See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/3259094

Water in oil-filled high-voltage equipment part II: Water content as

physicochemical tools for insulation condition diagnostic

Article  in  IEEE Electrical Insulation Magazine · October 2007

DOI: 10.1109/MEI.2007.4318270 · Source: IEEE Xplore

CITATIONS READS
19 503

2 authors, including:

Issouf Fofana
University of Québec in Chicoutimi
282 PUBLICATIONS   2,561 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Transformer oil reclamation by combining several strategies enhanced by the use of four adsorbents View project

CFD Modelling of power transformers View project

All content following this page was uploaded by Issouf Fofana on 29 November 2017.

The user has requested enhancement of the downloaded file.

 F E A T U R E A R T I C L E

Water in Oil-Filled High-Voltage

Equipment Part II: Water Content as
Physicochemical Tools for Insulation
Condition Diagnostic
Key words: water content, water saturation limit, criteria of limiting state, diagnostics of electrical
equipment, residual life

Introduction V. G. Arakelian
I
n power equipment, such as transformers, switchgear, bush- The All-Russia Electrotechnical Institute, Moscow,
ings, cables, and their accessories, insulating liquid func-
tions as a heat-transfer fluid as well as a dielectric material. This
Russia
insulating fluid, commonly mineral oil, is used alone as electrical
insulation only in regions in which, by design, voltage stresses I. Fofana
are relatively low. A solid insulating material is used in regions in Université du Québec à Chicoutimi, Québec,
which voltage stresses are high, or in which a particular physical
configuration is needed. The solid insulation materials com-
Canada
monly used as wrapping and spacers are cellulose papers and
boards. In-service power equipment life/aging is mainly related
to the degradation of the insulation that is caused dominantly
by the thermal upgrading of the insulating paper, together with
the decomposition of the paper. The by-products are water and Our investigations show the advantage
other substances of partly polarizable and ionizable character.
Moisture, the number one enemy of power equipment, worsens of diverse methods for measuring
the dielectric properties and accelerates the aging of solid/liquid humidity in insulating liquids, and a
insulation. The Achilles heel of cellulose comes from its affinity
for water and oil degradation products. Therefore, it is important
variety of analytical expressions for
to know the condition, by means of suitable diagnostic tools, of calculating the condition of liquid/
the oil-impregnated paper used as primary insulation in power solid insulation systems.
equipment.
In the first part of our investigations [1], we examined the
current state of knowledge of the effect of water on paper and
liquid-insulating materials. We demonstrated that the scientific
representation of the states of water in insulating liquids has a
direct practical application as a diagnostic tool for electrical equip- materials, followed by a critical review of up-to-date water quan-
ment, at least for the dielectric properties of the insulation. tification methods. Based on the physicochemical criteria of the
The work presented in this second part of these investigations is limiting state discussed in the third section, this article proposes
divided into four sections. The first section deals with the influence some analytical expressions that can help electrical engineers
of water on the electrical characteristics of the oil/paper insulating monitor and/or diagnose oil-filled, high-voltage equipment.

