WATER – IN– PAPER ACTIVITY: A NEW APPROACH FOR MOISTURE MANAGEMENT

IN TRANSFORMERS1

Oleg ROIZMAN
IntellPower, Australia
Valery DAVYDOV
Monash University, Australia
Barry WARD
EPRI

Abstract

Moisture in oil/paper transformer insulation can be expressed in many different terms. Among well
accepted terms are water in paper (%), water in oil in parts per million (ppm), relative saturation (%)
and dew point (oC) to name just a few. It was shown experimentally that, generally, the water
content of paper determined by the well known equilibrium charts is equal neither to the average
water content nor to the water content of any particular part of the solid insulation in a real power
transformer. Based on numerous tests with different types and thickness of paper insulation it was
suggested that the amount of water associated with the level determined by the equilibrium charts
corresponds to the amount of water available for exchange between solid insulation and oil, which
we call ‘active’ water.
A new concept of water-in-paper activity has been introduced in order to explain the discrepancies
between theoretically expected and experimentally obtained results. And the following definition for
water-in-paper activity in the sealed paper-oil system has been proposed: Water-in-Paper Activity
(Awp) in a sealed paper-oil insulation system is the measure of water available in cellulose for
exchange with surrounding oil.
The authors believe that the introduction of the water activity concept into monitoring and
diagnostics of electrical insulation will help to better understand and maintain the quality of
dielectric materials and electrical insulation in particular.
This paper aims to provide a definition, means of measurement, way of standardization, test
procedure and applications of a new concept of moisture management in transformers.

Introduction

Water presents a danger for reliable operation of a transformer and needs to be monitored and
controlled in a way to minimize its harmful effect on the transformer insulation system. Most water
is concentrated in solid insulation which is predominantly made of cellulose material (paper). There
is no direct method for moisture assessment in solid insulation and therefore different approaches to
evaluation of moisture levels have been proposed [1].

1
Presented at EPRI Substation Equipment Diagnostics Conference XIII
March 6-9, 2005, Royal Sonesta Hotel New Orleans, New Orleans, Louisiana
The most common approach used for moisture assessment in solid insulation was developed in the
1960s and has been used since then. This approach utilizes the idea of moisture equilibrium between
oil and paper in the insulation system of a transformer and makes it possible to assess roughly the
water content in paper by taking a single measurement of the water content in oil.
A set of moisture equilibrium curves relating water content of oil (WCO) to water content of paper
(WCP) applicable to transformer insulation has been known for decades [1-7]. Due to the fact that
transformers are rarely in equilibrium, the applicability of these curves in practice has always been
questionable. In spite of the precautions made by many practical engineers and the authors of the
papers [8,9,10], these curves are still being widely used for the estimation of dryness of transformers
during operational and storage stages.
It is important to understand that a number of assumptions are put in place when suggesting the use
of moisture equilibrium curves. These assumptions are often either not valid or misinterpreted and
therefore used incorrectly. Invalid assumptions made when using equilibrium charts for estimating
the water content of paper from a single moisture in oil test were discussed in [9-11] and
summarized as follows:
Assumption 1, Moisture equilibrium is readily attainable: Thus, water-in-paper can be evaluated
from the measurement of the water-in-oil content with the help of moisture equilibrium charts.
Assumption 2, WCO is uniformly distributed throughout: Thus, the top oil temperature can be
used when relating the moisture-in-oil, measured from an oil sample taken from the bottom drain
valve, to moisture in the cellulose.
Assumption 3, Water solubility of oil is the same for all new mineral oils and does not change
with aging: Thus, well-known moisture equilibrium charts are applicable to any mineral oil.
Assumption 4: The value of WCP is uniform and corresponds to the moisture content by dry weight
of the bulk cellulose insulation.
Assumption 5: Absorption and desorption of water from paper can be described by the same
equations. Hysteresis is ignored.
Assumption 6: The time constant to attain moisture equilibrium is the same as that of the time
constant to achieve temperature equilibrium.
Assumption 7: The sorption isotherms are universal for all kinds of cellulose material used in a
transformer. Thus .04 in. high density spacer insulation would have the same sorption isotherms as
.005 in. kraft turn insulation.
The estimated magnitude of errors for the above assumptions could achieve a vary high value and
was summarized in [11]

Equilibrium: ± 20%
Uniform oil distribution: ± 15%
Oil solubility: ± 100%
Uniform WCP distribution: ± 20%
Temperature increase or decrease: ± 20%
 Equal temperature & moisture time constants: ± 15%
All insulation has same sorption characteristics: ± 15%
If all of these errors are additive, it can be seen that the combined error could approach 200%, and
this does not take into account the method of oil sampling and accuracy of measurement of WCO.

