Experience with Capacitive Online Sensorsfor Moisture Evaluation in Transformer Insulation

Ivanka Atanasova-HöhleinSiemens AG, Germany

Maja Končan-Gradnik, Tim Gradnik, Biljana ČučekMilan Vidmar Electric Power Research Institute, Slovenia

Piotr Przybylek, Krzysztof SiodlaPoznan University of Technology, Poland

Knut Brede Liland SINTEF Energy Research, Norway

Senja LeivoVaisala Oyj, Finland

Qiang LiuThe University of Manchester, UK

This article presents new criteria for on-line moisture assessment in liquid filled electrical equipment

Key Words: capacitive sensors, moisture, transformer insulation, Karl Fischer measurement, temperature dependence, relative saturation, breakdown voltage, risk evaluation, mineral oil, ester, water content

1. Introduction

Moisture determination in the insulating liquid is a routine measurement for transformers and related electrical equipment to determine the moisture condition of insulation system and to assess the risk of failure occurrence. Up to now, majority of the gathered measurements on moisture in operating transformers are spot values, based on Karl Fischer (KF) measurements. These data do not give essential information about the temperature dependent distribution of moisture between solid and liquid insulation.

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 involves the following procedure to obtain a correct moisture profile of a transformer: proper placement of the sensor, gathering an adequate

measurement data set related to load and temperature and evaluation of the measured data by comparison to historical values. Availability of continuous moisture-in-transformer measurements and its relation to temperature opens up new diagnostic possibilities and approaches in comparison to conventional KF spot value measurements.

The introduction of online measuring methods requires new evaluation methodology which in many cases will not follow single spot evaluation techniques. The introduction of new measuring principles like capacitive sensors measuring % relative saturation requires acceptance by the user. Therefore a lot of questions arise, e.g.:

Are water content in liquid values (in mg/kg) as measured by Karl Fischer and relative saturation values (in %) of capacitive sensor readings convertible?

What are the uncertainties of such conversion? Can limits for relative saturation be used to improve transformer in-service risk

evaluation?This paper reports the results of round robin test (RRT) for determining moisture

saturation coefficients in different insulating liquids including mineral oils, ester liquids and a silicone oil. Three methods are evaluated by five independent laboratories. In addition, field experiences of online moisture monitoring on a number of in-service transformers are collected. By analyzing these data, it is aimed to make contribution to answering the above questions.

2. Convertibility of Karl Fischer values and sensor readings

As the conventional mg/kg measurements of water in oil (WCL) have been well established and with reference data readily available for condition assessment, there is a motivation to convert the new measurement data of relative saturation (RS) of oil with water 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 1 and equation 2. Equation 2 has an Arrhenius form as the moisture saturation in oil (S) is temperature dependent. The natural logarithmic form can also be expressed in its decimal logarithm form for convenience as shown in equation 3.

WCL= RS100

× S (1)

S=eA e−Be /T (2)

S=10A10−B10 /T (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, compositions and conditions [1]. As both natural logarithm and decimal logarithm forms of coefficients are used in the literature, equations (4) and (5) can be used to convert the moisture saturation coefficient values.

Ae=A10 ln (10 ) ;Be=B10 ln (10 ) (4)

A10=A e log (e ); B10=Be log (e ) (5)

2.1. Methods for determining moisture saturation coefficients

There is not available international standard for determining moisture saturation coefficients in insulating liquids and hence a large variety of methods has been introduced in the literature using different equipment and procedures. This becomes one of the major factors contributing to the variance of 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 oil sample to a known RS level at a temperature, 2) measure the absolute moisture content WCL of the oil sample, 3) calculate the saturation level S at that temperature based on equation (1),4) repeat the procedure at three or more temperatures to obtain the coefficients according to

equation (2) or equation (3). 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). In order to obtain results with sufficient accuracy the correlation coefficient of the linear regression should not be less than 0.995.

In this paper, three methods for determining moisture saturation coefficients used in the RRT are described in detail. 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 sensor measurements of insulating liquids. Method 2 introduces a dedicated test cell and conducts the measurements at an equilibrium state whereas Method 3 proposes a simple experimental setup and a quick procedure to conduct the measurements at a transient/quasi-equilibrium state.

