2008 Annual Report Conference on Electrical Insulation Dielectric Phenomena Particle Effect on Breakdown Voltage of Mineral and Ester Based Transformer Oils X. Wang and Z.D. Wang School of Electrical and Electronic Engineering, the University of Manchester, Manchester M60 1QD, UK zhongdong.wang@manchester.ac.uk Abstract- ac breakdown voltage tests are usually used for quality check of mineral oils during transformer operation. When studying ester-based oils, previous experimental research including the ones carried out at University of Manchester laboratories, were based on ‘as-received’ oil samples and particle effects were not quantified. Similarly, water effects on oils determined previously were also ambiguous since water can be combined with particles to reduce breakdown voltages. This paper used mineral and ester-based transformer oils, i.e. 10 Gemini X as mineral oil, Midel7131 as synthetic ester-based oil and FR3 as nature ester-based oil, in the forms of “as-received” and “processed”, and their AC breakdown strengths were comparatively studied. Weibull distributions of ac breakdown voltages of three oils were discussed. It can be seen from the results that particle effects on ac oil breakdown voltages are dominant for “as–received” mineral oil and the ‘processed’ mineral oil has a distinctly different breakdown characteristic from ‘as-received’ mineral oil in terms of the effect of water contents. Introduction Transformer oil is one of the most essential components for conventional oil immersed transformers as it acts both as electrical insulation and thermal coolant. The dielectric strength of the oil and more specifically the level of oil contamination determine the dielectric safety margin of the transformer insulation system. In addition, transformer oil may be used in power transformers for decades without changing, and the transformer oil, together with the insulation paper, can deteriorate due to ageing. This long term ageing produces water, acid and many other by-products, soluble or insoluble. Maintenance guides consider the oil to be a component that could be treated and managed separately from the oil-paper insulation system. Ac breakdown voltage tests, together with many other test techniques on oil samples, were used for quality check and control. Ester-based transformer oils have gained the success in distribution transformers at certain countries and regions, and they are currently under research and development for application in large power transformers. Ester-based oils have been shown to extend the remaining life of paper as it is more hygroscopic than mineral oil; more water is therefore absorbed and reacted by ester-based oils so the paper is kept dry. Ester oils are also advantageous compared to traditional mineral oil as these oils are less flammable, non-toxic and more biodegradable. However, using ester-based oils also has 978-1-4244-2549-5/$25.00 © 2008 IEEE certain disadvantages, i.e. that vegetable oils can more easily oxidize and hydrolyze. Especially in United Kingdom, large power transformers are typically of free breathing design so measures must be taken to control oxidation and hydrolysis. Research activities have been increasingly active on ester based transformer oils for recent years and many test results on ac breakdown voltages were published, including some statistics analysis results [1]. However, the oil samples used in those tests were often not filtered or, in another word, not particle controlled, leaving the ac breakdown test results more relevant to the particles or impurities rather the oil itself. In this paper, the withstand voltages of ‘as-received’ and ‘processed’ mineral and ester-based transformer oil were investigated. As the withstand voltage of a dielectric fluid is a statistically distributed quantity corresponding to a probability of failure, a large sample of breakdown voltages was analyzed and their statistical distributions compared. The mean withstand voltages, as well as the highest and lowest breakdown voltages were discussed. In addition, the water effect on withstand voltages for both ‘as-received’ and ‘processed’ mineral oil were studied. It can be found that the water needs to be combined with particles to have a significant effect on mineral oil ac breakdown voltages. Experimental procedures 2.1 Experiment samples A type of mineral oil, 10 Gemini X provided by National Grid, was used as traditional mineral oil. Midel7131 provided by M&I materials, was used as the synthetic ester-based oil, and FR3, provided by Cooper Power Systems, was used as the nature ester-based oil. For each type of oil, both the ‘as-received’ and ‘processed’ oil samples were tested. The ‘as-received’ oil here refers to oil samples processed through degassing and dehydrating only. The ‘process’ oil here refers to oil samples processed through degassing, dehydrating and filtering. The processing procedure will be described below. 