Parameters for moisture identification

There are many ways to measure the presence of water in a transformer insulation system. There are
accordingly many terms to describe the water state in the insulation media. In general transformer
insulation systems consist of three media in which water can be found. These are; the gas space, the
insulation liquid (usually mineral oil), and the solid insulation in the form of conductor paper,
pressboard, wood, and other cellulose material. Partial water vapor pressure, absolute humidity,
relative saturation, dew point, mixing ratio, water by dry weight, ppm – all describe moisture of a
three component system.
Most commonly used terms to describe the moisture state of the transformer insulation are water
content in oil (WCO) measured in ppm, water content of paper (WCP) measured in percent by dry
weight, and the relative saturation of water in oil (RS) in %. Each measurement should be
considered along with temperature at which that value is observed. It will be shown later that the
quantification of the water state of the insulation system by any of these parameters alone has its
drawbacks and limitations.
When water in oil reaches the level beyond which oil can hold no more water it is said that the oil is
saturated with water. There are four saturation formulas that apply to gas-liquid systems that have
only a single condensable component, and they are listed here:

Relative p H 2O
Saturation RS = 100 (1)
p HS 2O
Molal p H 2O
Saturation MS = (2)
P − p H 2O
Absolute p H 2O MWH 2O
Saturation AS = (3)
P − p HS 2O MWDRY
Percentage p H 2 O / (P − p H 2 O )
PS = 100
Saturation (
p HS 2O / P − p HS 2O ) (4)

Where,

P is the pressure of the total gas mixture in the oil

PH2O is the vapor (partial moisture) pressure
PSH2O is the saturated vapor pressure
MWH2O is the molecular weight of vapor
MWDRY is the molecular weight of total gas mixture without water
Out of the four definitions of saturation only the relative saturation (RS) has found practical use in
water assessment in transformers. This is the quantity most commonly used for moisture assessment
and also is measured by the commercial RS sensors. However with an increasing demand for
determination and measurement of other dissolved gases in transformer oil, the need to consider
other saturation parameters will surface.

Relative Saturation
As discussed above, moisture relative saturation (RS) is the ratio of the actual partial vapor pressure
of water in oil to the same at saturation. Also RS can be calculated as the ratio of the actual amount
of moisture in the oil expressed in mg/kgoil (ppm) to the maximum amount of moisture the oil can
hold:
WCO
RS = 100 , (5)
WCO S

where WCO is the water content in oil in ppm, and WCOS is the saturation limit at a certain
temperature for that oil. Therefore, a relative saturation of 100% means the oil can hold no more
water without forming emulsion, and a relative saturation of 0% indicates there is no moisture in the
oil. When the system is in its thermodynamic equilibrium the term Equilibrium Relative Saturation
is used (ERS).
Relative saturation, combined with oil temperature, can be used to estimate the actual amount of
moisture in the oil if the mass of oil is known.
For the system consisting of air and water instead of Relative Saturation the term Relative Humidity
(RH) is normally used.

Water Content in Paper

The water content in transformer paper (WCP) is usually given as a single percentage, e.g. a new
transformer will be expected to have a WCP less than 1% whereas an old transformer may have a
WCP greater than 3%. These figures represent the bulk water content in the winding insulation and
are the ratio of the mass of the total water content to the mass of dry paper. However it would be
expected and it was shown in [12] that WCP will vary at a given time throughout a transformer
insulation system and at a given insulation location over a period of changing transformer load.
In a similar way the water content of the pressboard (WCPB) in a transformer is expected to be non
uniform and to vary with load.
The water content is very often the only parameter used to define moisture conditions in a paper/oil
insulation system. It influences many of the physical, mechanical and electrical properties of
transformers. Moisture content is, and will remain, of great interest to both manufacturers and their
customers. However, moisture content alone does not address the majority of the problems
regarding stability and aging, loading capability, life expectancy, reduction in dielectric stress, and
rehabilitation of transformer insulation.

Water Content in Oil

Until recently water content in oil (WCO) has been the only parameter measurable in a transformer
oil/paper system. Naturally, electrical utility practitioners tried to utilize this parameter to predict the
moisture state of power transformers and to devise other quantities describing the effects of water on
different aspects of transformer operation and maintenance.
Traditionally WCO has been measured by taking oil samples and performing a so called Karl
Fischer (KF) titration test in the laboratory environment. The result is the mass of water in oil
expressed in ppm (g/ton, mg/kg, etc). Research and practical experience have shown that a single
(once at a time) measurement of moisture content is not sufficient for the assessment of the moisture
state in a transformer. Also, the KF method suffers from a lack of accuracy and repeatability which
may result in very unrealistic and misleading estimations. Nowadays, reliable measurement of water
in oil can be accomplished with use of relative saturation instruments [7,13], based on thin film
capacitor sensors. These instruments allow measurement of not just relative saturation of water in oil
(for which they are designed for) but also these instruments are able to output the value of WCO in
ppm, provided the water in oil solubility characteristic is known. WCO is important to know when
filling the new transformer with processed oil. WCO is also widely used for determination of the
water content in solid insulation when employing moisture equilibrium theory and corresponding
equilibrium curves. However; as mentioned in the introduction, the use of equilibrium theory in
operating transformers is very limited and prone to a high level of error.

