Moisture Measurement and Assessment in Transformer Insulation

741
MOISTURE MEASUREMENT AND ASSESSMENT

IN TRANSFORMER INSULATION –
EVALUATION OF CHEMICAL METHODS
AND MOISTURE CAPACITIVE SENSORS
WORKING GROUP
D1.52
AUGUST 2018
MOISTURE MEASUREMENT AND
ASSESSMENT IN TRANSFORMER
INSULATION – EVALUATION OF CHEMICAL
METHODS AND MOISTURE CAPACITIVE
SENSORS
WG D1.52
Members
I. ATANASOVA-HÖHLEIN, Convenor DE
P. AGREN FI C. BEAUCHEMIN CA
B. ČUCEK SI L. DARIAN RU
V. DAVIDOV AU L. DREIER CH
T. GRADNIK SI M. GRISARU IL
M. KONČAN-GRADNIK SI K. KRYCZYNSKI PL
S. LEIVO FI J. LI CN
K. BREDE LILAND NO Q. LIU UK
M. MARUGAN FR D. MIHAJLOVIC SR
T. PREVOST US P. PRZYBYŁEK PL
P. RAVILA FI O. ROIZMAN AU
A. SHKOLNIK IL K. SIODŁA PL
Corresponding Members
M. DAN AU
J. VAN PETEGHEM BE
B. NOIRHOMME CA
Copyright © 2018
“All rights to this Technical Brochure are retained by CIGRE. It is strictly prohibited to reproduce or provide this publication in
any form or by any means to any third party. Only CIGRE Collective Members companies are allowed to store their copy on
their internal intranet or other company network provided access is restricted to their own employees. No part of this
publication may be reproduced or utilized without permission from CIGRE”.
Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any responsibility, as to the
accuracy or exhaustiveness of the information. All implied warranties and conditions are excluded to the maximum extent
permitted by law”.
WG XX.XXpany network provided access is restricted to their own employees. No part of this publication may be
reproduced or utilized without permission from CIGRE”.
ISBN : 978-2-85873-443-6
Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any
MOISTURE MEASUREMENT AND ASSESSMENT IN TRANSFORMER INSULATION – EVALUATION OF CHEMICAL METHODS AND MOISTURE CAPACITIVE SENSORS
EXECUTIVE SUMMARY
In the last decade, capacitive polymer sensors have been increasingly used to evaluate moisture in power transformers.
Experience shows that effective integration of moisture sensors into on-line diagnostic systems requires proper
placement of the sensor, gathering an adequate measurement data set and evaluation of the measured data by
comparison to historical values in order to obtain a correct moisture profile. Availability of continuous moisture-in-
transformer measurements and its relation to temperature opens up new diagnostic possibilities in comparison to
conventional Karl Fischer spot value measurements. This brochure is a continuation of the work of Cigre Brochure 349
“Moisture equilibrium and moisture migration within transformer insulation systems”, 2008 and allows new insights into
the distribution of moisture between insulating liquids and solid insulation.
Main topics of the brochure are:
- OVERVIEW AND STATE OF THE ART OF MOISTURE MEASUREMENT BY MEANS OF CHEMICAL
METHODS (KARL FISCHER) AND POLYMER CAPACITIVE SENSORS
Chapter 1 deals with an overview of the most important characteristics of polymer capacitive sensors like sensitivity and
accuracy, stability and selectivity, response time, calibration, adjustment as well as the influence of these on the
measurement performance and uncertainty sources.
- CRITICAL REVIEW OF EXISTING STANDARDS FOR THE MEASUREMENT OF MOISTURE
The Working Group performed an extensive Round Robin Test (described in Chapter 2) for water determination by Karl
Fischer titration and by means of capacitive moisture sensors in different types of insulating liquids – new and service
aged mineral oil, synthetic and natural ester and silicone insulating liquid. The methods for the determination of moisture
in the solid insulation have been also evaluated. The results of the Round Robin Test with the corresponding statistical
evaluation allow the estimation of measurement uncertainties, the comparison between different sensors, as well as
recommendations for future standard revisions.
- POSSIBILITIES AND RESTRICTIONS FOR CONVERTIBILITY BETWEEN ABSOLUTE WATER
CONTENT DETERMINED USING KARL FISCHER TITRATION METHOD AND RELATIVE
SATURATION MEASURED USING CAPACITIVE SENSORS
As the conventional mg/kg measurements have been well established and with reference data readily available for
condition monitoring, there is a motivation to convert the new measurement data of RS into mg/kg equivalents and
vice versa. Conversion of moisture measurements between the capacitive sensors and Karl Fischer titration depends on
the moisture saturation curve of the insulation liquid. In a Round Robin Test the possibility and uncertainties of such
conversions for different insulating liquids have been evaluated (Chapter 3).
- MOISTURE DISTRIBUTION BETWEEN LIQUID AND SOLID INSULATION
A short description of the indirect methods used for the determination of moisture in the solid insulation, especially of
the methods based on evaluation of moisture equilibrium curves is done in Chapter 4. The influence of the structure
of solid and liquid insulation and its condition on the uncertainties of the results is elaborated.
- APPLICATION OF THIN FILM POLYMER CAPACITIVE SENSORS FOR ON-LINE MONITORING
OF OIL FILLED ELECTRICAL EQUIPMENT
Moisture determination of the insulating liquid is a routine measurement for transformers and related electrical
equipment used to infer the cellulosic insulation moisture. Experience shows that effective integration of moisture
sensors into on-line diagnostic systems involves the following elements to obtain a proper moisture profile of a
transformer: apropriate placement of the sensor, gathering an adequate measurement dataset and evaluation of the
measured data by comparison to historical values.
It is shown that it is possible to derive evaluation criteria for on-line monitoring of moisture in electrical equipment
based on the correlation between dielectric strength and relative moisture saturation in a dielectric liquid. In addition,
the relative saturation/temperature hysteresis plot is found to be an effective way to indicate the wetness of transformer
insulation based on practical examples in service (Chapter 5).
Availability of continuous moisture-in-transformer measurement by means of capacitive sensors and its relation to
temperature opens up new diagnostic possibilities in comparison to conventional KF spot measurements.
3
4
TABLE OF CONTENTS
EXECUTIVE SUMMARY................................................................................................................................................. 3
1 . IMPORTANCE, INFLUENCE AND MEASUREMENT OF MOISTURE ..................................................... 9

1.1 IMPORTANCE AND INFLUENCE OF MOISTURE ................................................................................................. 9
1.2 KARL FISCHER MEASUREMENTS - PRINCIPLE AND INTERNATIONAL STANDARDS ................................ 14
1.2.1 Karl Fischer titration technique principle ......................................................................................................................... 14
1.2.2 Existing international standards for KF water content measurements in various substrates ................................. 15
1.3 MEASUREMENT OF MOISTURE WITH CAPACITIVE SENSOR ...................................................................... 17
1.3.1 Metal oxide capacitive sensors ......................................................................................................................................... 17
1.3.2 Operating principle of aluminium oxide moisture sensor ............................................................................................ 17
1.3.3 Capacitive thin film polymer sensors ................................................................................................................................ 17
1.3.4 Measurement performance and uncertainty sources of capacitive thin film polymer sensors ............................. 19
1.3.5 Sensitivity and accuracy ...................................................................................................................................................... 19
1.3.6 Stability and selectivity ....................................................................................................................................................... 20
1.3.7 Response time ........................................................................................................................................................................ 21
1.3.8 Calibration of capacitive polymer sensors ..................................................................................................................... 22
1.3.9 Calibration traceability....................................................................................................................................................... 22
1.3.10 Laboratory(factory) calibration ............................................................................................................................... 22
1.3.11 Other possible calibration /verification methods of capacitive polymer sensors ......................................... 22
1.3.12 Two-point calibration ................................................................................................................................................. 22
1.3.13 One-point calibration in the gas phase.................................................................................................................. 23
1.3.14 One-point verification in the oil phase ................................................................................................................... 23
1.3.15 Common mistakes and error sources during humidity calibration .................................................................... 23
1.3.16 Calibration against Karl Fisher titration results?................................................................................................... 23
1.3.17 Adjustment ..................................................................................................................................................................... 24
1.4 REFERENCES TO CHAPTER 1 ............................................................................................................................... 25
2 . RRT ON KARL FISCHER AND CAPACITIVE MOISTURE SENSOR DETERMINATION OF WATER CONTENT 27
2.1 SCOPE OF THE RRT ............................................................................................................................................... 27
2.2 RRT SETUP ................................................................................................................................................................ 27
2.3 RRT PROTOCOL ..................................................................................................................................................... 28
2.4 ROUND ROBIN TEST PARTICIPANT LABORATORIES ..................................................................................... 29
2.5 RRT SAMPLES .......................................................................................................................................................... 29
2.5.1 Water in oil reference materials ...................................................................................................................................... 29
2.5.2 Insulating liquids samples .................................................................................................................................................... 29
2.5.3 New mineral oil sample containing insulating paper ................................................................................................... 30
2.6 RRT CAPACITIVE SENSORS.................................................................................................................................. 30
2.7 EVALUATION OF THE RRT RESULTS ................................................................................................................... 30
2.8 INFLUENCES ON KARL FISCHER MEASUREMENT RESULTS ......................................................................... 30
2.8.1 Quality assurance of participating laboratories ........................................................................................................... 30
2.8.2 Oil sampling influence on KF results ................................................................................................................................. 32
2.8.3 Measurement equipment influence on KF results ........................................................................................................... 33
2.8.4 Influence of oil sample transfer into titration cell on KF results .................................................................................. 33
2.8.5 Influence of sample amount on KF results ........................................................................................................................ 34
2.8.6 Influence of reference materials sample amount on KF results .................................................................................. 35
2.9 RRT KARL FISCHER TEST OF 4 INSULATING LIQUIDS. STATISTICAL EVALUATION ............................... 38
2.10 STATISTICAL EVALUATION OF RRT CAPACITIVE MOISTURE SENSOR TESTS .......................................... 40
2.10.1 Evaluation of sample temperature impact during RRT on capacitive sensor results ..................................... 41
2.11 LEARNINGS FROM RRT ON CAPACITIVE SENSORS...................................................................................... 43
2.11.1 Factory calibration and its verification................................................................................................................... 43
2.11.2 Sensitivity in lower RS region in insulating liquid ................................................................................................. 45
2.11.3 Response time............................................................................................................................................................... 47
2.11.4 Temperature dependence ......................................................................................................................................... 48
5
2.12 WATER CONTENT DETERMINATION OF OIL IMPREGNATED PAPER INSULATION................................ 50

2.12.1 Karl Fischer method IEC 60814 ............................................................................................................................... 50
2.12.2 Calculation of the water content in the solid insulation ....................................................................................... 50
2.12.3 Test results from the round robin on pressboard sample ................................................................................... 51
2.13 CONCLUSIONS FROM THE RRT RESULTS OF KARL FISCHER COULOMETRIC TESTING ACCORDING TO STANDARD
IEC 60814 .............................................................................................................................................................................. 52
2.14 CONCLUSIONS FROM THE RRT RESULTS OF RS% SENSORS TESTING ACCORDING TO RRT PROCEDURE 53
3. CONVERSION OF MOISTURE MEASUREMENTS BETWEEN CAPACITIVE SENSORS AND KARL FISCHER

TITRATION ...................................................................................................................................................... 55
3.1 INTRODUCTION ..................................................................................................................................................... 55
3.2 LITERATURE REVIEW ON MOISTURE SATURATION COEFFICIENTS ........................................................... 55
3.2.1 Mineral oils ............................................................................................................................................................................ 55
3.2.2 Alternative insulating liquids .............................................................................................................................................. 58
3.2.3 Comparison between the liquids ....................................................................................................................................... 60
3.3 METHODS FOR DETERMINING MOISTURE SATURATION COEFFICIENTS ................................................ 61
3.3.1 Method-1................................................................................................................................................................................ 62
3.3.2 Method-2................................................................................................................................................................................ 63
3.3.3 Method-3................................................................................................................................................................................ 64
3.4 MOISTURE SATURATION COEFFICIENTS OBTAINED FROM CIGRE RRT ................................................... 64
3.5 SUMMARY OF CHAPTER 3 .................................................................................................................................. 68
4. INDIRECT METHODS OF MOISTURE DETERMINATION IN THE SOLID INSULATION....................... 71

4.1 METHODS OF DETERMINATION OF WATER CONTENT IN SOLID INSULATION .................................... 71
4.1.1 Background ............................................................................................................................................................................ 71
4.1.2 Methods based on dielectric response ............................................................................................................................ 71
4.1.3 Indirect methods based on the water content measurement in electro-insulating liquid ....................................... 73
4.2 RESTRICTIONS IN THE APPLICATION OF INDIRECT METHODS .................................................................. 74
4.2.1 Kind and chemical structure of solid and liquid insulation ........................................................................................... 74
4.2.2 Ageing condition of liquid insulation ................................................................................................................................ 75
4.2.3 Ageing condition of solid insulation .................................................................................................................................. 76
4.2.4 Moisture equilibrium state .................................................................................................................................................. 76
4.2.5 Distribution of moisture in the solid insulation ................................................................................................................. 77
4.3 CONCLUSIONS TO CHAPTER 4 ......................................................................................................................... 79
5. APPLICATION OF THIN FILM POLYMER CAPACITIVE SENSORS FOR ON-LINE MONITORING OF OIL FILLED
ELECTRICAL EQUIPMENT ............................................................................................................................. 81
5.1 WHERE TO INSTALL A CAPACITIVE SENSOR .................................................................................................. 81
5.2 EXPERIENCE FOR MOISTURE EVALUATION WITH CAPACITIVE SENSORS .............................................. 83
5.2.1 Drying of oil ........................................................................................................................................................................... 83
5.2.2 Knowledge on seasonal moisture variation with temperature.................................................................................... 83
5.3 POSSIBLE EVALUATION CRITERIA FOR MOISTURE IN TERMS OF RS ........................................................ 85
5.3.1 Possible evaluation criteria for moisture in terms of RS ............................................................................................... 86
5.3.2 BDV Interpretation and alert values ................................................................................................................................ 86
5.4 USE OF TEMPERATURE/RS CURVES OF CAPACITIVE SENSORS................................................................. 87
5.4.1 Model experiments in a climate chamber with and without solid insulation. Difference between insulating oils and ester
liquids. Influence of the speed of heating/cooling .................................................................................................................. 87
5.4.2 Identification of a problem with sensor installation ...................................................................................................... 92
5.4.3 RS/Temperature curves from equipment in service ...................................................................................................... 94
5.4.4 Use of capacitive sensors for dew point measurements in transformers ................................................................101
5.5 CONCLUSIONS TO CHAPTER 5 ....................................................................................................................... 106
6
5.6 REFERENCES TO CHAPTER 5 ............................................................................................................................. 106
APPENDIX A. DEFINITIONS OF GENERAL TERMS USED IN THIS TB ([A1] – [A9]) .....................................109
7
8
1 . IMPORTANCE, INFLUENCE AND MEASUREMENT OF

MOISTURE
1.1 IMPORTANCE AND INFLUENCE OF MOISTURE

Moisture in the insulating liquid can be measured directly. This is seldom the case with moisture in the solid insulation.
Usually the moisture in the solid insulation of an aged transformer is not easy determined, since cellulose absorbs
moisture from the surrounding air during paper sampling. Thus, special precautions should be taken if a transformer is
being scrapped and solid insulation samples for moisture are to be taken. As the sampling process will take hours, the
risk of sample contamination is high. This is the reason, why so little evidence of direct Karl Fischer measurements of
moisture in paper are available.
Most evaluations of moisture in the solid insulation are determined through calculations based on the measured moisture
in the insulating liquid, using different methodologies, e.g. water in paper vs water in oil diagrams, water sorption
isotherms or moisture models. Whilst all these methods calculate moisture in paper, they do not take into consideration
that only part of the moisture is exchanged between the insulating liquid and paper.
The cellulosic materials in a transformer have different thicknesses [1.1]. Winding insulating paper is usually in µm-
thickness, pressboard barriers in mm-thickness. Clamping structures consisting mainly of laminated board or laminated
wood and are in cm-thickness. The exchange of moisture between thicker and denser structures and insulating liquid
is much slower.
Figure 1.1 Thick structures of the insulation in a transformer [1.1]

Thick insulation, is typically laminated wood or laminated board clamping elements and moisture is exchanged very
slowly. Thin cold insulation, are e.g. cylinder, angle rings and caps, operating at insulating liquid temperature and most
of the exchangeable moisture is stored in these. Thin hot insulation, is insulating paper, operating at the temperature of
the copper conductors and as such are usually dry.
Some examples of transformer insulation are presented in Figure 1.2.
Laminated pressboard Pressboard cylinders Insulation paper

Figure 1.2 Different kind of cellulosic materials
 The thickness of the insulation and the changes in its temperature strongly influence moisture migration
between the electro insulating liquid and cellulose. An increase in the insulation temperature causes water
migration from paper to oil, whereas a decrease results in migration in the opposite direction. Most of the
moisture in the oil-cellulose insulating system (even up to 99%) is contained in cellulose materials. The water
9
content, both in liquid and solid insulation, influences the life of a transformer in many ways [1.1].Reduction
of dielectric strength
There is a direct dependency between the breakdown voltage and the relative saturation of the insulating liquid – Figure
1.3. This characteristic is based on oil without solid insulation and is similar for different types of insulating liquids [1.2].
The combination of this graph with a graph describing % moisture in paper and % relative humidity (Figure 4.5 in
Chapter 4) would give the dependence of % moisture in paper to the breakdown voltage in oil under equilibrium
conditions – Figure 1.4 [1.3].
80
Breakdown voltage (kV/2,5 mm)
70
60
50
40
30
20
10
0
0 10 20 30 40 50 60 70 80 90 100 110
% Relative saturation (%RS)
Figure 1.3 Dependency between breakdown voltage and water content in insulating liquid [1.2]
7
0°C
6 20°C
40°C
% Moisture in paper (WCP)
5 60°C
100°C
4
0
30 40 50 60 70 80
Breakdown voltage (kV)
Figure 1.4 Dependence between the breakdown voltage and % moisture in cellulosic solid insulation at different
service temperatures [1.3]
From this diagram it is observed that at lower temperatures the breakdown voltage in oil does not significantly change
with the water content of the solid insulation. When evaluating the serviceability of transformers using dielectric strength
based on oil moisture, it is usual that the oil sample temperature > 40°C. From oil taken at lower temperatures, there
is a much higher risk of underestimating the impact on dielectric strength at a higher loads. The sampling temperature
is thus not only important for the evaluation of the water content, but also directly influences the breakdown voltage.
This dependency can be used as a criteria for serviceability and is well known by users. Most breakdowns of moist
transformers happen after switching on after a longer shut down, or during a sudden load reduction after a longer
period of higher load. In the first case a quick diffusion of moisture out of the moist paper causes a significant reduction
in the breakdown voltage at the boundary layers. In the second case warm oil can absorb a considerable amount of
moisture, however the diffusion constants of mineral oil to solid insulation are low (Figure 1.5) [1.4], therefore oil with
higher water saturation and low breakdown voltage can be the result. In such a case the water saturation limit in the
insulating liquid can be exceeded and free water appears (Figure 1.6) [1.5].
10
Figure 1.5 Diffusion time (days) for pressboard with thickness one millimetre; WCP – water content in paper [1.4]
120
100
% Relative Saturation
80
60 %RS with wet

insulation
40 %RS only oil
20
0
0 20 40 60 80
Temperature (°C)
Figure 1.6 An example of relative saturation changes as a result of temperature changes [1.5]
A service example for such an event is the following:
A 35 years old 375 MVA 150/20 kV generator step-up transformer (shell type) has been switched on after a long outage
period. Prior to switching the temperatures were below -15°C for several weeks. The transformer was directly connected
(without circuit breaker) to a generator and the voltage of the generator was slowly increased. A few hours later it
exploded. The most probable reason for the fault is the sudden exchange of water in the boundary surfaces between
solid insulation and oil. This will result in bubbles of free water in the high voltage stress area.
Faults in aged (and probably wet) instrument transformers (regretfully often with explosion and fire) have been reported
to occur in the early morning hours following a previous hot day. In this case the cooling down of the wet oil over night
may have resulted in oversaturation with moisture and initiated the breakdown.
The breakdown voltage, in the case of moist insulation, is very much dependent on the temperature gradient (Figures
1.7 and 1.8) [1.6]. In the case of slow increase/decrease in temperature, diffusion processes prevent moisture
enrichment in oil. When there is a quick increase/decrease in temperature, moisture increases rapidly and breakdown
voltage decreases.
11
Figure 1.7 Temperature and breakdown voltage in oil for different temperature gradients [1.6]
Figure 1.8 Breakdown voltage in oil against temperature gradient during cooling-down phase for initial moisture
content of 3.5% in transformerboard [1.6]
Furthermore, in the case of moist insulation, ignition of partial discharge can occur along with the growth of temperature
in the insulation, and the discharges may not extinguish despite a drop in the temperature (Figure 1.9). The authors of
[1.7] explain this results from phenomena occurring at the interface of the cellulose and oil. This comes about when
temperature increases, migration of water from the cellulose to the oil takes place. A consequence of the increase in
the moisture at the interface of the materials is a decrease in the surface resistivity of the cellulose. This leads to PD
ignition at the surface. A drop in temperature of the insulation does not quickly dry the interfacial area due to the slow
diffuse rate, which favours PDs.
12
Figure 1.9 The experimental results of partial discharges obtained for an insulation sample
with a moisture content of 2.5%; (a) water activity, temperature and moisture content of oil;
(b) PD apparent charge [1.7]
 Effect of the water content on cellulose aging
Aging of the paper is proportional to the water content (WCP). Water decomposes the cellulose structure hydrolytically,
whereby the long cellulose molecule (degree of polymerisation – DP approximately at least 1000) is separated into
smaller sections, as if cut with scissors. At a degree of polymerisation of 200, cellulose has almost no mechanical
strength and the the transformer is no longer short-circuit-proof. Experiments show that the decrease of the degree of
polymerisation is higher if the initial moisture of the paper is higher, i.e. its mechanical properties are reduced
proportionally – see Figure 1.10.
13
Figure 1.10 Cellulose depolymerisation speed dependence on moisture content in insulation paper for different
temperatures [1.8]
 Bubbles evolution
Bubbles evolution results from the release of water vapour from transformer insulation, mainly winding paper. The
bubbles of water vapour decrease the insulating liquid’s dielectric strength. Under the influence of an electric field these
bubbles can lead to partial discharges. Moreover, the bubbles may also cause a pressure increase. Bubble evolution is
enabled under conditions of substantial moisture and high insulation temperature. It is emphasized in several
publications [1.9-1.11] that the higher the water content in cellulose materials the lower the initiation temperature of
bubble evolution. Importantly, the lower the degree of polymerisation (DP) of cellulose the lower the initiation
temperature of this phenomenon (Figure 1.11) [1.12].
Figure 1.11 Dependence of temperature of bubble inception (°C) on the water content for new and aged cellulose
(for different degrees of polymerisation, DP: 1360, 670, 464, 272) [1.12]
1.2 KARL FISCHER MEASUREMENTS - PRINCIPLE AND INTERNATIONAL STANDARDS

1.2.1 Karl Fischer titration technique principle
The determination of water content is one of the most frequently used methods in laboratories around the world. In
addition to methods requiring complex apparatus, such as infrared reflection spectrometry, gas chromatography or
microwave spectroscopy, two methods have become preferred.
Drying methods - drying ovens, infrared lamps and infrared balances. These methods are commonly found in various
standards, but have some disadvantages, such as duration of the analysis and sensitivity to other volatile compounds.
14
Titration methods – Karl Fischer titration is an absolute method in analytical chemistry which determinates both free
and bonded water using volumetric or coulometric titration. Titration basically means to add a reagent of known
concentration (titre) to an unknown substance, until the concentrations are balanced.
Higher (absolute) water contents are preferably determined by volumetric titration, where the added volume is
measured and the water content is calculated by a stoichiometric equation. The sensitivity of volumetric titration is
limited, the lower determination limit is 50 - 100 ppm of water, so that is not applicable for dry insulation oils with
typical values 2 – 30 ppm (IEC 60422).
The reactions occurring during Karl Fischer titration are known to be complex, but are essentially of water with iodine
ion, sulphur dioxide, an organic base (RN) and an alcohol in an organic solvent.
Equation 1.1 illustrates the reactions taking place in a titration vessel [1.13].
H2O + I2 + [RNH]+SO3CH3- + 2RN → [RNH]+SO4CH3- + 2[RNH]+ I-