September/October 2007 — Vol. 23, No. 5 0883-7554/07/$25/©2007IEEE 15

 Influence of Water on the Electrical
Characteristics of Insulating Materials
Despite all its drawbacks, mineral insulating oil will continue
to be used as an insulating liquid, as long as natural reserves of
petroleum are not depleted. Therefore, considerable attention is
still being given to this natural product.
A. Dielectric Losses and Electrical Conductivity
of Liquid Dielectrics
Dielectric losses in insulating oil at power frequency are usu-
ally the result of the ionic and electrophoresis electrical conductiv-
ity of the oil and its impurities. Hydrocarbons in oils (paraffins,
isoparaffins, naphthenes, cyclo-aromatics, and aromatics) are
characterized by low dielectric losses. Thus, impurities (already
present in unused oil or formed during its storage or operation)
are responsible for an increase in the dissipation factor.
It has been shown experimentally that dissolved water, even at
high concentrations and at higher temperatures, does not increase
dielectric losses [2]. Dissolved water in unused oil is in a molecular
state, not dissociated into ions; consequently it does not influence
dielectric losses. Water in emulsion, contrary to dissolved water,
increases the dielectric losses at the expense of electrophoresis
conductivity. This is due to the movement of charged or polarized
micelles as emulsion water can be dissociated into ions. In used
oils, some amount of bound water is added due to interactions
with the resinous oxidation aging by-products, including salts of
naphthenic acids among others. In the colloidal polar particles
of oil, water connected to them is solubilized. Therefore, bound
water markedly increases dielectric losses.
Changes in the dielectric losses of oil have been monitored
during vacuum drying [2] (Figure 1). The first zone is probably
related to the removal of emulsion water. This leads to a sharp
decrease in dielectric losses. The second and third zones (this
division is given in [2]) are zones of low influence on dielectric
losses. These are caused by the removal of dissolved and physi-
Figure 1. Dissipation factor of oil (at 90°C) versus duration of
vacuum drying process for various transformer oils.
16 IEEE Electrical Insulation Magazine
cally bound water occurring at equilibrium. This process of water
removal takes place through a stage of dissolved water and is
accompanied not only by a decrease of dielectric losses, but also
by an increase in the specific volumetric resistance.
Generally, the substances having a low influence on the dielec-
tric losses of oil exert a high influence on the dielectric losses of
oil-impregnated paper, and vice versa. Low-molecular peroxides,
acids, and other polar substances dissolved in oil, which practi-
cally do not influence the dielectric losses of oil, are adsorbed by
paper and raise its dielectric losses. Substances forming colloids
in oil, which are the basic source of conductivity and dielectric
losses in oil, depict very low influence on the dielectric losses of
paper. This is a confirmation of the above-mentioned water equi-
librium, in which the distribution of water in a paper-oil system
goes through a state of intrinsic solution in oil.
B. Electric Strength of Liquid Dielectrics
The electric strength of liquid dielectrics is not directly related
to specific conductivity or dielectric losses; but it strongly depends
on the presence of impurities. Dissolved gases and compounds,
with a low boiling point, play an essential role during breakdown,
reducing the electric strength of dielectric liquids. Polar com-
pounds at the impurity level, forming intrinsic solutions in oil, do
not strongly influence the electric strength. However, if they are in
the colloidal or emulsion state, the low-boiling substances greatly
influence the electric strength of oil. This is probably related to
their boiling-off with the formation of gas cavities [2].
Water has a strong influence on electric strength. In a wide
range of relative humidity, the electric strength of clean insulat-
ing mineral oil, for temperatures higher than 0°C, falls practically
exponentially as relative humidity ϕ increases up to 1:
Ubr = 93 e –ϕ, for ϕ ≤ 1. (1)
It remains at a constant level of 20–25 kV in a small range of
supersaturated states (Figure 2) [2], [3]. A similar relationship is
Figure 2. Electric strength of insulating liquids versus relative
humidity.
observed for more simple chemical analogs of hydrocarbons, the
alkylbenzene oils:
Ubr = 102 e –ϕ – 14.5 ϕ , for ϕ ≤ 1. (2)
However, this is only a rough approximation. It is difficult to
rely on this relationship because the deviation in values of electric
strength is large [4]. This deviation depends highly on the nature
of the insulating liquids and for mineral oil, on its structural-group
composition. Due to the possible formation of charged micelles
and conducting bridges, water has more influence on the break-
down of oil when it is in the colloidal state. At low-water contents
in clean unused mineral oil, electric strength decreases slightly as
humidity increases. But, at a relative humidity of between 10 to
15%, it decreases more markedly as an exponential curve.
The influence of water on electric strength during vacuum dry-
ing is illustrated in Figure 3 [2]. The first zone is caused by the
removal of emulsion water and characterized by a sharp increase
in electric strength. The second and third zones, zones of weak
dependence on the electric strength, close to 60 kV in a standard
discharge gap, are characterized by the removal of both physically
bound and dissolved water at equilibrium.
Considering the dependencies presented in Figures 1 and 3, it
is possible to conclude that bound water can reduce the electric
strength of oil by approximately 40% and increase the dissipa-
tion factor by a factor of 2 or 3. Emulsion water can reduce the
electric strength approximately by a factor of 4, and increase the
dissipation factor by a factor of 1.5.
The influence of humidity on the dielectric strength of insulat-
ing liquids can be studied with the use of a more general parameter,
i.e., relative humidity. It is standardized, as a rule, by the different
contents of water at a given temperature. However, the influence
of temperature is also very specific and depends on water content
[2]. Therefore, despite an abundance of experimental data, it has
been impossible, until now, to generalize and to mathematically
Figure 3. Electric strength of oil versus duration of vacuum
drying. I, II, and III zones.
September/October 2007 — Vol. 23, No. 5 17
describe, for engineering use, the influence of this most important
general parameter of liquid dielectrics.
C. Electric Characteristics of Paper
The adsorption of polar components results in a change in the
electrophysical characteristics of paper, particularly, in the in-