Water-in-Paper Activity

Water-in-paper activity (Awp) is the measure of water available for exchange between the cellulose
material and the surrounding oil. The term was introduced by the authors in [8] and is being used for
moisture assessment and its effects on transformer insulation. Water-in-paper activity reflects the
active part of total moisture content by dry weight or free water which, under normal circumstances,
can be exchanged between paper and oil. Water activity by itself is not a new concept and has been
used in food, biomedical, chemical, and paper industries for a long time. The use of Awp for
assessment and control of integrity of insulation proved to be advantageous over many traditional
indicators of moisture in transformers.
Numerically, water-in-paper activity is equal to the equilibrium relative saturation (ERS) divided
by 100.

Awp=ERS/100; (6)
or
Awp=p/pS (7)
In order to illustrate the concept of Awp consider the following example:
Four 31 mm thick pressboard blocks (total dry mass of 745 grams) of 0.5% water content were
added to the paper/oil system containing 345 grams (dry mass) of thin paper insulation of 4.6%
water content. After conducting a number of temperature rises and then keeping the system at the
constant temperature of 80 °C for 7 days, it came to the state where no significant moisture
exchange occurred. This state is known as moisture equilibrium. Using the equilibrium curves the
active water content of insulation of approximately 3% was determined. The calculated average
water content by dry mass for this system is equal to (745*0.005+345*0.046)/(745+345) = 1.8%,
which is significantly different from the value of 3% obtained using the equilibrium charts. The
reason for the difference lies in the fact that migration of water within cellulose insulation is driven
by two different physical processes. First, a fast adsorption of moisture into the surface layers of the
solid insulation occurs, and second, a slow diffusion of water from the outer to the inner layers takes
place. The lower the temperature the less water the oil can hold—temperature decrease forces water
into paper. On the other hand the lower the temperature the slower the diffusion rate of water into
the inner layers of paper. Even one week of the experiment under the constant temperature of 80 °C
was far from enough to reach the expected uniformly distributed 1.8% of water content in paper by
dry mass. As it will be shown later (see Figure 1) the equilibrium RS corresponding to the active
water content of 3.0% is 22% resulting in the water activity of 0.22.
It has been demonstrated experimentally that, generally, the water content of paper determined by
equilibrium charts is equal to neither the average water content nor to the water content of any
particular part of the solid insulation in a real power transformer. Based on numerous tests with
different types and thickness of paper insulation it was suggested that the amount of water which
migrates from paper to oil and visa versa corresponds to the free water available for exchange with
oil. In a real transformer this ‘active’ part of water is almost always less than the total amount of
water in solid insulation as determined by the ratio of water accumulated in paper to the dry weight
of paper insulation. This active part of water corresponds to water activity and can be found from
sorption isotherms relating active water content and water activity.
Water-in-paper activity or equilibrium relative saturation measures the water-vapor pressure
generated by the water present in cellulose material.
Paper being a hygroscopic product may absorb water in different ways: sorption and chemical
reaction, sorption and formation of a hydrate, water binding by surface energy, diffusion of water
molecules in the material structure, capillary condensation formation of a solution etc. [14].
Therefore, water molecules present in a paper material are not all retained by the paper with the
same strength. In fact, only part of the total moisture content can be exchanged between the paper
and its surrounding oil under normal conditions.
Moisture content includes both an immobilised part (eg, water of hydration) and an active part. This
explains the lack of accuracy and reliability of most methods for measuring and determination of
water content. The quantity of water found in a paper sample depends on both the method of
measurement and on the way it is applied. The active part of moisture content - or free water - is
usually measured in terms of its vapour pressure, expressed as percent relative saturation generated
in equilibrium with the paper sample in a closed system at constant temperature.
As it will be shown later the water activity indicates the degree of freedom of the water absorbed in
a paper and shows dimensions, structure, cohesion properties as well as electrical and chemical
properties better than moisture content does. The effect of this water activity on physical properties,
such as differences in water-vapour pressure, and not moisture content govern the interchange of
water between a paper and its surrounding oil. Therefore, mobility of moisture in a paper/oil system
is determined by any difference between Awp and RS/100 in the surrounding oil.

Water-in-Paper Mobility

Water-in-paper mobility (Mwp) is a new concept which the authors introduce in this paper. As was
mentioned in the previous section mobility of moisture in a paper/oil system can be determined by
any difference between Awp and RS/100. A normalised response of RS to temperature fluctuations
in an operating transformer may largely deviate from the benchmarked value of Awp. The spread of
that deviation can serve as a good indicator of a quantity of moisture migrating from paper to oil and
back during the daily load and temperature cycle. This means that not only the amount of water
which is available for exchange would matter, but also the amount of water which actually has been
exchanged between paper and oil is of great interest.
It is known that not only moisture, heat and O2 harmfully affect the insulation life, but more
importantly, a stress caused by transients results in an even greater danger to the insulation.
Therefore the deviation of RS/100 from Awp expressed in mathematical terms can be utilised as a
measure of the stress applied to the oil/paper insulation system.
For calculation purposes it is proposed to utilise the following equation to estimate the water
mobility:
t2
Mwp = ∫ ( Awp − RS (t ) / 100)dt (8)
t1

The lower the value of Mwp the lower water mobility and the less dangerous effect of moisture on
the insulation life.
The authors have observed that usually at high levels of Awp there is a high value of Mwp. However,
this could not always be the case. For example, when a transformer with forced oil circulation
system operates with a relatively high level of Awp at a relatively constant load the water-in-paper
mobility could be low.
As will be shown later in the paper, water-in-paper mobility along with water-in-paper activity can
form the two key factors upon which moisture assessment and management is dealt with.