2.2 Preprocessing procedures The oil samples were filtered through a 0.2µm membrane filter unit for 3 cycles. It was found that after 3 cycles, the oil is clean enough and the particle number in oil samples could not decrease further even more filter cycles were given. The particle number in oil samples was decided by HIAC 8011 liquid particle counting system, which can detect particles with 598 diameter from 1µm to 200µm. The particle counting results are shown in Figure 1. Figure 1.a, 1.b and 1.c show the results for mineral oil, FR3 and midel7131 separately. 1000000 'As-received' mineral Particle number (in 100ml oil) 100000 'Processed' mineral 10000 1000 100 10 1 0 10 20 30 40 Particle diameter (µm) 50 60 1000000 'As-received' FR3 Particle number (in 100ml oil) 100000 'Processed' FR3 10000 2.4 Water uptake procedures 1000 In order to decide the water effect of ‘as-received’ and ‘processed’ mineral oil, a desiccator and glycerol solution were used to control the relative humidity in the transformer oil. According to the Raoult’s law, the relative humidity of the air in the desiccator can be controlled by the proportion of glycerol solutions. The Raoult’s Law is expressed as P=xaPa, where P is the resulting vapor pressure, x is the mole fraction of component a in solution and Pa is the vapor pressure of pure a. Oil sample container was deposited in the desiccator with a certain relative humidity for one week. After conditioning, the container was sealed and stored for another week before testing to allow further diffusion of moisture to create a high degree of consistency in the sample. Care was taken during the water uptake procedure for ‘processed’ oil sample in order not to contaminate the sample. The moisture content of each oil sample was determined by Karl Fischer titration method. A metrohm 648 Coulometer with oven were used to provide accurate moisture reading. The dry oil samples with water content less than 5% RH were achieved by the drying method as specified in preprocess procedures. 100 10 1 0 10 20 30 40 50 60 Particle diameter (µm) 100000 'As-received' midel7131 Particle number (in 100ml oil) measurements with partial sphere electrodes and 1mm gap distance, following the ASTM D1816 test standard. 2 series of 5 breakdowns are required by the standard, but in order to compare the statistical distributions of different oils, 8 series of 5 breakdowns were taken to give 40 breakdown voltages in total. It is believed that the size of 40 breakdowns were big enough to give a reasonably accurate distribution to estimate the failure voltage at a low risk rate. It needs to be noted that ester-based transformer oils have higher viscosities and lower interfacial tensions than mineral oil, leading to slow ejection of gas bubbles created during the breakdown. Therefore, it seems sensible for ester-based transformer oils to have a longer standing time after each breakdown. However Cooper Power System published a test guide which only requests extra standing time after pouring the oil sample into the test cell and before the start of the tests, therefore 1 minute standing time between two successive breakdowns specified by ASTM 1816 was used for all the oil samples. 10000 'Processed' midel7131 1000 100 10 1 0 10 20 30 40 Particle diameter (µm) 50 60 Figure 1. Compare of particle counting results The ‘clean’ oil is defined by CIGRE working group 12.17 as the particle content with a diameter larger than 5µm is down to 300 per 100ml in a oil sample [2]. As can been seen from the filtering results, the particle numbers in ‘processed’ oil samples were almost reaching the number given in the definition of ‘clean’ oil. Subsequently, the oil samples were individually degassed and dried at less than 1kPa for 2 days at 80°C, then were given a further day to cool down to ambient temperature under vacuum conditions. It was found that the relative water contents in the oil samples were all less than 3% RH after drying, which could be considered as very dry oil. This applies to all three oils, i.e. mineral oil, Midel7131 and FR3. 2.3 AC withstand voltage tests A Baur DPA75 was used for AC withstand voltage Experimental Results and Discussions Previous research has drawn the conclusion that ester-based oil has comparable dielectric strength as mineral oil, with similar withstand voltage which is defined as the breakdown voltage at a specified low risk of failure [3]. In order to gain a better understanding on how the oil properties were influenced by particles, a test was carried out to compare the AC withstand strengths of ‘as-received’ and ‘processed’ mineral and ester-based transformer oils. Researches have also been conducted concerning the water content effect on the breakdown voltage of mineral oil. It is generally considered that the breakdown voltages are decreased with the increase of relative water content, a significant reduction of breakdown voltage was noticed by many researchers for the relative water content between 5 to 10 599 percents [4]. However, little research has investigated the