crease of its dielectric losses. The dependencies of the dissipation
factor at 80°C and the dielectric permittivity of oil-impregnated
paper εr with water content W, in dry weight percentage, are ex-
pressed by these (generalized) equations [2]:
tgδ = 0.0021 + 0.0083 W 2, (3)
εr = 3.8 + 0.41W. (4)
The presence of moisture in oil-impregnated paper also influ-
ences the inception of partial discharges (PD): the resistance of
paper to PD decreases with moisture uptake [5].
Water Quantification Methods and
Their Potentialities
A. Physical Methods
ASTM Method D 4056-01 [6] is the standard method for evalu-
ating the solubility of water in hydrocarbons and aliphatic ester
lubricants. Formulas given in the document, allow the calculation
of water concentration up to 1000 mg/kg in hydrocarbons and up
to 30,000 mg/kg in oxygen-containing compounds, as a function
of density, refraction, and molecular weight. ASTM Method D
1315-63 (1973) describes the direct determination of water in
insulating oil by extraction.
B. Chemical Methods
These methods involve the use of specific reagents such as so-
dium, hydrides, carbides, or azides of metals (Mg3N2). This group
of reagents interacts with water, forming gaseous components,
the volume of which can be measured.
2H2O + 2Na = 2NaOH + H2,
2H2O + CaH2 = Ca(OH)2 + 2H2,
2H2O + CaC2 = Ca(OH)2 + C2H2,
6H2O + Mg3N2 = 3Mg(OH)2 + 2NH3,
4H2O + LiAlD4 = Al(OH)3 + LiOH + 4HD.
All these methods (reactions) allow determining the dissolved
water. The use of calcium hydride gives the greatest effect: for
one mole of water, one mole of hydrogen is formed. This method
has been widely used for some time and standardized in Russia
(GOST 7822-75 [7]), but it is practically not used anymore. The
sensitivity is about 10 mg/kg, which is not acceptable for engi-
neering purposes. The use of the lithium-aluminum-deuterium
compound presents the advantage of using mass-spectrometer
detection [8].
Currently, the most widely used technique is the Karl Fischer
(KF) method. This method is based on the chemical oxidation-
reduction reaction of iodine with sulfur dioxide in the presence
 of water and in an excess of pyridine. The basic chemical reac-
tion is:
H2O + I2 + SO2 = 2HI + SO3.
With volumetric titration, iodine is added as a titrant. The
method uses an automatic titrator [9]. With coulometric titration,
iodine is generated electrochemically by anodic oxidation on a
generating electrode. As long as there is water, generated iodine
will react chemically with it. Therefore the amount of electricity
generated is a direct measure of the water content. The method
provides a determination from 2 to 100 mg/kg. Coulometric
titration is performed with commercial instruments and is stan-
dardized as IEC Standard 60814 [10] and ASTM Method D 1533
[11]. IEC 814 (1985) is used for new and used insulating liquids.
Most commercial instruments and chemicals are suitable for this
analysis.
Both the Karl Fischer and the hydride calcium methods give
results that are too high due to the reaction with hydroperoxides,
acids, and other impurities [12]. The authors in [12] consider
that the Karl Fisher and hydride calcium methods involve the
determination of not only dissolved water but also of bound and
emulsion waters. Authors in [13] consider that GOST 7822-75
and IEC 60814 determine dissolved and emulsion waters, but not
bound water. In our opinion, the Karl Fischer reagent is a pow-
erful means of destroying bound water. Consequently, it should
be considered as a method for the determination of total water
content, with due regard for its reaction with peroxides and other
impurities. Borsi [14] shares the same opinion, specifying that
the Karl Fischer method determines both dissolved and chemi-
cally bound water.
C. Chromatographic Methods
It is likely that direct injection of an insulating oil sample into
the injection port of a chromatograph at 250 to 300°C will detect
all the water contained in this sample, regardless of its state. This
is generally achieved with a precolumn to separate the oil. A 2.5
cm × 3 mm precolumn, filled with a polymeric packing material,
allows the collection of up to 0.6 ml of insulating oil and the analy-
sis of up to 6 oil samples without forming asymmetrical peaks of
water on the chromatogram after 2 minutes. This is performed at a
helium flow rate of 25 ml/minute, in a 2 m × 3 mm column filled
with 15% of PEGA on “Polysorb-1”, at an injection temperature
of 180°C, with a column and thermoconductivity detector tem-
perature of 120°C, and for an injected sample volume of 100 µl.
It is possible to raise the injection temperature to be sure that all
the water in the oil sample is analyzed. However, it is impossible
to use sample syringes directly, and it is necessary to replace the
precolumn or purge it at high temperature.
Another way to determine water content in insulating oils is
based on its preliminary extraction in a gas phase and its determi-
nation by chromatography or coulometry. The amount of extracted
water will depend on the extraction conditions.
Extraction of water also takes place during the extraction of
gases from transformer oil for the purpose of their chromatograph-
ic analysis. In the automatic device with dynamic thermovacuum
extraction of gases [15], water is detected in the thermal conductiv-
18 IEEE Electrical Insulation Magazine
ity detector before hydrocarbons C4. In a two-channel recording,
the water peak appears as the fourth peak in the chromatogram of
the simple gases, after hydrogen, oxygen, and nitrogen. The water
peak is asymmetrical, and its asymmetry increases with the water
content of the oil. One of the reasons is the nonlinearity of the iso-
therms of adsorption of water on Porapack N. Another reason, and
probably the most important one, is the large amount of extracted
water and the inadequate gas connections for the analysis of such
a polar compound. However, the integration system easily copes
with the processing of asymmetrical peaks, and it is not a handicap
for the quantitative calculation of water content. The advantage of
this method is that data on water content appear simultaneously
with those of diagnostic gases. The detection limit is less than 1
mg/kg; but, the upper limit is probably limited by the design of
the extraction system. As the water content calculated in micro-
liters of water vapor per liter of oil is very high, a water content
of 1 mg/kg corresponds approximately to 1000 µl/l.
For the gas chromatographic determination of water content
as an independent task, it is necessary to achieve a symmetric
peak of water first. For this purpose, it is necessary to use a
column filled with Porapack Q and up to 10% of PEG, with an
average molecular weight of 1000 (PEG-1000), and to install all
gas connections as water vapor inlets in a thermostat at a higher
