23rd International Conference on Electricity Distribution Lyon, 15-18 June 2015

Paper 0374

TRANSFORMER’S MOISTURE ASSESSMENT WITH ONLINE MONITORING

Senja LEIVO Jukka LEPPÄNEN

Vaisala Oyj – Finland Vaisala Oyj – Finland
senja.leivo@vaisala.com jukka.leppanen@vaisala.com

from ambient air moisture or moisture released from

ABSTRACT sampling system’s materials in contact with oil. Even a
small amount of water contamination can have a
This paper discusses the application of online moisture significant effect on final results.
monitoring of operating loaded transformers. Real-time
monitoring offers many benefits the fleet operator can The sample handling at a laboratory has to be performed
achieve that are discussed below and shown as data with care. However, there is always some moisture
ingress from atmosphere during the sample preparation.
examples from operating transformers. The data
Therefore KF results tend to have a positive bias to
collected in this paper also shows the varying moisture higher moisture content especially with dry oil-samples.
patterns of different transformers, proving them to be Also different reagents may give somewhat different
very individual depending on their loading and results. Round robin tests have shown significant
maintenance history. Water solubility of two aged oil differences in results for identical samples. The results
samples was defined to illustrate the effect of oil ageing. had a bigger variation with dry oil samples [5].

The many uncertainties related to the traditional method The fundamental drawback of the method based on oil
to define paper moisture from oil samples by applying samples is the long time constants to establish the
equilibrium between oil and paper especially at lower
moisture equilibrium charts are discussed on a general
temperatures [6]. This leads to a situation that the
level. samples taken from transformers are often times not
representative of equilibrium, because due to the load and
INTRODUCTION temperature variations, equilibrium is hardly ever
achieved in a loaded transformer (Figure 1).
Power transformers are among the most valuable and
important assets in electrical power networks. Knowing
their condition is essential to meet the targets of MOISTURE DYNAMICS IN TRANSFORMER
maximizing the return on investment and lowering the Temperature fluctuation by load drives moisture
total costs related to a transformer’s lifetime operation. exchange between oil and the solid insulation. At the
temperature increase water is released from paper into oil
Moisture is one dominant factor in solid insulation and vice versa with the temperature decreasing [7,8]. The
degradation. Traditionally moisture assessment of the same moisture pattern is shown in Figure 1.
insulation paper has been conducted from oil samples
applying moisture equilibrium charts. However, this
method includes many factors that may increase the
assessment uncertainty significantly. There are various
error sources related to the sampling and sample
handling, which are discussed later in this paper.

It is also widely known that moisture in a transformer

increases the risk of bubble formation and decreases the
dielectric strength of insulation oil [1,2,3,4].

PAPER MOISTURE FROM OIL-SAMPLE

Moisture assessment of insulation paper of an operating
transformer is commonly conducted from oil samples
applying moisture equilibrium charts. Oil samples are Figure 1 Online monitoring data: oil moisture, top oil
taken from the transformer and sent to a laboratory to be temperature and load from a 30/60 MVA transformer
analyzed for moisture with KF titration (coulometric). (ON/OF).
Such a method is susceptible to various factors that may
affect the results. Desorption is a faster process than absorption, indicated
by hysteresis in online moisture data (Figure 2).
The sampling procedure is crucial for reliable moisture Determining water content in paper is difficult with
results and thus should be performed with great care. traditional sampling due to the hysteresis effect of the
Especially dry insulation oils are prone to contamination
moisture migration between paper and oil. The moisture

CIRED 2015 1/5

 23rd International Conference on Electricity Distribution Lyon, 15-18 June 2015

Paper 0374

ppm results may vary significantly depending on where relative saturation and temperature, when the water
in the loading cycle the sample is taken. solubility characteristic of oil is known (Figure 1).
Structure of capacitive moisture sensor and its
operating principle
A capacitive moisture sensor is basically a parallel plate
capacitor. At least one of electrodes is porous to water
and allows water to permeate inside the dielectric
polymer layer. Absorbed water molecules increase
permittivity and this can be measured as increased
capacitance of the sensor element. The sensor is very
selective to water and almost no interfering effects of
other oil molecules are observed.