Method 1: This method uses a closed chamber exposing the stirred insulation liquid to a constant environment with pre-set RS and temperature. Such environment can be achieved in different manners including climate chamber, saturated salt solution chamber or controlled mixing of wet and dry air into a closed chamber. The absolute moisture content of the liquid at equilibrium state is measured through Karl Fischer titration. Moisture saturation coefficients are calculated based on results obtained at multiple temperatures.

Experimental Design

It is important to maintain a constant and reliable environment. This can be achieved via installing relative saturation and temperature sensors in the chamber for verification. It is also important to avoid contamination of the sample (i.e. salt or other contaminants) which might be an issue choosing the salt solution controlling the environment. In addition, unnecessary

overlong humidification period should be avoided to minimize the effect of oil oxidation/hydrolysis as polar by products could affect the moisture saturation level. The choice of the method depends on the equipment and resources available to the user. Climate chamber was used in the Method 1 for the RRT.

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 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 level off value and two successive measurements in steady state should not differ more than ± 5%. The duration of this process for the oil humidity to reach steady state depends on the stirring speed of oil, 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.

Method 2: This method 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 details in [2].

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 1. The glass vessel should be air tightly 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 gas-tight pass-through should be placed on the top lid for insertion of the moisture and temperature sensor as well as a needle pass-through used for liquid sampling. During conditioning, the insulating liquid should be mixed using a magnetic stirrer. Capacitive and temperature sensors should be placed directly above magnetic stirrer.

Figure 1. Sketch and photo of the test vessel used in Method 2 [2].

Test Procedure

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 in range 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 conditioning process, after achievement of preset oil temperature, the relative humidity (RH) of the air in climatic chamber should be set close to the relative saturation RS of the oil to prevent water migration in case of leakage in the measurement setup. After achievement of set oil temperature it is also recommended to release the potential pressure built in the vessel by using the needle valve. The time of liquid conditioning, for each temperature level, should be long enough, that the setup can reach the state close to moisture and temperature equilibrium. Usually, three hours conditioning time for each temperature level is enough. Time necessary to achieve equilibrium state strongly depends on many factors like: level of relative saturation of liquid, operation of climatic chamber or measurement setup tightness. After equilibrium state 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.

Method 3: This method also uses both Karl Fischer and capacitive sensor measurements of the insulating liquids with a simple setup and a quick conditioning process.

Experimental Design

An aluminum made bottle (commercially available size of 1.2 liters) with a stirring bar is used as a test cell for Method 3, as shown in Figure 2. The insulating liquid is filled in the cell up to about 1 cm from the upper edge.

Figure 2. Photo of the test vessel used in Method 3.

Test Procedure

After filling in 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 2). The capacitive sensor head should be in the area of oil circulation. When the temperature is cooled down 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 to the range near to room temperature. The measurements should be carried out at least at three temperatures.

2.2. Moisture saturation coefficients obtained from round robin test (RRT)

Tables 1-2 and Tables 4-6 show the mean values of moisture saturation A and B coefficients found in the RRT with respect to the natural logarithmic form in equation 2 and decimal logarithmic form in equation 3. The oil types considered are a new mineral oil (a mixture of hydrocarbons, naphthenic type), a service aged mineral oil (interfacial tension: 22 mN/m, acidity: 0.03 mgKOH/goil, color: 3.5, dissipation factor at 90C: 0.058), a new synthetic ester (pentaerythritol tetra ester), a new natural ester (soybean oil) and a new silicone liquid (di-alkyl silicone polymer). The moisture saturation curves are also plotted in Figure 3 to Figure 7. The three measurement methods introduced in section 2.1 were employed. Method 3 has been used by five laboratories, so mean values and standard deviation values are evaluated.

It is observed that Method 1 yields higher moisture saturation values than the other two methods. For the new mineral oil and the new silicone oil, there is only a slight difference in the moisture saturation across different measurement methods. As for the aged mineral oil, the synthetic ester and the natural ester, the difference in the evaluated moisture saturation is significant, particularly for the synthetic ester.