Equation 1.1
Lower (absolute) water content samples require coulometric titration, which uses an electric current to generate the
reagent. The current releases the stoichiometrically corresponding amount of iodine from the iodine containing Karl
Fischer reagent by electrolysis (according to Faraday’s law), as shown in equation 1.2.
2 I- - 2 e → I2 (2)
Equation 1.2
One mole of iodide reacts with one mole of water stoichiometrically, so that 1 mg of water is equivalent to 10.72
coulombs. Based on this principle it is possible to determine the amount of water directly from the quantity of electricity
(number of coulombs) required for the electrolysis.
The detection limit for coulometric titration is in the μg/g of water, which is convenient for measurement of water
content in insulating transformer oils.
1.2.2 Existing international standards for KF water content measurements in various substrates
Absolute water content in a variety of substrates (lubricating oils, additives, solvents, etc.) with KF titration can be
determined according to different international standards, which are listed in the table 2.3. The majority of the in RRT
participating laboratories determine water content according to IEC 60814 standard. Only one laboratory stated ASTM
D 1533-12 and one GOST 24614-81 standard. Comparison of the individual requirements of standards are presented in
the table below. It is observed that there are no common criteria for the determination of repeatability and
reproducibility even with similar liquids.
15
Table 1.1 International standards for KF water measurements

Standard: Measurement Suggested sample Repeatability (r), 95% Reproducibility
range amount results differ less (R), 95% results
then: differ less then:
IEC 60814
0.60√x
Insulating liquids – Oil
Optimum sample size 1.5√x
impregnated paper and x=average of the
pressboard – Determination of > 2 mg/kg has been found 5 cm3 duplicate x=average of
water by automatic for range 2-100 mg/kg determinations in one means of two labs
coulometric Karl Fischer lab
titration
ASTM D1533-12 No suggestion in
standard only, ’follow
Standard test method for 14 mg/kg in the 0
the manufacturer’s 7 mg/kg in the 0 to 50
Water in Insulating Liquids by 0 to 50 mg/kg to 50 mg/kg
instruction for mg/kg range.
Coulometric Karl Fischer range.
suggested specimen
titration
size’
GOST 24614-81 0,3 for <1 mg/kg
0.0001 to 100% 1 for <5 mg/kg
Coulometric method for water
5-10 g for <10 mg/kg 2 for <10 mg/kg
determination in liquids and (1-106 mg/kg) not specified
10 for <50 mg/kg
gases which do not react with 1-5g for 10-100 mg/kg
20 for <100 mg/kg
the Karl Fischer reagent
100 for <500 mg/kg
200 for <1000 mg/kg
ISO 12937
0.,01874x0,5
1ml-2ml
Petroleum products – 0.003%-0.1% 0.06877x0,5
x=average of the test
Determination of water – for petroleum product
(30-1000 mg/kg) values being
Coulometric Karl Fischer samples
compared
titration method
ASTM D6304-07
Standard Test Method for Volumetric
Determination of water in Volumetric injection: injection:
Petrolium Products, Lubricating
3 cm3 0,08852x0,7 0,5248x0,7
oils, and Additives by Karl
Fischer Titration 10 mg/kg to 2,5% Mass injection: Mass injection:
for expected water
concentrations: 0.03813x0,6 0.4243x0,6
GOST 54281-2010 (10-25000 mg/kg)
Petroleum products, 10-100 mg/kg
x=mean of duplicate x=mean of
lubricating oils and additives.
measurements duplicate
Test method for determination
measurements
of water by coulometric Karl
Fischer titration
ASTM D1364-02
0.03√x
25 ml 0.06√x
Standard test method for <0,5% to 10%
x=average of the
Water in Volatile Solvents (Karl for 0,5% water in x=average of the
(<5000-105 mg/kg) duplicate
Fischer Reagent Titration volatile solvents two means
determinations
method)
Only first three standards in Table 1.1 (IEC 60814, ASTM D1533-12, GOST 24614-81) are suitable for moisture
determination in transformer oils. All others are not suitable as their measurement ranges are too large and have a high
lower limit. These were intended for other materials with typically higher water in oil values.
16
1.3 MEASUREMENT OF MOISTURE WITH CAPACITIVE SENSOR

Capacitive sensors have been used for decades for the determination of moisture in electrical equipment. One
method measures the water vapour pressure in the gas phase and calculates the moisture level in the solid insulation
(provided an equillibrium is established). Here capacitive polymer and oxide sensors are used. In the case of electrical
equipment, usual applications are in factory commissioning tests or on site and also for the case of repair or long term
storage without insulating liquid. Another method assesses moisture in the liquid insulation and uses moisture models
to calculate the moisture content in the solid insulation. This mostly uses capacitive thin film polymer sensors.
As the use of online moisture sensors enables a thurough understanding of moisture dynamics between the solid and
liquid insulation, these sensors are the focus of this brochure.
Special attention is paid to:
- Measurement performance and uncertainty sources

- Calibration
- Optimal sensor location
- Correlation to chemical methods, e.g. Karl Fischer measurements
- Diagnostic guidelines and limit values
1.3.1 Metal oxide capacitive sensors

Metal oxide sensors, including aluminum oxide technology, are designed for very low dew point measurement in
industrial processes. While the types of materials used in construction can vary, the sensor structure and operating
principle generally remain the same. These capacitive sensors are built in a layered structure, sandwiching together a
substrate base layer, a lower electrode, a hygroscopic metal-oxide middle layer, and a water permeable upper electrode.
The capacitance across the upper and lower electrode changes based on the amount of water vapor absorbed by the
metal oxide layer (the dielectric of the capacitor), which is a function of dew point.
While providing excellent low dew point measurement accuracy to -100°C in very dry condititions, they tend to offer
poor long-term stability in processes with varying dew points at the higher ranges (e.g. refrigerant dried systems). Metal
oxide sensors can also be easily damaged by high humidity levels and condensation. This drift in the output reading
means frequent calibration, including adjustment, which can typically be done only by the sensor manufacturer.
The aluminium oxide (Al2O3) sensor provides accurate determination of dew point, ppm, or relative humidity in most
industrial gases. The aluminium oxide sensor is, by function, a capacitor. The electronics are mounted internally and
process the signal from the sensor locally.
1.3.2 Operating principle of aluminium oxide moisture sensor

An aluminium probe is chemically treated to form a thin porous coating of aluminium oxide. The aluminium oxide is
then coated with a thin, permeable layer of gold. The gold layer and the aluminium probe form the the sensor electrodes.
The gold layer is permeable to moisture and conductive. This forms the second electrode of a capacitor. The pores in
the aluminum oxide layer absorb water vapor from the gas stream in amounts directly proportional to the moisture
content of the gas stream. The absorption of water molecules changes the capacitance of the sensor. The capacitance
of the sensor is measured which is then converted to the moisture value.
The aluminium oxide sensor is suitable for use in virtually any application where moisture measurements are required.
However, they tend to offer poor long-term stability in processes with varying dew points at the higher ranges. The
moisture probe is not a mass sensor but responds only to water vapor pressure, therefore it is also pressure dependent
as is dewpoint as well. Requirements for a dew point determination are a quick response from dry to moist and from
moist to dry, with no dependency on the flow rate.
1.3.3 Capacitive thin film polymer sensors

Capacitive polymer sensors are widely used in monitoring of moisture in HV equipment, as well as in automated
equipment for oil processing and refurbishment (reclaiming). The use of capacitive sensor instruments is a mature
technology and has been used since 1970’s in various applications to measure moisture in gas. In late 1990’s the same
technology was introduced to measure relative moisture saturation in oils.
17
Figure 1.12 Structure of a capacitive thin polymer sensor

A capacitive moisture sensor is basically a parallel plate capacitor (Figure 1.12). At least one of the electrodes is
permeable to water vapor and allows water molecules to diffuse into the dielectric polymer layer. Absorbed water
molecules increase the permittivity and this can be measured as increased capacitance of the sensor element. The
sensor is very selective to water, with almost no observed response to other molecules in oil.
Overall absorption forces and the water content of solution in the equilibrium state are determined by a Gibb's energy
of mixing (Equation 1.3).
Gmix = niI
Equation 1.3
and the driving force is chemical potential (Equation 1.4 and Equation 1.5)
 = o + RT ln a
Equation 1.4
-o = RT ln a
Equation 1.5
ni = Amount of agent i in moles

 = Chemical potential in solution
o= Chemical potential of pure water
R = Universal gas constant
T = Temperature in K
a = Water 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 (Equation 1.6).
aw = ERH = *(ppm/ppmsat)
Equation 1.6
if  = 1 then
aw = (ppm/ppmsat) = %RS
aw = Water activity
ERH = Equilibrium relative humidity %
ppmsat = Saturation ppm
RS = Relative moisture saturation %
18
The absorption forces at the molecular level are binding forces between the water molecules and the molecules in oil.
1.3.4 Measurement performance and uncertainty sources of capacitive thin film polymer sensors
How accurate is the measurement? While this question seems simple enough, the answer may not always be. Choosing
the most suitable measurement instrument requires an understanding of the factors that contribute to the uncertainty
of a measurement. This in turn provides an understanding of what is stated in the manufacturer’s specifications.
The performance of a measurement is defined by its dynamics (measurement range, response time), accuracy
(repeatability, precision, and sensitivity), and stability (tolerance for aging and harsh environments). Of these, accuracy
is often considered to be the most important quality; it is also one of the most difficult to specify.
The minimum required information for a thin film polymer sensor is suggested in Table 1.2:
Table 1.2 Minimum requirements for a capacitive thin polymer sensor
Measurement range (moisture)
aw or 0-1
RS (%) 0-100 %
-xx - +xx °C
Measurement range (temperature °C) Shall cover the whole operating temperature
range typical in transformers
± xx or ±% of the reading.
Accuracy (including non-linearity, hysteresis, Manufacturer shall specify the operating
repeatability, temperature dependence) temperature range, where the accuracy
specification applies.
Response time (90%) at 20°C (min) xx minutes.
A capacitive sensor should be accompanied by a calibration certificate, which should show the difference between
observed and reference humidity and temperature in the desired ranges. Additionally, it should be specified whether
the calibration is traceable to international references or not. For an instrument giving readings in moisture ppm, the
formula used to convert relative saturation to ppm, standard coefficients and instructions on how to change coefficients
are needed.
1.3.5 Sensitivity and accuracy

Sensitivity is the relationship between a change in measurement output and a change in reference value. Ideally, this
relationship is linear, but in practice, all measurements involve some imperfections or uncertainty. The agreement
between the measured value and the reference value is often called “accuracy”, but this term can be used somewhat
vaguely by manufacturers.
Specified accuracy usually includes repeatability, which is the capability of the instrument to provide a similar result
when the measurement is repeatedly presented with the same conditions (Figure 1.13). However, it may also include
hysteresis, temperature dependency, non-linearity, and long-term stability. Repeatability alone is often a minor source
of measurement uncertainty, and if the accuracy specification does not include other uncertainties, it may give the
wrong impression of the actual performance.
19
Figure 1.13 Repeatability Figure 1.14 Transfer Function
Deviations from the linear relationship can manifest as non-linearity (Figure 1.15). This effect is often emphasized at
the extremes of the measurement range. It is therefore useful to check if the accuracy specification includes the non-
linearity, and whether the accuracy is specified for the full measurement range. If it is not, this gives reason to doubt
the measurement accuracy near the extremes. For monitoring moisture in transformer oil, sensor performance at the
low end of the moisture measurement range is important i.e. at dry end, a few %RS only.
Figure 1.15 Non-linearity Figure 1.16 Hysteresis
Hysteresis is the change in measurement sensitivity that depends on the direction of the change in the measured
variable (Figure 1.16) e.g. from dry to moist and back to dry, the sensor output may show different characteristics. This
may be a significant cause of measurement uncertainty in the case of some humidity sensors, which are manufactured
from material that bonds strongly to water molecules. If the specified accuracy does not indicate whether hysteresis is
included, this source of measurement uncertainty will be left unspecified. In addition, if the calibration is performed in
only one direction, the effect of hysteresis will not be visible during calibration. Thus, inclusion of hysteresis in a
specification is necessary.
Ambient conditions such as temperature also affect the accuracy of a measurement. If the temperature dependency is
not specified and the operating temperatures change significantly, repeatability may be compromised. The specification
may be given for the full operating temperature, or for a specific, limited, or “typical” operating range. Specifications
expressed in this way leave other temperature ranges unspecified.
1.3.6 Stability and selectivity

The sensitivity of a measurement device may change over time due to aging. In some cases, this effect may be
accelerated by interference from chemicals or other environmental factors. If the long-term stability is not specified, or
20
if the manufacturer is unable to provide recommendations for the typical calibration interval, the specification only
actually indicates the accuracy at the time of factory calibration.
A slow change in sensitivity (sometimes referred to as drift) is harmful, because it may be difficult to observe and it
might affect the long-term monitoring of trends. Selectivity is defined as the instrument’s insensitivity to changes in
factors other than actual moisture. For instance, moisture measurement performed in oil containing certain chemicals
may be affected by the chemicals. This effect may be reversible or irreversible. The response to some chemicals may
be exceedingly slow, and this cross-sensitivity to the chemical can easily be confused with drift. An instrument with
good selectivity is not affected by changes in any factors other than the actual moisture. In transformer applications it
is important that the manufacturer of the sensor confirms that the sensor materials can tolerate oil, typical additives
and aging products.
1.3.7 Response time

Response time is typically defined in manufacturers’ datasheets as T60 or T90, which means the time it takes the sensor
to reach 60% or 90% of the final value. However, when measuring in oil, the oil properties and movement have a clear
effect on the response time. If there is no oil flow i.e. no oil exchange on the sensor surface, the response time becomes
strongly dependent on water diffusion in oil. This is a very slow process and obscures the performance of the sensor.
To determine moisture sensor response time, only a simple setup is required. Since relative oil moisture (%RS) is a
strongly temperature dependent parameter, it is important to keep a stable temperature, where all equipment used are
in thermal equilibrium. Temperature differences impact the time to reach thermal equilibrium due to variations in
thermal mass and can become the dominant factor, concealing the true moisture response time.
In such a test, one needs to have oil samples at least at two different moisture levels (dry and moist), but avoid samples
with free water i.e. with free water at the bottom of the vessel. Each sample must be in equilibrium regarding moisture.
Thus it is recommended that samples be prepared at least the previous day and then kept at a stable temperature. The
samples must be tightly sealed to avoid any moisture exchange with ambient air. In addition, the sample container
materials must be hydrophobic, preferably glass or metal.
Figure 1.17 Schematic response time curve of a moisture sensor when moved from dry oil to moist (1) and back to
dry (2)
When performing the actual test, the sensor is immersed in moving dry oil. Oil flow can be generated by using a
magnetic stirrer. Again, careful sealing is required to avoid moisture contamination with ambient air. The output reading
of the sensor is monitored and recorded when the reading stabilizes. Then the sensor is removed from the dry oil and
and immersed into the moist oil. The output reading of the sensor is again monitored and recorded when the reading
stabilizes. Finally, the sensor is immersed back in dry oil and the readings recorded as before. To minimize the effect of
sensor probe introduced moisture or cross contamination of the two samples during the test, the oil volume should be
at least 200 ml. If absolute values of the sensor readings are of interest, then calibrated references must be available
in both oil samples, which increases complexity.
21
The response-time can be calculated from the recorded measurement data. Typically, response time from dry to moist
is clearly shorter than from a moist to dry condition. In the example shown in Figure 1.17, T90 (dry-moist) is the time
between the point (1) and time when 90% of the response is reached (blue). Respectively T90 (moist-dry) is the time
between the point (2) and time when 90% response is reached (red).
1.3.8 Calibration of capacitive polymer sensors

Measurement devices should be calibrated regularly to maintain their long-term stability. Always refer to the
manufacturer’s instructions and recommendations in relation to calibration intervals.
1.3.9 Calibration traceability

The working standard is the first element in the traceability chain, which contains a series of calibrations and references
up to the primary standard. Whereas many instruments calibrated against a certain reference may be accurate in
relation to each other (high precision), the absolute accuracy with regards to the primary standard cannot be verified if
the calibration uncertainty is not specified.
Traceability of calibration means that the chain of measurements, references, and related uncertainties up to the primary
standard is known and documented. This allows calculation of the uncertainty of the calibration reference and
determination of the instrument’s overall accuracy.
1.3.10 Laboratory(factory) calibration

Laboratory (or factory) calibration is the most accurate method for calibrating equipment. Laboratory calibration offers
lower uncertainties than onsite calibration. As the environmental effects are minimized and the number of factors
influencing the calibration is reduced significantly. If a sensor is calibrated with a traceable certificate it does not need
to be calibrated to each specific insulating fluid used, because relative moisture saturation (%RS) is a parameter
independent to fluid type. It must, however, be ensured that the sensor materials are compatible with the fluid, which
should be confirmed by the manufacturer.
Onsite verification (not a proper calibration) is a quick and easy way of checking measurement equipment without
having to remove it from the site where it is normally used, remembering that proper stabilization times are allowed for
both the measurement device and the working standard. Uncertainty of a calibration performed onsite may be very
difficult to define.
1.3.11 Other possible calibration /verification methods of capacitive polymer sensors

As these sensors are measuring relative humidity, then traceable relative humidity standards should be able to be used
as calibration references. Since relative humidity (%RH) is a temperature dependent parameter, it is extremely
important that the whole calibration system is in temperature equilibrium, as even small temperature differences
between the sensor and the reference can cause significant errors (Table 1.3).
Table 1.3 Example of the effect of 2 °C temperature difference between the calibration reference and the sensor at
different moisture levels at 20°C.
Reference humidity at 20 °C Sensor reading at 22 °C %RH difference due to

temperature difference
10 %RH 8.8 %RH 1.2 %RH

20 %RH 17.7 %RH 2.3 %RH
30 %RH 26.5 %RH 3.5 %RH
1.3.12 Two-point calibration

Sensors tend to have good linearity so a two-point calibration can suffice. This requires an ability to generate two
different humidity conditions with at least 30%RH difference. This can be achieved e.g. using saturated salt solutions
and/or dry gas (air, nitrogen) (Table 1.4). It must be possible to remove the sensor from the oil system and that it can
22
operate as a stand-alone instrument. To have an optimal response time and to avoid any contamination of the calibration
equipment, the probe head and sensor should be cleaned before calibration. To prevent sensor damage, follow
manufacturers’ recommendations for cleaning agents/solvents.
The dry end calibration can be achieved using nearly 0 %RH with properly dried silica and dry air or nitrogen. The
higher humidity point can be achieved using saturated salt solutions, for example 33%RH (MgCl) or 75%RH (NaCl).
Table 1.4 The following table gives corresponding relative humidity (%RH) values at ambient pressure and
different temperatures if a dry gas with known moisture content is used as calibration reference.
Temperature Gas moisture Gas moisture Gas moisture

(°C) Frostpoint / ppmV Frostpoint / ppmV Frostpoint / ppmV
-20°C / 1025 ppm -40°C / 190 ppm -60°C / 20 ppm
15 6.1 %RH 0.75 %RH 0.064 %RH
20 4.2 %RH 0.55 %RH 0.047 %RH
25 3.3 %RH 0.41 %RH 0.035 %RH
30 2.4 %RH 0.30 %RH 0.026 %RH
35 1.8 %RH 0.23 %RH 0.019 %RH
Confirm any unit specific instructions on calibration conditions from the sensor manufacturer, as well as any adjustment
options if significant deviations are found.
1.3.13 One-point calibration in the gas phase

One-point calibration in the gas phase can also be done at ~0 %RH with dry air/nitrogen (see Table 1.4). If adjustment
is needed, offset correction would be more beneficial than gain adjustment to ensure accuracy in dry conditions, though
gain drift is more typical in the sensor.
1.3.14 One-point verification in the oil phase

Calibration is performed with another (calibrated) sensor simultaneously under the same oil conditions. Adjustments
should be avoided / carefully considered, because it might be impossible to maintain the conditions stable enough over
the whole calibration/adjustment process. Comparisons of the readings of those two instruments should be done in aw
(%ERH, %RS), not in ppm unless the same calculation model and the same solubility coefficients are used for both
instruments.
The sensor only measures the state of oil at its surface. Hence good oil flow is crucial to ensure that both sensors
measure the same oil, so as to avoid water diffusion effects (see Chapter 5, Figure 53). For the same reason, it must
be confirmed that there is good oil exchange through any protective filter around the sensor.
1.3.15 Common mistakes and error sources during humidity calibration

 Thermal equilibrium not reached after inserting the sensor into the measurement environment.
 Stabilization time is not sufficient after changing the humidity level of the reference in two-point calibration.
 Hysteresis not considered (some sensors may have significant hysteresis behavior).
 There exists temperature difference between the reference sensor (working standard) and the sensor under
calibration.
1.3.16 Calibration against Karl Fisher titration results?

It is only recommended to adjust the readings of capacitive sensors using references which can produce the same
primary parameter being measured by the sensor (i.e. %RH). Such devices typically also have as an output parameter
absolute water content in oil (ppm). However, that parameter is calculated based on predefined and programmed oil
solubility. If one is trying to make an offset correction for the ppm reading of the device (or a system) at one condition
23
based on KF result, it may increase the errors significantly at other conditions over the measurement range, especially
at the dry-end of the range.
See also chapter 3.
1.3.17 Adjustment
If the measurement readings deviate from the reference, the readings of the instrument can be corrected. This is
referred to as adjustment. Adjustments performed at one point are often referred to as offset correction; two-point
adjustment is a linear correction for both offset and gain (sensitivity) (Figure 1.18). Refer to the sensor manufacturer’s
instructions when offset correction and gain corrections are applied in one or two point adjustment; these may vary
depending on the manufacturer.
For transformer oil monitoring, the most important range is the dry end i.e. < 20%RH (<20 %RS). Special attention
should be paid when making offset correction at higher moisture levels, because the same correction applies to the
whole range, the adjustment should be made in as dry condition as possible.
a)
b)
c)
Figure 1.18: Principles for one-point adjustment with offset correction (a) and gain correction (b) and two-point
adjustment with both offset and gain correction (c) of a measurement device
24
Once the instrument has been adjusted, it should be calibrated once more to verify its performance after adjustment.
In metrology, the term calibration is often confused with adjustment, and only means comparing the measured value
with a known reference, called a working standard.
1.4 REFERENCES TO CHAPTER 1

[1.1] Cigre Brochure 349 “Moisture equilibrium and moisture migration within Transformer insulation
systems”, 2008.
[1.2] I. Fofana, V. Wasserberg V., H. Borsi and E. Gockenbach, “Challenge of mixed insulating liquids
for use in high voltage transformers, Part 1: Investigations of mixed liquids”, IEEE Electrical
insulation magazine 18(2002), Nr. 3, 18 – 25.
[1.3] I. Atanasova-Hoehlein, “Determining the Water Content in Transformers Possibilities and

Application Limits of Moisture Distribution Curves”, VGB Power Tech 5, 2005, 103 – 106.
[1.4] 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
[1.5] Atanasova-Hoehlein I, Contribution to Cigre D1.52.
[1.6] B. Bauerschaper B., O. Kleboth-Lugova, T. Leibfried, “The electrical strength of transformer oil
in a transformer board-oil system during moisture non-equilibrium”, Conference on Electrical
Insulation and Dielectric Phenomena Albuquerque, New Mexico, 2004
[1.7] W. Sikorski, K. Walczak, P. Przybylek, “Moisture migration in oil-paper insulation system in

relation to online partial discharges monitoring of power transformers”, Energies, Vol. 9, Issue 12,
pp. 1-16, 2016.
[1.8] P. Koestinger ., E. Aronsen, P. Boss, G. Rindlisbacher, Practical experience with the drying of
power transformers in the field applying the LFH technology, Cigre Session Paris, Paper A2-205,
2004.
[1.9] T.V. Oommen, S.R. Lindgren, “Bubble evolution from transformer overload”, in Proc. of:
Transmission and Distribution Conference Exposition, IEEE/PES, No. 1’(2001), pp. 137-142,
2001.
[1.10] P. Przybylek, “The influence of cellulose insulation aging degree on its water sorption properties
and bubble evolution”, IEEE Transactions on Dielectrics and Electrical Insulation, Vol. 17, No. 3,
pp. 906-912, 2010.
[1.11] M. Sakai, K. Maki, Y. Taniguchi, N. Daikuhara and S. Yanabu, “A research of critical temperature
on bubble formation under overloaded oil-immersed transformer”, Trans. Inst. Electr. Eng., No.
(121-B)2001, pp. 500–506. 2001.
[1.12] P. Przybylek, “A comparison of bubble evolution temperature in aramid and cellulose paper”, in
Proc. of 2013 IEEE International Conference on Solid Dielectrics, Bologna, Italy, pp. 983-986,
2013.
[1.13] M. Koch, S. Tenbohlen, J. Blennow, I. Hoehlein; Reliability and Improvements of Water Titration
by the Karl Fischer Technique
[1.14] Metrohm Monograph - P.Bruttel, R. Schlink; Water Determination by Karl Fischer Titration
25
26
2 . RRT ON KARL FISCHER AND CAPACITIVE MOISTURE