Sorption Isotherms

Cellulose material is hydroscopic and therefore has its so called sorption characteristics. In terms of
water activity, the sorption characteristic is the relationship of water content WCP in percent and
Awp. Figure 1 shows a typical sorption isotherm for oil impregnated cellulose at 80°C.

Sorption Isotherm for 1mm pressboard at 80 C

10

6
WCPa, %

0
0 0.2 0.4 0.6 0.8 1
Aw p
 Figure 1
Experimentally obtained sorption isotherm for 1 mm oil impregnated pressboard
Sorption isotherm is a key characteristic which can be used for determination of water-in-paper
active (WCPa) from measured Awp.
Due to the complexity of sorption phenomena, sorption isotherms cannot be calculated in advance
and have to be experimentally determined for each single product. The practical usefulness of
sorption isotherms is severely limited by several factors:
• Sorption isotherms are valid only for a single material(e.g. one type of cellulose)
• Sorption isotherms are affected by any variation in material ( e.g. aging)
• Sorption isotherms show a difference between absorption and desorption curves (sorption
hysteresis)
• Sorption isotherms are experimentally determined; unfortunately most methods used for
measuring water content are neither very accurate nor reliable.

From the above it should be clear that moisture content very often cannot be used to accurately
determine Awp and vice versa. The practical usefulness of sorption isotherms is limited to products
whose previous moisture history is known and which have a defined composition and a stable
temperature.
The slope of sorption isotherms is such that a small variation in paper water content corresponds to a
large variation in Awp. Due to this fact, moisture content cannot be used to make a rough estimate of
Awp. However, Awp can often be used for moisture checks such as the check of a moisture content
deviation from factory values.
In practice, moisture content and water activity should be considered as two independent
parameters.
Sorption isotherms for cellulose material are typically sigmoid in nature. There are a number of
known mathematical expressions used for describing sorption isotherms. Among those is frequently
used the Fessler equation [5].
WCP = A exp( B / T )( p S (t ) Awp) C (9)
where A, B and C are the constants determined experimentally.
This type of equation was also used in building a set of the Monash equilibrium curves [8].
The Fessler equation (9) has been criticized for not following the sigmoid shape of natural sorption
isotherms for cellulose materials [14]. A much better approximation may be presented by the
Guggenheim-Anderson-deBoer (GAB) model:
WCP0 ⋅ C ⋅ K ⋅ Awp
WCPa = (10)
(1 − K ⋅ Awp)(1 − K ⋅ Awp + C ⋅ K ⋅ Awp)
where: C - constant
K= constant, in the range 0.7-1
Awp - water activity
WCPo - GAB monolayer moisture
WCPa – active moisture water content
The C and K constants related to the energies of interactions between the first and distant sorbed
molecules at the individual sorption sites.
This equation (10) has a similar form to another frequently used model called the BET (Brunauer,
Emmet & Tetter) equation, but has an extra constant, K. BET is actually a special case of GAB, with
K=1.
The modified GAB equation replaces C with C/T, where T is the temperature in °C. This enables
isotherms to be estimated for any temperature, based on data measured at one temperature.
However, the accuracy of this is approximate only, as it assumes all types of cellulose are affected
by temperature identically.
It is important to understand that solid insulation in transformers can be of different types, thus of
different sorption characteristics. For example, according to specification provided by Weidman Co
laminated hi-lam kraft transformer board has maximum water content of 10 % where as TX
laminated precompressed pressboard is of 3%.

Measuring Awp

Generally, the water activity of a product can be determined from the relative saturation of the oil
surrounding the paper insulation when the oil and the adjacent paper are at equilibrium.
When a paper is surrounded by oil, any change in temperature automatically causes the paper to
exchange moisture with oil, until the partial water vapor pressures at the surface of the paper and in
the oil are again equal. Regarding the measurement of Awp this sets the requirement of maintaining
the constant temperature of the paper/oil system for a certain period of time, or at least limiting
temperature fluctuations around a certain temperature level to a feasible minimum.
In an operating transformer this condition can not be achieved readily. Therefore a special technique
would be needed in order to detect or devise Awp from on-line measurement of RS.
A fuzzy logic based identification technique has been developed to detect the water-in-paper activity
from on-line measurement of temperature and relative saturation in oil. Using this method it is
imperative that the history of temperature and RS values were recorded for as many loading
conditions as possible [8].

Awp and WCP

The sorption curves determine the relationship between Awp and WCP. When using these curves for
an oil/paper system it should be understood that the WCP corresponds to the active water content
WCPa and is not normally equal to the total water content by dry weight.
Figure 2 shows that these curves are of different shape compared to the sigmoid type relationship
discussed above. The reason for that is the Fessler approximation, which is not accurate for the Awp
range of 0.7 – 1.0.
 25

20
T= 0

15 20
wcp, %

40
10
60
80
10 0
5

0
0 0 .1 0 .2 0 .3 0 .4 0 .5 0.6 0.7 0 .8 0 .9 1

Awp
Figure 2
Moisture equilibrium curves relating water-in-paper activity to the water content of paper
based on the Fessler equation [5].