Table 1. Moisture saturation coefficients for a new mineral oil from RRT.

Method

Moisture Saturation CoefficientsDecimal logarithm equation (3) Natural logarithm equation (2)

A10 B10 Ae Be

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 (5 labs) 6.89 0.58 1503 181 15.88 1.33 3461 416

Average 7.09 1574 16.33 3625

Figure 3. Moisture saturation versus temperature for a new mineral oil from RRT results.

Table 2. Moisture saturation coefficients for a service aged mineral oil from RRT.

Method

Moisture Saturation CoefficientsDecimal logarithm equation (3) Natural logarithm equation (2)

A10 B10 Ae Be

Mean Std. Dev. Mean Std. Dev.

Mean Std. Dev. Mean Std. Dev.

Method 1 7.37 - 1636 - 16.97 - 3768 -Method 2 6.90 - 1508 - 15.89 - 3473 -

Method 3 (5 labs) 6.60 0.52 1379 183 15.20 1.20 3174 422

Average 6.96 1508 16.03 3473

The condition of the service aged sample is presented in Table 3.

Table 3. Parameters of the service aged oil sample

Parameter Color Acidity Dissipation factor at 90°C Interfacial tension

Value 3.5 0.03 mg KOH/g Oil 0.058 22 mN/m

Standard ISO 2049 IEC 62021-2 IEC 60247 EN 14210

Figure 4. Moisture saturation versus temperature for a service aged mineral oil from RRT results.

Table 4. Moisture saturation coefficients for a new synthetic ester from RRT.

Method

Moisture Saturation CoefficientsDecimal logarithm equation (3) Natural logarithm equation (2)

A10 B10 Ae Be

Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev.Method 1 5.96 - 768 - 13.73 - 1769 -Method 2 5.32 - 608 - 12.25 - 1400 -Method 3

(5labs) 4.98 0.42 511 120 11.46 0.97 1177 276

Average 5.42 629 12.48 1449

Figure 5. Moisture saturation versus temperature for a new synthetic ester from RRT results.

Table 5. Moisture saturation coefficients for a new natural ester from RRT.

Method

Moisture Saturation CoefficientsDecimal logarithm equation (3) Natural logarithm equation (2)

A10 B10 Ae Be

Mean Std. Dev.

Mean Std. Dev. Mean

Std. Dev. Mean Std. Dev.

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

Method 3(5labs) 5.10 0.36 634 114 11.80 0.86 1460 261

Average 5.33 693 12.27 1596

Figure 6. Moisture saturation versus temperature for a new natural ester from RRT results.Table 6. Moisture saturation coefficients for a new silicone liquid from RRT.

Method

Moisture Saturation CoefficientsDecimal logarithm equation (3) Natural logarithm equation (2)

A10 B10 Ae Be

Mean Std. Dev. Mean Std. Dev.

Mean Std. Dev. Mean Std. Dev.

Method 1 5.76 - 1005 - 13.27 - 2315 -Method 2 6.30 - 1194 - 14.51 - 2750 -

Method 3 (5 labs) 5.73 0.24 1008 76 13.19 0.56 2321 175

Average 5.93 1069 13.66 2462

Figure 7. Moisture saturation versus temperature for a new silicone liquid from RRT results.

2.3. Comparison of moisture saturation curves between different insulating liquids

The moisture saturation curves of the oils studied in the RRT can be calculated from the A and B (mean values from tables 1-2 and tables 4-6) coefficients according to equation 1 and equation 2 or equation 3, as shown in Figure 8.

Figure 8. Moisture saturation versus temperature for the mineral oils, synthetic and natural esters and silicon liquid examined in the RRT.

It is clear that moisture saturation depends not only on temperature or ageing condition, but also the type and composition of the insulating liquid. The lowest water saturation levels can be seen in the mineral oil, and then followed by the silicone liquid, the natural ester and the synthetic ester. This difference in water saturation has been known to be attributed to the liquid polarity, of which mineral oil and silicone liquid are non-polar or weakly polar. Both ester liquids have higher moisture saturation due to the presence of the polar ester functional groups. Generally, polar or ionic compounds will dissolve only in polar solvents, while nonpolar solutes dissolve in nonpolar solvents. The general rule of thumb is, "like dissolves like." As water molecules are polar, the presence of polar functional groups will favour the water dissolution in ester liquids.