water effect on very clean oil. Therefore tests were carried out to investigate the difference of water effect on both ‘as-received’ and ‘processed’ mineral oil. 99.5 Probability (%) 95 3.1 AC withstand strength comparison 70 40 10 1 20 25 30 35 40 45 50 55 60 Breakdown Voltage (kV) Figure 2. Distribution for breakdown voltages of mineral oil 'Processed' FR3 oil 'As-received' FR3 oil Probability (%) 95 70 40 10 1 25 30 35 40 45 50 55 Breakdown Voltage (kV) Figure 3. Distribution for breakdown voltages of nature ester oil 95 Probability (%) The AC withstand strength results for ‘as-received’ ester-based oil were reported, previously produced at the same laboratory at the University of Manchester [4]. In order to verify the result, a successive of 40 breakdown tests was carried out using the same configuration for ‘as-received’ oils. The breakdown tests were also carried out on ‘processed’ oils to give a distribution of breakdown voltages. Figure 2 through Figure 4 show the breakdown voltage distributions for ‘as-received’ and ‘processed’ mineral oil, synthetic ester and nature ester oils. The mean breakdown voltage is given in the figures by the dot lines. According to the weakest link theory, their breakdown voltage distributions are fitted by 2 parameter Weibull distribution given by equation (1) (1) Where x: electric field λ: scale parameter k: shape parameter For all the 3 types of transformer oils, the breakdown voltage distributions follow the Weilbull distribution well. The Weibull fitting distributions are given in the figure by the solid lines. From the results, the dielectric strengths of ‘processed’ oils were increased significantly comparing to those of ‘as-received’ oil, in terms of mean breakdown voltage, lowest and highest breakdown voltages. The increase of lowest breakdown voltages, however, was larger than that of highest breakdown voltage, i.e. the lowest breakdown voltage was increased by 9.4 kV and the highest breakdown voltage was increased by 3.2kV for FR3 oil. The higher dielectric strength for ‘processed’ oils is caused by fewer particles suspended in the oil. The fitting parameters of different oil are given in Table 1. From the results, it can be easily seen that both the scale and shape parameter are increased after oil processing and the ‘processed’ oil follows the Weilbull distribution much better than ‘as-received’ oil. Especially at low probabilities, the breakdown voltages for ‘processed’ oil almost spot on the fitting curve, while the ‘as-received’ oil breakdown voltage was far from the fitting curve, usually lower than the fitting breakdown voltage. This deviation may be explained by the theory that breakdown is caused by weak links. It is easy to understand that the breakdown characteristic of ‘processed’ oil is steady because there are fewer particles in the oil, which means less weak links. For ‘as-received’ oil, the more the particles with different sizes, the more possibilities for different sized particles to be in the path between the two electrodes to initiate streamers, the less predictable the breakdown voltage may be. This results in the deviation between the measured breakdown voltage and the fitting value at low probabilities. 'Processed' mineral oil 'as-received' mineral oil 'Processed' Midel7131 'As-received' Midel7131 70 40 10 1 0.1 30 35 40 45 50 55 Breakdown Voltage (kV) Figure 4. Distribution for breakdown voltages of synthetic ester oil Table 1. Weibull distribution parameters for three oils – particle effect Λ k Mineral 6.3 38.8 ‘As-received’ oil FR3 Midel7131 8.2 12.7 43.0 43.4 Mineral 13.4 49.6 ‘Processed’ oil FR3 Midel7131 11.9 12.9 46.5 46.9 The distribution parameters of oil dielectric strength are given in Table 2. Compared with the previous results reported in [4], all the mean breakdown results for mineral, synthetic and nature esters are a little smaller. However, this difference is within the toleration level, when considering the results were based on a successive 100 breakdowns. It is suggested by many researchers that successive breakdown could increase the test results, attribute to the removal of surface irregularities or heating up the local oil [5]. Comparing the distribution parameters of ‘as-received’ and ‘processed’ oil samples, conclusion can be drawn that the ‘processed’ oils have higher dielectric strengths and less 600 standard deviations than ‘as-received’ oils. It is well known that breakdown is initiated by the ‘weak link’ in the oil, such as gas bubbles, particles or surface defects. Since gas contents and electrode surfaces have been pre-controlled well, less particles in the ‘processed’ oil produce less ‘weak links’, therefore it is more difficult to initiated breakdown in the ‘processed’ oils and consequently, the breakdown voltages are higher and more consistent. Breakdown Voltage (kV) 60 Table 2. Breakdown voltages & distribution parameter for different oil samples ‘As-received’ oil Mean breakdown voltage (kV) Lowest breakdown Voltage (kV) Highest