Awp and WCO

Ingress of water in oil is caused by a few different mechanisms. Firstly, moisture desorbs from the
paper during temperature increase and represents the active water available for exchange with
surrounding oil. Secondly, moisture can enter the oil from outside of transformer tank due to poor
sealing. Thirdly, water can be generated as a byproduct as a result of chemical processes in oil and
paper during the life of the transformer.
As follows from Equation (5) the WCO is directly proportional to RS and can easily be determined
from values of Awp as:
WCO = Awp ⋅ WCO S (11)
Figure 3 shows the set of curves WCO = f(Awp) drawn at different temperatures. When building the
relationship between Awp and WCO it should be understood that moisture solubility in oil must be
known to avoid any kind of errors. Despite the validity of the relationship (11) it also should be
noted that water content in oil can be calculated at any time given that the values of RS, T and
solubility characteristic of oil are all known at that instance of time.
 900

800

700
T=100
600

ppm500
400
80
300

200 60
100
40
0 0 ºC
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Awp
Awp

Figure 3
Moisture equilibrium curves relating water-in-paper activity to the water content of oil.

Awp and RS

Numerically water activity is equal to equilibrium relative saturation (ERS) divided by 100.
However, there is a difference at a physical level which is important to understand. ERS is a
moisture saturation in liquid regardless of whether another material is present in the system or not.
ERS does not reflect the moisture content of paper for example and serves as a stand-alone
characteristic of moisture in oil. On the contrary, Awp is the measure of active water primarily in
paper available for exchange with the surrounding oil.
Both water activity (paper) and relative saturation (oil) are referred to the saturation pressure (ps) or
partial pressure of water vapor above pure water. The saturation pressure (ps) is strongly dependent
on temperature. At room temperature for example, (ps) increases by about 6.2% for a 1°C increase in
temperature. The relative saturation of oil is strongly temperature dependent. At 50 %RS and 60 °C
of the oil temperature, an increase of 10 °C results in a relative saturation decrease of about 15 %. At
90% RS, the same temperature increase causes relative saturation to decrease by even more – about
25 %.
The water-in-paper activity is not as strongly dependent on temperature. At 1% of water content of
paper and 60°C, for a temperature increase of 10°C the water-in-paper activity varies only by 0.006
Awp, or 0.6 % in terms of RS.
The water-in-paper activity is more sensitive to the changes in paper moisture content. At 70 °C an
increase of 1% in WCP results in increase of Awp by about 6.5% in terms of RS. This makes Awp a
good indicator of the solid insulation conditions.

Factors affecting moisture migration

Water activity equilibrium (thermodynamics) and rate of diffusion (dynamics of mass transfer) are
the two main factors influencing moisture migration in an oil/paper system. Multi-domain cellulose
material with regions formulated to different water activities cause the whole system to be in a non-
equilibrium state. This will result in moisture migration from the higher water activity (higher
chemical potential) to the lower.
Equation (12) relates the thermodynamic chemical potential of water vapor to the Awp:
wp = wp 0 RT ln Awp , (12)
where wp = chemical potential in the sample of water vapor;
wpo = chemical potential for pure water vapor;
R = ideal gas law constant;
T = temperature in K.
Water potential is the measure of the potential energy of water at a point in a system relative to the
potential energy of pure water.
In the multi component system such as air-oil-paper-pressboard after equilibration for the water,
wp(thin pressboard)= wp(oil) = wp(thick pressboard), and thus water activity of thin pressboard is
equal to water activity of oil and equal to water activity of thick pressboard. In other words, the
chemical potentials for the water are the same when the systems are brought to equilibrium at a
constant relative saturation. Even though the percent moisture was not the same for the thin and
thick pressboard, the water activities were the same, thus no change in moisture content occurs when
they are put together unless other chemical reactions took place.
The differences in vapor pressure of water between two regions also affects the rate of change in
WCP . On a simplistic basis this can be described by Fick’s First Law of diffusion
d (WCP) D
= ( p1 − p 2 ) (13)
dt l
where d(WCP)/dt = amount of moisture exchange per unit of time,
D = effective diffusivity of water,
l = the path length for diffusion,
p1 = water vapor pressure in system 1, and
p2 = water vapor pressure in system 2.
It is obvious that in real systems, as moisture is exchanged, the difference between p1 and p2 gets
smaller and thus the rate of moisture transfer decreases.