temperature. We are using the automatic analysis system for water
in insulating liquids shown in Figure 4.
The conditions for the analysis are: a column of 180 cm × 1/8
in., filled with Porapack Q of 80 to 100 mesh and 10% of PEG-
1000, argon carrier-gas at 20 ml/minute, a thermal conductivity
detector, a detector temperature of 150°C, a temperature of the
column and extraction system of 120°C. The extraction time of
water vapor is 60 to 80 seconds; contact time (intensive mass-
exchange in a thin layer) is 2 to 3 seconds. The retention time of
the symmetric peak of water is 3 to 3.5 minutes. The cycle for the
analysis is 6 minutes. The detection limit is 0.2 mg/kg. The repeat-
ability of results is 4%. Figure 5 shows a chromatogram for the
Figure 4. Automatic gas chromatographic analysis of insulating
oils for water content: 1 - small three-way valve for installing
the oil syringe; 2, filter; 3, measured vessel; 4, device for the
thermovacuum extraction of gases and water.
water content in insulating oil. Calibration is done using a solution
of water in mineral oil prepared with the help of ultrasound.
The method of calibration influences the results; the method
determines the absolute water content. As measures for intensive
mass-exchange and linear isotherms of water content (as we shall
see later) are developed, the method provides correct results using
constant calibration factors in all ranges of concentrations.
D. Sensor Controls
Nowadays, there is an increasing interest in using transformer
monitoring systems. One of the key variables to be monitored
is moisture content in the oil, as moisture is recognized to be
harmful to the insulation. Moisture in oil measurements using
thin film capacitance sensors have been available for some time
[16], [17]. The development of these sensor systems reacting to
moisture has reached such a level that it is now possible to use
them without endless checking and recalibration. The thin-film
transducers work over a wide range of relative humidity, and they
are directly dependent on dissolved water or water partial pressure.
The transducers cannot react with other states of water.
The main advantage of sensor systems is the possibility of
using them for monitoring. Sensors are calibrated with water
vapor and attached directly onto operating equipment. Usually,
the humidity transducer is combined with the temperature trans-
ducer. Thus, the sensors will give information about the relative
humidity and temperature of the oil in a continuous mode that
provides a way of controlling humidity and, therefore, the electric
characteristics of the solid insulation.
E. Correlation Between Methods used to Quantify
Water in Dielectric Liquids
Each of the above listed methods has specific features in rela-
tion to the complex water/insulating oil system. Therefore, it is
not surprising that the correlation between these methods is poor;
it should be so. It is necessary to define what is expected from
the analysis of water in oil. For the determination of total water
content, probably only the method of direct injection of oil at high
temperature is reliable. The Karl Fischer method will determine
all the water present in unused insulating liquids and will prob-
ably overestimate it because of impurities. Gas chromatographic
analysis with the dynamic thermovacuum extraction system will
provide correct determination of all water except chemically
bound water. For the determination of relative humidity above
Figure 5. Chromatogram for the analysis of water in mineral
oil.
September/October 2007 — Vol. 23, No. 5 19
the insulating liquid, only the thin-film transducers are suitable.
Also, it is necessary to recognize that there is no reliable method
for determining dissolved water.
F. Determination of Water in Layered Dielectrics
IEC Standard 60733 [18] describes a method for the deter-
mination of humidity, not only in insulating oils but also in oil-
impregnated paper and cardboard. The method can determine
water in oil-impregnated paper and cardboard in a range of 0.1
to 20 % by weight. The basis for this method is the Karl Fischer
titration. The use of the Karl Fischer titration for the analysis of
water in paper shows the capacity of the method to determine
bound water in oil indirectly, but with accuracy.
Physicochemical Criteria of the Limit State
For diagnostic purposes, it is necessary to have a criterion for
the limiting state for decision-makings concerning the mainte-
nance and the residual life of power equipment. For the reliability
of diagnostic actions, legal documents should be submitted; a
report on the normative and criterion parameters [19]. The norma-
tive parameters — the basic technical requirements of technical
and design documentation — are intended for the qualification of
the equipment condition as either serviceable or faulty.
Criterion parameters [19] focus on the requirements describ-
ing the equipment in terms of its ability to perform its functions,
as efficient, partly efficient or disabled. Criteria of the limiting
state are the quantitative characteristics of a limiting state, i.e.,
a condition in which further operation is not acceptable or not
efficient.
The electric characteristics should be ensured during all ser-
vice operation. They are the critical parameters. By comparison,
the physicochemical characteristics are not as critical in service
[20]. The oil characteristics may vary widely. These variations
are not a reason to remove the equipment from service. When
switching the equipment off for other reasons, the differences in
the physicochemical characteristics of the oil, as compared to the
normative parameters, will serve as justification for replacing or
regenerating the oil. Actually, from the long list of physicochemi-
cal parameters, only two can be considered as critical parameters.
These are the partial pressure of the gases and water for sealed
electrical equipment, and the humidity of the solid insulation. The
latter is measured as paper moisture content, dissipation factor
or conductivity.
Normalized water contents in oil are meaningless because
equilibrium conditions change with temperature. The same value
of moisture in oil can correspond to either a satisfactory or an
unsatisfactory condition of paper. For example, critical values
for moisture in paper used in capacitors are related to critical PD
inception levels [21]. With the current technology of paper drying
(at 120°C, down to a residual pressure of 1 mbar), the residual
moisture of paper in the capacitor is one thousandth of a percent.
At W < 0.5 %, the critical level of PD appearance, Ecr, is reduced
by 10%, at W = 3% by a factor of 2. At W > 3% the moisture in
the solid insulation, as indicated in Table 1, corresponds to filling
a monolayer, which should be considered as not acceptable.
Acceptance criterion values are given in Table 1.