SENSOR DETERMINATION OF WATER CONTENT
Cigre D1.52 focused on performing a Round Robin Test (RRT) on Karl Fischer measurements of insulating liquid and
solid insulation, as well as on capacitive moisture sensors in order to evaluate the factors influencing these
measurements and to determine measurement uncertainties.
2.1 SCOPE OF THE RRT
1. To evaluate influences on Karl Fischer measurements in respect to:
 quality assurance
 sampling and sampling transfer
 measurement equipment
 sample amount
2. To analyze repeatability (r) and reproducibility (R) of Karl Fischer measurements of water content in insulating liquids
3. To analyse repeatability (r) and reproducibility (R) of capacitive sensors measurements of water content in insulating
liquids
4. To analyse repeatability (r) and reproducibility (R) of measurements of water content in pressboard.
5. User experience with RRT
2.2 RRT SETUP

Laboratories received a questionnaire about their daily work routines:
 Questionnaire
 test method used in laboratory
 laboratory accreditation acc. to ISO/IEC 17025
 measuring equipment
• type of titrator
• titration techniques used
• reagents used
• type of electrodes used
• permissible drift
 transformer oil sample transfer
• how they transfer the sample into the titration vessel
• how many times they flush the syringe before they inject the sample into the vessel
• sample amount and its accuracy of weighing
 verification and control of test method used
• use of standard reference materials
• if they use control charts
• relative standard deviation (RSD)
• repeatability (r)
• reproducibility (R)
• measurement uncertainty (M.U.)
• if they participate in Interlaboratory proficiency tests and how well they perform
Together with samples of insulating fluids, certified reference materials and capacitive sensors, they received detailed
instructions on how to perform RRT:
 Instruction procedure for RRT

 RRT parts:
• Karl Fisher measurement of reference materials and insulating liquids
• Measurement with capacitive sensors in insulating fluids
• Karl Fisher measurement of moisture in solid insulation
27
• Voluntary test – capacitive measurement and Karl Fisher measurement of oil with solid insulation
at different temperatures
 Sequence of the measurements
• Karl Fisher measurements of water in insulating fluids (coulometric, volumetric)
• Capacitive sensor measurements in insulating fluids
• Karl Fisher measurement of moisture in solid insulation (stripping procedure or methanol extraction)
 Laboratory humidity and temperature conditions
 Stirring settings during the capacitive sensor measurements
 Rinsing instructions for capacitive sensors before measurements
 Exact timing for recording the capacitive sensor readings.
2.3 RRT PROTOCOL

Procedure of the RRT is indicated in Figure 2.1.
RRT constituents
Questionnaire
Capacitive sensors
Certified Reference
water in oil materials
Water content in oil Test sample NMO % relative saturation

measurements by Karl and temperature
(new mineral oil)
Fischer measurements by
capacitive sensors and
WCL calculations
Test sample AMO
(aged mineral oil)
Test sample SE
(synthetic ester)
Water content in Test sample NE

pressboard (natural ester)
measurements by Karl
Fischer or methanol
extraction Test sample NMO + P
(new mineral oil with
paper)
Figure 2.1 CIGRE D1.52 RRT protocol
28
2.4 Round Robin Test Participant laboratories

21 participants from 16 countries participated in RRT, 19 of them reported their results (see Table 2.1).
Table 2.1 RRT participant laboratories list
Laboratory: State:
ABB AS Laboratory Norway
ABB Oy Service Finland
ABB Sp. z o.o. Poland
Alstom Grid | PTR - TICC | R&D France
CJSC «Technical Inspection UES» Russia
EFACEC ENERGIA, S.A. Chemical Laboratory Portugal
Electric Power Technology SINTEF Energy Research AS Norway
Milan Vidmar Electric Power Research Institute Slovenia
Institut Nikola Tesla Serbia
KONČAR-Electrical Engineering Institute Croatia
Labelec - Estudos, Desenvolvimento e Actividades Laboratoriais, S.A. Portugal
Laborelec GDF Suez Belgium
Nynas AB Sweden
Poznan University of Technology, Institute of Electrical Power Engineering Poland
Sea Marconi Technologies Italy
Siemens TR Germany
TERNA RETE ITALIA Italy
The Israel Electric Corp. Ltd. Israel
TJ|H2b Analytical Services USA
Vaisala Oyj Finland
Weidmann Electrical Technology AG Switzerland
2.5 RRT samples
2.5.1 Water in oil reference materials

Two certified reference water in oil materials were added:
 Hydranal water standard oil no. 34694 with 6.0 ± 0.8 mg/kg water in oil content, obtained by direct coulometric
determination
 NIST reference material 8506a with 18.3 ± 1.9 mg/kg water in oil content, obtained by direct coulometric
determination and 21.2 ± 1.7 mg/kg water in oil content obtained by interlaboratory determination
with a view to evaluate results of water in oil determined using Karl Fischer for each participating laboratory.
2.5.2 Insulating liquids samples

Four insulating liquid samples were distributed to RRT participants in sealed 150 ml (for Karl Fischer measurements)
and 300 ml (for capacitive sensor measurements) aluminum bottles:
 NMO - new mineral oil
 AGO - aged mineral oil
29
 SE - synthetic ester
 NE - natural ester
All samples were stored in a climate chamber with 25 % relative humidity for 14 days at 20 °C.
2.5.3 New mineral oil sample containing insulating paper
One sample for water in paper RRT Karl Fischer determination with oven or methanol extraction was prepared. Water
in oil content was first determined by Karl Fischer titration. Volunteer Tests with capacitive sensors at different
temperatures were carried out to observe the influence of humidity migration from the solid insulation to oil, at different
temperatures and its detection by capacitive sensors and Karl Fisher measurements.
2.6 RRT Capacitive sensors
Each of the RRT participants received three capacitive moisture sensors with instruction manual and manufacturer
specifications (Table 2.2):
Table 2.2 Manufacturer specifications of the RRT capacitive sensors

Sensor 1 Sensor 2 Sensor 3
Measurement Range Rel. saturation 0 % … (100 %) 0 % … 100 % 0 % … 100 %
Temperature -40 ... +180 °C -40 to +150 °C -20...+100 °C
Accuracy Rel. saturation ± 2 % (0-90 % RS) < ± 2 % (5-95 % RS) <±2%
Temperature ± 0.2 °C ± 0.2 °C ± 3 °C
2.7 Evaluation of the RRT results

Statistical manipulation of KF measurements was carried out by applying the "Q test". Outlier data was identified and
removed from further statistical evaluation.
After the “Q test”, the following values were calculated:
 Xaverage average RRT value
 Stdev RRT standard deviation
 RSD RRT relative standard deviation
 Rcalc = 2.8 ∙ RRT Stdev calculated reproducibility from ISO 13521 (95% confidence interval)
 RIEC60814 = 1.5 √RRT Xaverage reproducibility from IEC 60814 (95% confidence interval)
2.8 Influences on Karl Fischer measurement results

Investigation on influences on Karl Fischer measurement results was performed by evaluation of returned questionnaires
from 19 RRT participating laboratories.
Additionally, the influence on KF results relating to transformer oil sampling procedures was evaluated by one accredited
laboratory.
The influence of sample size on KF results was evaluated by comparison of RRT performed on different oil sample sizes
by some laboratories.
2.8.1 Quality assurance of participating laboratories

7 of 19 in the RRT participating laboratories were accredited to ISO/IEC 17025 for 2 – 15 years.
Laboratories indicated different test methods for the determination of moisture in oil and paper insulation. 14 of them
stated IEC 60814, 1 laboratory ASTM D1533-12 and 1 GOST 24614-81 as their standard test method.
16 laboratories used a variety of mechanisms for insuring quality, including use of reference standard materials (RM)
from different producers (Merck, Riedel de Haën, Fluka, Atlantis standard, NIST, Sigma-Aldrich) with a very wide range
of concentrations (Table 2.3).
30
Table 2.3 Concentrations of reference materials used by laboratories
WCL of RM used in
Lab. No.
mg/kg (ppm)
Lab. 1 100
Lab. 2 15, 30
Lab. 3 10
Lab. 4 15, 30
Lab. 5 6, 20
Lab. 7 6, 100
Lab. 8 4, 8, 1000, 10000
Lab. 9 6
Lab. 10 -
Lab. 11 7, 20, 300
Lab. 12 6, 1000
Lab. 13 10, 1000
Lab. 14 -
Lab. 16 5
Lab. 17 6
Lab. 18 5, 20
Lab. 19 100
The frequency of use of standard reference materials between laboratories also varies widely (Figure 2.2).
Frequency
of use in Frequency of use of standard reference
Frequency of days per materials per year
Lab. No. use year
Lab. 1 daily 250 Lab. 14
Lab. 2 1x/week 52 Lab. 10
Lab. 3 1x/year 1 Lab. 8
Lab. 4 1x/3months 4 Lab. 19
Lab. 5 2x/2weeks 52 Lab. 3
Lab. 7 1x/month 12 Lab. 16
Lab. 13
after replacing ?
Lab. 12
Lab. 8 reagents
Lab. 9
Lab. 9 1x/3 months 4 Lab. 4
Lab. 10 - - Lab. 11
Lab. 11 1x/month 12 Lab. 7
Lab. 5
Lab. 12 1x/6 months 2
Lab. 2
Lab. 13 1x/6 months 2 Lab. 18
Lab. 14 - - Lab. 1
Lab. 17
Lab. 16 1x/6 months 2
Lab. 17 2x/day 500 0 50 100 150 200 250 300 350 400 450 500
Lab. 18 1x/day 250 n determinations of RM per year
Lab. 19 1x/year 1
Figure 2.2 Frequency of use of standard reference materials per year
31
Participating laboratories ensured quality in different ways, but only accredited laboratories used control charts and
have calculated their repeatability, relative standard deviation (RSD) and measurement uncertainty (Table 2.4).
Table 2.4 Repeatability, RSD and measurement uncertainty of accredited laboratories
RSD
labs repeatibility labs measurement uncertainty
Accredited labs from Control Chart
3% ≤ 5% ± 4,7%
Lab. 2
Lab. 3 2,6% - min 0,5 ppm or 4,5%

for RM 15 3%
Lab. 4 3,5% for RM 29 1,6% ± 12%
<10 mg/kg ±15%+0,5mg/kg
Lab. 5 7,3% 10% ≥10 mg/kg ±15%
Lab. 11 - 0,6.√average 8,5%

at 5 mg/kg: 15% < 5 ppm: 30%
Lab. 18 at 20 mg/kg: 7% - > 5 ppm: 15%
Lab. 19 - - 3,9%
The RSD values from control charts in accredited labs are in range from 3% to 7%, for very low WCL values 15%. As
any RSD value depends on WCL concentration range, this may account for the observed spread of daily chart RSD
values.
The reported measurement uncertainties of the 7 accredited labs are from 4% to 15%. For WCL less then 10 ppm,
30% measurement uncertainty is reported.
2.8.2 Oil sampling influence on KF results
One of the RRT participants, accredited to ISO/IEC 17025, carried out parrallel sampling to IEC 60475 of twelve
operating transformers (110 kV) with different water in oil contents, into 250 ml glass syringes and 1 liter glass bottles.
Water in oil content was measured to IEC 60814, with approximately 500 mg sample amounts (Table 2.5).
Table 2.5 Statistics of bottle and syringe comparative sampling
WCL S WCL B difference MU
TR
[mg/kg] [mg/kg] WCL S - WCL B [mg/kg]
no.
(Syringe) (Bottle) [mg/kg]
1 6 5 1 1.3
2 6 5 1 1.3
3 7 6 1 1.5
4 7 7 0 1.6
5 9 9 0 1.9
6 9 8 1 1.8
7 9 9 0 1.9
8 11 10 1 1.6
9 11 9 2 1.5
10 12 10 2 1.7
11 12 10 2 1.7
12 14 13 1 2.0
Laboratory expanded MU measurement uncertainty (k=2) is:

MU for WCL ≥ 10 mg/kg: ± 15%
MU for WCL < 10 mg/kg: ± (15% + 0.5 mg/kg)
32
Average difference in KF water content in oil between syringe and bottle is in all cases lower or comparable with
laboratory MU. The differences do not exceed significantly the laboratory MU, which shows that the impact of sampling
is not substantial.
2.8.3 Measurement equipment influence on KF results
Laboratories reported on the maker and type of instruments they use; Methrom, Mettler, SI Analytics Titroline, Denver
Instruments 3
Producer of Karl Fischer instruments used
5% (1 lab)
5% (1 lab) D
C
17% (3 labs)
B
72% (13 labs)

A
Figure 2.3 Statistics of Karl Fischer instrument producer used
Titration techniques used: 12 labs use coulometric determination with direct injection, 5 laboratories reported they have
both coulometer with direct injection, and coulometric – heating with oven determination.
Various configurations of titration cells use different type of electrodes (generator electrode with or without diaphragm,
etc.) and suitable titration reagents. The multiplicity of variables made it difficult to reach a conclusion on what elements
were affecting the accuracy of results.
All laboratories, except one, use non-pyridine reagents from different manufacturers (GS Chemicals, Sigma-Aldrich,
Fluka, Merck).
In case of water determination by coulometer heating oven, air or nitrogen are used, at temperatures 150°C or 180°C.
2.8.4 Influence of oil sample transfer into titration cell on KF results
Laboratories normally obtained oil samples in both syringes and bottles. About a third of them use only bottles, and a
third use syringes, and the other third use both. As they all received the same samples, the sampling method was not
part of the analysis.
The laboratories transferred the oil sample from the sampling vessel into the titration cell mostly by drawing it with a
syringe.
The participating laboratories discharged the samples as set out in Figure 2.4.
33
Flushing the syringe and needle before measurements
6%, directly from syringe
12%, new syringe for each

sample
29%, three times
24% once
29%, twice
Figure 2.4 Statistics of flushing the needle before measurements – daily practice
There were three groups of labs observed regarding the flushing frequency of syringes and needles before KF
measurements. As only 4-5 labs are present in each group, any statistical evaluation of syringe flushing effect is
impossible.
2.8.5 Influence of sample amount on KF results
Standard IEC 60814 states that the optimum sample size has been found to be 5 cm 3 for most types of insulating
liquids, new and used, with water contents of between 2 mg/kg and 100 mg/kg.
Variations in injected sample and the accuracy of weighing the sample (suggested ± 0.1 g), are reported in Figure 2.5.
Daily practice of 19 laboratories

Injected sam ple am ount and its accuracy of weighing
9 0,01 Sample amount (g) Accuracy (g)

8
7
6 0,0001
5 0,005
0,1 8,5
4 0,0001 0,0001
3 0,01
5 0,0001 0,01 0,1
2 4 0,0001 0,2 3,5 0,0001 3,5 0,0001 4,35
1 0,0001 2 0,001 2,15 1,75 2
1,5 0,0001 1 1 1,5
0 0,5 0,6
LAB. LAB. LAB. LAB. LAB. LAB. LAB. LAB. LAB. LAB. LAB. LAB. LAB. LAB. LAB. LAB. LAB.
1 2 3 4 5 7 8 9 10 11 12 13 14 16 17 18 19
Daily practice Weighing

Sample amount Accuracy
Lab. 1 3-5 g 0.1 g
Lab. 2 1-2 g 0.1 mg
10g (<10 mg/kg),
7g (10-50 mg/kg), 0.01 g (oil)
Lab. 3 4g (>50 mg/kg) 0.1mg (pap.)
34
5g(<100 mg/kg),
Lab. 4 1g(>100 mg/kg) 0.1 mg
Lab. 5 ~ 0.5 g 0.1 mg
Lab. 7 2 0.1 mg
depending on the
Lab. 8 expected WCL 0.1 mg
Lab. 9 1ml 0.2 ml
Lab. 10 3-4 g 0.1 mg
Lab. 11 0.6 g 1 mg
Lab. 12 2.1-2.2 g 0.01 g
Lab. 13 1.000 g 0.1 mg
Lab. 14 3-4 g 0.1 mg
Lab. 16 1.5-2 g 0.01 g
Lab. 17 2g 0.1 g
Lab. 18 1-2 g 0.1 mg
Lab. 19 4.27-4.43 g ± 0.005 g
Figure 2.5 Routine practice of 19 laboratories - injected sample amount and the accuracy of weighing
2.8.6 Influence of reference materials sample amount on KF results
Two series of KF measurements on two reference materials with low and medium values were performed.
The first oil samples were tested using routine laboratory practice (Figure 2.5). A second run used a smaller sample <
1 gram, enabling evaluation of the influence of sample size on KF results.
2.8.6.1 Low water content value - tested material HYDRANAL Nr. 34694 (Reference material value = 6.0 ± 0.8
mg/kg)
15 laboratories participated in 57 measurements in the first and second RRT. All the measurements from both RRT
were statistically evaluated. A comparison was made between the reproducibility as declared by the relevant standard
and the reproducibility as found for the group of RRT participating laboratories. The average result, Stdev, RSD,
calculated reproducibility (Rcalc) and reproducibility (RIEC 60814) from IEC are stated in next table. Finally, RRT accuracies
were calculated as a percent deviation from the certified reference material value (Figure 2.6).
KF results of Reference material Hydranal Nr.34694 All measurements from 1st and 2nd RRT
13
(15 labs)
12
Reference material KF result / ppm
11 RRT Xaverage (ppm) 6.4

10
9 RRT Stdev (ppm) 1.41
8
7
RRT RSD (%) 22.0
6 Rcalc (ppm) 3.95
5
4 R(IEC 60814) (ppm) 3.79
3
2
1
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Labs no.
KF results (sample a. <1 g) KF results (sample a. >1 g)
RM value (6,0±0,8 ppm) avr
avr ± 3s avr ± 2s
Figure 2.6 RRT statistics of KF results for low water reference material
35
Values were calculated (Equation 2.1 and Equation 2.2):

𝑅𝑐𝑎𝑙𝑐 = 2.8 ∙ 𝑅𝑅𝑇 𝑆𝑡𝑑𝑒𝑣 (from ISO 13521) [2.1]
Equation 2.1 Calculated Reproducibility ISO 13521
R (IEC 60814) = 1.5 ∙ √RRT Xaverage (from IEC 60814)

Equation 2.2 Reproducibility IEC 60814
Rcalc was 3.95 and is comparable to RIEC 60814 which was 3.79. We can conclude that Rcalc of CIGRE WG D1.52 RRT as
performed by 18 laboratories on Reference material with low water content (6.0 ppm), met IEC 60814 reproducibility
requirements.
The influence of sample size was evaluated by comparison of statistical parameters of the two groups of measurements
and is reported in Table 2.6.
Table 2.6 Comparison of different sample amounts on RRT Karl Fischer statistics for low water reference material
Sample amount Sample amount

less than 1 g more than 1 g
(9 labs) (9 labs)
RRT Xaverage (ppm) 6,2 6.5
RRT Stdev (ppm) 1,38 1.45
RRT RSD (%) 22,2 22.3
Accuracy (%) 3,3 8.3
Accuracy was calculated (Equation 2.3):

(𝑅𝑅𝑇xaverage − XCRM )
𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 (%) =
XCRM
Equation 2.3 RRT Karl Fisher accuracy
Where
XCRM – water content of certified reference material in ppm
In cases of low water content reference material, the RSD of KF results with sample amounts less than 1 gram (22.2%)
are nearly equal to RSD for all measurements (22.3%).
2.8.6.2 Medium water content value - tested material NIST 8506a

(Reference material value = 18.3 ± 1.9 mg/kg)
13 laboratories participated in a second RRT with 36 measurements (Figure 2.7). Results from first RRT were not
considered because the shipment of reference material was inhomogeneous.
36
All measurements from 2nd RRT

KF results of Reference material NIST 8506a
32 (13 labs)
30
Reference material KF result / ppm
28
26 RRT Xaverage (ppm) 17.3
24
22
20 RRT Stdev (ppm) 3.51
18
16
14 RRT RSD (%) 20.3
12
10 Rcalc (ppm) 9.83
8
6
4 R(IEC 60814) (ppm) 6.24
2
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Labs no.
KF results (sample a. <1 g) KF results (sample a. >1 g)

avr avr ± 3s
avr ± 2s RM value (18,3±1,9 ppm)
Figure 2.7 RRT statistics of KF results for medium water reference material
Rcalc was 9.83 and was a bit larger than RIEC 60814 which was 6.24. The higher reproducibility was observed in four
international proficiency tests conducted in 2012, 2013, 2014 and 2015 by IIS- Institute for Interlaboratory Studies
Spijkenisse, the Netherlands [2.2][2.3][2.4][2.5][2.6][2.7], with 45 – 59 participating labs. Results are presented
in Table 2.7.
Table 2.6 Comparison of CIGRE D1.52 RRT and IIS statistics
IIS – last four years results of Proficiency test CIGRE D1.52

RRT
Transformer Transformer Transformer Transformer Transformer Transformer (CRM NIST

oil (used) oil (fresh) oil (used) oil (fresh) oil oil 8506a)
2015 [2.7] 2015 [2.6] 2014 [2.5] 2014 [2.4] 2013 [2.3] 2012 [2.2] 2013
No. of labs 56 45 59 48 53 50 13
participating
RRT Xaverage (ppm) 28.0 17.9 30.7 16.80 19.6 19.2 17.3
RRT Stdev (ppm) 2.94 2.94 3.99 2.73 3.13 3.39 3.51
RRT RSD (%) 10.5 16.4 13.0 16.2 16.0 17.7 20.3
Rcalc (ppm) 8.21 8.22 11.17 7.64 8.76 9.50 9.83
R(IEC 60814) (ppm) 7.94 6.34 8.31 6.15 6.64 6.57 6.24
The RSD from CIGRE D1.52 RRT with medium WCL reference material NIST 8506a, is comparable with the IIS
proficiency tests RSD results with respect to similar WCL samples. In general, the RSD values from IIS are slightly lower,
which follows from having a higher number of participating laboratories.
We can conclude that Rcalc in all seven cases is slightly higher than IEC 60814 requirement.
The sample size influence on the evaluation of NIST 8506a reference material is presented on Table 2.7:
Table 2.7 Comparison of different sample amounts on RRT Karl Fischer statistics for medium water reference
material
37
Sample amount Sample amount

less than 1 g more than 1 g*
(9 labs) (4 labs)
RRT Xaverage (ppm) 17.0 18.3
RRT Stedev (ppm) 2.63 4.85
RRT RSD (%) 15.5 26.5
*The calculated values for sample amounts more than 1 gram, are not representative, because only four
laboratories participated
In case of medium water content reference materials, the RSD of KF results with sample amounts less than 1 gram
(15.5%) were lower than RSD of all sample amounts measured (20.3%).
The RSD from CIGRE D1.52 RRT with medium WCL reference material NIST 8506a, is comparable with the IIS
proficiency tests RSD results with respect to similar WCL samples. In general, the RSD values from IIS are slightly lower,
which follows from having a higher number of participating laboratories.
2.9 RRT KARL FISCHER TEST OF 4 INSULATING LIQUIDS. STATISTICAL EVALUATION

Absolute water content in four insulating liquid samples was measured by each of the RRT participants according to the
procedure described in paragraph 2.3. Three KF measurements per sample were performed and ambient temperature
was noted (Figure 2.8 – Figure 2.11).
RRT KF RESULTS FOR NMO

(16 labs, no outliers)
RRT Xaverage (ppm) 10.2

RRT Stdev (ppm) 1.1
RRT RSD (%) 10.8
RRT Rcalc (ppm) 3.0
RRT Rcalc (%) 30
RIEC60814 (ppm) 4.8
Figure 2.8: RRT statistics of KF results for new mineral oil
Rcalc was 3.0 ppm (Figure 2.8) which was lower than RIEC 60814 which was 4.8 ppm.
We can conclude that Rcalc of CIGRE WG D1.52 RRT, performed by 16 laboratories on new mineral oil test samples,
meets IEC 60814 reproducibility requirements.
RRT KF RESULTS FOR AMO