Overall absorption forces and the water content of

Figure 2 Online monitoring data from a 30/60 MVA
transformer: oil moisture plotted as function of temperature solution in the equilibrium state are determined by a
Gibb's energy of mixing.
In a wet transformer the variation in oil moisture is Gmix = nii
clearly wider and thus the effect of sampling moment is
even bigger. The data in Figure 3 shows that the moisture and the driving force is chemical potential
content of an oil sample taken at 35°C can vary between
 = o + RT ln a.
30 ppm and 50 ppm depending on the previous loading
history of the transformer before the oil sample was -o = RT ln a.
taken. If those values were used to define paper moisture
using the Oommen’s equilibrium curve the results would ni = Amount of agent i in moles
be 4.3% and 6% respectively. This clearly shows how  = Potential in solution
misleading the results can be if moisture of solid o= Potential of pure water
insulation is defined only from an oil sample which is R = Gas constant
taken during the dynamic condition of a transformer. T = Temperature
a = Activity

This is connected to the mole fraction so that the activity

a = *x, x is a mole portion of the saturation and
coefficient  is a function of pressure, temperature and
concentration and it is specific for each system.

aw = ERH = *(ppm/ppmsat)

if  = 1 then
aw = (ppm/ppmsat) = RS

Figure 3 Online monitoring data from a 10 MVA aw = Water activity

transformer (ONAN): oil moisture (ppm) plotted as function ERH = Equilibrium relative humidity %
of temperature (°C). Variation of moisture at 35°C is ppmsat = Saturation ppm
highlighted with a red arrowed line. RS = Relative moisture saturation %

CAPACITIVE MOISTURE SENSOR The absorption forces at the molecular level are binding
forces between the water molecules and the molecules in
Capacitive sensors are used for online measurement of
oil’s relative moisture saturation (%RS) and temperature. oil.
Water activity is the driving force for moisture migration
between oil and paper. In equilibrium relative saturation WATER SOLUBILITY OF OIL
can be used as a reference for determining paper
moisture. However, there is hardly ever equilibrium due Water solubility of oils is exponentially temperature
to temperature fluctuations and the thermal mass of dependent. The warmer the oil the more water it can hold.
insulation. Also, relative saturation is a temperature This is opposite to the characteristics of solid insulation.
dependent parameter and sensitive to the location of the Figure 4 presents the average water solubility of mineral
sensor. transformer oils i.e. their saturation curve. Solubility is
Water concentration in ppm can be calculated from affected also by additives and aromatic compounds of oil.

CIRED 2015 2/5

 23rd International Conference on Electricity Distribution Lyon, 15-18 June 2015

Paper 0374

However, commercial mineral transformer oils typically

have a small variation of the amount of additives or
aromatic compounds and thus their water solubility is
very similar. It’s important to note that other dielectric
liquids like synthetic esters or natural esters have
significantly higher water solubility than mineral oils.

Figure 5 Water solubility curves (saturation) of aged and

new oils calculated using the coefficients shown in Table 1.

MOISTURE AND DIELECTRIC STRENGTH

OF OIL
Figure 4. Average water solubility of mineral transformer Dielectric strength is one of the most important
oil presented as average curve. Data is from several parameters of the transformer oil. It has been reported
published curves as well as Vaisala own laboratory analysis that the breakdown voltage is affected by several factors,
of commercial oil samples. The variation (-+) between such as moisture, particles, acidity and pressure [9,14,15].
solubility values is shown in the graph. A recent study shows that it is rather the relative moisture
saturation (%RS) of oil than the absolute moisture (ppm)
Aged and contaminated oil has been reported to have that affects the dielectric strength of oil. The results were
typically higher solubility than new or like new oil practically similar for all the tested oils. When the
[9,10,11,12]. Solubility increases as the amount of polar moisture was expressed relative to saturation, the
molecules from chemical reactions of oil and cellulose breakdown voltage remained high at the moisture
ageing processes increase, in practice the chemical saturation that was below 20 %RS but then decreased
composition of oil changes. However, it is possible that significantly with increasing moisture [16].
the saturation level of aged oil is lower than new oil as
shown in Figure 5, which has also solubility curves of ON-LINE MOISTURE MONITORING
aged mineral oils. Tr1 is an old still operating distribution
transformer; Tr2 scrapped, old distribution transformer Online moisture monitoring allows an operator to see the
and Tr3 operating, old industrial transformer. It is moisture migration trend over a long period term as a
assumed that less exponential temperature dependency of transformer-specific pattern.
Tr2 and Tr3 oils is due to their chemical composition. Moisture in oil may have a significant effect on the
Unfortunately, it was not yet possible to make a full loading capacity of a transformer because a high moisture
analysis of those. level (%RS) reduces the dielectric strength of oil [16],
thus the operational loading limit of a transformer with
Table 1 Solubility coefficients (A, B) for new, unused oil high moisture may differ from its nominal limit.
(IEC60422, Vaisala) and three of aged, serviced oils. *[13]

IEC60422* Vaisala Aged oil 1 Aged oil 2

A 7.0895 7.3694 6.835 6.379
B -1567 -1662.7 -1486.4 -1347.4

If the water content in oil increases so that its amount is

equal or bigger than the maximum oil can hold at that
temperature saturation occurs. Oil is also saturated with
water when temperature decreases below the saturation
temperature and the actual water content (ppm) in oil
becomes higher than the maximum at this new lower
temperature.
Figure 6 Monitoring data, 7days, from a 10 MVA
transformer (ONAN): temperature and oil moisture as
%RS and ppm. Sensor located after the cooling radiator.