Establishing Awp benchmark

It is possible to determine the value of Awp, which could be specified as a nameplate parameter.
Let’s assume that a transformer has been dried to WCP equal to 0.5% by dry weight. Then at 65°C
temperature rise the ERS will be in the vicinity of 1.5-3% for the standard transformer oil.
In order to establish the Awp benchmark the manufacturer should install a moisture sensor and
perform the temperature rise test to obtain the RS response to the temperature change under rated
current conditions. Calculation of WCPa then should be based on detailed design knowledge
available to the manufacturer.
In each test case it is important that the manufacturer has data readily available to demonstrate that
the calculated values are supported by experimental testing (use of paper samples is one of the
solutions).
Top oil temperature monitoring should be conducted in order to predict RS at a level of top oil.
A model for determination of water activity should consider the following:
• The oil flow within the winding ducts
• The oil interactions with paper and coolers
• The distribution of temperature within thin paper insulation
• The initial water content of cellulose insulation
Loading a transformer beyond the nameplate rating during the temperature rise test can help to
determine the ultimate value of water content in oil. That could be useful to evaluate the possibility
of oil saturation during shutdown and a risk associated with energizing of a transformer in case of
cold load pickup.
The above test should be performed in accordance with IEEE Std C57.12.00-2000.
Figure 4 shows the results of the factory temperature rise test with relevant measurement of RS. As it
can be seen from this graph the Awp is about 0.018 which would correspond to approx. 0.6 % water
content in paper. For the new transformer we can assume that WCP=WCPa and the theory of
moisture equilibrium apply as the temperature was kept long enough to establish the state of
thermodynamic equilibrium.
Similar tests were performed in the laboratory where water content in paper was controlled from 0.8
% by dry weight to 7%.
 80 8

70 7

60 6
T
50 5

RS, %
T, C

40 4

30 3
RS
20 2

10 1

0 0
05

05

05

05

05

05

05

05

05

05
4

5
:0

:0

:0

:0

:0

:0

:0

:0

0:

1:

2:

3:

4:

5:

6:

7:

8:

9:
16

17

18

19

20

21

22

23

Time, hours

Figure 4
Response of RS to temperature change for new transformer

100 100

90 90

80 80
T
70 70

60 60
T, deg C

RS, %

50 50

40 40

30 30

20 20
RS
10 10

0 0
0 10 20 30 40 50 60 70
Time, hour

Figure 5
Response of RS to temperature rise: WCP= 0.8%
Examining the two responses of Figures 5 and 6 leads to the conclusion that performing the
temperature rise on the 2.3% coil results in the much higher Awp = 0.13-0.16, than the same
temperature rise on the 0.8% coil with the Awp = 0.045-0.055. Also the water in paper mobility is
much higher for the coil with 2.3% of water.
From these two figures it is also evident that it is possible to establish the reference value for Awp at
a rated temperature rise. Indeed, Awp does not change significantly with change in temperature
nearly of 30 °C. Change in Awp is mainly caused by change in WCP.

100 100

90 90
T
80 80

70 70

60 60
T, Deg C

RS, %
50 50

40 40

30 30
RS
20 20

10 10

0 0
0 10 20 30 40 50 60 70
Time, hour

Figure 6
Response of RS to temperature rise: WCPa= 2.3%
The purpose of the temperature rise test should be to develop a model for RS change in response to
winding and oil temperature change. Given that in a new transformer the ppm of water in oil leaving
the winding should be equal to the ppm of oil entering the windings from the radiator, allows to
predict the RS response at the top oil level based on measurement of RS and T at the bottom and T at
the top. That would result in possibility to determine water distribution along the height of the
transformer.
The response, not the absolute single value of the RS, is indicative of the Awp. The more RS varies
with the temperature change the higher the Awp for the particular transformer and the higher the
water-in-paper mobility.

Water activity and risk of bubble formation

There must be special conditions present for a bubble to form. Formation of bubbles in the paper/oil
media is a complex phenomenon and many factors should be taken into consideration when
predicting the state at which a bubble forms. Among the variables effecting conditions under which
a bubble evolves are temperature, temperature gradient across insulation, temperature derivative and
direction of change, water content in solid (thin) insulation and oil media, blanket gas pressure,
static head of oil, electrical stress, quality of cellulose material and chemical properties of cellulose
and oil. Although it is difficult to account for each of these factors or their combination, the Awp,
being a measure of water available for exchange with surrounding oil, is definitely able to serve as a
very convenient parameter for determination of bubble evolution.
A bubble forms in a liquid as a result of water vapor pressure unbalance when the system attempts to
follow a sudden change toward a new equilibrium condition. The unbalance is determined by the
difference in partial vapor pressures inside the paper, and same outside the paper, i.e. in oil. The
Awp clearly reflects this active water pressure, which should be in excess of static pressure over oil.
If there was a possibility of measuring RS in the proximity of conductor paper surface there would
be a clear signature indicating of bubbling.
According to its definition, Awp is a “two edge sword” which defines not only the amount of water
available for transfer from paper to oil but also it defines an amount of water which could be
absorbed by the conductor paper during transformer shutdown. If then during fast temperature
increase (energizing at condition close to rated) RS decreases and then increases above the level at
which it was before the temperature change then we can assume that bubbling is taking place.
An even more accurate prediction can be made by measuring the RS in the space above the oil and
comparing it to that of in the oil.