Figure 9 compares the moisture saturation characteristics of the different insulating liquid types as recorded from RRT and literature (mineral oil: [3]–[18], synthetic ester: [5], [10], [14], [17], [20], natural ester – [5], [9], [14], [17], [21], [22], silicon oil – [5], [19], [23]).

Figure 9. Moisture saturation versus temperature for the mineral oils, synthetic and natural esters and silicon liquid for results from literature and RRT.

The moisture saturation curves from Figure 9 have been calculated from the average values of A and B 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 9, the difference between the literature and 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.

3. Possible evaluation criteria for moisture in terms of RS

Breakdown in the liquid insulation because of insufficient dielectric strength due to moisture is one of the biggest threats in moist transformers. A number of tests with correlation between AC breakdown strength and moisture in insulating liquid (mineral oil and esters) without solid insulation are described in the literature [23]−[31].

Dependency curves of breakdown voltage to absolute water content can be transferred to breakdown voltage to relative saturation using moisture saturation values given in Section 2.

Figure 10 illustrates the behavior of AC breakdown voltage (BDV) with relative saturation (RS) at room temperature aggregated from literature [23]−[31]. 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 [24]. Similarly for BDV, values recorded in 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 is to account for different electrode gap distances used in different sources [24].

Figure 10. BDV versus %RS at room temperature aggregated from [23]−[31].

3.1. 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 with the highest voltage rating is suitable for service with breakdown voltage in oil > 50 kV/2.5 mm [32]. This corresponds to 30% reduction compared to the initial breakdown voltage threshold (70 kV/2.5mm). According to the lower boundary line in Figure 10, the 30% reduction in breakdown voltage corresponds to relative saturation value of about 30%. This allows to use RS of 20% as an early alert for evaluating moisture in insulating liquid by means of capacitive sensors at service operating temperatures.

In addition to the early alert threshold, a RS to temperature (RS/T) plot is proposed to indicate the hysteresis loop of moisture migration 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 single measurement of relative saturation of oil does not give a full picture of the real moisture level of transformer insulation system. Only the analysis of the relative saturation results recorded in a wide temperature range allows to classify the transformer into one of three groups: moist transformers, transformers fit for service, dry transformers, what was explained in details in Section 3.2.

The sensor location should be planned carefully to provide a sufficient oil flow, since this affects its response time [16], [33]. The proper oil flow also confirms that sensor is measuring representative oil. The proposed RS/T hysteresis loop evaluation is valid if the capacitive sensor is positioned in the oil flow inside tank or cooler pipe and thus records the moisture and the temperature in the bulk oil. In case integrated capacitive sensors in monitoring equipment with tempered sensors are being used, a calculation model provided by sensor manufacturer might be useful for assesment of moisture level at critical location at different operational temperatures.

3.2. RS/T hysteresis curves for transformers with different moisture levels

A summary of available data on moisture measurements by capacitive sensors directly connected to oil flow in transformers (in the tank or cooler pipes) from a few cases of operating transformers is presented on Figure 11. The examples below concern mineral oil filled transformers. The RS/T hysteresis curves can be used to classify these transformers into three characteristic moisture levels.

1) 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 Tr P1 (moisture sensor placed in a valve at bottom of the cooling block).

2) 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 – cases of transformers Tr P3 and Tr E (moisture sensor placed in a valve at top of the cooling block).

3) Dry transformers:

RS/T hysteresis curve values are below 5% RS and the hysteresis loop is very narrow – case of transformers New and Tr H dried (moisture sensor placed in a valve at top of the cooling block).

Figure 11. RS/T hysteresis curves for transformers with different moisture levels.

3.3. Influence of moisture sensor placement on measurements

While evaluating transformer moisture level, it is important to consider the positioning and the accessibility of the moisture sensor to oil flow along with its relative saturation and temperature records. In case 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, which is shown also in the following example.

Tr P2 (ONAN cooled transformer) – Moisture sensors have been placed in cooler pipes at tank top and tank bottom locations. Figure 12 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 the 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, compared to dry transformer.