CIRED 2015 3/5

 23rd International Conference on Electricity Distribution Lyon, 15-18 June 2015

Paper 0374

Knowing the moisture-induced momentary change in the

breakdown voltage could be useful regarding the
transformer operation. By monitoring relative saturation
of oil, an operator can get a warning, if moisture
increases to level where there will be a dramatic drop in
the dielectric strength of oil, which increases the risk of
dielectric faults. This may happen, for example, after
heavy loading of a wet transformer, where increasing
temperature drives water from solid insulation to oil
(Figure 6). When this oil then cools down relative
saturation can peak at a high level because absorption of
water back to paper may take even days. Figure 7. Example of a moisture sensor probe installed in
the oil cooling circulation line (top)
The monitoring provides the operator with real-time data
to detect early signs of faults and correct them. The Moisture assesment of solid insulation
transformer-specific moisture dynamics pattern received
from online moisture monitoring can be used as input for There are publications [9,10] but not any standardized
operational loading guidance. method to derive paper moisture from the online
measured relative moisture saturation of insulation oil.
Effect of sensor location
The moisture sensor should be installed in a location If the relative saturation is rather stable and long
where it is in direct contact with truly representative oil averaging is used, moisture of paper can be roughly
of the transformer. A sensor in the oil cooling circulation estimated from the average temperature and relative
line has proven to be the most beneficial for two reasons. moisture saturation. From the data shown in Figure 1 the
First, the oil is certainly representative and secondly, the average top oil temperature and relative moisture
oil flow improves the sensor response time significantly, saturation values are 32°C and 13.4 %RS respectively.
so that the measurement values are representative and The moisture sorption isotherms for new kraft-paper give
real-time. an estimated paper moisture of ~3.8% [9,10]. Fessler
approximation gives roughly the same paper moisture
The very bottom of the transformer tank should be level. However, the reference temperature used in this
avoided if it is not evident that there is true oil exchange example is not exactly the winding temperature, but top
present. Without the oil flow at the bottom, the sensor oil temperature, thus the derived paper moisture value is
might be measuring just still sludge not the real just a rough estimate. Also the transformer in question is
conditions of the sludge. already over 40 years old, and it is not clear how well the
Relative moisture saturation (%RS) is strongly isotherms of the kraft-paper apply to aged solid
temperature dependent, which must be taken into account insulation.
when choosing the sensor location. If the operator wants
to have the most representative figure on the conditions Oil moisture in ppm as a function of temperature gives a
inside the upper part of the transformer tank e.g. relative better view on the amount of water that is available for
moisture saturation to define paper moisture of the the moisture exchange process between the paper surface
transformer, the sensor location in the top cooling pipe as and oil. This moisture exchange pattern of a transformer
close to the tank as possible gives the most realistic can be better utilized as input in conversion algorithms of
paper moisture [17].
results. Again if the operator wants to avoid high relative
saturation, then locate the sensor after the cooler/radiator. Also usefulness and effectiveness of online dryers can be
evaluated from the amount of water available in oil
One set of moisture and temperature sensors at the top during varying loading conditions.
and another set at the bottom can be used to estimate
dynamically temperature and moisture gradient during Benefits of online monitoring
transformer operation under varying loading conditions. Benefits of online moisture monitoring with capacitive
This may be especially useful for older ONAN sensors:
transformers with significantly varying loading 1. Capacitive sensors measure the relative moisture
conditions. saturation in oil real-time.
2. Measured relative moisture saturation is inherently
If the solubility characteristic of oil in use is well known valid also in other insulation liquids like synthetic or
also momentary saturation temperature can be calculated natural esters.
in the system to avoid water saturation and thus a 3. Oil ageing and its effect on moisture saturation is
dielectric breakdown. It is also possible to utilize winding negligible, because it is intrinsically included in the
temperature if available and model relative moisture initial measurement value.
saturation deviation close by the windings. 4. The sensors are easy to integrate into a larger

CIRED 2015 4/5

 23rd International Conference on Electricity Distribution Lyon, 15-18 June 2015

Paper 0374

monitoring system. transformer paper-oil systems” IEEE Electr. Insul.