Table 1. Criteria of limiting state for physicochemical diagnostics based on
water content in solid insulation.
Capacitors Other equipments
Paper humidity, %, maximum 3 4
Dissipation factor, %, maximum 1.6
Specific volumetric resistance, Ω.m, minimum 2⋅109
Diagnostics Based on Water Analysis
A. Calculation of the Humidity Content
of Solid Insulation
Time domain spectroscopy or frequency domain spectroscopy
techniques are used by power utilities for assessment of power
transformer insulation wetness [22], [23]. For accurate results,
the transformer has first to be held at a constant temperature for
several days [22]. Otherwise, the most effective use of these
techniques is for trending purposes only and not for an indication
of absolute insulation wetness.
Moisture equilibrium characteristics between oil and paper
insulation are well established and often used [24], [25]. Moisture
in paper can be determined from the moisture measured in oil
in those periods when the moisture in oil has attained its steady
state [25], [26]. Thermodynamically, equilibrium is reached when
macroscopic observables parameters (pressure, temperature, etc.)
have ceased changing with time [26]. Water migration/dynamics
between the paper and oil are well-known to be a slow process.
So even with a constant temperature, a moisture steady state/
equilibrium requires a long time to be attained. In addition, trans-
former load usually varies, and ambient temperature describes
daily cycles, so steady state (thermodynamic equilibrium) is
never completely attained. This is particularly true in transform-
ers that have load profiles which are cyclical and have periods
of rapid change.
Nevertheless, a local thermodynamic equilibrium (LTE) can
occur. This happens when the macroscopic observables are vary-
ing in space and time, but so slowly that, for any point, one can
assume thermodynamic equilibrium in some neighborhood about
that point [27]. Therefore, LTE can take place when a transformer
operates at a constant load for a sufficient time. Equilibrium dia-
grams, therefore, can be used provided their validity is limited to
areas with LTE. However, at low temperatures, LTE are hardly
A transducer placed in oil, but calibrated from vapor water in
1
air, provides values of relative humidity of a gas phase at equi-
librium, even though the gas phase is not present.
20 IEEE Electrical Insulation Magazine
reached because time constants are very long. Moisture diffusion
time for 1 mm-thick oil-impregnated pressboard is 6 hours at 70°C
and 333 hours at 20°C [24].
Diagnostics based on water content consist of determining the
moisture content of the solid insulation, which can be calculated
from the partial pressure of water (p) and temperature (t) (5) [1],
[2] or directly from the relative humidity (ϕ) above the insulating
liquid (6) [1], [2]. The terms A, B, h, q, n, and m are defined in
(12) and (13) and Tables 6 and 7 in Part 1 [1]. These parameters
are directly related to dissolved water.
W = A e –B t p h + q t, (5)
W = (4.61 + lnϕ)/(n – mϕ). (6)
There are two ways of obtaining data about moisture in the
liquid insulation of operating equipment: by monitoring the
relative humidity1 and temperature, or by laboratory analysis of
oil samples. The laboratory analysis can be performed by gas
chromatography, the Karl Fischer method, or with a humidity
transducer.
B. Use of Transducers on the Equipment
The humidity of insulating oil in operating equipment can
be measured with a permanently installed, thin-film transducer,
providing relative humidity values calibrated with water vapor
[16], [17]. The values of ϕ and t thus obtained can be directly used
for the calculation of the characteristics of paper as the humidity
transducer provides only the relative humidity responsible for the
humidity in the solid insulation (6). Using (5), the partial pres-
sure of water vapor at saturation above pure water is described
by the equation:
ln(ps) = 8.0589 – 1729.9875/(t + 233.856), (7)
where ps (in mmHg at temperatures higher than 0°C ) is calculated
from the reading of the temperature transducer. The water partial
pressure (in mmHg) is calculated from the combined reading of
the transducer:
p = ϕ10 (8.0589 – 1729.9875)/(t + 233.856). (8)
It is necessary to install the transducer in the upper, hotter layers
of the oil; but the temperature of paper is always higher. If this
difference is known as a function of loading and oil temperature,
it should be used for recalculating the relative humidity and partial
pressure. A local temperature rise will raise the local partial pres-
sure of water, the solubility of oil, and the relative humidity. As
shown previously, this will increase the concentration of dissolved
water due to the dissolution of a part of the dispersed and bound