Figure 12. RS/T hysteresis curves in case of moisture sensors placed at tank top and tank bottom in a moist transformer.

4. Conclusions

This paper evaluated several methods for calculating A and B coefficients which will theoretically allow the conversion of Karl Fischer values in mg/kg into % relative 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.

In scope of measurement procedure for determination of saturation coefficients it is recommended at least three temperatures with adequate intervals should be chosen and the RS condition should be aimed in the range of 5% RS to 75% RS.

Measured data from capacitive sensors in operating transformer recording the actual bulk oil temperature in oil flow can be used directly for assessment of moisture level and related risks. Otherwise a calculation model may be needed to recalculate the critical RS values.

The shape of RS/T hysteresis curve 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 case of transformers filled with mineral oil, the following criteria for quantification of risk associated to moisture levels have been derived:

Shape (broadness) of the hysteresis curve – on basis of the recorded temperature interval, in dry transformers the RS difference between min and max RS value is not more than 15%. In moist transformers slow increase/decrease of operating temperatures is favorable to avoid oversaturation in oil and to reduce the risk of dielectric breakdown.

Maximum RS value of the hysteresis curve is depending on transformer operational 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 a sign of alert. These RS values correspond to the guidelines given in the informative Annex B of IEEE C57.106:2015. In accordance with Annex A of IEC 60422:2013 relative saturation of oil in the range from 20 to 30% means wet cellulose insulation. The RS of oil above 30% is considered as extremely wet insulation.

Acknowledgment: This article presents a part of the results of Cigre D1.52 working group study, which would not have been possible without the active contribution of Patrik Agren (ABB, FI), Claude Beauchemin (Tjh2b, CA), Leonid Darian (CJSC, RUS), Valery Davidov (AU), Lars Dreier (Weidmann, CH), Marius Grisaru (IEC, IL), Krzysztof Kryczynski (ABB, PL), Jian Li (CQU, CN), Marielle Marugan (GE, FR), Draginja Mihajlovic (IEENT, SR), Bernard Noirhomme (IREQ, CA), Thomas Prevost (Weidmann, USA), Pekka Ravila (Vaisala, FI), Oleg Roizman (AU), Anatoly Shkolnik (IL), Julie Van Peteghem (Elia, BE).

REFERENCES[1] Y. Du, A. V. Mamishev, B. C. Lesieutre, M. Zahn, and S. H. Kang, "Moisture solubility for

differently conditioned transformer oils," IEEE Transactions on Dielectrics and Electrical Insulation, vol. 8, no. 5, pp. 805−811, October 2001.

[2] P. Przybylek, K. Siodla, “Application of capacitive sensor for measuring water content in electro-insulating liquids,” Eksploatacja i Niezawodnosc – Maintenance and Reliability, vol. 18, no. 2, pp. 181−185, 2016.

[3] A. Shkolnik, "Determination of water content in transformer insulation," in Proceedings of 2002 IEEE 14th International Conference on Dielectric Liquids, 2002, pp. 337−340.

[4] P. J. Griffin, V. Sokolov, and B. Vanin, "Moisture equilibrium and moisture migration within transformer insulation systems," 2004.

[5] 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, August 2016.

[6] T. V. Oommen, "Moisture equilibrium in paper-oil insulation systems," in Proceedings of EIC 6th Electrical/Electronical Insulation Conference, 1983, pp. 162−166.

[7] 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.

[8] "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.

[9] 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, no. 5, pp. 1615−1622, October 2012.

[10] V. G. Aralkellian and I. Fofana, "Water in Oil-Filled, High-Voltage Equipment, Part I: States, Solubility, and Equilibrium in Insulating Materials," IEEE Electrical Insulation Magazine, vol. 23, no. 4, pp. 15−27, July 2007.

[11] 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, no. 1, pp. 394−402, February 2014.

[12] "Vaisala HUMICAP® Moisture and Temperature Transmitter for Oil MMT330," Vaisala, 2015.[13] "Hydran M2 Instruction Manual," GE Energy, July 2005. [14] M. Koch, S. Tenbohlen, and T. Stirl, "Diagnostic Application of Moisture Equilibrium for Power

Transformers," IEEE Transactions on Power Delivery, vol. 25, no. 4, pp. 2574−2581, October 2010.