(16 labs, 1 outlier)
RRT Xaverage (ppm) 16.0

RRT Stdev (ppm) 3.0
RRT RSD (%) 18.7
RRT Rcalc (ppm) 8.4
38
RRT Rcalc (%) 53

RIEC60814 (ppm) 6.0
Figure 2.9: RRT statistics of KF results for aged mineral oil
Rcalc was 8.4 ppm (1 outlier) (Figure 2.9) which was higher than RIEC 60814 which was 6.0 ppm.
We can conclude that Rcalc of CIGRE WG D1.52 RRT, performed by 15 laboratories on aged mineral oil test samples,
does not meet IEC 60814 reproducibility requirements.
RRT KF RESULTS FOR SE

RRT Xaverage (ppm) 350

RRT Stdev (ppm) 7.5
RRT RSD (%) 2.0
RRT Rcalc (ppm) 21.0
RRT Rcalc (%) 6
RIEC60814 (ppm) 28.1
Figure 2.10: RRT statistics of KF results for synthetic ester
Rcalc was 21.0 ppm (1 outlier) (Figure 2.10) which was lower than RIEC 60814 which was 28.1 ppm.
We can conclude that Rcalc of CIGRE WG D1.52 RRT, performed by 15 laboratories on synthetic ester test samples,
meets IEC 60814 reproducibility requirements.
RRT KF RESULTS FOR NE

39
RRT Xaverage (ppm) 170

RRT Stdev (ppm) 5.3
RRT RSD (%) 3.1
RRT Rcalc (ppm) 14.8
RRT Rcalc (%) 9
RIEC60814 (ppm) 19.6
Figure 2.11: RRT KF results for natual ester
Rcalc was 5.3 ppm (1 outlier) (Figure 2.11) which was lower than RIEC 60814 which was 19.6 ppm.
We can conclude that Rcalc of CIGRE WG D1.52 RRT, performed by 16 laboratories on natural ester test samples, meets
IEC 60814 reproducibility requirements.
The RRT statistics of the results in the four insulating liquids is presented on Table 2.9.
Table2.8 Summary of RRT statistics of KF WCL results on the four insulating liquids
2,8 ∗ 𝑅𝑅𝑇 𝑆𝑡𝑑𝑒𝑣 1,5 ∗ √𝑅𝑅𝑇 𝑋𝑎𝑣𝑒𝑟𝑎𝑔𝑒
𝑅𝑅𝑇𝑋𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑅𝑅𝑇𝑋𝑎𝑣𝑒𝑟𝑎𝑔𝑒
RRT Xaverage RRT Stdev RRT RSD RRT Rcalc RIEC60814
(ppm) (ppm) (%) (%) (%)
NMO 10.2 1.1 10.8 30 47

AMO 16.0 3.0 18.7 53 38
NE 170 5.3 3.1 9 12
SE 350 7.5 2.0 6 8
2.10 STATISTICAL EVALUATION OF RRT CAPACITIVE MOISTURE SENSOR TESTS

Before each test, each of the three capacitive sensors was rinsed with the remaining fluid from the KF measurements.
The liquid from KF tests was poured in a small beaker and mechanically stirred using the capacitive sensor. Afterwards,
the sensor was inserted into the freshly opened 300 ml bottle and positioned on the magnetic stirrer plate. During tests,
the liquids were stirred by a magnetic stirring bar, which was pre-inserted before sealing the test bottles. The stirring
was set to low speed, assuring oil circulation around the sensor and avoiding turbulence in the fluid. After the sensors
were inserted in the bottles, reading of relative saturation (RS %) were immediately taken, again after 3 minutes, after
10 min and the sensor temperature after 10 min.
All the three sensors included in RRT were tested in a row to minimize the effect of ambient temperature and humidity
changes during the 10 min measurements in opened bottles.
Results are presented in the Table 2.10.
40
Table2.9 Statistical evaluation of the capacitive moisture sensor Round Robin test results
RRT measurement temperature

conditions at temeratures RS% after 3 min results RS% after 10 min results
between 19...26 °C
Sample Sensor Average St.Dev. RSD (%) Average St.Dev. RSD (%)
CS-1 25.0 5.2 21% 24.1 4.1 17%

CS-2 24.0 7.4 31% 23.4 6.8 29%
New mineral oil CS-3 24.5 7.0 29% 24.4 6.3 26%
Average (17 labs, 3
sensors) 24.5 6.5 27% 23.9 5.7 24%
CS-1 25.4 4.6 18% 24.5 3.7 15%

CS-2 25.7 6.1 24% 25.8 6.1 23%
Aged mineral oil CS-3 25.1 6.5 26% 24.9 5.6 23%
Average (17 labs, 3
sensors) 25.4 5.7 23% 25.1 5.1 20%
CS-1 22.0 2.7 12% 21.3 1.7 8%

CS-2 21.8 3.2 15% 21.0 2.9 14%
Synthetic Ester CS-3 22.2 5.2 23% 20.0 1.6 8%
Average (17 labs, 3
sensors) 22.0 3.7 17% 20.8 2.1 10%
CS-1 22.4 3.2 14% 21.2 2.3 11%

CS-2 21.4 3.6 17% 19.8 2.1 10%
Natural Ester CS-3 22.2 5.4 25% 21.6 4.4 20%
Average (17 labs, 3
sensors) 22.0 4.1 19% 20.9 2.9 14%
Stabilisation of sensors signals was observed within 10 minutes for samples NMO and AMO.
For sensors in more viscous samples of NE and SE, stabilised did not occur within 10 minutes and this can be observed
from the low average values of RS% for these two samples.
2.10.1 Evaluation of sample temperature impact during RRT on capacitive sensor results
To evaluate the effect of oil sample temperature on RS% results for each laboratory, results with recalculated to 20oC
using equation 2.4 with solubility coefficients A and B shown in Table 2.10. Variations in the RRT participant’s sample
temperatures ranged from 19 °C to 26 °C.
1 1
𝑅𝑆%20°𝐶 = 𝑅𝑆%𝑥°𝐶 ∗ 10𝐵 ( − )
𝑇2 𝑇1
Equation 2.4 Recalculation of the RRT results to a temperature of 20°C
Table 2.11 Solubility coefficients used for the recalculation of the RRT results to a temperature of 20°C
RRT INSULATING LIQUIDS SOLUBILITY

COEFFICIENTS
Solubility coefficient
Sample
A B
New mineral oil 6.89 1503
Aged mineral oil 6.60 1379
Synthetic Ester 4.98 511
Natural Ester 5.1 634
Comparison of original RSD (%) and recalculated RSD (%) is presented in Table 2.12.
41
Table 2.12 Statistical evaluation of the capacitive moisture sensor Round Robin test results
RRT measurement temperature RS% after 10 min

conditions at temeratures RS% after 10 min results
recalculated results to 20°C
between 19...26 °C
Sample Sensor Average St.Dev. RSD (%) Average St.Dev. RSD (%)
Average (17 labs, 3
New mineral oil sensors) 23.9 5.7 24% 25.9 5.4 21%
Average (17 labs, 3
Aged mineral oil sensors) 25.1 5.1 20% 27.1 5.0 19%
Average (17 labs, 3
Synthetic Ester sensors) 20.8 2.1 10% 21.4 1.8 9%
Average (17 labs, 3
Natural Ester sensors) 20.9 2.9 14% 21.7 2.7 12%
The recalculation shows that the slight temperature differences (19 oC …26 oC) between the test conditions in the
different laboratories during the RRT, did not increase RSD (%) substantially.
Statistical evaluation of sensors measurements for the four liquid samples in the 17 laboratories are presented in Figure
2.12, Figure 2.13, Figure 2.14 and Figure 2.15.
50
45 RRT CS RESULTS AFTER 10 MIN
40
FOR NMO
(17 labs)
CS NMO results (RS%)
35
30 RRT Xaverage (RS%) 23.9
25 RRT Stdev (%) 5.7
20
RRT RSD (% RS%) 23.8
15
10 RRT Rcalc (RS%) 16.1
5 RRT Rcalc (% RS%) 67
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
CS results avr avr ± 2s avr ± 3s Lab No.
Figure 2.12 RRT CS results after 10 min for NMO
50
RRT CS RESULTS AFTER 10 MIN
45 FOR AMO
40 (17 labs)
CS AMO results (RS%)
35
RRT Xaverage (RS%) 25.1
30
RRT Stdev (RS%) 5.1
25
20 RRT RSD (% RS%) 20
10 RRT Rcalc (%RS%) 57
5
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
CS results avr avr ± 2s avr ± 3s Lab. No.
Figure 2.13 RRT CS results after 10 min for AMO
42
50 RRT CS RESULTS AFTER 10 MIN

FOR SE
45
(17 labs, 2 outliers)
40
35
CS SE results (RS%)
30 RRT Stdev (RS%) 2.1

25 RRT RSD (% RS%) 10.0
20
RRT Rcalc (RS%) 5.8
15
RRT Rcalc (% RS%) 28
10
5
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
CS results avr avr ± 2s avr ± 3s (Q test outliers in red) Lab. No.
Figure 2.14 RRT CS results after 10 min for SE
RRT CS RESULTS AFTER 10 MIN

50 FOR NE
45
(17 labs, no outliers)
40
35
CS NE results (RS%)
RRT Stdev (RS%) 2.9

30
25 RRT RSD (% RS%) 13.9
15 RRT Rcalc (% RS%) 39
10
5
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
CS results avr avr ± 2s avr ± 3s Lab no.
Figure 2.15 RRT CS results after 10 min for NE
2.11 LEARNINGS FROM RRT ON CAPACITIVE SENSORS
2.11.1 Factory calibration and its verification

All three sensors had a calibration region according to the certificates (Table 2.13).
Table 2.13 Capacitive sensors specification data by manufacturer
Region rH Precision rH Precision Temp. Stability Response

Sensor 1 0 – 100% ±1.5% ±0.20°C at 20°C no data 10 min at 20°C (90%
rH)
Sensor 2 5 – 95% < ±2% ±0.2°C at 25°C ± < 10 sec (90% rH)
1%/year
Sensor 3 0 – 100% ± 3% ±2% (-20 – 120°C) no data no data
43
The calibration of the sensors has been tested in air and using salt solutions with known RS% at a constant temperature
(22°C) with different equilibrating time (10 min, 30 min, 60 min). The results of the sensor readings %RS to the
theoretical %RH for the sensors 1, 2 and 3 are shown in Figures 2.16 – Figure 2.18.
Figure 2.16 Theoretical %RH to %RS readings for sensor 1
44

It is clear that sensor 3 is only linear above 10% RS and cannot reliably measure a value below ~10%.
Additional tests in insulating liquid showed similar behavior.
2.11.2 Sensitivity in lower RS region in insulating liquid
The sensors 1, 2 and 3 have been tested in natural ester hermetically sealed container, with volume of 2.6 l (Figure
2.19) and in a climate chamber at different temperatures (Figure 2.20).
Following conditions were tested:

Temperature variation: 20° – 60 °C
3 repeated cycles - 20°C (24h), exponential increase until 60°C (takes 8 hours), 60°C (24 h)
3 repeated cycles – 20°C (24h), immediate increase until 60°C, (takes 5 minutes), 60°C (24 h)
The different cycle durations were used to investigate the influence of load on moisture saturation curve hysteresis.
Moisture of the natural ester by Karl Fischer titration: 82 mg/kg
45
Figure 2.19 Container Figure 2.20 Temperature program of the climate chamber
with different moisture
probes in a climate
chamber
70 70
60
60
50
relative humidity [%]
50
40
Temperature [°C]
30 40
20 30
10
20
0
10
-10
-20 0
28/08/2015 31/08/2015 03/09/2015 06/09/2015 09/09/2015 12/09/2015
Time
RS Sensor 2 RS sensor 3 RS sensor 1

Temperature sensor1 T_climate chamber
Figure 2.21 Development of %RS of the 3 sensors with time under the temperature program of Figure 2.20 in a
natural ester with an absolute moisture of 82 mg/kg
Figure 2.21 demonstrates the insensitivity below the 10% measuring range, but maybe also influence due to
incompatibility to ester, sensor 3 delivers negative values. At a temperature of 60°C, sensor 1 and sensor 2 indicate the
same relative saturation, whereas the indication of both sensors differ by about 10% at a temperature of 20°C.
46
2.11.3 Response time

The response time of sensor 1 has been investigated in different insulating liquids – in new mineral oil (Figure 2.22), in
aged mineral oil (Figure 2.23), in synthetic ester (Figure 2.24) and natural ester (Figure 2.25).
Figure 2.22 Response time of sensor 1 in new mineral oil
Figure 2.23 Response time of sensor 1 in aged mineral oil
47
Figure 2.24 Response time of sensor 1 in synthetic ester
Figure 2.25 Response time of sensor 1 in natural ester
These graphs show, that the equilibration time of 10 minutes seems adequate for stabilizing the readings in the case
of mineral oil, in the case of esters the equilibration time appears longer.
2.11.4 Temperature dependence

The temperature dependence of the three sensors has been checked in air in a climate chamber working on the
psychrometric principle, under following conditions (Figure 2.26):
5% - 65% relative humidity up and down (in 5% steps, each step held 2h)
Temperature slowly rising from 40°C to 60°C in 5°C steps (each step held 24h)
48
Figure 2.26 Temperature program for testing the temperature dependence (air, climate chamber)
Figure 2.27 %RS reading of sensor 1 – sensor 3, depending on temperature and time
Temperature dependence test in air (Figure 2.27) shows that sensors 2 and 3 drift in their RS readings with increasing
temperature. Only sensor 1 shows stable readings in all temperature regions.
49
2.12 WATER CONTENT DETERMINATION OF OIL IMPREGNATED PAPER INSULATION
2.12.1 Karl Fischer method IEC 60814

The following methods for the determination of the water content in solid insulation are described in the IEC 60814:
 Determination of water after previous extraction with methanol
 Evaporative stripping method
2.12.2 Calculation of the water content in the solid insulation
There are different calculation methods for the water content described in the standards IEC 60814 and ISO 287. Table
2.14 gives an overview about the calculation methods:
Table 2.14 Available calculation methods for the determination of water content in the solid insulation
Standard Method description Formula Basis

IEC 60814 Water content in solid 𝑚(𝐻2 𝑂) Dried
𝑊𝑎𝑡𝑒𝑟 𝑐𝑜𝑛𝑡𝑒𝑛𝑡 (% 𝑏𝑦 𝑚𝑎𝑠𝑠) = ∙ 100 %
insulation by Karl 𝑀 sample
Fischer
where
m(H2O) is the quantity of water evaporated in the oven or
extracted by means of methanol in grams
M is the mass of dried, de-oiled solid insulation sample grams
ISO 287 Paper and board - 𝑚0 − 𝑚1 Moist
𝑊𝑎𝑡𝑒𝑟 𝑐𝑜𝑛𝑡𝑒𝑛𝑡 (% 𝑏𝑦 𝑚𝑎𝑠𝑠) = ∙ 100 %
Determination of 𝑚0 sample
moisture content of a where
lot - Oven-drying
method m1 is the quantity of water titrated in grams
m0 is the mass of moist, solid insulation sample in grams
The main difference between the different methods is the basis for the calculation. IEC 60814 for solid insulation
calculates the water content using the dried, de-oiled sample. ISO 287 (referenced also in IEC 60641 “Pressboard and
Presspaper for Electrical Purposes) uses a moist sample. The difference in the results is relatively small at low water
contents and becomes larger at higher water contents. Figure 1. shows the relationship between the two calculation
methods.
50
Figure 1.28 Moisture content based on moist or dry sample using two standard methods
2.12.3 Test results from the round robin on pressboard sample

In the round robin test, the water content was calculated with the two above mentioned methods (Equation 2.4 and
Equation 2.5):
𝑚𝐻2𝑂
𝑐(𝐻2 𝑂)∗ = ∙ 100 %
𝑚𝑑𝑟𝑦,𝑑𝑒𝑜𝑖𝑙𝑒𝑑 𝑐𝑒𝑙𝑙𝑢𝑙𝑜𝑠𝑒
Equation 2.4 Calculation of water content based on the dry sample
𝑚𝐻2𝑂
𝑐(𝐻2 𝑂)∗∗ = ∙ 100 %
𝑚𝑑𝑟𝑦,𝑑𝑒𝑜𝑖𝑙𝑒𝑑 𝑐𝑒𝑙𝑙𝑢𝑙𝑜𝑠𝑒 + 𝑚𝐻2𝑂
Equation 2.5 Calculation of water content based on the moist sample
2.12.3.1 Evaporative stripping oven method

The results of the RRT on water content in pressboard, obtained with the evaporative stripping oven method are
presented in Table 2.15.
Table 2.15 Results of the RRT on water content in pressboard obtained with the evaporative stripping oven method
Property Lab. 3 Lab. 4 Lab. 10 Lab. 11 Lab. 18 Lab. 16 Lab. 7

Oven
150 180 180 180 - 200 200
temperature °C
Carrier Gas nitrogen nitrogen nitrogen nitrogen - nitrogen air
Mass of oiled
204.3 185.8 36.2 - 124.9 130 99.4
sample (mg)
c(H20) *
3.36 5.77 5.15 5.59 5.46 5.90 5.97
(%)
c(H20) **
3.25 5.46 4.90 5.29 5.18 5.56 5.64
(%)
51
2.12.3.2 Methanol extraction

The results of the RRT on water content in pressboard, obtained with the methanol extraction method are presented in
Table 2.16.
Table 2.16 Results of the RRT on water content in pressboard obtained with the methanol extraction method
Property Lab. 1 Lab. 8 Lab. 9 Lab. 12 Lab. 13

Mass of oiled
669.6 328.9 - 559.6 3560.0
sample (mg)
c(H20) *
5.50 5.56 5.68 3.48 2.92
(%)
c(H20) **
5.2 5.3 5.4 3.4 2.8
(%)
2.12.3.3 Reproducibility of both methods.

A statistical evaluation of both methods for all 13 results on the dried samples shows no outliers and following data
(Table 2.17):
Table 2.17 Overall statistical evaluation
Method average Standard deviation Relative standard R

c(H20)* (%) deviation
Evaporative 5,15 1,11 21,6% 60,5%
stripping oven
method and
methanol
extraction
However, it must be kept in mind, that these results are valid only for the concentration measured. Lower water
concentrations in paper may result in higher reproducibility values.
2.13 CONCLUSIONS FROM THE RRT RESULTS OF KARL FISCHER COULOMETRIC TESTING
ACCORDING TO STANDARD IEC 60814
The coulometric KF determination of absolute water content in insulating liquids is well known and used worldwide in
transformer oil diagnostic laboratories. The following lists the learnings from the RRT.
 A very small mass of oil sample, less than 1 g, is enough for the test. This should be considered at the next
revision of IEC 60814.
 Verifying of accuracy of WCL determination is possible using commercially available Certified Reference
materials with various defined water in oil concentrations.
 The Karl Fischer method is applicable not only for mineral oils, but also for other insulating liquids.
 The reproducibility of KF measurements depends strongly on dissolved water in insulating liquid concentration.
 There is no need to specially calibrate coulometric apparatus for KF measurements.
 The formula for calculation of R in IEC 60814 should be revised.
The IEC 60814 reproducibility requirement R (IEC 60814) = 1.5 ∙ √RRT Xaverage was met in four cases out of the six RRT
samples. For very low WCL, 6 ppm sample and 16 ppm mineral oil, the reproducibility was found higher. As the IIS
proficiency testing carried out during the last 5 years demonstrates similar results, we conclude that the reproducibility
criterium of this test method for water concentrations lower than 30 ppm, is too stringent and should be reviewed for
a revision of standard.
52
2.14 CONCLUSIONS FROM THE RRT RESULTS OF RS% SENSORS TESTING ACCORDING TO
RRT PROCEDURE
The measurement of relative saturation of oil (RS%) by use of a capacitive thin film polymer sensor is not a standardized
test method for laboratory use. Each sensor needs to be factory calibrated before use.
Relative saturation measurement of insulating liquids by means of sensors, is mostly used in on-line monitoring
apparatus for supervision of operating transformers. To gain insight into the reproducibility of this measurement when
using different commercial sensors, the WG developed a laboratory RRT procedure.
 Results presented in the Table 1.10 show that the three tested sensors differed in their performance and
correspondingly in the relative standard deviation, although they had been factory calibrated and had similar
specification.
 Capacitive sensors have an intrinsic response time of 10 minutes or longer for esters.
 Oil flow is necessary in the measurement process.
 Since the RRT was performed with a single RS and at a single temperature, statistical evaluation is not
representative.
 Additional tests showed differences in the sensitivity (especially in the lower humidity range), response, drift,
temperature dependence and compatibility.
The user must be aware of the importance of all these parameters for reliable sensor behavior and request the necessary
information from the manufacturer. Capacitive thin film polymer sensors have response times ≥ 10 min, which is
acceptable in their application as continuous on-line monitoring devices.

[2.1] ISO 13528:2015 International Standard, second edition; Statistical methods for use in proficiency testing by
interlaboratory comparison
[2.2] IIS Institute for Interlaboratory Studies Spijkenisse, The Netherlands; Results of Proficiency test, Transformer Oil,
October 2012
[2.3] IIS Institute for Interlaboratory Studies Spijkenisse, The Netherlands; Results of Proficiency test, Transformer Oil,
November 2013
[2.4] IIS Institute for Interlaboratory Studies Spijkenisse, The Netherlands; Results of Proficiency test, Transformer Oil
(fresh), November 2014
(used), November 2014
(fresh), November 2015
(used), November 2015
53
54
3. CONVERSION OF MOISTURE MEASUREMENTS BETWEEN

CAPACITIVE SENSORS AND KARL FISCHER TITRATION
3.1 INTRODUCTION
Moisture in insulating liquid is one of the most widely used parameters for transformer insulation condition monitoring
[3.1]. Historically, coulometric Karl Fischer titration was used for measuring the absolute moisture content in insulating
liquid, with units of mg/kg, μg/g or parts per million (ppm) [3.2, 3.3]. In recent years, attention has progressively been
paid to the application of new measurement techniques like capacitive sensors. These provide the relative saturation
(RS) of a sample, expressible as the ratio of absolute moisture to moisture saturation of a sample, with unit of
percentage (%) [3.2, 3.3].
As the conventional mg/kg measurements have been well established and with reference data readily available for
condition monitoring, there is a motivation to convert the new measurement data of RS into mg/kg equivalents and
vice versa. Conversion of moisture measurements between the capacitive sensors and Karl Fischer titration depends on
the moisture saturation curve of the insulation liquids. From a mathematical point of view, this can be done through
the relationships expressed in Equation 3.1 and Equation 3.2. Note Equation 3.2 has an Arrhenius form as the moisture
saturation in insulating liquid (S) is temperature dependent. The natural logarithmic form can also be expressed in its
decimal logarithm form for convenience as shown in Equation 3.3.
𝑅𝑆
𝑊𝐶𝐿 = ×𝑆
100
Equation 3.1
𝑆 = 𝑒 𝐴𝑒 −𝐵𝑒⁄𝑇
Equation 3.2
𝑆 = 10𝐴10 −𝐵10⁄𝑇
Equation 3.3
where 𝑊𝐶𝐿 is the absolute water content in liquid (mg/kg), 𝑅𝑆 is the relative saturation (%), 𝑆 is the moisture saturation
level (mg/kg), 𝑇 is the temperature in Kelvin, whereas 𝐴𝑒 , 𝐵e, 𝐴10 and 𝐵10 are moisture saturation coefficients in natural or
decimal logarithm forms that depend on fluid characteristics, composition and condition [3.4]. As both natural logarithm and
decimal logarithm forms of coefficients are used in the literature, equations 3.4 and 3.5 can be used to convert the moisture
saturation coefficient values.
𝐴𝑒 = 𝐴10 ln(10) ; 𝐵𝑒 = 𝐵10 ln(10)
Equation 3.4
𝐴10 = 𝐴𝑒 log(𝑒) ; 𝐵10 = 𝐵𝑒 log(𝑒)
Equation 3.5
3.2 LITERATURE REVIEW ON MOISTURE SATURATION COEFFICIENTS