water. As there is no direct communication between the transducer
and paper through a gas space, equilibrium of partial pressures is
impossible. In a stationary mode at constant temperature, a static
equilibrium will take place and the concentration of dissolved
water will become identical in different temperature zones, despite
the complicated moisture dynamics in oil-impregnated paper. For
oil around the transducer (9):
p = K cH2O for an isotherm t = toil, (9)
and for oil around paper:
p’ = K’ cH2O for an isotherm t = tpaper (10)
thus:
p’/K’ = p/K
(11)
Taking into account that Henry’s factor (K) is equal to the ratio
of saturation pressure to solubility (1[a]) in Part 1 [1], it yields:
p’ c*H2O’ / pH’ = p c*H2O / pH,
ϕ’ = ϕ c*H2O / c*H2O’, (12)
where values related to paper temperature are indicated by a prime.
Water solubility values in mineral oil at different temperatures
(permanent characteristic for a given grade of oil, against water
saturation limit) are obtained from Figure 7 of Part 1 [1], (with
the tolerance which will be leveled due to ratio) as:
ln(c*H2O ) = 7.33 – 1521.3 /(t + 233.856). (13)
Water vapor pressure at a given paper temperature is obtained
from (8) for ϕ’ and t = tpaper, then inserted into (5) and (6).
C. Use of Transducer in the Laboratory
Data necessary to calculate paper humidity can be obtained
by directly measuring the relative humidity of an oil sample
with a transducer, at the temperature corresponding to the paper
temperature in the equipment, with the subsequent calculation
of partial pressure from (8), and paper humidity W from (5) and
(6). In this case, it is not necessary to use otherwise unreliable
solubility data. In this way, the task of calculating W can be solved
without approximation.
D. Use of Other Humidity-in-Oil Data
from Laboratory Investigations
Humidity (H) can be obtained from the laboratory analysis of
oil by chromatography or by the Karl Fischer method. For diag-
nostic purposes, it is necessary to determine the relative humidity
of air at equilibrium ϕ. Converting H into ϕ is possible only with
water content isotherms (similar to those in Figure 8 of Part 1 [1])
at all operating temperatures.
The values available at 23 and 60°C, shown in Table 4 of Part 1
[1], can be used for the analytical representation of isotherms. By
analogy with the adsorption isotherms of Lengmure and Brunauer-
Emmett-Teller [28], the correlation can be expressed as:
September/October 2007 — Vol. 23, No. 5 21
C ϕ A/B
H/H* = –––––––––––– (14)
1+ (C–1) ϕ ABϕ
which, with a certain combination of coefficients, allows the de-
scription of all the variations of water content isotherms: direct,
convex, concave, and S-shaped. Coefficient C varies from 0 to 1.
Coefficients A and B should be positive and less than the estab-
lished limit. Their ratio determines the slope of an S-shaped curve.
When C = 1 and A = B, the relationship becomes linear. The values
for coefficients A, B, and C shown in Table 2 are calculated from
the relative humidity values found in Table 4 Part 1 [1].
The 60°C isotherm from (14) with the above chosen coeffi-
cients correlates well with experimental results. The isotherm at
20°C does not, and for the coefficients we have preferred the data
obtained at relative humidities of 0.2 and 0.5. As with temperature
rise, the isotherm should be straightened, the coefficient C should
be close to 1, and coefficients A and B should become closer.
For synchronizing coefficients A and B, their ratio is used. The
A/B ratio should not exceed 1. To provide the specified bound-
ary values of the coefficients, it is necessary to use the following
functions varying linearly with temperature t in °C, using values
given in Table 2:
ln(B / (20 – B)) = – 2.0904 + 0.03808 t,
ln(C / (1 – C)) = – 0.97097 + 0.02622 t, (15)
ln(A/B / (1 – A/B)) = – 0.85458 + 0.02742 t.
Expression (14) significantly deforms the initial area up to ϕ
= 0.2 between 10 and 35°C and overestimates it. For this range,
we shall take advantage of the Langmuir equation:
H/H* = N ϕ/( M + ϕ). (16)
Using data from (14) for relative air humidities of 0.2 and 0.3,
the relationship of factors N and M with temperature t in °C is:
N = 1.147 – 0.025 t. (17)
M = 0.032 + 0.003 t.
Expressions (14) and (15) (for ϕ ≥ 0.2) and equations (16) and
(17) (for ϕ < 0.2 and t < 35°C) allow us to calculate the relative
humidity of oils between 10 and 120°C (Figure 6), and to select
values of relative humidities ϕ corresponding to the values H
determined by analysis, taking into account the water saturation
limit H* [1]:
ln H* = 6.41 – 1217 / (t + 233.856) (17a)
Table 2. Parameters A, B, and C depicted in (14) for mineral oil.
A B C A/B
23°C 0.43 1.15 0.3 0.3739
60°C 10.5 12.2 0.8 0.8607