Effect of Water activity on dielectric strength

Any increase in water activity reduces dielectric breakdown strength. Research has shown that there
is a nearly linear dependence between the dielectric breakdown voltage of transformer insulation and
RS [13]. This might be arguable as RS is effected by two parameters - absolute water content and
temperature. This may result in the situation when an increase in temperature causes decrease in
relative saturation due to an increase in water solubility limit. On the other hand it may also cause a
decrease in RS due to desorbtion of water from paper and therefore increasing water content of oil.
For that reason neither WCP nor RS are appropriate means for monitoring and determination of
dielectric breakdown risk relative to those parameters. On the contrary water activity presenting
water available for exchange is “actively” directly taking part in the reduction of dielectric strength.
Awp is much less dependant on temperature than RS and therefore only changes in WCPa will cause
respective changes in Awp, which in turn will affect the dielectric integrity of the oil/paper system.
Although the relevant curves are yet to be constructed it is now clear that the Awp, not RS or WCP,
should be chosen to estimate the reduction in dielectric strength of both oil and paper.

Water activity and transformer drying process

Water activity may be a very important factor in controlling dryness of transformer insulation.
Indeed, water activity--not the water content--determines the lower limit of water available for
exchange with surrounding oil. Reducing water activity by keeping the oil dry is the way to maintain
transformer insulation in a good condition. The diffusion of water from the paper is a slow process
and therefore drying of transformer insulation is a time and energy extensive and costly exercise.
The molecular activity of water-vapor (ie, pressure) in hygroscopic materials like cellulose increases
with temperature at roughly the same rate as the increase in pressure of saturated water vapor. In fact
water activity shows only limited temperature dependence.
In a conventional hot air/vacuum method increasing the temperature of air in a dryer neither
decreases its water-vapor pressure nor increases its 'hygroscopicity'. Warm air increases the water-
vapor pressure generated by the paper to be dried by increasing its temperature. The drying enthalpy
and therefore the energy necessary to dry solid transformer insulation is strongly influenced by
water activity. For pure water (ie, Awp = 1,0) the evaporation enthalpy amounts are from 2300 to
2500 kJ/kg. These high values of enthalpy explain why the last part of the moisture content is so
time consuming to extract when drying.
The speed at which water can be removed by evaporation from cellulose is directly proportional to
the active surface of the parts of solid transformer insulation and to the difference in water-vapor
pressure between the product and its surroundings. Differences in moisture content have nothing to
do with this! Calculating and controlling drying process requires the knowledge of the relationship
between Awp and %WCP (sorption isotherms). As mentioned in previous sections, this relationship
is especially hard to determine. In practice it is necessary to consider that sorption isotherms
correspond to conditions of static equilibrium. Therefore, differences with theoretical calculations
based on sorption isotherms will occur.
Optimizing a drying process can only be done by experimentation. As already shown, moisture
content measurement alone does not give an accurate image of the drying process itself.
As was mentioned above drying a transformer cellulose insulation to a given moisture content can
be a time consuming exercise. From the point of view of energy conservation it is worth asking if
water activity rather than moisture content should be considered in the definition of the product
moisture condition after drying. Whenever product quality and stability are the main factors to
determine what a dryer should do, the answer should be positive. Significant energy savings as well
as better quality control will result from such a decision.
Understanding the Awp concept in relation to drying becomes of paramount importance when the on
load drying is concerned. This is where the process may take years and optimization of its critical
parameters can be of a great importance.
The Monash University research team had been involved in work associated with drying of a French
utility transformer [17]. The transformer under study was a 22MVA GSU power plant unit showing
high water content.
The Pall HTP-70 dehydrator was the system used to perform the on-load dryout of the transformer.
The system was designed to remove to some extend water, particles, air and other gases dissolved in
the transformer oil. The principle was based on a mass transfer by evaporation under partial vacuum
coupled with particle filtration. There was no extra heating device mounted in the HTP. The oil flow
rate was about 70 liters per minute. A Pall water sensor installed at the entrance of the system
continuously monitored the water-in-oil relative saturation (%RS) and oil temperature. The data was
recorded at a rate of one acquisition per hour.
Changes in WCPa during and after dryout have been continuously monitored. The Monash Water-
in-Paper algorithm [8] based on the Awp principle and run by the TMM software has been used to
estimate the WCPa using the Vaisala transmitter outputs since 30 January 2002 until removal of the
transmitter on 3 April 2003. The WCPa prior to January 2002 has been estimated from the Karl
Fischer WCO measurements. Figure 3 shows the WCPa profile from December 2001 to March
2003.
 3.5

3.0
3

2.5

2
WCPa, %

1.73

1.45 1.45
1.5 1.39
1.29

0.73

0.5

0
Dec , 2001 Jan 31 – Mar 25, May 18 – Jun 1, Jul 9 – Aug 5, 2002 Nov 22 – Dec 9, Dec 18– Jan 7, Feb 18 - Mar 7,
2002 2002 2002 2003 2003

Figure 7
Change in WCPa during and after Dryout
Prior to dryout the initial WCPa has been estimated as 3.0%. Based on the TMM estimation this
value has decreased to 0.73% at the end of drying. Then, as it was expected, the “active” moisture
content started to increase slowly and after one year has reached the value of 1.73%, which was still
well within the safety 2% of WCPa margin. Assuming that there was no moisture ingress into the
transformer tank since the dryout was completed, the increase in WCPa was caused by re-
distribution of moisture within cellulose insulation.
The above example is a real life demonstration of applicability of Awp concept to on-load drying
process.