[15] 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 Proceedings of IEEE 18th International Conference on Dielectric Liquids (ICDL), 2014, pp. 1−5.

[16] 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, no. 3, pp. 9−16, August 2016.

[17] 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 Proceedings of IEEE International Conference on Dielectric Liquids, 2011, pp. 1−5.

[18] D. Gan, F. Liu, L. Du, and Y. Liu, "Analysis of water solubility in transformer oil using least square fitting method," in Proceedings of International Conference on High Voltage Engineering and Application, 2010, pp. 93−95.

[19] T. Gradnik, M. Koncan-Gradnik, N. Petric, and N. Muc, "Experimental evaluation of water content determination in transformer oil by moisture sensor," in Proceedings of IEEE International Conference on Dielectric Liquids, 2011, pp. 1−4.

[20] P. Przybylek, „Water solubility in insulating liquids in aspect of moisture investigation using capacitive probe,” Przeglad Elektrotechniczny, vol. 11a, pp. 347-350, November 2012

[21] R. Villarroel, D. F. Garcia, B. Garcia, and J. C. Burgos, "Moisture diffusion coefficients of transformer pressboard insulation impregnated with natural esters," IEEE Transactions on Dielectrics and Electrical Insulation, , vol. 22, no. 1, pp. 581−589, February 2015.

[22] "IEEE Guide for Acceptance and Maintenance of Natural Ester Fluids in Transformers," IEEE Std C57.147-2008, pp. 1−31, 2008.

[23] 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, no. 3, pp. 18−31, May-June 2002.

[24] 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, no. 6, pp. 1916−1925, December 2012.

[25] 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.

[26] Vaisala White Paper, "The Effect of Moisture on the Breakdown Voltage of Transformer Oil", Retrieved 25 July 2014.

[27] Cooper Power Systems, "Envirotemp FR3 Fluid Testing Guide", 2008. [28] 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.

[29] K. Miners, "Particles and Moisture Effect on Dielectric Strength of Transformer Oil Using VDE Electrodes," IEEE Transactions on Power Apparatus and Systems, vol. PAS-101, no. 3, pp. 751−756, March 1982.

[30] M. Krins, H. Borsi, and E. Gockenbach, "Influence of carbon particles on the breakdown voltage of transformer oil," in Proceedinfs of 12th International Conference on Conduction and Breakdown in Dielectric Liquids, 1996, pp. 296−299.

[31] M. Beyer, W. Boeck, K. Möller, and W. Zaengl, Hochspannungstechnik: Theoretische und praktische Grundlagen, Springer Berlin Heidelberg, 2013.

[32] 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.

[33] Vaisala White Paper, “The Effect of Moisture Sensor Location on Reliable Transformer Oil Monitoring,” Available: https://www.vaisala.com/en/media/12931.

BIOGRAPHIES

Ivanka Atanasova-Hoehlein has been responsible for the Siemens chemical-physical laboratory for transformer materials in Nuremberg, Germany since 2004. Main topics of interest are properties of conductive and insulating materials for transformer application. She is engaged in Cigre D1 and IEC TC 10 activities. (ivanka.hoehlein-atanasova@siemens.com)

Maja Končan Gradnik received her University Diploma in Chemical Engineering in 1975 and Post-graduate Diploma in Chemistry in 1983 from Faculty of Natural Science and Engineering , the University of Ljubljana. In 1985 she joined the Milan Vidmar Electric Power Research Institute in Ljubljana where she is leading Physical-Chemical Transformer Diagnostics Department. The field of her research is transformer condition monitoring, insulating materials test techniques, transformers risk management and residual lifetime evaluation technique.From 1992 Mrs.