Figure 6. Relative water content isotherms of mineral oil,
computed from (14)–(17).
The partial pressure of water is calculated from (8) and the
paper humidity from (5) and (6).
E. Calculation of the Electric Characteristics
of Paper
The dissipation factor and specific volumetric resistance of
paper at a given level of paper humidity W (5) and (6) are calcu-
lated [2] using equations:
tg δ = ( p + q W r ) exp (− 2700 / T ), (18)
ρv = s exp [ h / T − n ( W − m )d], (19)
where tg δ is in absolute units, ρv in Ω × m, W in %, T is the
temperature of paper, in K; p, q, r, s, h, n, m, d are the tabulated
data from Table 3.
F. Oil Temperature Acceptable Limits Computation
From the obtained value of oil humidity, it is possible not only
to calculate the values of paper humidity, dissipation factor and
specific conductivity, but also to determine the lower acceptable
limit of oil temperature. The notion of an acceptable temperature
of oil arises in relation with the redistribution of gases and water
under the action of temperature between two or three phases of
the electrical equipment insulating system: gas, liquid, and solid.
The change of temperature of the oil because of changes of load
or ambient temperature can result in the adverse redistribution of
water regarding the achievement of the limiting state for the solid
insulation humidity. The estimation of an acceptable lower limit
of oil temperature based on the analysis of humidity is especially
important after repeated operation start-ups of cold equipment.
Before being energized, the equipment should be warmed-up to
the allowable lower limit of temperature calculated above.
22 IEEE Electrical Insulation Magazine
Table 3. Empirical factors for calculating the dissipation factor and specific
resistance of paper using (18) and (19).
Cellulose material
Cable Paper Capacitor paper
Coefficients “K-120” “K-2-10”
p 2.68 3.8
q 6 7
r 1.85 2.3
s 2.7.105 1.4.106
h 5800 5600
n 2.8 2.3
m 0.1 0.03
d 0.8 1
A humidity/temperature transducer permanently installed on
the equipment will give direct information on a developing situa-
tion. In its absence, the task can be solved by using the laboratory
analysis of oil-water content.
The total amount of water G (in kg) in a sealed apparatus is the
amount of water in the solid insulation (paper) Gp, the transformer
oil, Goil, and the gas space Gg:
G = Goil + Gp + Gg = 10–3HQoil + 10WQp + 10–3αVg, (20)
where the value of the oil water content H (in mg/kg) is obtained
by analysis (GC, KF), or calculated from the relative humidity
(obtained from a transducer and (14) or (16) and water saturation
limit (17a), [1], at the temperature of measurement. Humidity
of the solid insulation W (in %) is calculated from (5) or (6),
the weight of transformer oil Qoil and paper Qp, in tons, and the
volume of the gas space Vg, in m3. The humidity of the gas above
the transformer oil α (in g/m3) is derived from the partial pressure
[in mm Hg, (8)], as:
α = 0.98529 pH2O, (21)
or from the relative humidity and temperature, using hygrometric
tables.
From the calculated total amount of water in the equipment,
one can determine the paper humidity, the humidity of the oil and
of the gas space at equilibrium at all operating temperatures in an
iterative way. Other parameters related to humidity are determined
using (18) and (19).
G. Residual Life Prediction
Expression (20) for the equilibrium of water is the basis for the
determination of residual life. Assuming that the total amount of
water in electrotechnical equipment increases continuously and
linearly, neither paper humidity nor dissipation factor, nor specific
resistance will have a linear relationship with temperature, which
makes the calculation of residual life directly from these param-
eters impossible. Therefore, at operation start-up, it is necessary
to calculate the total amount of water in the equipment from the
results of the analysis of humidity in oil, and to use it as a starting
reading point: ΣGbeg. Then, it is necessary to calculate the limiting
values of humidity, dissipation factor, and specific resistance at all
operating temperatures and the total amount of water in the equip-
ment in the limiting state. Then, from the values obtained in the
limiting state, the most efficient and proven final value has been
chosen: ΣGend. From the water content (Gτ) accumulated during
time τ between present time and the beginning of operation in
years, calculated from the analysis of oil humidity in service, the
residual life rres, expressed in material (accumulation of water),
in percentage, is:
rres = 100 (ΣGend – ΣGτ)/(ΣGend – ΣGbeg), (22)
or expressed in time, in years:
rres = τ(ΣGend – ΣGbeg)/(ΣGτ –ΣGbeg) – τ. (23)
If the initial value ΣGbeg is not known, the residual lifetime is
calculated based on regular analyses as:
rres = ∆τ(ΣGend – ΣGτ2)/(ΣGτ2 – ΣGτ1), (24)
where τ2 and τ1 designate the calculation of the total amount of
water during the current and previous measurements, and ∆τ is
the interval of time, in years, between measurements.
Thus, both monitoring and laboratory data of humidity of
insulating oils become the important diagnostic components of
oil-filled equipment in service [29].
Conclusions
Our investigations, based on scientific representations of the
states of water in insulating liquids; show the advantage of di-
verse methods for measuring humidity in insulating liquids, and
a variety of analytical expressions for calculating the condition
of solid insulation. Solid insulation is the most vulnerable link
of all the insulations, as it is most susceptible to the degrading
influence of water. We propose the analysis of water content of oil
either by direct or indirect methods. In addition, we also propose
analytical expressions for automated decision making for the tasks
in the operative diagnostics of oil-filled equipment.
The proposed analytical relations allow automating the deci-
sion making of tasks in the operative diagnostics of the oil-filled
equipment for use in a system of monitoring.
Acknowledgment
The authors express deep gratitude to Michel Duval for revis-
ing the English text and for his help in finding suitable terminol-
ogy.
References
[1] V. G. Arakelian and I. Fofana, “Water in oil-filled high-voltage
equipment part I: States, solubility and equilibrium in insulating
materials,” IEEE Elect. Insul. Mag., vol. 23, no. 4, pp. 15–27,
2007.
September/October 2007 — Vol. 23, No. 5 23
[3] F. Ishi and M. Ueba. Trans. IEE Pap. no. 3, part A, pp. 154–158,
1972.
[4] I. Fofana, H. Borsi, E. Gochenbach. “Fundamental investigation on