5. The monitoring data of moisture and temperature can Mag., Vol. 15, Jan.–Feb. 1999, 11–20.
be used as input for example to transformer ageing [3] A. M. Emsley and G. C. Stevens, "Review of
models, thermal models, bubbling prediction models chemical indicators of degradation of cellulosic
and moisture assessment of solid insulation. electrical paper insulation in oil-filled transformers,"
6. Errors related to oil sampling, transportation and IEEE Proc.-Sci. Technol., Vol. 141, No. 5, Sept.
handling in a laboratory and titration are excluded. 1994, 324-334.
[4] T.V. Oommen and S.R. Lindgren, ”Bubble evolution
CONCLUSIONS from transformer overload”, Transmission and
Distribution Conference and Exposition, 2001
Besides slow moisture ingress processes, such as IEEE/PES, Vol.1, 28 Oct.-2 Nov. 2001, 137 – 142.
cellulose degradation, there may be considerable short- [5] M. Koch, S. Tenbohlen, J. Blennow, I. Hoehlein,
term variation in the moisture of the transformer oil 2007, “Reliability and Improvements of Water
driven by changes in the transformer loading and ambient Titration by the Karl Fischer Technique”, 15th ISH
temperature. 27-31 August 2007 Ljubljana, Slovenia
[6] O. Roizman, V.G. Davydov, 2003, “New procedure
The change in the relative moisture saturation is caused for classification and ranking of operating power
by moisture exchange between the oil and cellulose and transformers on basis of water contamination” EPRI
the temperature-dependency of water solubility. Substation Diagnostic Conference 2003.
[7] Colin Feely, “Transformer Moisture Monitoring and
As the dielectric strength of insulation oil is affected by Dehydration – Powercor Experience”, Proceedings
higher relative moisture saturation, also the dielectric TechCon Asia Pacific 2006
strength of oil may fluctuate during the transformer [8] P.J. Griffin, “How to prevent rain in power
operation. Knowing the moisture-induced momentary transformers”, ASTM Standardization News, Nov
change in the breakdown voltage could be useful 1991
regarding the transformer operation and maintenance. [9] CIGRE working group A2.30, “Moisture equilibrium
and moisture migration within transformer insulation
By monitoring relative moisture saturation of oil, an systems”, CIGRE Brochure 349 2008
operator can get a warning if moisture increases to a level [10] M. Koch, “Improved Determination of Moisture in
where there will be a dramatic drop in the dielectric Oil-Paper-Insulations by Specialised Moisture
strength of oil, which increases the risk of dielectric Equilibrium Charts”, Proceedings of the XIVth
faults. The monitoring provides the operator with real- International Symposium on High Voltage
time data to detect early signs of faults and correct them. Engineering, Beijing, China, 2005 p.508
[11] D. Susa, S. Paasimaa, J. Palola, J. Kalliokoski, K.
The online moisture monitoring gives the operator the Heinonen, “Real-Time Moisture Monitoring Of
true moisture patterns on the transformers of the utility’s Power Transformer”, Conference Proceedings
fleet. This information can be used both for maintenance TechCon Asia Pacific 2006.
as well as operational purposes like applying online [12] T. Gradnik, M. Koncan-Gradnik, N. Petric, N. Muc,
drying and as input to loading guidance. “Experimental evaluation of water content
determination in transformer oil by moisture sensor”,
Moisture and temperature data can also be used to IEEE International Conference on Dieletric Liquids
estimate moisture content and even distribution in the 2011, Trondheim, Norway
solid insulation. The authors do not have an opinion [13] International Standard IEC 60422 Edition 4.0 2013-
whether %RS or moisture in ppm is more suitable for 01
such an assessment, because both parameters have their [14] M. Koch, M. Fischer, S. Tenbohlen, “The breakdown
pros and cons. More studies are needed to get a better voltage of insulation oil under the influences of
understanding. However, moisture values received from a humidity, acidity, particles and pressure”,
longer period of online monitoring are likely to give a International Conference APTADM, 2007, Wroclaw,
more representative figure on the insulation moisture than Poland
just single oil samples taken once a year. [15] P. J. Griffin, “Water in transformers – so what!”,
National Grid Conference on Condition Monitoring
REFERENCES in High Voltage Substations, Dorling, 1995
[1] T. K. Saha, "Review of Modern Diagnostic [16] J. Leppänen, S. Leivo, M. Aronniemi, 2014, ”
Techniques for Assessing Insulation Condition in Transformer oil moisture and dielectric strength”
Aged Transformers", IEEE Trans. on Dielectrics and Proceedings of TechCon® Asia Pacific, 119-131
Electrical Insulation, Vol. 10, No. 5, 2003, 903-917. [17] O. Roizman, “Intelligent Sensor Networks for
[2] Y. Du, M. Zahn, B. C. Lesieutm, A. V. Mamishev, Transformer Health Monitoring”, 2014, Proceedings
and S. R. Lindgren, “Moisture equilibrium in Euro TechCon Conference, Warwickshire, UK

CIRED 2015 5/5