation Symposium, Nord-Is 03, Tampere,
Finland, 2003, pp. 397-404.
21 M. Koch, S. Tenbohlen, J. Blennow, I. Hoehlein - “Reliability and Improvements of
Water Titration by the Karl Fischer Technique”, 15th International Symposium on
High Voltage Engineering, ISH07, Ljubljana, Slovenia, 2007, p. 427.
22 V. Sokolov et al. - “Moisture Equilibrium and Moisture Migration within Transformer
Insulation Systems” Cigré Working Group A2.30, Technical Brochure, No. 359, Paris,
France, 2008.
23 J. Blennow, A. Dernfalk, K. Walczak, C. Ekanayake, P. Koetsinger, L. Karlsson, U.
Gäfvert, S. Gubanski, A. Bartnicki - "Sequential Comparative Study of Dielectric
Response and Analyses of Oil and Paper from a Power Transformer Undergoing
Repair", 14th Intern. Symp. on High Voltage Engineering, ISH05, Beijing, China, 2005,
p. 505.
24 J. Blennow, C. Ekanayake, K. Walczak, B. Garcia, S.M. Gubanski - "Field
Experiences with Measurements of Dielectric Response in Frequency Domain for
Power Transformer Diagnostics", IEEE Trans. on Power Delivery, Vol. 21, No. 2,
2006, pp. 681-688.
25 J. Blennow, K. Walczak, A. Dernfalk, B. Garcia, A. Bartnicki, S. Gubanski,
O. Samuelsson, U. Karlsson - “Comparative study of dielectric spectroscopy
measurements and analyses of oil and paper samples from a transformer undergoing
repair” Nordic Insulation Symposium, Nord-IS 05, Trondheim, Norway, 2005, pp. 35-
38.
26 T. V. Oommen - “Moisture Equilibrium Charts for Transformer Insulation Drying
Practice” IEEE Trans. on Power Apparatus and Systems, Vol. PAS-103, No. 10, Oct.
1984, pp. 3063-3067.
27 T. Leibfried, A.J. Kachler, W.S.Zaengl, V. Der Houhanessian, A. Küchler, B.
Breitenbauch - “Ageing and Moisture Analysis of Power Transformer Insulation
Systems”, CIGRÉ Session 2002, Paris, France, Paper 12-101.