Conclusion

A new concept of water-in-paper activity has been proposed to overcome many problems
associated with moisture assessment and management in transformer insulation.
It was shown that the water-in-paper activity, not the relative saturation, reflects harmful effect of
water on the dielectric strength and the risk of bubbling. It was also demonstrated that the water-in-
paper activity, not the water content in oil, should be used for estimation of the active part of water
content in paper.
Water content in paper corresponding to the Awp represents the active part of total water content,
which can be determined with the assistance of relative saturation sensors.
It is possible to benchmark the Awp by performing the temperature rise test at a factory and then use
it for further identification of moisture ingress or effectiveness of drying of transformer insulation.
The new concept and parameter of the water-in-paper activity can be standardized and used as a
measure of moisture state and quality assurance for both paper and oil in transformers.

References

1. Du, Y.; Zahn, M.; Lesieutre, B.C.; Mamishev, A.V.; Lindgren, S.R., ”Moisture equilibrium in
transformer paper-oil systems” ; Electrical Insulation Magazine, IEEE , Volume: 15 , Issue: 1
, Jan.-Feb. 1999, Pages:11 – 20.
2. J.D. Piper, “Moisture Equilibrium Between Gas Space and Fibrous Materials in Enclosed Electric
Equipment”,AIEE Transaction, Vol. 65, pp. 791-797, December 1946.
3. J. Fabre and A. Pichon, "Determining Process and Products of Paper in Oil. Application to
Transformers," CIGRE, Paper 137, 1960.
4. T V Oommen, “Moisture Equilibrium Charts for Transformer Insulating Drying Practice”, IEEE
Trans. on Power Apparatus and Systems, Vol. PAS-103, 10, 1984, pp. 3063-3067
5. W J McNutt, W A Fessler, T O Rouse and O R Compton, “A Refined Mathematical Model for
Prediction of Bubble Evolution in Transformers”, IEEE Trans. on Power Delivery, Vol. 4, No. 1,
January 1989, pp. 391-399.
6. B. Pahlavanpour; M. Martins; Eklund; “Study of moisture equilibrium in oil-paper system with
temperature variation”, Properties and Applications of Dielectric Materials, 2003. Proceedings of
the 7th International Conference on , Volume: 3 , 1-5 June 2003, pp:1124 – 1129, vol.3.
7. T V Oommen, “On-Line Moisture Monitoring in Transformers and Oil Processing Systems”,
CIGRE Symposium, Berlin 1993, Paper 110-03, pp. 1-4
8. V.G. Davydov, O. Roizman and W.J. Bonwick, “Moisture Evaluation in Oil and Paper for Sealed
Transformer Insulation System”, Proceedings of EPRI Substation Equipment Diagnostics
Conference VII, New Orleans, 1999, p. II-5-1 to II-5-15.
9. O.Roizman and V.G.Davydov, “New Procedure for Classification and Ranking of Power
Transformers on the Basis of Water Contamination”, Proceedings of EPRI Substation Equipment
Diagnostics Conference XI, 14 p., New Orleans, LA, February 2003.
10.B. Ward, IEEE Tutorial: “Moisture Estimation in Transformer Insulation”, IEEE/PES
Transformers committee meeting. San Diego, CA, March 9, 2004.
11.PC57.106 Revision Working Group Meeting, “IEEE Guide for Acceptance and Maintenance of
Insulating Oil in Equipment”, IEEE Insulating Fluids Subcommittee, Las Vegas, NV, October 26,
2004.
12.V. G. Davydov, O. M. Roizman and W. J. Bonwick, “Evaluation of Water Content in
Transformer Insulation Systems ”, Proceedings of EPRI Substation Equipment Diagnostics
Conference VI, New Orleans, 1998, 15 p.
13.P A von Guggenberg and J R Melcher, “An Immersible Relative Saturation Moisture Sensor with
Application to Transformer Oil”, Proceedings of the 3rd International Conference on Properties
and Applications of Dielectric Materials, Tokyo, July 8-12, 1991, pp. 1258-1261.
14.B. Vanin, “Oil-Impregnated Cellulosic Insulation; Moisture Diffusion and Equilibrium in View
of Intefacial Adsorption; Water vapor by Cellulose in Insulation Microcapillaries.”, CIGRE 2000,
Paper 15-204.
15.V. Sokolov and B. Vanin, "In-service Assessment of Water Content in Power Transformers",
Proc. of the 62nd Annual International Conference of Doble Clients, 1995.
16.T.V. Oommen, S. Lindgren, “ Bubble evolution from Transformer Overload”, Transmission and
Distribution Conference and Exposition, 2001 IEEE/PES ,Volume: 1 , 28 Oct.-2 Nov. 2001.
17.P. Guuinic, M Pristchepa, B Ward, V Davydov , A Reykherdt, O Roizman, “On-line Drying of
Transformer Insulation at EdF: Lessons Learned”, Proceedings of EPRI Substation Equipment
Diagnostics Conference VI, New Orleans, 2004.