Končan-Gradnik has been an active member of CIGRE D1, WG D1.01, IEC TC 10 and many IEC TC10 and CIGRE D1 working bodies in the field of transformer insulation and diagnostics. She is also the president of slovenian standardization body SIST TPD (Liquid and gaseous dielectrics). (maja.koncan-gradnik@eimv.si)

Tim Gradnik received MSc. in Electrical Engineering from the University in Ljubljana, Slovenia in 2004. He joined Milan Vidmar Electric Power Research Institute (EIMV) High-Voltage Laboratory department working of development and calibration of HV laboratory measurment equipment. In 2006 he joined the EIMV Physical-Chemical Transformer Diagnostics Department, focusing on research of power transformer thermal modeling, assesment of existing and and development of

new transformer on-line monitoring and diagnostic applications. He is actively participating in CIGRE workgroup D1.52 and is a member of IEC - TC 14/MT 60076-7 and IEC TC 57 WG 10. He is secretary of CIGRE SC A2 and representative of the CIGRE National Committee A2. (tim.gradnik@eimv.si)

Biljana Čuček received MSc. in chemistry from the University of Ljubljana, Slovenia in 1992. Since 2007 she has been employed at the Milan Vidmar Electric Power Research Institute (EIMV) as a researcher at Physical-Chemical Transformer Diagnostics Department. The field of her research is transformer condition monitoring, insulating materials test techniques, transformers risk management and residual lifetime evaluation technique. She is a CIGRE D1.52 workgroup member.

(biljana.cucek@eimv.si)

Piotr Przybyłek was born in Poland in 1979. He received the M.Sc. degree in high voltage engineering in 2003 and the Ph.D. degree in electrical engineering in 2007, both from the Faculty of Electrical Engineering in Poznan University of Technology. In 2017 he received Doctor Habilitatus degree from this same University. He has participated in several research projects, both international and national. He has authored or coauthored over 100 publications, mainly in the field of diagnostics of

power transformer and physicochemical properties of oil-paper insulation. (piotr.przybylek@put.poznan.pl)

Krzysztof Siodła was born in Poznan, Poland. He received M.Sc. diploma in electrical engineering in 1980 and Ph.D. in 1989 from Poznan University of Technology, Poznan, Poland. In 2005 he received Doctor Habilitatus degree from this same University. He is currently a professor at Poznan University of Technology in Poznan, Poland. He is a Head of the Division of High Voltage and Electrotechnic Materials and vice-director of the Institute of Electric Power Engineering. He was a member of CIGRE Study Committee D1 Materials and Emerging Test Techniques

and CIGRE Working Group WG D1.37. Currently he is a member of WG D1.52. Author and co-author of 4 technical books and about 180 papers. His main research interest is: high voltage engineering, HV measurement technique, insulating materials, diagnostics and monitoring of high voltage electric power equipment. (krzysztof.siodla@put.poznan.pl)

Knut Brede Liland received the M.Sc. (Cand Scient) degree in plasma physics from the University of Tromsø, Norway in 1991. The Ph.D. degree in gas discharge physics from the University of Tromsø, Norway (Doctor Scient) and the Université Paul Sabatier, Toulouse, France (Doctorat de l’Universite Paul Sabatier) in 1997. He has been with the SINTEF Energy Research since 1998, now working with power transformer and cable research topics. (knut.b.liland@sintef.no)

Senja Leivo is the Senior Industry Expert at Vaisala in Finland. Her professional focus is on condition monitoring of power transformers. At Vaisala with her 20 years of experience she is responsible for identifying the strategic trends and new monitoring needs within the industry, as well as bringing customers’ voice close to the Vaisala’s product development. As part of her expert role she is active member in CIGRE working groups. Senja holds a Master of Science degree in Materials

Engineering. (senja.leivo@vaisala.com)

Qiang Liu (S’08-M’12) obtained the B.Sc. degree in electrical engineering (2005) and the M.Sc. degree in high voltage and electrical insulation (2008) from Xi’an Jiaotong University in China, and the Ph.D. degree in electrical power engineering (2011) from The University of Manchester in UK. Currently he is a Senior Lecturer at the Power and Energy Division of the School of Electrical and Electronic Engineering at The University of Manchester. His research interests are on pre-breakdown and breakdown phenomena in liquids, ester transformer liquids,

streaming electrification, aging of insulating materials, transformer asset management and high voltage testing. (qiang.liu@manchester.ac.uk)