3.2.1 Mineral oils
With respect to the water saturation equation of the natural and decimal logarithm form, Table 3.1 records the moisture
saturation coefficients, A and B for new mineral oils as adapted from literature. Although they all belong to mineral oils,
the exact compositions vary, so the oil brand/model name is given whenever it is achievable. In addition, due to lack
of further details, the measurement method information is only broadly classified into two groups (I - Karl Fischer
Titration with Climate Controlled Conditioning; II - Karl Fischer and Capacitive Sensor Measurements). As new oil
samples are considered in all the reference work, the condition of the oil samples should reasonably be the same, i.e.
fresh condition. This allows reasonable statistic evaluation of the saturation coefficients.
Figure 3.1, illustrates the moisture saturation expressed in mg/kg with temperature for new mineral oils based on the
A and B coefficients given in Table 3.1. It is clear that the moisture saturation level increases with temperature and
variance in moisture saturation is larger at higher temperatures. At 100 °C, the difference in the moisture saturation
level could be as large as ~500 mg/kg.
55
Table 3.1: Moisture saturation coefficients for new mineral oils as adapted from literature
Moisture
Saturation Measurement
Coefficients Oil Type Reference
Method
A B
6.20 1320 Nynas Nytro 4000X II [3.19]
6.50 1400 Nytro 4000X II [3.19]
6.50 1400 New mineral oil II [3.19]
6.50 1400 Nynas Libra II [3.20]
6.78 1475 Lyra X II [3.21]
6.80 1500 Diala S4-ZX-I II [3.19]
6.80 1520 Ergon Highvolt III II [3.19]
6.93 1520 Nynas Nytro 3000 II [3.18]
7.06 1570 Hyvolt II [3.20]
7.09 1567 Not specified I [3.11]
7.12 1537 Aromatic content 21 % - [3.8]
7.20 1598 Shell Diala D II [3.18]
7.28 1653 Shell Diala D I [3.14]
7.29 1647 Nynas Taurus II [3.9]
7.30 1630 Shell Diala D I [3.12]
7.35 1647 Naphthenic oil I [3.13]
7.37 1662 Not specified II [3.16]
7.37 1662 Not specified II [3.22]
7.42 1670 Refined naphthenic oil I [3.10]
7.44 1686 Nynas Taurus I [3.15]
7.71 1783 Highly refined - [3.17]
7.86 1836 Not specified - [3.7]
7.11 1585 Average

0.40 123 Standard deviation
6.20 1320 Minimum
7.86 1836 Maximum
56
Figure 3.1: Water solubility (S) versus temperature for new mineral oil as adapted from literature
Table 3.2 reports the findings from literature on moisture saturation coefficients for aged mineral oils. Note that if
available from the literature, details of the oil samples (such as the state of ageing) are also recorded. With respect to
the A and B coefficients in Table 3.2, the moisture saturation versus temperature curves are shown in Figure 3.2. One
of the challenges of comparing moisture saturation curves of aged oil samples is the difference in the ageing state.
Without a basis of a common condition, average value of the coefficients will not be evaluated here.
Although the detailed condition of the aged oil samples is unclear, with reference to the average coefficients obtained
for new mineral oils, aged samples typically exhibit higher moisture saturation. This is most likely due to the presence
of low molecular weight acids (LMA) that are polar and hydrophilic, thereby assuming the role of emulsifying agents in
oil which increases moisture solubility of the oil [3.18, 3.25].
Table 3.2: Moisture saturation coefficients for aged mineral oils as adapted from literature
Moisture
Coefficients Oil Type and Details Reference
Method
A B
5.50 1030 Dry Technol Y3000, acidity 0.16 mg KOH/g oil II [3.23]
5.80 1130 Wet Technol Y3000, acidity 0.08 mg KOH/g oil II [3.23]
6.08 1143 Shell Diala K, acidity 0.49 mg KOH/g oil II [3.18]
6.20 1270 Dry Modriča Traafomol S, acidity 0.02 mg KOH/g oil II [3.23]
6.20 1270 Wet Modriča Traafomol S, acidity 0.05 mg KOH/g oil II [3.23]
6.35 1262 54-year-old 10 MVA Tx with naphthenic mineral oil II [3.20]
6.47 1325 75-year-old 4 MVA transformer with uninhibited mineral oil II [3.20]
6.54 1349 33-year-old 225 MVA Tx with Shell Diala AX II [3.20]
6.58 1329 21-year-old 22.5 MVA Tx with Shell Diala S II [3.20]
7.01 1477 Aged oil - [3.24]
7.63 1732 11-year-old 8 MVA Tx with naphthenic mineral oil II [3.20]
8.42 1921 Oil with 0.30 mg KOH/g oil - [3.7]
57
Figure 3.2: Water solubility (S) versus temperature for aged mineral oil as adapted from literature
3.2.2 Alternative insulating liquids

With the advent of alternative liquids, it is important to investigate the properties of these liquids for application in
power system equipment. In this section, the moisture saturation coefficients for three alternative liquid types including
synthetic ester, natural ester and silicone insulating liquid will be reported. These liquids are in new or unaged condition.
The moisture saturation coefficients for new synthetic esters from literature are shown in Table 3.3. Based on the
coefficients, Figure 3.3 illustrates the temperature dependent moisture saturation curves. The same sequence of
tabulating the moisture saturation coefficients and plotting the moisture saturation curves with temperature is used for
new natural ester (Table 3.4 and Figure 3.4) as well as for new silicone insulating liquid (Table 3.5 and Figure 3.5).
Assuming similar fresh condition of the samples, the average, maximum and minimum values of the coefficients were
evaluated. As can be seen, the work in evaluating moisture saturation for alternative liquids is not as much as for
mineral oils. Particularly for the silicone insulating liquid, there were only three sets of data found during the study.
Table 3.3: Moisture saturation coefficients for new synthetic ester as adapted from literature
Moisture
Coefficients Insulating liquid Type Reference
Method
A B
5.32 608 Midel 7131 II [3.9]
5.41 582 Midel 7131 I [3.14]
5.47 652 Midel 7131 II [3.26]
5.60 655 Midel 7131 II [3.18]
5.66 696 Midel 7131 II [3.21]
5.49 639 Average

5.32 582 Minimum
5.66 696 Maximum
58
Figure 3.3: Water solubility (S) versus temperature for new synthetic ester as adapted from literature
Table 3.4: Moisture saturation coefficients for new natural ester as adapted from literature
Moisture
Method
A B
5.34 707 FR3 II [3.9]
5.53 764 FR3 II [3.18]
5.71 802 Not specified - [3.28]
5.74 808 Biotemp I [3.27]
5.75 807 Rapeseed insulating liquid I [3.13]
6.15 937 FR3 II [3.21]
5.70 804 Average

5.34 707 Minimum
6.15 937 Maximum
59
Figure 3.4: Water solubility (S) versus temperature for new natural ester as adapted from literature
Table 3.5: Moisture saturation coefficients for new silicone insulating liquid as adapted from literature
Moisture
Method
A B
5.30 875 Silicone insulating liquid II [3.23]
5.45 799 Silicone insulating liquid I [3.29]
6.30 1194 Silicone insulating liquid II [3.9]
5.68 956 Average

5.30 799 Minimum
6.30 1194 Maximum
Figure 3.5: Water solubility (S) versus temperature for new silicone insulating liquid as adapted from literature
3.2.3 Comparison between the liquids

This section covers the comparison of moisture saturation levels among mineral oil and alternative liquids pertaining to
new or unaged insulating liquid samples. Figure 3.6 shows the moisture saturation curves based on average A and B
values, corresponding to new or unaged samples of the different liquids. Moisture saturation is seen to depend on
temperature and ageing condition, as well as the type and composition of the insulating liquid. The lowest water
saturation levels can be seen for mineral oil, and then followed by silicone insulating liquid, natural ester and synthetic
ester. This difference in water saturation has been known to be attributed to the liquid polarity, of which mineral oil
and silicone insulating liquid are non-polar or weakly polar [3.9]. Both ester liquids have higher moisture saturation due
to the presence of the polar ester linkages (2-4 per molecule for synthetic esters and 3 for natural esters) [3.9, 3.30].
As water molecules are polar, the presence of those polar ester linkages will hence help attract water, thereby increasing
moisture saturation of the ester liquids.
In the previous sections, the record of measurement methods has also hinted towards the influence it has on the
moisture saturation of liquids. Owing to the difficulty of pinpointing the exact measurement setup and procedure
followed in the literature, the comparison among different methods would be less meaningful. In the following sections,
the measurement methods used in the CIGRE RRT are clearly defined, making the assessment of impacts on moisture
saturation clear.
60
Table 3.6: Average moisture saturation coefficients for new or unaged insulating liquid types as adapted from
literature
Moisture Saturation Moisture Saturation
Coefficients (Decimal Coefficients (Natural
List logarithm Equation 3.5) logarithm Equation 3.4) Insulating liquid Type
𝐴10 𝐵10 𝐴𝑒 𝐵𝑒
MOAve 7.11 1585 16.38 3650 New Mineral Oil
SEAve 5.49 639 12.65 1470 New Synthetic Ester
NEAve 5.70 804 13.13 1852 New Natural Ester
SOAve New Silicone
5.68 956 13.09 2201
Insulating liquid
Figure 3.6: Water solubility (S) versus temperature for mineral oil and alternative liquids
3.3 METHODS FOR DETERMINING MOISTURE SATURATION COEFFICIENTS

There is no international standard for determining moisture saturation coefficients in insulating liquids and hence a large
variety of methods have been introduced in the literature, using different equipment and procedures. This is the major
factor contributing to the variance in observed moisture saturation coefficients. In addition, the details of experimental
design and procedure for determining moisture saturation coefficients are rarely given in the literature.
The principle for determining the moisture saturation coefficients can be generalized as:
1) condition the insulating liquid sample to a known 𝑅𝑆 level at a temperature
2) measure the absolute moisture content WCL of the insulating liquid sample
3) calculate the saturation level S at that temperature based on equation 3.1
4) repeat the procedure at three or more temperatures to obtain the coefficients according to equation 3.2 or equation 3.3
In this section, three methods for determining moisture saturation coefficients used in the RRT are described in detail,
of which the results are given in section 3.4. Method-1 uses Karl Fischer measurements of insulating liquids conditioned
in a controlled climate chamber at an equilibrium state; both Method-2 and Method-3 use Karl Fischer and capacitive
61
sensor measurements of insulating liquids. Method-2 introduces a dedicated test cell and conducts the measurements
at equilibrium whereas Method-3 proposes a simple experimental setup and a quick procedure to conduct the
measurements in a transient/quasi-equilibrium state.
The calculation procedure applied in the study comprises the least squares method for a linear regression, where the
coefficient Ae represents the intercept and Be the slope in a linear regression with X-axis = 1/T and Y-axis =
ln(WCL∙100/RS). To obtain results with sufficient accuracy the correlation coefficient of the linear regression should not
be less than 0.995.
3.3.1 Method-1
Method-1 uses a closed chamber exposing the stirred insulation liquid to a constant environment with preset RS and
temperature. Such an environment can be achieved using a climate chamber, saturated salt solution chamber or with
a controlled mixing of wet and dry air into a closed chamber. The absolute moisture content of the liquid at equilibrium
is measured using Karl Fischer titration. Moisture saturation coefficients are calculated based on results obtained at
multiple temperatures.
3.3.1.1 Experimental design
Several methods are available for controlling the environment above a liquid sample so as to maintain a constant and
reliable environment. This can be achieved via installing relative saturation and temperature sensors in the chamber for
verifying the environment. It is also important to avoid contamination of the sample (i.e. salt or other contaminants).
In addition, extended periods of humidification should be avoided to minimize the effect of insulating liquid
oxidation/hydrolysis as polar by-products could affect the moisture saturation. Note that the choice of the method
depends on the equipment and resources available to the user. A climate chamber was used in the Method-1 for the
RRT.
3.3.1.2 Climate Chamber
If a climate chamber is available, this would be the most effective and the least laborious method to humidify a liquid
sample to equilibrium state. If the chamber is designed with small hatches, it will be even more beneficial, particularly
for access into the chamber whenever needed during the humidification process without excessively disturbing the
climate.
3.3.1.3 Saturated Salt Solution Chamber
This method involves the use of a saturated salt solution (salt in deionised water) in a glass container with a large
surface area. A magnetic stirrer should also be used in the solution. In addition, to provide an evenly distributed
environment, a fan is used to disperse the air above the solution. As for temperature control, the chamber should be
either put into an incubator or otherwise with the insulating liquid temperature being regulated. Generally, the use of a
saturated salt solution provides good environment control of relative humidity. With close reference to [3.6], note that
the usage of different salts in a closed environment and at certain temperatures will yield different relative humidities.
Caution needs to be exercised to avoid salt splutter during stirring and general contamination of the insulating liquid
samples. Alternatively, glycerol-water solutions can also be used to control relative humidity of air in a closed
environment [3.31].
3.3.1.4 Gas Flow Chamber
The third option is to mix dry gas with moist gas. This gas can be air or nitrogen, moistened through a reservoir of
deionised water preheated to a desired temperature. The mixed gas is subsequently directed (e.g. using Teflon tubes)
into a closed chamber (plastic box) with small open holes to avoid over-pressure. For temperature control, the plastic
box is placed in an incubator. The gas flow can be regulated by using a relative humidity sensor in the climate volume
and adjusting the gas flow rates of dry and wet air. Although it can be challenging to find the exact regulation position
for the valves, this method can provide a very stable condition. Apart from that, this method is particularly advantageous
for avoiding insulating liquid degradation if nitrogen is used.
62
3.3.1.5 Test Procedure

The liquid samples are put into small open beakers and placed into the climate controlled chamber at a low temperature
and relative humidity to avoid direct water condensation into the insulation liquid before the real experiment starts.
Dried syringes are used to sample the liquids and Karl Fischer titration with direct injection is used to measure the
absolute moisture content of the liquid samples. Three samples per measurement are needed for a repeatability check.
Regular sampling and measurements are continued until a stable moisture content reading is reached. The climate
chamber should be stable and not vary more than ± 0.2°C and the relative humidity not more than 0.5% in steady
state. The Karl Fischer measurements should have reached a levelled off value and two successive measurements in
steady state should not differ more than ± 5%. The duration of this process for the insulating liquid humidity to reach
a steady state depends on the stirring speed of insulating liquid, temperature and the relative humidity of the air in the
chamber. The repetition of the procedure for at least three temperatures is necessary to derive the moisture saturation
coefficients.
3.3.2 Method-2
Method-2 uses both Karl Fischer and capacitive sensor measurements of the insulating liquids, conditioned in an airtight
test cell at three temperatures. This method is described in detail in [3.32].
3.3.2.1 Experimental Design
The main part of the measurement setup is a glass vessel with a volume of about one litre filled with insulating liquid to
be tested, as shown in Figure 3.7. The glass vessel should be airtight and sealed by the lid made of
polytetrafluoroethylene. Instead of glass and polytetrafluoroethylene other materials may be used which are
characterized by very low hygroscopicity and do not react chemically with the investigated liquid. A probe equipped with
capacitive and temperature sensors and a needle used for liquid sampling through the lid. During conditioning, the
insulating liquid should be mixed using a magnetic stirrer. Capacitive and temperature sensors should be placed directly
above the magnetic stirrer.
Figure 3.7 Sketch and photo of the test vessel used in Method-2

The glass vessel is filled with the insulating liquid to be tested with about 10% headspace to allow thermal expansion.
The conditioning process of the liquid should be carried out for at least three temperature levels chosen from the range
of 20°C to 60°C. The temperature difference between two sequent levels should be at least 15°C, whereas the measured
values of investigated liquid saturation RS, in above mentioned temperature range, should be contained from 75% (for
low temperature) to 15% (for high temperature). Therefore, it is necessary to initially moisten the liquid sample. It is
recommended to prepare the sample of relative saturation equal to 60% ± 15% at the temperature 20°C. During the
conditioning process, after achieving the preset insulating liquid temperature, the relative humidity (RH) of the air in
climatic chamber should be set close to the relative saturation RS of the insulating liquid to prevent water migration in
case of a leakage in the measurement setup. After achieving the set insulating liquid temperature, it is also
recommended to release the potential pressure built up in the vessel by using the needle valve. The time for conditioning
the liquid, at each temperature level, should be long enough that the setup can reach a state close to moisture and
temperature equilibrium. Usually, three hours conditioning time for each temperature level is enough. The time
necessary to achieve equilibrium strongly depends on many factors like: level of relative saturation of liquid, operation
of climatic chamber or measurement setup tightness. After equilibrium is achieved at each temperature, RS and
temperature of the liquid are recorded; liquid samples are taken and water content of liquid is measured by means of
Karl Fischer titration method.
63
3.3.3 Method-3
Method-3 also uses both Karl Fischer and capacitive sensor measurements of the insulating liquids with a simple setup
and a quick conditioning process.
3.3.3.1 Experimental Design
An aluminium made bottle (commercially available size of 1.2 litres) with a stirring bar is used as a test cell for Method
3, as shown in Figure 3.8. The insulating liquid is filled in the cell up to about 1 cm from the upper edge.
Figure 3.8 Photo of the test vessel used in Method-3

After filling with the liquid sample, the test cell is tightly closed and put into an air circulating oven at 80°C for
approximate 3 hours. The cell is then taken out of the oven and placed immediately on a magnetic agitator. The stirring
velocity should be on the lower end (e.g. 50 rpm). A probe equipped with capacitive and temperature sensors is fitted
in the test cell (Figure 3.8). The capacitive sensor head should be in the area of insulating liquid circulation. When the
temperature is reduced to the range between 60°C and 70°C, RS and temperature readings from the capacitive sensor
are recorded and a liquid sample for absolute moisture measurement using Karl Fischer titration is taken (three
measurements). The procedure is repeated when the temperature decreases to the range between 30°C to 40°C and
finally at room temperature. The measurements should be carried out at least at three temperatures.
3.4 MOISTURE SATURATION COEFFICIENTS OBTAINED FROM CIGRE RRT

This section reports findings from RRT. Table 3.7 - 3.8 and Table 3. show the mean values of moisture saturation A
and B coefficients, again with respect to the decimal logarithmic form in Equation 3.4. The insulating liquid types
considered as the same as the previous sections, which are new mineral oil, aged mineral oil, new synthetic ester, new
natural ester and new silicone insulating liquid. The moisture saturation curves are plotted in Figure 3.9 to Figure 3.13.
Note that the three measurement methods introduced in section 3.3 were employed. Method-3 has been used by
multiple labs, so mean and standard deviation are evaluated.
It is observed that Method-1 yields higher moisture saturation values than the other two methods. For new mineral oil
and new silicone insulating liquid, there is only a slight difference in the moisture saturation across different
measurement methods. As for the aged mineral oil, synthetic ester and natural ester, the difference in the evaluated
moisture saturation is significant, particularly for synthetic ester. Assuming that the accuracy of Karl Fischer
measurements is similar across the three methods; the greater difference across the three methods implies that the
need for achieving equilibrium is higher with increasing liquid moisture saturation (liquid type or ageing condition). This
could potentially be achieved by Method-1 that involves better exposure of the liquid to the controlled air environment
and longer conditioning duration, hence a better chance of truly achieving equilibrium.
Table 3.7: Moisture saturation coefficients for new mineral oil from CIGRE RRT
Moisture Saturation Coefficients
Method
Decimal logarithm equation (3.3) Natural logarithm equation (3.2)
64
Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev.
Method 1 7.31 - 1642 - 16.8 - 3782 -
Method 2 7.08 - 1577 - 16.3 - 3632 -
Method 3
6.89 0.58 1503 181 15.88 1.33 3461 416
(5labs)
Average 7.09 1574 16.33 3625
Figure 3.9: Water solubility (S) versus temperature for new mineral oil from RRT results
Table 3.8: Moisture saturation coefficients for service aged mineral oil from CIGRE RRT
Method
Method 1 7.37 - 1636 - 16.97 - 3768 -
Method 2 6.90 - 1508 - 15.89 - 3473 -
Method 3
6.60 0.52 1379 183 15.20 1.20 3174 422
(5labs)
Average 6.96 1508 16.03 3473
The condition of the service aged sample is presented in Table 3.9.
Table 3.9 Condition (oil values) of the service aged mineral oil sample
Parameter Color Acidity Dissipation factor Interfacial tension

at 90°C
Value 3,5 0,03 mg KOH/g oil 0,058 22 mN/m
Standard ISO 2049 IEC 62021-2 IEC 60247 EN 14210
65
Figure 3.10: Water solubility (S) versus temperature for service aged mineral oil from RRT results
Table 3.10: Moisture saturation coefficients for new synthetic ester from CIGRE RRT
Method
Method 1 5.96 - 768 - 13.73 - 1769 -
Method 2 5.32 - 608 - 12.25 - 1400 -
Method 3
4.98 0.42 511 120 11.46 0.97 1177 276
(5labs)
Average 5.42 629 12.48 1449
Figure 3.11: Water solubility (S) versus temperature for new synthetic ester from RRT results
Table 3.11: Moisture saturation coefficients for new natural ester from CIGRE RRT

Method
Method 1 5.54 - 739 - 12.76 - 1702 -
66
Method 2 5.34 - 707 - 12.30 - 1628 -

Method 3
5.10 0.36 634 114 11.80 0.86 1460 261
(5labs)
Average 5.33 693 12.27 1596
Figure 3.12: Water solubility (S) versus temperature for new natural ester from RRT results
Table 3.12: Moisture saturation coefficients for new silicone liquid from CIGRE RRT
Method
Method 1 5.76 - 1005 - 13.27 - 2315 -
Method 2 6.30 - 1194 - 14.51 - 2750 -
Method 3
5.73 0.24 1008 76 13.19 0.56 2321 175
(5labs)
Average 5.93 1069 13.66 2462
Figure 3.13: Water solubility (S) versus temperature for new silicone liquid from RRT results
Figure 3.14 compares the moisture saturation characteristics of the different insulating liquid types as recorded from
literature and CIGRE RRT. The moisture saturation curves have been calculated from the average values of A and B
67
coefficients. As a common condition cannot be used as a basis for aged mineral oil results from literature, it will be
omitted here. From Figure 3.14, the difference between the literature and CIGRE RRT curves for new or unaged liquids
is minimal, except for the silicone insulating liquid. This could be due to the lack of literature results. Other than that,
it is again evident that moisture saturation is influenced by insulating liquid type, with synthetic ester having the highest
moisture saturation, followed by natural ester, silicone insulating liquid and mineral oil. In addition, ageing also increases
the moisture saturation, as illustrated by the higher moisture saturation of the CIGRE RRT aged mineral oil curve
compared with the CIGRE RRT new mineral oil curve.
Figure 3.14: Water solubility (S) versus temperature for results from literature and CIGRE RRT
3.5 SUMMARY OF CHAPTER 3

1. Cigre WG D1.52 evaluated several methods for the calculating of A and B coefficients which will theoretically
allow the conversion of Karl Fischer values in mg/kg into % saturation and vice versa. Statistical evaluation of
the results shows that the uncertainty of measurement in case of Karl Fischer and capacitive sensors will reflect
also on the uncertainty of A and B values. In addition, impact of different measurement procedures
(determination in transient or equilibrium state) would bring extra uncertainties. Reproducible A and B values
can be derived only with sensors showing no temperature dependence, having a quick response and an
adequate sensitivity ≤ 5 % RS and being compatible with the respective fluid being measured.
2. Meaningful A and B values can be derived only with sensors showing no temperature dependence, having a
quick response and an adequate sensitivity ≤5% and compatible in the corresponding fluid (see chapter 1.3).
3. In the measurement procedure for determination of saturation coefficients it is recommended that at least three
temperatures with adequate intervals should be chosen and the RS condition should be determined in the range
of 5% RS to 75% RS.
4. The comparison of the water solubility of new and aged mineral oil over the usual temperature interval shows
that aged mineral oils have higher saturation levels than new mineral oils.
5. Temperature dependence of saturation levels are different for different types of insulating liquids. Synthetic
ester>natural ester>silicone insulating liquid>mineral oil