some transformer liquids under various outdoor conditions,” IEEE
Trans. Dielec. Elect. Insul., , vol..8, no.6, pp. 1040–1047, 2001.
[5] S. Okabe, M. Koto, Y. Fukuda, Y. Hayashi, T. Nara, and K. Taka-
hashi, “A study on the deterioration characteristics of power capaci-
tors with oil-impregnated paper,” Elect. Eng. Jpn., vol. 115, no. 8,
pp. 1–11, 1995.
[6] ASTM D4056-01, Replaced by ASTM D4056-01(2006), Standard
Test Method for Estimation of Solubility of Water in Hydrocarbon
and Aliphatic Ester Lubricants, Aug 2001.
[7] GOST 7822-75 Oils petroleum: A method of determination of the
content of dissolved water (in Russian), 1975.
[8] L. Szepes, K. Torkos, R. Dobo and A.Szekely, “A new analytical
method for the determination of the water content of transformer
oils,” IEEE Trans. Elect. Insul., vol. E1-17, no. 4, pp. 345–349,
1982.
[9] K. Fischer, „Neues Verfahren zur Vassanalitischen Bestimmung
des Wassergehaltes von Flussigkeiten und festen korpern,“ Angew.
Chem., vol. 48, p. 394, 1935.
[10] IEC 60814, “Insulating liquids–Oil-impregnated paper and
pressboard–Determination of water by automatic coulometric Karl
Fischer titration”, Aug 1997.
[11] ASTM D1533, “Standard Test Method for Water in Insulating
Liquids by Coulometric Karl Fischer Titration”, 2005.
[14] H. Borsi, “Gassing behavior of different insulating liquids for
transformers,” Electra vol. 188, pp. 21–41, 2000.
[15] V. G. Arakelian, “The long way to the automatic chromatographic
analysis of gases dissolved in insulating oil,” IEEE Elect. Insul.
Mag., vol. 20, no. 6, pp. 8–25, 2004.
[16] Vaissala Company, “Vaissala HMP228: Moisture and temperature
transmitter for qil,” User’s Guide, M210282en-A, Helsinki, Finland,
May 2002.
[17] Doble Engineering Company, “Domino and IDD® Intelligent
Diagnostic Device for On-Line Moisture in Oil Analysis,” MKT-
SL-IDDMO-09/04, Watertown, MA 02472-4037, U.S.A.
[18] IEC 60733, “Method for determination of water in insulating oils,
and in oil-impregnated paper and pressboard”, 1982.
[21] T. Oshi, S. Tsukao, T. Ishii, S. Itahashi, and M. Kamba, “Properties
of aged oil impregnated paper power capacitor,” in Proc. 7th Inter.
Conf. Properties Appl. Dielect. Mat., vol.1, pp. 246–249, June 1–5,
2003.
[22] A. Seytashmehr, I. Fofana, A. Akbari, H. Borsi and E. Gockenbach,
“Effects on temperature on the dielectric response of transformer,”
in Proc. 15th Int. Symp. High-Voltage Eng., ISH 2007, Ljubljana,
Slovenia, Paper T8-537, Aug. 27–31, 2007.
 [23] W.S. Zaengl, “Application of dielectric spectroscopy in time and 1997 to August 1999) at the Schering Institute of High Voltage
frequency domain for HV power equipment,” IEEE Elec. Insul. Engineering Techniques (University of Hanover, Germany). He
Mag., vol. 19 no. 6, pp. 9–22, Nov./Dec. 2003.
joined the Université du Québec à Chicoutimi (UQAC) as an as-
[24] Y. Du, M. Zahn, B.C. Lesieutre and A.V. Mamishev and S.R. Lind-
gren “Moisture equilibrium in transformer paper-oil systems,” IEEE
sociate researcher in 2000. He is currently associate professor at
Elect. Insul. Mag., vol. 15, no. 1, pp. 11–20, 1999. the UQAC, Quebec, Canada.
[25] J. Aubin and B. D. Sparling, “Moisture in paper assessment from Since 2005, Dr Fofana has been the Chairholder of the Canada
continuous monitoring of moisture in oil,” Weidmann ACTI 4th Research Chair, tier 2, on Insulating Liquids and Mixed Dielec-
Annu. Tech. Conf., San Antonio, TX, November 28, 2005. trics for Electrotechnology (ISOLIME). He has been appointed
[26] M. Koch, S. Tenbohlen and T. Stirl, “Advanced online moisture to the Technical Committee on Energy and Power Systems of
measurements in power transformers,” CMD 2006 Int. Conf. Condi- the International Association of Science and Technology for
tion Monitoring Diagnosis, Changwon, Korea, April 2–5, 2006. Development (IASTED), a member of the IEEE Task Force on
[27] http://en.wikipedia.org/wiki/Thermodynamic_equilibrium#Local_
Atmospheric Icing performance of line insulators, and a member
thermodynamic_equilibrium (June 15th 2007).
[28] R. A Meyers, “Encyclopedia of physical science and technology”,
of the ASTM Committee D27 on Electrical Insulating Liquids and
Orlando Academic Press, 1992. Gases. He authored/co-authored more than 80 scientific publica-
[29] V. G. Arakelian, “Effective diagnostics for the condition of oil-filled tions and has three patents.
equipment,” IEEE Elect. Insul. Mag., vol. 18, no. 6, pp. 26–38,
2002.
Erratum
In the feature article, Water in Oil-Filled High-Voltage Equip-
ment Part I: States, Solubility, and Equilibrium in Insulating
Vadim Gareginovich Arakelian is a Materials, published in the July-August issue of IEEE Electrical
leading research worker of All-Russia Elec- Insulation Magazine, vol. 23, no. 4, pp. 22 and 23, Figures 3 and
trotechnical Institute, Moscow. In 1961 he 5 had errors. The figures should appear as follows:
graduated from Moscow Institute of Fine
Chemical Technology (M. V. Lomonosov).
In 1969 he defended his dissertation in the
field of theory and practice of gas chroma-
tography. When he joined the All-Russia
Electrotechnical Institute in 1969, his scien-
tific interests were reoriented from analytical
chemistry into the area of electrical engineering, especially in the
development of bases of creation of SF6 electrotechnical equip-
ment and in the application of gas chromatography for diagnostics
of damage to oil-filled electrotechnical equipment.
Dr. Arakelian has generated a new branch of knowledge in
electrical engineering, which was the basis of his doctor’s dis-
sertation “Physical Chemistry of Electrotechnical Equipment,”
which he defended in 1995. This was the first dissertation in the
field of physical chemistry of electrotechnical equipment. He Figure 3. Relative water content isotherms of mineral oil.
is the author of more than 100 published works, more than
80 of those in the field of electrical engineering. He is au-
thor of a monograph,
2002 (“Physical
chemistry of SF6 electrotechnical equipment,” Moscow, 2002).
He may be contacted at: arakelva@mtu-net.ru.

Issouf Fofana received an electrome-

chanical engineering degree in 1991 from
the University of Abidjan (Côte d’Ivoire),
a master and Ph.D. degrees, respectively in
1993 and 1996, from the Ecole Centrale de
Lyon, France, where he was a postdoctoral
researcher in 1997. From 1998 to 2000, he
was a researcher (Fellow of the Alexander Figure 5. Relative amount of excess water (versus the total
von Humboldt Stiftung from November amount of water) in mineral oil versus the relative humidity of
air.

24 IEEE Electrical Insulation Magazine

View publication stats