[3.1] S. Tee, Q. Liu, Z. Wang, G. Wilson, P. Jarman, R. Hooton, et al., "Seasonal Influence on Moisture Interpretation for
Transformer Ageing Assessment," Electrical Insulation Magazine, IEEE, 2016.
[3.2] "IEC 60422 Mineral insulating oils in electrical equipment - Supervision and maintenance guidance (Edition 4.0)," ed:
International Electrotechnical Commission - Fluids for Electrotechnical Applications Technical Committee, 2013, p. 93.
[3.3] Y. Du, M. Zahn, B. C. Lesieutre, A. V. Mamishev, and S. R. Lindgren, "Moisture equilibrium in transformer paper-oil systems,"
Electrical Insulation Magazine, IEEE, vol. 15, pp. 11-20, 1999.
[3.4] Y. Du, A. V. Mamishev, B. C. Lesieutre, M. Zahn, and S. H. Kang, "Moisture solubility for differently conditioned transformer
oils," Dielectrics and Electrical Insulation, IEEE Transactions on, vol. 8, pp. 805-811, 2001.
[3.5] K. B. Liland, "Description of a method for finding the water solubility in insulation liquids," ed: SINTEF, 2014.
68
[3.6] L. Greenspan, "Humidity fixed points of binary saturated aqueous solutions," Journal of research of the national bureau of
standards, vol. 81, pp. 89-96, 1977.
[3.7] A. Shkolnik, "Determination of water content in transformer insulation," in Proceedings of 2002 IEEE 14th International
Conference on Dielectric Liquids. ICDL 2002 (Cat. No.02CH37319), 2002, pp. 337-340.
[3.8] P. J. Griffin, V. Sokolov, and B. Vanin, "Moisture equilibrium and moisture migration within transformer insulation systems,"
2004.
[3.9] P. Przybylek, "Water saturation limit of insulating liquids and hygroscopicity of cellulose in aspect of moisture determination
in oil-paper insulation," IEEE Transactions on Dielectrics and Electrical Insulation, vol. 23, pp. 1886-1893, 2016.
[3.10] T. V. Oommen, "Moisture equilibrium in paper-oil insulation systems," in 1983 EIC 6th Electrical/Electronical Insulation
Conference, 1983, pp. 162-166.
[3.11] P. J. Griffin, C. Bruce, and C. J., "Comparison of water equilibrium in silicone and mineral oil transformers," vol. 10-9, ed:
Minutes of the Fifty-Fifth Annual International Conference of Doble Clients, 1988.
[3.12] "Shell Diala Oils," vol. Shell Lubricants Technical Bulletin SOC: 39-92, ed. Shell Oil Company: One Shell Plaza, 900 Louisiana
Street, Houston, Texas 77002, (800)23 1-6950.
[3.13] J. Li, Z. Zhang, S. Grzybowski, and M. Zahn, "A new mathematical model of moisture equilibrium in mineral and vegetable
oil-paper insulation," IEEE Transactions on Dielectrics and Electrical Insulation, vol. 19, pp. 1615-1622, 2012.
[3.14] V. G. Aralkellian and I. Fofana, "Water in Oil-Filled, High-Voltage Equipment, Part I: States, Solubility, and Equilibrium in
Insulating Materials," Electrical Insulation Magazine, IEEE, vol. 23, pp. 15-27, 2007.
[3.15] R. Villarroel, D. F. Garcia, B. Garcia, and J. C. Burgos, "Diffusion coefficient in transformer pressboard insulation part 2:
mineral oil impregnated," IEEE Transactions on Dielectrics and Electrical Insulation, vol. 21, pp. 394-402, 2014.
[3.16] "Vaisala HUMICAP® Moisture and Temperature Transmitter for Oil MMT330," V. Oyj, Ed., ed, 2015.
[3.17] "Hydran M2 Instruction Manual," G. I. Power, Ed., ed.
[3.18] M. Koch, S. Tenbohlen, and T. Stirl, "Diagnostic Application of Moisture Equilibrium for Power Transformers," Power Delivery,
IEEE Transactions on, vol. 25, pp. 2574-2581, 2010.
[3.19] T. Gradnik, B. Čuček, and M. Končan-Gradnik, "Temperature and chemical impact on determination of water content in
dielectric liquids by capacitive moisture sensors," in 2014 IEEE 18th International Conference on Dielectric Liquids (ICDL) ,
2014, pp. 1-5.
[3.20] D. Martin, T. Saha, C. Perkasa, N. Lelekakis, and T. Gradnik, "Fundamental concepts of using water activity probes to assess
transformer insulation water content," IEEE Electrical Insulation Magazine, vol. 32, pp. 9-16, 2016.
[3.21] M. Jovalekic, D. Kolb, S. Tenbohlen, L. Bates, and R. Szewczyk, "A methodology for determining water saturation limits and
moisture equilibrium diagrams of alternative insulation systems," in 2011 IEEE International Conference on Dielectric Liquids,
2011, pp. 1-5.
[3.22] D. Gan, F. Liu, L. Du, and Y. Liu, "Analysis of water solubility in transformer oil using least square fitting method," in 2010
International Conference on High Voltage Engineering and Application, 2010, pp. 93-95.
[3.23] T. Gradnik, M. Koncan-Gradnik, N. Petric, and N. Muc, "Experimental evaluation of water content determination in
transformer oil by moisture sensor," in 2011 IEEE International Conference on Dielectric Liquids, 2011, pp. 1-4.
[3.24] E. Mladenov, S. Staykov, and G. Cholakov, "Water saturation limit of transformer oils," IEEE Electrical Insulation Magazine,
vol. 25, pp. 23-30, 2009.
[3.25] L. E. Lundgaard, W. Hansen, and S. Ingebrigtsen, "Ageing of Mineral Oil Impregnated Cellulose by Acid Catalysis," Dielectrics
and Electrical Insulation, IEEE Transactions on, vol. 15, pp. 540-546, 2008.
[3.26] P. Przybylek, (2012) Water solubility in insulating liquids in aspect of moisture investigation using capacitive probe. Przeglad
Elektrotechniczny.
[3.27] R. Villarroel, D. F. Garcia, B. Garcia, et al., "Moisture diffusion coefficients of transformer pressboard insulation impregnated
with natural esters," Dielectrics and Electrical Insulation, IEEE Transactions on, vol. 22, pp. 581-589, 2015.
[3.28] "IEEE Guide for Acceptance and Maintenance of Natural Ester Fluids in Transformers," IEEE Std C57.147-2008, pp. 1-31,
2008.
[3.29] I. Fofana, V. Wasserberg, H. Borsi, and E. Gockenbach, "Challenge of mixed insulating liquids for use in high-voltage
transformers.1. Investigation of mixed liquids," IEEE Electrical Insulation Magazine, vol. 18, pp. 18-31, 2002.
[3.30] S. Tenbohlen and M. Koch, "Aging Performance and Moisture Solubility of Vegetable Oils for Power Transformers," Power
Delivery, IEEE Transactions on, vol. 25, pp. 825-830, 2010.
[3.31] Ch. Forney, D. Brandl, "Control of Humidity in small controlled-environment chambers using glycerol-water solutions",
HortTechnology, January-March, 1992, 2(1), 52-55.
[3.32] P. Przybylek, K. Siodla, "Application of capacitive sensor for measuring water content in electro-insulating liquids",
Esploatacja i Niezawodnosc - Maintenance and Reliability, vol. 18, No. 2, 181-185, 2016.
69
70
4. INDIRECT METHODS OF MOISTURE DETERMINATION IN THE

SOLID INSULATION
4.1 METHODS OF DETERMINATION OF WATER CONTENT IN SOLID INSULATION
4.1.1 Background
The most popular methods of moisture content determination can be divided into two groups. The first group is based
on the dielectric response. The second group of methods are based on the water content measurement in electro-
insulating liquid. The moisture of electro-insulating liquid is measured by means of a capacitive sensor (CS) or Karl
Fischer titration method (KFT). On the basis of water content in electro-insulating liquid, temperature of insulation and
by means of water sorption isotherms or moisture equilibrium curves it is possible to determine the water content in
solid insulation expressed in percent by weight.
Recently a laboratory specimen of a device for continuous monitoring of moisture content in oil-paper insulation of high-
voltage oil-filled electrical equipment has been developed [4.1]. The humidity sensor is mounted directly at the oil-filled
HV electrical equipment. Sensor operation depends on the humidity of the capacitor, the dielectric plate of the capacitor
being made of cellulose materials.
Both groups of methods, based on dielectric response and based on moisture equilibrium, have their advantages and
restrictions.
4.1.2 Methods based on dielectric response

4.1.2.1 Polarization and depolarization current (PDC)
Polarization and Depolarization Current (PDC) measurement is one of the time-domain based dielectric measurement
methods. When a DC voltage is applied across an insulation material, the current flowing through the dielectric on DC
excitation is expressed as the sum of the polarization current and the conduction current. Polarization current
characterizes the adjustment of the polarizing species to the applied field and decreases to zero after infinitely long
time. The conduction current is due to the movement of free charges across the dielectric and has a constant value as
long as the dielectric is stressed [4.2]. By removing the step-voltage source and short-circuiting the insulation material,
depolarization current appears as negative values and decreases towards zero. Figure 4.1 shows an example of the
polarization and depolarization current measurements. The polarization and depolarization current and the difference
between polarization and depolarization current curves at long measuring times are related to the conductivity of the
paper/pressboard part of the insulation system and its water content [4.3].
Figure 4.1 An example of PDC measurement results [4.3]
4.1.2.2 Recovery voltage measurement (RVM)

Recovery Voltage Measurement (RVM) is also a time domain based dielectric measurement method. RVM is based on
established knowledge of polarization of oil-paper insulation. Figure 4.2 demonstrates a typical measurement procedure
[4.4].
71
Figure 4.2 Sketch of RVM measurement (test object is charged with Uo from 0  t  t1, grounded from t1  t  t2 and
for t  t2 the recovery voltage is measured at open circuit condition) [4.4]
A step voltage U0 is applied over a completely discharged test object with geometric capacitance C0. During the charging
period the polarisation current, ip(t), flows through the test object. After the charging period, the test object is short-
circuited (grounded) for a certain time (usually shorter than the charging time) and the depolarisation current id(t)
flows. Then the grounding is removed, and the recovery voltage UR(t) is measured under open-circuit conditions [4.3].
The source of the recovery voltage is the relaxation processes inside the dielectric material, giving rise to an induced
charge on the electrodes of the test object.
The polarization spectrum is often used, which is established by performing a series of recovery voltage measurements
with stepwise increasing charging time tc and short-circuit time tg, usually with the ratio tc/tg=2. The peak recovery
voltage URmax and the initial rate of rise of the recovery voltage dUR/dt are recorded and plotted versus the charging
time used [4.5]. The moisture content in the oil/paper insulation would influence the polarisation phenomena and hence
the recovery voltage. By comparing the recovery voltage or polarization spectrum, the insulation conditions e.g. different
level of moisture are indicated.
4.1.2.3 Frequency domain spectroscopy
In addition to the time domain methods, an equivalent method in the frequency domain, called Frequency Domain
Spectroscopy (FDS) or Dielectric Frequency Response (DFR), is used to investigate the polarization responses from
sinusoidal excitations at different frequencies. When an AC sinusoidal voltage U(ω) is applied, the complex impedance is
measured. In this method, imaginary and real parts of the capacitance and dielectric dissipation factor (tanδ) of an
insulation system are estimated over a wide frequency range [4.6]. The advantage of using tanδ is that it is independent
of the test object geometry. Figure 4.3 shows the typical features of FDS measurement results and the common
interpretation of the results in the different frequency bands [4.4].
Figure 4.3 An example of FDS measurement results [4.3]
72
As shown in Figure 4.3, dielectric methods are not only influenced by moisture but also by different factors – these can
be ageing and conductivity of oil and solid insulation, insulation geometry and temperature. The evaluation is based on
models and algorithms.
4.1.3 Indirect methods based on the water content measurement in electro-insulating liquid
The indirect methods are based on the water content measurement in electro-insulating liquid. Using the water content
in liquid, temperature of insulation and by means of water sorption isotherms or moisture equilibrium curves, it is
possible to determine the water content in the solid insulation.
Water sorption isotherms are characteristics for different temperature values which present the dependence of water
content in cellulose material as a function of relative humidity of the air (Figure 4.5). If a capacitive sensor is installed
in the transformer, relative saturation of the liquid and its temperature are measured, and by means of water sorption
isotherms, water content in solid insulation can be determined.
In turn, moisture equilibrium curves are a group of characteristics describing the dependence of water content in
cellulose material as a function of water content in electro-insulating liquid, expressed in ppm by weight (Figure 4.6).
They are prepared for different temperatures. The water content in liquid is determined by means of the Karl Fischer
titration method. It is also possible to convert the relative saturation of liquid measured with the use of capacitive sensor
into the water content expressed in ppm by weight. To determine the water content in cellulose insulation it is necessary
to know insulation temperature and use appropriate moisture equilibrium curves.
Moisture equilibrium curves can be constructed by a direct method – storing the paper and oil at various moisture values
and measuring the moisture in paper and oil or indirectly by combination of two curves – oil and air % relative humidity
(Figure 4.4) with paper and air % relative humidity (Figure 4.5). A combination of these curves results in a
representation (Oommen curves, Figure 4.6) of moisture in paper (%) versus moisture in oil (ppm).
Figure 4.4 Dependence between moisture in new oil Figure 4.5 Dependence between % moisture in
(ppm) and % relative humidity in air [4.7] paper and % relative humidity in air [4.8]
73
Figure 4.6 Moisture equilibrium curves [4.9]
4.2 RESTRICTIONS IN THE APPLICATION OF INDIRECT METHODS

The application of the moisture equilibrium curves and water sorption isotherms requires consideration of the following:
 kind and chemical structure of solid and liquid insulation
 ageing condition of liquid insulation
 ageing condition of solid insulation
 moisture equilibrium state
 distribution of moisture in the solid insulation
 temperature of insulation
4.2.1 Kind and chemical structure of solid and liquid insulation
In power transformer insulation two kinds of paper are used – cellulose and aramid. Both materials have quite different
hygroscopicity. The measurement results, presented in [4.10, 4.11] indicate that the hygroscopicity of cellulose is higher
than aramid material. Therefore, water sorption isotherms and moisture equilibrium curves will be different for cellulose
and aramid materials.
The shape of moisture equilibrium curves is influenced also by other parameters, like kind of insulating liquid and its
chemical structure. Water solubility may be significantly different in various electro-insulating liquids. The lowest water
solubility oil is mineral and silicon oil and is much higher in natural and synthetic esters [4.12, 4.13, 4.14]. Different
water solubility can be explained by different polarity of molecules of these liquids. Different solubility of water affects
the shape of moisture equilibrium curves. In Figure 4.7 are compared the moisture equilibrium curves for new cellulosic
paper impregnated with different electro-insulating liquids at 50C. Figure 4.8 shows a comparison of moisture
equilibrium curves for mineral oil and natural ester over a wide temperature range.
Figure 4.7 Comparison of moisture equilibrium curves for new cellulose paper impregnated with different electro-
insulating liquids at 50C; figure prepared using data presented in [4.14]
74
Figure 4.8 Comparison of moisture equilibrium curves for paper and pressboard with mineral oil and with ester oil
[4.15]
4.2.2 Ageing condition of liquid insulation
The aging of oil-paper insulation causes an increase in polar acidic products in the electro-insulating liquid. These
products are responsible for an increase in water solubility. Some experience shows that aged oils can absorb up to
three times much water than new oils. Examples of the influence of ageing on mineral oil moisture equilibrium curves
are shown in Figure 4.9 and Figure 4.10.
Figure 4.9 Moisture equilibrium curves for new and aged mineral oil [4.16]
Figure 4.10 Moisture equilibrium curves for cellulose of polymerisation degree equal to 464, impregnated with new
and aged mineral oil at 50C; figure prepared using data presented in papers [4.14, 4.17]
75
4.2.3 Ageing condition of solid insulation

As a result of aging processes, there are structural changes of the cellulose fibers and molecules. In the process of
water adsorption the degree of cellulose crystallinity and the number of free hydroxyl groups play an important role.
Both of these parameters change as a result of aging taking place in cellulose [4.14, 4.18]. During thermal aging, the
OH groups of cellulose chains become attracted to one another and are less able to attract water molecules [4.19]. It
was shown in [4.20, 4.21] that the aging process of cellulose reduces its hygroscopicity. Figure 4.11 presents the water
sorption isotherms for cellulose of different DP obtained for temperature equal to 50C. In turn, in Figure 4.12 moisture
equilibrium curves obtained for cellulose with different polymerisation degree are compared.
Figure 4.11 Water sorption isotherms Figure 4.12 Moisture equilibrium curves of cellulose of
of cellulose of different polymerisation different polymerisation degree, impregnated with new
degree at 50C [4.14] mineral oil at 50C; figure prepared using data presented in
[4.14]
4.2.4 Moisture equilibrium state

In order to correctlty determine the moisture content in cellulose insulation by means of indirect methods, it is necessary
to achieve equilibrium of moisture in oil-paper insulation. This occurs when the insulation temperature does not vary
significantly for a long time. An increase in insulation temperature causes water migration from paper to oil, whereas
its decrease results in the opposite. Water migration in transformers in service is caused by:
 load changes of the transformer,
 ambient temperature changes,
 non-uniform temperature distribution inside the insulation.
Mmoisture equilibrium in oil-paper insulation of a transformer is difficult to obtain [4.22].
76
80 1 2 3 4 5
70
60
Humidity
°C ;%Rel.Feuchte
50
Temp. Inlet [°C]
rel. Moist. Inlet [%]
40 Temp. Outlet [°C]
rel. Moist. Out [%]
30
Test period: lower
Versuchsphase:
20 temperature; bypass
niedrige Temp., Bypass
10
0
18.07.2003
28.07.2003
07.08.2003
17.08.2003
27.08.2003
06.09.2003
16.09.2003
26.09.2003
16.10.2003
08.07.2003
06.10.2003
Date
Datum
Figure 4.13 Oil drying on a 16 MVA transformer by means of molsieve cartridges [4.27]
Figure 4.13 shows a practical drying example where moisture equilibrium requires a long time to be established, if at
all.
Phase 1 and 2: Cartridge with molsieve are used. The relative moisture content at the outlet has decreased
considerably.
Phase 3: Temperature decrease causes a decrease in relative moisture saturation in oil, since the equilibrium is in favor
of solid insulation in paper.
Phase 4: The cartridge has been bypassed. This results in the same temperature in inlet and outlet. Although the oil
temperature is high enough, it takes about 4 weeks until a relatively steady moisture reading is obtained.
Phase 5: The molsieve cartridge is put into operation again.
4.2.5 Distribution of moisture in the solid insulation

It can be found in the literature [4.23, 4.24] that the decisive factor affecting moisture distribution is temperature
distribution in the insulation system. For regions with higher temperature (the upper part of the transformer tank), the
water content in the cellulose insulation is lower compared to regions with lower temperature (the lower part of the
tank). The decrease of the amount of the water stored in cellulose insulation accompanied by temperature increases
results from the value of hydrogen bond energy compared to the energy of cellulose particle vibration. The increase in
temperature causes the increase in vibration energy and this reduces the probability of water particle addition to free
hydroxyl groups of a cellulose molecule and increases the probability of release of water which was bonded [4.25].
It should be noted that not only temperature but also polymerisation degree changes hygroscopicity of cellulose (Figure
4.11). Both factors may significantly affect the moisture distribution in a large power transformer, as shown in [4.26].
In this experiment, papers of different polymerisation degree were investigated. The polymerisation degree of new
paper was equal to 1360 and for degraded paper was equal to 670, 464, and 272. Three sets of samples were prepared
for the research. Each set contained four paper samples with different polymerisation degrees. All samples were dried
and then were immediately impregnated with new mineral oil. The sample sets prepared in this way were wound onto
a heater, each set at a different height (Figure 4.14). Due to convection, each set was exposed to different temperature.
Moreover, two strips of moistened paper were wound onto the heater. The water content in the moistened paper (WCP)
was 6.2%. The system was conditioned to achieve moisture equilibrium between oil and paper wound onto the heater.
Moisture, through the oil, migrated from the moistened paper to the samples dried before. This procedure lasted 32
days with the goal of achieving moisture equilibrium. The water content in all samples was determined by means of
Karl Fischer method. Results of water content measurements are presented in Figure 4.15.
77
Figure 4.15 Water content in paper samples after conditioning in

function of cellulose polymerisation degree for three different
temperature values [4.26]
Figure 4.14 Placement of paper
samples with different polymerisation
degrees DP [4.26]
For example, consider when the top part of the test system had a paper DP of 464 and compare that to the un aged
lower part with DP of 1360, this shows a moisture difference of 3.03%. On the other hand, the moisture difference
between top and bottom for new samples was 2.16%. This example shows how cellulose aging can enhance moisture
non-uniformity [4.26].
In paper [4.27] the moisture distribution in a 31.5 MVA, 110/6.6 kV transformer (Figure 4.16) with oil-directed (OD)
cooling system is shown. The authors of this paper took samples of cellulose from transformer insulation to measure
its water content (Figure 4.17) by means of Karl Fischer titration method (KFT). It should be understood, that the results
may be overstated, due to the time taken for sampling with an open transformer.
Figure 4.16 31,5 MVA Figure 4.17 Results of KFT analysis

transformer [4.27] and position of the samples [4.27]
The results obtained in laboratory models and real transformers illustrate how much the values between the highest
and lowest moisture content in insulating paper can vary. These values can considerably differ from mean values of
insulation moisture determined by means of indirect methods.
78
4.3 CONCLUSIONS TO CHAPTER 4

 Moisture migration between liquid and solid insulation is a dynamic process, dependent on the type and ageing
of the insulation, and on temperature distribution etc. A reliable evaluation of moisture can be performed in
the liquid insulation. By means of indirect methods an evaluation of moisture in the solid insulation can be done
under certain restrictions, e.g. kind and chemical structure of solid and liquid insulation, ageing condition of
liquid and solid insulation, moisture equilibrium state in oil and paper insulation, distribution of moisture in the
solid insulation etc.
 The application of capacitive sensor and water sorption isotherms for evaluation of moisture level of solid
insulation is more reasonable than using the equilibrium curves due to at least two reasons. The first one is
that by means of the capacitive sensor, changes of moisture and temperature over time are monitored. Thus it
is possible to determine the time when the insulating system is at or close to equilibrium. The other reason is
that it is not necessary to include the type of liquid, its chemical structure or ageing state as in the case of
moisture equilibrium curves. In both cases, for water sorption isotherms and moisture equilibrium curves,
cellulose aging should be considered.
 A important parameter for determining serviceability of a liquid insulation is breakdown voltage. The relationship
between relative moisture saturation to breakdown voltage, which is independent of the type and ageing status
of the liquid, allows the evaluation of safety margins.
[4.1] L.A. Darian, A.A. Kozlov, A.V.Kozlov, V.P. Polistchook, A.V. Shurupov, N.A. Matveev, “Diagnostic
control of oil-paper insulation based on method of direct measurement of paper moisture content”,
D1-302, CIGRE 2016
[4.2] A. Kumar and S. Mahajan, “Time domain spectroscopy measurements for the insulation diagnosis
of a current transformer”, IEEE Transactions on Dielectrics and Electrical Insulation, Vol. 18, pp.
1803-1811, 2011.
[4.3] Brochure CIGRE 414 WG D1.01, Dielectric Response Diagnoses for Transformer Windings,
2010.
[4.4] Brochure CIGRE 254 WG A2.18, Dielectric Response Methods for Diagnostics of Power
Transformers, 2004.
[4.5] S. Gubanski, P. Boss, G. Csepes, V. Der Houhanessian, J.F.P. Guuinic, U. Gäfvert, V. Karius, J.
Lapworth, G. Urbani, P. Werelius and others, “Dielectric response methods for diagnostics of
power transformers”, IEEE Electrical Insulation Magazine, Vol. 19, pp. 12-18, 2003.
[4.6] A. Setayeshmehr, I. Fofana, C. Eichler, A. Akbari, H. Borsi, and E. Gockenbach, “Dielectric

spectroscopic measurements on transformer oil-paper insulation under controlled laboratory
conditions”, IEEE Transactions on Dielectrics and Electrical Insulation, Vol. 15., pp. 1100-1111,
2008.
[4.7] T. V. Oommen, J. Thompson and B. Ward, Moisture estimation in transformer insulation, Panel
Session IEEE Transformer Committee, March 2004.
[4.8] R. Jeffries, “The sorption of water by cellulose and eight other textile polymers”,
J. Textile Institute Trans., Vol. 51, No. 9, pp. 339-374, 1960.
[4.9] T.V. Oommen, T.A. Prevost, “Cellulose insulation in oil-filled power transformers: Part II –
Maintaining insulation integrity and life”, IEEE Electrical Insulation Magazine, Vol. 22, No. 2, pp.
5-14, 2006.
[4.10] M. Jovalekic, D. Kolb, S. Tenbohlen, L. Bates, R. Szewczyk, “A methodology for determining

water saturation limits and moisture equilibrium diagrams of alternative insulation systems”, in
Proc. of IEEE International Conference on Dielectric Liquids (ICDL), pp. 1-5, 2011.
[4.11] P. Przybylek, “A comparison of bubble evolution temperature in aramid and cellulose paper”, in
Proc. of 2013 IEEE International Conference on Solid Dielectrics, Bologna, Italy, pp. 983-986,
2013.
79
[4.12] CIGRE Brochure 436, “Experiences in service with new insulating liquids”, 2010.
[4.13] I. Fofana, V.G. Arakelian, “Water in oil-filled high-voltage equipment, Part I: States, solubility, and
equilibrium in insulating materials”, IEEE Electrical Insulation Magazine, Vol. 23, Issue 4, pp. 15-
27, 2007.
[4.14] P. Przybylek, “Water saturation limit of insulating liquids and hygroscopicity of cellulose in aspect
of moisture determination in oil-paper insulation”, IEEE Transactions on Dielectrics and Electrical
Insulation, Vol. 23, No. 3, pp. 1886-1893, 2016.
[4.15] V. Vasovic, J. Lukic, C. Perrier, M.-L. Coulibaly, “Equilibrium charts for moisture in paper and
pressboard insulations in mineral and natural ester transformer oils”, IEEE Electrical Insulation
Magazine, Vol. 30, No. 2, pp. 10-16, 2014.
[4.16] I. Atanasova-Hoehlein, U. Thiess, “Moisture evaluation in transformers”, Elektrizitätswirtschaft,

Vol. 103, Nr. 25, 2004 pp- 57-61.
[4.17] A. Shkolnik, “Determination of water content in transformer insulation”, in Proc. of 14th

International Conference on Dielectric Liquids (ICDL 2002), pp. 337-340, 2002.
[4.18] S. Prosinski, “Wood chemistry”, Państwowe Wydawnictwo Rolnicze i Leśne, Warsaw, 1984 (in
Polish).
[4.19] M. Koch, S. Tenbohlen, T. Stirl, “Advanced online moisture measurements in power

transformers”, in Proc. of International Conference on Condition Monitoring and Diagnosis, CMD,
2006.
[4.20] R. Liao, Y. Lin, P. Guo, H. Liu, H. Xia, “Thermal aging effects on the moisture equilibrium curves
of mineral and mixed oil-paper insulation systems”, IEEE Transactions on Dielectrics and
Electrical Insulation, Vol. 22, pp. 842-850, 2015.
[4.21] P. Przybylek, “The influence of cellulose insulation aging degree on its water sorption properties
and bubble evolution”, IEEE Transactions on Dielectrics and Electrical Insulation, Vol. 17, No. 3,
pp. 906-912, 2010.
[4.22] IEEE Std C57.152-2013, “IEEE Guide for Diagnostic Field Testing of Fluid-Filled Power
Transformers, Regulators, and Reactors”.
[4.23] FIST 3-30, “Transformer Maintenance”, United States Department of the Interior Bureau of
Reclamation, Denver, Colorado, October 2000.
[4.24] Cigre Brochure 349 “Moisture equilibrium and moisture migration within Transformer insulation
systems”, 2008.
[4.25] J. Slowikowski, “Factors influencing proper cellulose insulation moisture’s level determination
based on measurement contents of water dissolved in oil in paper-oil transformer”, Przeglad
Elektrotechniczny - Electrical Review, No. 1k, Vol. 2003, pp. 195-198, 2003 (in Polish).
[4.26] P. Przybylek, „The influence of temperature and aging of cellulose on water distribution in oil-
paper insulation”, IEEE Transactions on Dielectrics and Electrical Insulation, Vol. 20, Issue 2, pp.
552-556, 2013.
[4.27] K. Walczak, A. Graczkowski, J. Gielniak, S. Gubanski, H. Moscicka-Grzesiak, “Assessment of

insulation state of power transformer after 35 years of operation using different diagnostic
methods”, in Proc. XIVth International Symposium on High Voltage Engineering, Beijing, China,
paper 422, G-086, 2005.
80
5. APPLICATION OF THIN FILM POLYMER CAPACITIVE

SENSORS FOR ON-LINE MONITORING OF OIL FILLED
ELECTRICAL EQUIPMENT
Moisture dramatically decreases the dielectric strength of insulating liquids and accelerates ageing of solid insulation.
On-line continuous moisture monitoring has become possible with the introduction of thin film capacitive sensors. They
are used as
- a part of a complex monitoring systems or
- stand alone devices with datalogger
Thin film polymer capacitive sensors can measure moisture in insulating liquid, in terms of RS at the corresponding
temperature.
In some monitoring systems they are integrated in a temperated environment and can deliver calculated values for
moisture using a moisture model. In the case of a standalone device installed in the oil stream, a thin film polymer
sensor delivers realtime data on the RS of the insulating liquid and the corresponding temperature. This allows the
assessment of changes in moisture with load and cooling. The evaluation of the resulting hysteresis curves enables an
understanding of the dynamics of moisture in the insulating liquid and its temperature dependent distribution between
solid and liquid insulation.
5.1 WHERE TO INSTALL A CAPACITIVE SENSOR
When selecting a location for a moisture sensor, one has to consider the purpose of the measurement. For example, is
it to interpret paper moisture or to see whether there’s high relative moisture saturation in the oil system. Figure 5.1
shows examples of installations where the sensor is inserted in the oil flow. Whatever the target is, it is important to
confirm that the oil measured is truly representative of the oil inside the transformer tank. The cooling circulation pipe
after radiators/coolers is an optimal place to assess the risk of decreased dielectric strength due to high relative moisture
saturation of oil. The use of moisture sensors in the flow of hot and cold oil will provide additional useful information.
However, often it is the position of valves that determines where sensors are located. It is crucial, that the sensor is
positioned in the oil flow, especially in case of bypass installations. Figure 5.2 shows the installation of a capacitive
moisture sensor in a 300 MVA, 380 kV transformer, filled with a natural ester.
Figure 5.1 Examples of sensor locations where the sensor have been inserted into oil flow
81
Figure 5.2 Installation of a capacitive mositure sensor at a 300 MVA 380/110 kV

transformer filled with natural ester
Avoiding stagnant oil is important, as only the oil at the surface of the sensor is measured. For sensors that are in
stagnant locations the response is strongly governed by the water diffusion charateristics of oil. This is a very slow
process i.e. hours or even days, depending on temperature (Figure 5.3) [5.1]. The actual diffusion rate of water
molecules in oil, is also affected not only by the distance but also the cross sectional area of the pipe in which the
sensor is located.
Figure 5.3 Example of time taken to reach representative

moisture readings with sensor distance from oil flow
As temperature fluctuations occur in service, so do the readings of moisture in oil. If the sensor is not in good oil flow,
then the extremes of the moisture fluctuations may not be visible, compromising the utility of the measurements.
Sometimes there have to be compromises when installing a moisture sensor, simply because there are not any suitable
connection points (e.g. valves) in the optimal locations of the transformer. Adding new connectors/valves may not be
feasible other than during a comprehensive maintenance. A drain valve should be used as the last option, as the oil is
likely not representative. Even installation into a radiator is a better option than in a drain valve.
82
5.2 EXPERIENCE FOR MOISTURE EVALUATION WITH CAPACITIVE SENSORS

5.2.1 Drying of oil
An on-line moisture monitoring system was used during oil reconditioning. Figure 5.4 shows the sensor readings with
time during the process [5.2].
Oil reconditioning: The oil flows from; oil feed pump, moisture sensor, heat exchanger, vacuum chamber 1, vacuum
chamber 2, particle filter and then to the second oil feed pump.
In Figre 5.4, the reconditioning has been divided into three stages, separated by two solid horizontal lines. In stage 1
the oil was pumped from tank 1 to tank 2 via the reconditioning plant. All measured values are more or less stable,
which shows that the oil in tank 1 is rather homogeneous. In stage 2 the reconditioned oil from tank 2 is fed to the
reconditioning process, therefore, the difference between tank 1 and tank 2 becomes visible. A steep decline in aw
values and a steep increase in temperature is observed, while the measured ppm values are more or less stable. At
stage 3 the oil is recirculated in tank 2. We observe a continuous decrease in aw values and an increase in temperature.
However, the rate of change is decreasing as a function of time. Additionally, we see a small decrease in calculated
ppm values. To summarise, on-line moisture measurements during transformer oil reconditioning provides valuable
information, enabling optimization of the process time and monitoring process quality values.
Figure 5.4 Moisture sensor readings as a function of time during oil reconditioning process of transformer oil [5.1].
5.2.2 Knowledge on seasonal moisture variation with temperature

Figure 5.5 shows five years of on-line data for an industrial rectifier transformer (22 MVA, 13,8kV/300V water cooled)
with stable load. The top oil temperature as well as the ppm values show a clear seasonal moisture variation [5.3].
83
Figure 5.5 Seasonal moisture variation with temperature of 22 MVA water cooled rectifier transformer [5.2]
Similar results have been recorded with a 20 MVA, 65 kV transformer, manufactured in 2007. Which shows the seasonal
dependence of water content, since the oil temperature will be influenced by the ambient temperature in ONAN mode
(Figure 5.6).
Figure 5.6 Seasonal moisture variation with temperature at 20 MVA ONAN network transformer
Such trends, from installed sensors, give a much clearer picture of moisture dynamics than can be practically obtained
with laboratory Karl Fischer methods [5.4].
84
5.3 POSSIBLE EVALUATION CRITERIA FOR MOISTURE IN TERMS OF RS
The relationship between humidity and breakdown voltage has been extensively studied. In the early papers on this
subject, moisture was measured in terms of mg/kg Karl Fischer values. Different insulating liquids have different
amounts of moisture when expressed as mg/kg at the same relative humidity and temperature, because their ability to
hold moisture is different – see Figure 5.7 [5.5].
Figure 5.7 Water solubility in different insulating liquids at 23°C in mg/kg [5.5]
The AC breakdown voltage (BDV) test in a quasi-uniform electric field, such as defined in IEC 60156, is commonly used
for quality checking and acceptance tests for insulating liquids. It is sensitive to oil contamination from particles and
moisture Figure 5.8 [5.6]. There is extensive experience on the evaluation of oil parameters for electrical equipment,
which is reflected in international standards, e.g. IEC 60422, IEEE C57.106 and others. Moisture is also covered in these
standards, however, the dynamic nature of its movement between solid and liquid insulation makes it difficult to apply
a singe value.
80 60
Breakdown voltage (kV/2,5 mm)
70
50
Water content (mg/kg)
60
40
50
40 30 Breakdown voltage
(kV/2,5 mm)
30
20 Water content
20 (mg/kg)
10
10
0 0
0 50 100
Oil temperature (°C)
Figure 5.8 Development of oil breakdown voltage with dissolved moisture in a 46 MVA, 30 kV transformer,
manufactured 1970
85
5.3.1 Possible evaluation criteria for moisture in terms of RS

Breakdown in liquid insulation because of insufficient dielectric strength as a result of moisture, is one of the biggest
threats in moist transformers. A number of tests which correlate the dielectric strength with moisture (mineral oil and
esters) and without solid insulation are described in the literature [5.6 – 5.14].
Breakdown voltage, however, does not depend on the absolute amount of water in an insulating liquid, but on the
relative saturation at a given temperature. Dependency curves of breakdown voltage/water content can be translated
to curves of breakdown voltage/relative saturation, using A and B values described in Chapter 3.
Figure 5.9 illustrates the behaviour of BDV with RS at room temperature, aggregated from literature [5.4]. The number
of sets of results pertaining to each oil type is also displayed. To facilitate aggregation and collation of data, moisture
measurements used in all sources were expressed in RS [5.4]. Similarly for BDV, values recorded for all sources were
expressed as relative BDV in per unit (pu), by simply dividing the BDV measured in kilovolt (kV) by the BDV measured
at the lowest RS. This per unit BDV representation accounts for different electrode gap distances used by different
sources [5.4].
RS (%)
Figure 5.9 BDV versus RS aggregated from [5.6 - 5.14]
The findings from all the sources were further trended (red curve in Figure 5.9) to show the overall behaviour between
BDV and RS.
5.3.2 BDV Interpretation and alert values
Insulating liquid after processing (drying and degassing) has a relative saturation below 3% and a breakdown voltage
>70 kV/2,5 mm. According to statistical data, which is reflected in international maintenance standards, e.g. IEC 60422,
electrical equipment even with the highest rating is serviceable wrt BDV in oil ≥50 kV/2,5 mm [5.15]. According to the
lower boundary line in Figure 5.9, the 30% reduction in BDV corresponds to a relative saturation value of about 30%.
This allows to use of RS of 20% as an early alert level.
In addition to the early alert threshold, a RS to temperature (RS/T) plot is proposed to indicate the moisture hysteresis
between solid and liquid insulation. Broad hysteresis loops of the RS/T curves are typically observed for moist
transformers. They reflect the low diffusion time constant of moisture migration from oil to solid insulation during
transformer cool-down period.
The sensor location should be planned carefully to provide a sufficient oil flow, since this affects its response time [5.1].
The proper oil flow also confirms that the sensor is measuring representative oil and the RS/T hysteresis loop evaluation
is valid. In the case of integrated capacitive sensors in monitoring equipment with tempered sensors are being used, a
86
calculation provided by the sensor manufacturer might be useful for assesment of moisture level at a critical location
and at different operational temperatures.
Existing examples concern mainly mineral oil filled transformers. For ester filled transformers, more data should be
gathered before limit values can be established. A concern for ester filled transformers is the potential for hydrolysis
reactions, especially at higher operating temperature.
5.4 USE OF TEMPERATURE/RS CURVES OF CAPACITIVE SENSORS

The RS/T hysterisis curves are indicated as a new tool for evaluating operational risk for BDV. Since this kind of
evaluation has not been used very broadly, additional research and data were gathered.
5.4.1 Model experiments in a climate chamber with and without solid insulation. Difference between
insulating oils and ester liquids. Influence of the speed of heating/cooling
Aim was the identify the influence of load on moisture hysteresis curves. For this purpose the temperature program in
Figure 5.10 was used:
Temperature 20° – 60 °C
3 repeated cycles - 20°C (24h), exponential increase until 60°C (takes 8 hours), 60°C (24 h)
repeated cycles – 20°C (24h), immediate increase until 60°C, takes 5 minutes), 60°C (24 h)
Figure 5.10 Test setup and repeated cycles of a temperature program with different slopes for temperature
increase/decrease
5.4.1.1 Only Oil

New mineral oil with moisture of 27,7 mg/kg Karl Fischer was used and the RS was recorded by means of a capacitive
sensor, using the temperature program as described in Figure 5.10. The variation of temperature and RS with time are
shown in Figure 5.11.
87
Figure 5.11 Curves of temperature and RS for Figure 5.12 Curves of temperature and RS for mineral oil with
mineral oil subjected to the program in Figure 5.10
undried pressboard subjected to the program in Figure 5.10
5.4.1.2 Oil and undried pressboard
The experiment was done using the same temperature program, but in the presence of undried pressboard. The
conditions were:
Volume reactor: 2,6 Liter
Pressboard (1 mm thickness) weight 25,08 g
Starting moisture content in pressboard: 6,96 %
The temperature and RS with time are shown in Figure 5.12.
The dependence of RS with temperature for both pure oil and oil with pressboard are shown in Figure 5.13:
Hysteresis height with a steep

temperature slope
Hysteresis height with a flat

temperature slope
Figure 5.13 Hysteresis loops of RS and temperature for the tests with mineral oil described in 5.4.1.1 and 5.4.1.2
88
It is observed that,
1) The absolute max value of the hysteresis curve is different depending on pressboard dryness. The oil reaches a much
higher RS in the presence of moist solid insulation.
2) Broadness of the loop is a charateristic. Due to the low diffusion coefficients of moisture for solid insulation in mineral
oil, it takes a long time after temperature decreases for moisture to be absorb back in the insulation.
3) Slower increase/decrease of temperature results in lower RS values and lower broadness of the loop, compared to a
rapid temperature increase/decrease.
5.4.1.3 Only natural ester
New natural ester with moisture of 82 mg/kg Karl Fischer has been used and the RS has been recorded by means of a
capacitive sensor using the temperature program described in Figure 5.10. The variations in temperature and RS with
time are shown in Figure 5.14.
89
Figure 5.14 Temperature and RS for natural ester subjected to the program in Figure 5.10
Figure 5.15 Temperature and RS for natural ester with undried pressboard subjected to the program in Figure 5.10
90
5.4.1.4 Natural ester and undried pressboard
This experiment was done under the same temperature program as above, but in the presence of undried pressboard.
The conditions were:
Volume Reactor: 2,6 Liter
Pressboard Weight (1 mm thickness) 25,081 g
Starting % Pressboard: 6,96 %
The Temperature and RS with time are shown on Figure 5.15.
Figure 5.16 Hysteresis loops of RS vs temperature for the tests with natural ester described in 5.4.1.3 and 5.4.1.4
The RS dependence with temperature in Figure 5.16 reveals interesting insights,

1) the absolute value of RS in the presence of moist pressboard increases significantly in comparison to the ester
itself,
2) the hysteresis curves are much narrower than the similar curves with mineral oil. No hysteresis loops are
present. This is probably due to the much higher diffusion cofficients of moisture in ester to solid insulation.
3) In case of a quick heating/cooling the absolute values of the curves are higher, but the difference in comparison
to lower cooling rates is not as explicit as in the case of mineral oil – Figure 5.13.
5.4.1.5 Comparison of the behavior between mineral oil and natural ester with moist insulation
In the case where solid insulation is present, there is a continuous temperature driven dynamic process of moisture
exchange between solid and liquid insulation.
It is important to know that moisture diffusion coefficients for natural esters and solid insulation are much higher than
for mineral oil and solid insulation [15]. This leads to smaller time constants for moisture exchange and therefore to
very narrow hysteresis curve (relative saturation (RS) /temperature dependence) – Figure 5.17.
91
120
100
RS (%) 80
60
40
20
0
0 10 20 30 40 50 60 70
Temperature (°C)
oil+undried pressboard natural ester+undried pressboard
Figure 5.17 Comparison of hysteresis curves RS/Temperature for mineral oil and natural ester, both with undried
pressboard and with temperature program represented on Figure 5.10
5.4.2 Identification of a problem with sensor installation

Two sister 410 kV transformers with OFAF cooling, TR0 (manufactured and inservice since 1985) and TR1
(manufactured 1987, inservice since 1989) with the same operating mode, have been equipped with capacitive moisture
sensors. Moisture in oil has been additionally monitored over the years by Karl Fischer titration. In case of one of the
transformers there was a very good correlation between the calculated values of moisture and those reported with RS
for new oil. For the other transformer, there was a significant shift in the measured Karl Fischer values to the calculated
ones. During inspection it was found that the sensor is not placed in the circulating oil. This can easily seen on
temperature/RS graphs. Both transformers are new and dry. In the case of the first transformer the temperatures cover
the whole operating region and the saturation is in a narrow band (change 2 – 5%) (Figure 5.18). In the second case,
the temperature interval is very restricted, because the sensor is placed in stagnant oil, which does not reflect the
realtime moisture distribution (Figure 5.19). The direct comparison of both RS/temperature curves helped to identify
the problem [5.17].
92
22
20 Moisture (% RS)
Moisture (ppm)
18
KF, ppm
16
14
WCO, ppm
RS, %
12
10
0
60 54 48 55 56 60 58 66 60 59 59 63 67 57 66 58 58 62 53 49 52 59 52 40 59 63 59 43 61 68 65 59 75 68 64 59 66 56 57 54 54 51 49 50 53 56
Top oil Temperature, ºC
Figure 5.18 Moisture data for transformer TR0 as RS/T hysteresis (above) and chronologically RS, calculated
ppm values, as well as measured Karl Fischer values (below). Top oil temperature is marked for information.
93
20
Moisture (%RS)
18
Moisture (ppm)
KF, ppm
16
14
12
WCO, ppm
RS, %
10
0
31
27
25
44
46
45
49
46
57
41
59
47
51
41
44
47
48
44
36
40
45
39
50
44
57
66
49
53
59
48
50
47
49
52
57
51
58
40
49
47
47
48
57
58
47
47
43
51
48
43
48
54
50
55
50
58
54
45
Top oil Temperature, ºC
Figure 5.19 Moisture data for transformer TR1 as RS/T hysteresis (above) and chronologically RS,
calculated ppm values, as well as measured Karl Fischer values (below). Top oil temperature is marked
for information.
5.4.3 RS/Temperature curves from equipment in service
5.4.3.1 20/27 MVA, 66/11 kV, 38 year old substation transformer, failed 2005
In this case, the suggested main cause of the failure was assocaited with foggy (oversaturated) oil around the bottom
windings, caused by very wet insulation. Moisture of 4,2% (Karl Fischer) was measured in the inner insulation. High
moisture in the oil was measured during a few hot summer days, peaking at 50-55 ppm.
Before the failure there was a sudden change of the ambient conditions due to a stormy weather: ambient air
temperature dropped from 36 to 9 °C within 12 hours as a consequence of heavy cold rain. Rapid cooling of oil inside
the cooler resulted in foggy oil.
94
The established mechanism of failure was:

Condensation of water from oversaturated oil onto the insulation surface of bottom windings, followed by inception of
corona discharge on the wet surface and escalation of the corona discharge into a disc-to-disc arcing failure.
The position of the moisture sensors is shown on Figure 5.20. Figure 5.21 shows the temperature recorded by the
sensors. The development of %RS and the calculated ppm water is represented on Figure 5.22 and Figure 5.23 [5.18].
Figure 5.20 Positioning of the moisture probes at the tank and coolers
Figure 5.21 Temperature at the sensors
95
Figure 5.22 Development of % S
Figure 5.23 Development of calculated water content
96
Figure 5.24 %RS vs temperature (combination of Figures 5.21 – 5.23)
The graph of RS vs temperature (Figure 5.24) gives a clear picture of the moisture dynamics. Tank top and cooler to
tank top operate at a higher temperature. The RS is well above 20% and partially exceeds 30%. As elaborated in
Section 5.3, this is a clear indication of a dangerous condition, because of reduced dielectric strength in oil. Further,
looking at the cooler bottom trace, at low temperatures the saturation exceeds 60%. These are all very clear indications
of oil being oversaturated.
The RS/T curves of tank top, cooler top and bottom, exhibit expected moisture hysteresis curves. The hysteresis curve
of the tank bottom is untypical because the sensor is positioned in stagnat oil, thus not recording the actual moisture
of the bulk oil, but nevertheless indicating an oversaturated oil.
5.4.3.2 10 MVA, aged wet transformer, probe connected to tank bottom

A moisture capacitive sensor was connected to the tank bottom of an aged 10 MVA transformer. On Figure 5.25 the
developed RS, calculated moisture in mg/kg and temperature with time, are presented. The same data, again presented
as RS and moisture in ppm vs temperature is presented in Figure 5.26. This kind of presentation shows that the warning
limit of 20% at higher operating temperature (section 5.3) is exceeded. Also, the hysteresis loop of RS at 20°C is very
broad. At lower temperatures, RS values of 60% are reached, which points to oil moisture oversaturation and thus to
a danger of a possible oil breakdown.
97
Figure 5.25 %RS, calculated water content and temperature with time in a 10 MVA moist aged
transformer
Figure 5.26 Hysteresis loops %RS and calculated water content vs temperature in a 10 MVA moist aged
transformer
5.4.3.3 40 MVA refurbished transformer (OFAF), probe connected before radiator - special loading
test.
In the case of a refirbished 40 MVA transformer with OFAF cooling, a moisture sensor was connected before the radiator
and a loading test was performed. Figure 5.27 shows the development of RS, calculated moisture in mg/kg and
98
temperature over time. The same data presented as RS and moisture in ppm vs temperature is presented in Figure
5.28. This kind of presentation shows that the warning limit of 20% is not exceeded even at low operating temperature
(section 5.3) and the RS hysteresis loop at 20°C is narrow. This demonstrates low moisture levels, with no danger of
oil oversaturation.
Figure 5.27 Development of %RS, calculated water content and temperature with time in a 40 MVA
refurbished transformer during a special loading test
Figure 5.28 Hysteresis loops %RS and calculated water content vs temperature in a 40 MVA refurbished
transformer during a special loading test
5.4.3.4 Temperature/RS hysteresis curves for transformers with different moisture levels
A summary of the available data on moisture measurements by capacitive sensors, directly connected to oil flow in
transformers (in the tank or cooler pipes) for a few cases of operating transformers, is presented on Figure 5.29. The
99
examples below concern mineral oil filled transformers. The RS/T hysteresis curves can be used to classify these
transformers into three characteristic moisture levels.
Moist transformers:
RS/T hysteresis curve values exceed 20% RS at the lowest operating temperature and have a very broad
hysteresis loop at 20°C – case of transformer P1 (moisture sensor placed in a valve at bottom of the cooler.
Transformers fit for service, but with elevated moisture in comparison to the new condition :
RS/T hysteresis curve values do not exceed 20% RS even at low operating temperature and has a relatively
narrow hysteresis loop – case of transformers P3 and E1 (moisture sensor placed in a valve at top of the cooler.
Dry transformers:
RS/T hysteresis curve values are below 5% RS and the hysteresis loop is very narrow – case of transformers
New and H1 dried (moisture sensor placed in a valve at top of the cooler).
Figure 5.29 Hysteresis loops of %RS vs temperature in transformers with different moisture
While evaluating transformer moisture, it is important to consider the position and the accessibility of the moisture
sensor to oil flow, along with its relative saturation and temperature records. In cases of transformers with natural oil
circulation, a typical vertical oil temperature gradient can reach 20K – 30K at full load. Due to the difference between
the top and bottom oil temperature, a notable difference between the measured RS values can be expected, and is
shown in the following example.
Tr P2 (ONAN cooled transformer) – Moisture sensors have been placed in cooler pipes at tank top and tank bottom.
Figure 11 shows RS/T hysteresis curves from tank top and tank bottom in a moist transformer. Both hysteresis curves
exceed 20% RS. The tank bottom curve reaches very high values at lower temperature, suggesting a risk of oil
oversaturation with mositure and breakdown. It can be seen that the hysteresis loop is significantly broader and steeper
at lower temperatures, which reflects moisture migration between liquid and solid insulation due to temperature
dynamics in a moist transformer, as compared to a dry transformer.
100
Figure 5.30 Hysteresis loops %RS vs temperature for case of moisture sensors placed at tank top and tank bottom
in a moist transformer.
5.4.4 Use of capacitive sensors for dew point measurements in transformers

The temperature at which dew, or condensation, forms on cooling a gas is known as Dew Point (Td). This is the
temperature at which gas becomes saturated with water vapour. Td is used in atmospheric chemistry and process
engineering. It is a measure of absolute water vapour content and does not dependent on temperature. In electrical
equipment it is used for determining surface moisture content of cellulose insulation, during transformer installation,
after long term storage without insulating liquid and at the factory before shipment. The procedure is described in an
informative annex of C57.93 (IEEE Guide for installation and maintenance of liquid immersed power transformers).
Some transformer companies have developed their own correlations.
The procedure, described in IEEE C57.93 guide is a two-step procedure:
The dew point of the gas in the tank is measured along with the temperature of the transformer insulation and the
pressure inside the tank. From these measurements the moisture remaining in the insulation is estimated using available
curves (Figure 5.31 and Figure 5.32). If the measurement is within the acceptable range on the dew point limit curve,
then proceed with vacuum filling. If the measurement is not within the acceptable range, then additional processing is
required to remove moisture.
101
Figure 5.31 Conversion from dew point or frost point to vapor pressure according to IEEE C57.93
102
Figure 5.32 Moisture equilibrium chart [Piper 1946] according to IEEE C57.93
This procedure has also been adopted in IEEE C57.152 (IEEE Guide for diagnostic field testing of fluid filled power
transformers, regulators and reactors).
In these procedures the moisture equilibrium charts are based on the Piper curves. It has been reported, that the Piper
charts result in higher moisture content than experiment [5.19]; there is proposed a new chart where dew point is
directly proportional to the water content in the solid insulation (Figure 5.33). This chart has not been implemented in
standards.
103
Figure 5.33 Reference chart for moisture estimation from dew point measurements. Gas phase is at one atm
pressure [5.19].
The fundamental dew point measurement technology is a chilled mirror, where the sensor surface is cooled until
dew/frost forms on the sensor surface and the temperature of the surface is measured. Chilled mirror technology can
offer the highest accuracy over a wide range of dew points. Due to the chilled mirror’s optical measurement principle,
the sensor is sensitive to the presence of dirt, oil mist, dust, salt and any other contaminants on the mirror surface.
Similarly, accurate chilled mirror devices tend to be expensive and are often employed when absolute accuracy is
essential and frequent maintenance and cleaning can be performed [5.20].
On the market, there are basically two types of capacitive sensors which can and are being used for dew point
measurement in electrical equipment:
1. Thin-film polymer sensors. The capacitance of polymer sensors is directly proportional to the relative
humidity of the surrounding environment, from which water vapor pressure and thus dew point can be
calculated as sensor temperature is known.
2. Aluminum oxide (ceramic) sensors. The capacitance of AlOx sensors is directly proportional to vapor
pressure of the surrounding environment, from which dew point can be calculated.
It’s very important is to follow correct sampling procedures and allow an adequate time for equilibrium of the gas and
solid insulation moisture in the tank.
It is also crucial to consider that materials used in sampling may significantly affect the dew point measurement. Any
hygroscopic materials should be avoided. Many materials contain moisture in their structure, particularly organic
104
whether natural or synthetic. Even tiny quantities of absorbed water vapour on the surfaces of sampling lines can have
a significant effect. The lower the humidity the more critical are these precautions.
Impermeable materials should be chosen to avoid inward moisture diffusion through sampling tubes. PTFE (“Teflon”)
is relatively impermeable and usually satisfactory for dew points above -20°C. Materials like PVC and rubber are relatively
permeable and not suitable at low dew points. Steel and other metals are preferable materials used in sampling of dry
gases (Td <-20°C) as they are practically impermeable and their surfaces can be polished to reduce surface porosity
trapping water vapour (Figure 5.34).
If the measurement is performed in a still gas, then the importance of the sampling materials’ quality becomes even
more significant.
Figure 5.34 Illustration of moisture released by different tubing materials when flushed with very dry gas after
being at ambient humidity [5.21]
To avoid any sampling related error sources and minimize uncertainties, one should measure as close to the transformer
tank as possible, preferably directly inside the tank. One option is to measure though the openings for the radiator
pipes of the tank, before the radiators are installed. The blind plate of a flange can be equipped with a suitable access
point, like a ball valve, for the installation of a dew point probe (Figure 5.35.)
Figure 5.35 A dewpoint probe installed onto a tank radiator pipe flange
105
5.5 CONCLUSIONS TO CHAPTER 5

RS /Temperature hysteresis curves reliably show the moisture saturation in oil and thus can be used for an evaluation
of the moisture condition in electrical equipment. For evaluation of moisture by means of capacitive sensors, the
following should be taken into account :
 Measured data from capacitive sensors in an operating transformer, recording the actual bulk oil temperature
and in a good oil flow, can be used directly for assessment of moisture level and related risks. Otherwise, a
calculation may be needed to recalculate the critical RS values.
 The shape of RS/T hysteresis curves is characterized by the temperature driven moisture migration process
which is dependent on the moisture content in transformer insulation, on the rate of increase/decrease of
temperature, as well as on the type of insulating liquid.
 The dependence of RS and breakdown voltage delivers a valuable tool for risk assessment of breakdown in the
liquid insulation. From evaluation of RS/T hysteresis curves in the case of transformers filled with mineral oil,
the following criteria for quantification of risk associated with moisture levels have been derived:
- Shape (broadness) of the hysteresis curve – on the basis of the recorded temperature interval, in dry
transformers the RS difference between min and max RS value is ≤15%. In moist transformers a slow
increase/decrease in operating temperature is favourable to avoid oversaturation in oil and to reduce the
risk of dielectric breakdown.
- Maximum RS value of the hysteresis curve depends on transformer temperature: At oil operating
temperatures higher than 40°C a RS of 20% and at low operating temperatures RS of 30%, should be
treated as signifcant.

[5.1] Vaisala White Paper, “The Effect of Moisture Sensor Location on Reliable Transformer Oil Monitoring,” Available:
https://www.vaisala.com/en/media/12931.
[5.2] P. Agren, Communication to D1.52
[5.3] C. Beauchemin, “Example of moisture in oil”, IEEE Transformer committee, San Diego, April 2011.
[5.4] S. Tee, Q. Liu, Z. Wang, G. Wilson, P. Jarman, R. Hooton, et al., "Seasonal Influence on Moisture Interpretation
for Transformer Ageing Assessment," IEEE Electrical Insulation Magazine, May/June Issue, vol. 32, pp. 29-35,
2016.
[5.5] See Chapter 3.
[5.6] X. Wang and Z. D. Wang, "Study of dielectric behavior of ester transformer liquids under ac voltage," IEEE
Transactions on Dielectrics and Electrical Insulation, vol. 19, pp. 1916-1925, 2012.
[5.7] S. Tenbohlen, M. Koch, D. Vukovic, A. Weinlader, S. Barker, J. Baum, et al., "Application of vegetable oil-based
insulating fluids to hermetically sealed power transformers", Stuttgart: CIGRE, 2008.
[5.8] Vaisala, "The Effect of Moisture on the Breakdown Voltage of Transformer Oil", Retrieved 25 Juy 2014. Available:
http://www.vaisala.com/Vaisala%20D ocuments/White%20Papers/CEN-TIA-power-whitepaper-Moisture-and-
Breakdown-Voltage-B211282EN-A-LOW.pdf
[5.9] Cooper Power Systems, "Envirotemp FR3 Fluid Testing Guide", 2008.
[5.10] I. Fofana, V. Wasserberg, H. Borsi, and E. Gockenbach, "Challenge of mixed insulating liquids for use in high-
voltage transformers.1. Investigation of mixed liquids," IEEE Electrical Insulation Magazine, vol. 18, pp. 18-31,
2002.
[5.11] E. Gockenbach and H. Borsi, "Natural and Synthetic Ester Liquids as alternative to mineral oil for power
transformers," in Annual Report Conference on Electrical Insulation and Dielectric Phenomen ( CEIDP) , pp. 521-
524, 2008.
[5.12] K. Miners, "Particles and Moisture Effect on Dielectric Strength of Transformer Oil Using VDE Electrodes," IEEE
Power Engineering Review, vol. PER-2, pp. 36-36, 1982.
[5.13] M. Krins, H. Borsi, and E. Gockenbach, "Influence of carbon particles on the breakdown voltage of transformer
oil," in 12th International Conference on Conduction and Breakdown in Dielectric Liquids, ICDL '96., pp. 296-
299, 1996.
[5.14] M. Beyer, W. Boeck, K. Möller, and W. Zaengl, Hochspannungstechnik: Theoretische und praktische Grundlagen :
Springer Berlin Heidelberg, 2013.
106
[5.15] "IEC 60422 Mineral insulating oils in electrical equipment - Supervision and maintenance guidance (Edition
4.0)," International Electrotechnical Commission - Fluids for Electrotechnical Applications Technical Committee,
p. 93, 2013.
[5.16] R. Villarroel, D. Garcia, B. Garcia, J. Burgos, “Moisture diffusion coefficients of transformer pressboard insulation
impregnated with natural esters”, IEEE Transaction on Dielectrics and Electrical Insulation, Viol. 22, Nr. 1, pp.
581 – 589, 2015.
[5.17] D. Mihajlovic, Communication to D1.52
[5.18] V. Davydov, “Post-Failure evaluation of dielectric performance of winding of 38-y.o transformer enhanced by
on-line monotoring”, A2-302, Cigre Session Paris, August 2016.
[5.19] Oommen, IEEE Transactions of power apparatus and systems, Vol. 103, Nr.10, 1984, p 3063
[5.20] Dew point sensing technologies and their performance differences in compressed air applications, Vaisala
Application note 2010
[5.21]Guide to the Measurement of Humidity, National Physical Laboratory and Institute of Measurement and Control,
London 1996, ISBN 0 904457 24 9
107
108
APPENDIX A. DEFINITIONS, ABREVIATIONS AND SYMBOLS

APPENDIX A. DEFINITIONS OF GENERAL TERMS USED IN THIS TB ([A1] – [A9])
Acronym Phrase Definition
Absolute humidity The amount of water vapor present in a unit volume of air,
[kg/m3] usually expressed in kilograms per cubic meter
(μw) Chemical potential For a mixture of water and oil, the partial derivative of the
of water in oil Gibbs energy G with respect to the amount (number of moles)
of water, with temperature, pressure, and amounts of all other
constituents held constant.
Absorbent A substance which absorbs moisture often to prevent moisture

ingress in transformers or used in oil drying equipment.
Absorbent is different from adsorbent in that it takes moisture
evenly through the material, while an adsorbent takes moisture
to the surface.
Adsorbent A substance which adsorbs moisture, e.g. silica gel.

Calibration An operation that, under specified conditions, as a first step,
establishes a relation between values with measurement
uncertainties provided by measurement standards, and
corresponding indications with associated measurement
uncertainties. In a second step, uses this information to
establish a relationship for using measurement results [A5]
CRM Certified Reference Reference material characterized by a metrologically valid
Material procedure for one or more specified properties, accompanied
by a certificate that provides the value of the specified
property, its associated uncertainty, and a statement of
metrological traceability. [A5]
DP Degree of The number of monomeric units in

Polymerization a macromolecule or oligomer molecule, a block or a chain
[B2]. In case of cellulose, it is a number of anhydrous-ᵦ-
glucose monomers C6H10O5 in the cellulose molecule [A8].
Desiccant A hygroscopic substance used as a drying agent
Dew Point That temperature at which water vapor present in a gas is just
sufficient to saturate it. When air is cooled below the dew point,
the excess of water vapor appears as tiny droplets or crystals
of ice depending on the temperature of the air [A2].
Diffusion coefficient A proportionality constant between the molar flux of water due
of water to molecular diffusion and the gradient in the concentration of
water molecules.
Enthalpy of water- The enthalpy change associated with the dissolution of water
in-oil solution in oil at constant pressure resulting in infinite dilution. The
enthalpy of solution is most often expressed in kJ/mol at
constant temperature.
109
TDE Equilibrium A state of a system in which the macroscopic properties of each

(Thermodynamic phase of the system become uniform and independent of time.
equilibrium) If the temperature is uniform throughout the system, a state
of thermal equilibrium has been reached; if the pressure is
uniform, a state of hydrostatic equilibrium has been reached;
and if the chemical potential of each component is uniform, a
state of chemical equilibrium has been reached. If all these
quantities become uniform, the system is said to be in a state
of complete thermodynamic equilibrium [A4].
EMC Equilibrium The amount of moisture a material contains when it has

moisture content reached equilibrium with its environment.
%ERH Equilibrium relative A percent of relative humidity at equilibrium.

humidity
%ERS Equilibrium relative A percent of relative saturation at equilibrium.
saturation
Henry Law’s A coefficient proportionality in Henry’s Law – one of the
constant representaitng of solubility of water in oil.
Humidity Water in gas form/phase.
Hygrometer Instrument that measures humidity.
KF Karl Fischer A method in analytical chemistry that uses coulometric or

titration volumetric titration to determine water content in a sample. It
was introduced by German chemist Karl Fischer in 1935.
LOD Loss on drying A test method to determine the moisture content of a sample
(e.g. oil or paper).
Moisture absorption A physicochemical process in which water molecules enter

head space, insulating liquid or solid insulation material. This
is a different process from moisture adsorption (see below).
Moisture adsorption An adhesion of water molecules to a surface. This process

creates a thin film of the adsorbate (water) on a surface of
adsorbent (cellulose).
An increase in the concentration of water at the interface of a
gas/solid or liquid/solid insulation due to the operation of
surface forces.
Moisture content of See definition of “Water content of insulating liquid”.

oil
Moisture content of See definition of Water content of paper.
paper
Moisture desorption Release of adsorb or absorb water.
Moisture A relationship between water content of paper (e.g. cellulose)

equilibrium curves and water content of insulating liquid WCL (e.g. mineral oil) at
a given temperature at equilibrium.
Used for moisture assessment in solid insulation.
Moisture Sorption The dependence of water content of solid insulation not only
Hysteresis on the current value of water activity, but on the history of its
values. The deviation of water desorption isotherm from water
adsorption isotherm.
110
Moisture Sorption A relationship between moisture content of the material (e.g.

Isotherms cellulose) and water activity at a given temperature at
equilibrium. Used for moisture assessment in solid insulation,
such as cellulose (e.g. insulation paper, pressboard).
Moisture diffusion The movement of water molecules from an area of higher

concentration to an area of lower concentration within one
phase.
A, B Water saturation The coefficients characteristic for the given insulating liquid,
coefficients which allow the conversion of relative saturation into the water
content in liquid and vice versa.
Moisture sensor A sensor measuring relative humidity in a headspace or relative
saturation of insulating liquid.
Note: There are absolute humidity sensors, however these are
not commonly used in moisture measurement for
transformers.
Moisture transducer A device that converts electrical signal (e.g. output voltage of
a moisture sensor) to another signal (e.g. mA) representing
one of the moisture parameters (e.g. %RS).
Moisture probe Part of an instrument that houses the main body of the sensor
NV Neutralization value The acidity (neutralization value) of insulating liquid is a
(acidity) measure of the acidic constituents or contaminants in the oil
[mgKOH/goil] [B9].
%RH Relative humidity The ratio, expressed as a percentage, of the partial pressure
of water in the atmosphere at some observed temperature, to
the saturation vapor pressure of pure water at this temperature
[A2].
%RS Relative saturation The ratio, expressed as a percentage, of the partial pressure
(percent relative of water in oil at some observed temperature, to the saturation
saturation) vapor pressure of pure water at this temperature.
The degree or extent to which water is dissolved in oil or
absorbed by paper compared with the maximum possible,
usually expressed as a percentage.
RM Reference Material A material, sufficiently homogeneous and stable with reference

to specified properties, which has been established to be fit for
its intended use in measurement or in examination of nominal
properties [A5].
RRT Round Robin Test An interlaboratory test (measurement, analysis, or experiment)

performed independently several times.
Saturation Water The pressure exerted by a pure water (at a given temperature)
Vapor Pressure in a system containing only the water vapor and condensed
phase (liquid or solid) of water, adapted from [A2].
S Solubility of water A quantity expressing the maximum concentration of water

in oil that can exist in insulating liquid (e.g. mineral oil) at
thermodynamic equilibrium at specified temperature and
pressure.
Common measures of solubility include mass fraction, mole
fraction, molality, molarity, and others. Adapted from [A4].
111
Titration A technique to determine the concentration of a substance in

solution by adding to it a standard reagent of known
concentration in carefully measured amounts until a reaction
of definite and known proportion is completed, as shown by a
color change or by electrical measurement, and then
calculating the unknown concentration [A6].
Aw; aw Water activity A measure of the “effective concentration” of a water in oil,

meaning that the water‘s chemical potential depends on the
activity of a real water in oil solution, in the same way that it
would depend on concentration for an ideal solution.
A measure of water available for exchange between oil and
paper (water-in-paper activity, awp) or between oil and
headspace (water-in-oil activity).
WCL Water content of The quantity of water contained in insulating liquid expressed
insulating liquid (by as the ratio of masses of water and insulating liquid. Normally
mass) the letter is taking before drying and the ratio is expressed in
parts per million (ppm).
WCP Water content of A ratio of water weight and dry weight of a paper sample
paper expressed as a percentage
Water Vapor The gaseous phase of water.
WVP Water vapor The partial pressure of water vapor in any gas mixture in
pressure equilibrium with solid or liquid water.
The pressure at which water vapor is in thermodynamic
equilibrium with its condensed state.
[A1] CRC Handbook of Chemistry and Physics, 97th Edition, edited by William M. Haynes, 2017
[A2] IUPAC. Compendium of Chemical Terminology, 2nd ed. (the “Gold Book”), compiled by A. D. McNaught and A.
Wilkinson, Blackwell Science, Oxford (1997). XML on-line corrected version: http://goldbook.iupac.org created
by M. Nic, J. Jirat, B. Kosata
[A3] IUPAC Physical Chemistry Division. Quantities, Units and Symbols in Physical Chemistry, 3rd ed. (the “Green
Book”), prepared for publication by E. R. Cohen, T. Cvitaš, J. G. Frey, B. Holmström, K. Kuchitsu, R. Marquardt,
I. Mills, F. Pavese, M. Quack, J. Stohner, H. L. Strauss, M. Takami, A. J. Thor, RSC Publications, Cambridge
(2007).
[A4] Glossary of Terms Related to Solubility (IUPAC Recommendations 2008), Pure Appl. Chem., Vol. 80, No. 2, pp.
233–276, 2008.
[A5] International Vocabulary of Metrology, 3d eddition http://jcgm.bipm.org/vim/en/index.html
[A6] EC 60814 ED. 2.0 B:1997. Insulating liquids - Oil-impregnated paper and pressboard - Determination of water
by automatic coulometric Karl Fischer titration
[A7] Technical Brochure 349 (2008) Moisture Equilibrium and Moisture Migration within Transformer Insulation
Systems, WG A2.30, www.e-cigre.org
[A8] IEC 60450:2004:2004, Measurement of the average viscometric degree of polymerization of new and aged
cellulosic electrically insulating materials
[A9] IEC 60422:2013, Mineral insulating oils in electrical equipment – Supervision and maintenance guidance
112

Moisture Measurement and Assessment in Transformer Insulation

Uploaded by

Document Informationclick to expand document informationTB 741 - MOISTURE MEASUREMENT AND ASSESSMENT IN TRANSFORMER INSULATION – EVALUATION OF CHEMICAL METHODS AND MOISTURE CAPACITIVE SENSORS

Document Informationclick to expand document information

Copyright:

Available Formats

Moisture Measurement and Assessment in Transformer Insulation

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Moisture Measurement and Assessment in Transformer Insulation

Uploaded by

Copyright:

Available Formats

741

MOISTURE MEASUREMENT AND ASSESSMENT

1 . IMPORTANCE, INFLUENCE AND MEASUREMENT OF MOISTURE ..................................................... 9

2.12 WATER CONTENT DETERMINATION OF OIL IMPREGNATED PAPER INSULATION................................ 50

3. CONVERSION OF MOISTURE MEASUREMENTS BETWEEN CAPACITIVE SENSORS AND KARL FISCHER

4. INDIRECT METHODS OF MOISTURE DETERMINATION IN THE SOLID INSULATION....................... 71

5.6 REFERENCES TO CHAPTER 5 ............................................................................................................................. 106

APPENDIX A. DEFINITIONS OF GENERAL TERMS USED IN THIS TB ([A1] – [A9]) .....................................109

1 . IMPORTANCE, INFLUENCE AND MEASUREMENT OF

1.1 IMPORTANCE AND INFLUENCE OF MOISTURE

Figure 1.1 Thick structures of the insulation in a transformer [1.1]

Laminated pressboard Pressboard cylinders Insulation paper

60 %RS with wet

40 %RS only oil

 Effect of the water content on cellulose aging

1.2 KARL FISCHER MEASUREMENTS - PRINCIPLE AND INTERNATIONAL STANDARDS

H2O + I2 + [RNH]+SO3CH3- + 2RN → [RNH]+SO4CH3- + 2[RNH]+ I-

Table 1.1 International standards for KF water measurements

1.3 MEASUREMENT OF MOISTURE WITH CAPACITIVE SENSOR

- Measurement performance and uncertainty sources

1.3.1 Metal oxide capacitive sensors

1.3.2 Operating principle of aluminium oxide moisture sensor

1.3.3 Capacitive thin film polymer sensors

Figure 1.12 Structure of a capacitive thin polymer sensor

ni = Amount of agent i in moles

Table 1.2 Minimum requirements for a capacitive thin polymer sensor

Measurement range (moisture)

Response time (90%) at 20°C (min) xx minutes.

1.3.5 Sensitivity and accuracy

Figure 1.13 Repeatability Figure 1.14 Transfer Function

Figure 1.15 Non-linearity Figure 1.16 Hysteresis

1.3.6 Stability and selectivity

1.3.7 Response time

1.3.8 Calibration of capacitive polymer sensors

1.3.9 Calibration traceability

1.3.10 Laboratory(factory) calibration

1.3.11 Other possible calibration /verification methods of capacitive polymer sensors

Reference humidity at 20 °C Sensor reading at 22 °C %RH difference due to

10 %RH 8.8 %RH 1.2 %RH

1.3.12 Two-point calibration

Temperature Gas moisture Gas moisture Gas moisture

1.3.13 One-point calibration in the gas phase

1.3.14 One-point verification in the oil phase

1.3.15 Common mistakes and error sources during humidity calibration

1.3.16 Calibration against Karl Fisher titration results?

See also chapter 3.

1.4 REFERENCES TO CHAPTER 1

[1.3] I. Atanasova-Hoehlein, “Determining the Water Content in Transformers Possibilities and

[1.5] Atanasova-Hoehlein I, Contribution to Cigre D1.52.

[1.7] W. Sikorski, K. Walczak, P. Przybylek, “Moisture migration in oil-paper insulation system in

2 . RRT ON KARL FISCHER AND CAPACITIVE MOISTURE

2.2 RRT SETUP

 Instruction procedure for RRT

2.3 RRT PROTOCOL

Water content in oil Test sample NMO % relative saturation

Water content in Test sample NE

Figure 2.1 CIGRE D1.52 RRT protocol