breakdown Voltage (kV) Standard deviation (kV) Coefficient of variation ‘Processed’ oil Mineral FR3 Midel 7131 Mineral FR3 Midel 7131 36.0 40.5 41.6 47.7 44.5 45.1 23.8 25.6 27.9 35.8 35 36.2 48.9 49.9 48 58.2 53.1 53.3 6.3 5.8 4.5 4.1 4.56 4.3 17.5% 14.3% 10.5% 8.6% 10.3% 9.7% 50 'As-received' mineral 40 30 20 10 0 0 20 40 60 80 Relative Water content (%) 100 Figure 5. Breakdown voltages as a function of relative water content for 'as-received' mineral oil Table 3. Distribution characteristics for 'as-received' mineral oil The particle contents are quite different for different types of ‘as-received’ oils, therefore it is more meaningful to compare the dielectric strengths of ‘processed’ oils with similar particle contents. For ‘processed’ oil samples, it can be noticed that the ester-based oils have comparable dielectric strengths with mineral oil. The coefficient of variation for ester-based oil is a little larger than that of mineral oil, indicating the breakdown for ester-based oil is more unpredictable and there are more chances for ester-based oil to breakdown at lower voltages for a low risk. Water content (%RH) Mean breakdown voltage (kV) Standard deviation (kV) Coefficient of variation 3% 10% 25% 30% 35% 55% 35 29 20 15 16 7 4.0 4.5 4.1 3.8 4.9 2.9 11% 16% 21% 25% 31% 41% It can be seen from the result that the standard deviation, for different relative water content, was kept around 4 kV. However, as the mean breakdown voltage decreased the coefficient of variation increased from 10% at water content of 3% RH to nearly 40% at water content of 55% RH. b) Water effect on ‘processed’ mineral oil The same test was carried out on the ‘processed’ mineral oil. 3.2 Water effect on withstand strength The ‘processed’ mineral oil is clean oil with less than 500 Water affects the breakdown voltages of oil by influencing particles with diameter larger than 5µm in 100ml oil sample. the conductivity of microscopic particles suspended in the oil. The particle counting result could be seen in Figure 1. Instead These particles could create larger discharges in approaching of 20 breakdowns, 40 breakdown voltages were measured for a an electrode when they are more conductive, therefore increase more precise evaluation of distribution performance. The the likelihood to initiate gas bubbles and lead to breakdown. It average value of breakdown voltages, with the highest and has been proved that the type of particles suspended in the oil lowest voltages, as a function of relative water content is given also affect the water influence on the oil as the water and in Figure.6. particles have combined effect, which could amplified the The mean breakdown voltage decreasing tendency, as can reduction of dielectric strength of oil [6]. be seen from the results, can be divided into 3 parts to discuss. First, the breakdown voltage does not change much when the a) Water effect on ‘as-received’ mineral oil For each oil sample of different water content, 20 water content is less than 40% RH. Then, it gradually drops as breakdowns were measured to give a distribution. The average the water content is increased between 40% and 60% RH, and value of breakdown voltages, with highest and lowest finally it reduces to a plateau when the water content is larger breakdown voltage, as a function of water content expressed in than 70% RH. This trend is quite different from that of the ‘as-received’ mineral oil. relative humidity is given in Figure 5. The lowest and highest breakdown voltage could also be The mean breakdown voltage decreased significantly with divided into two parts. When the water content in oil is less water content between 3% and 10% RH. The reduction in the than 40% RH, the highest breakdown voltage is kept at about dielectric strength of oil is exponential for water content less 50 kV and the lowest breakdown voltage is reduced as water is than 40% RH. When water content in oil is larger than 40% RH, increased. This means although the mean breakdown voltage is the breakdown voltage gradually reduced to a plateau. Both the kept nearly the same at this stage, there is more chance for the lowest and highest breakdown voltage decreased with the oil with larger water content to breakdown at lower voltage. increase of water content in oil. This result is in consistence When the water content is larger than 40% RH, both the with previous research, bearing in mind that the particle highest and lowest breakdown voltage are reduced as the water number and size is first time quantified here. The distribution content is increased. Consequently, it emphasizes the necessity characteristics of oil breakdown voltages are given in Table 3. to use the highest and lowest breakdown voltage, not just the 601 mean breakdown voltage, to evaluate the dielectric strength of oil. created, resulting in the significant drop of dielectric strength. Conclusions Breakdown Voltage (kV) 60 a) The ‘as-received’ oil samples were processed to become clean oil. The dielectric withstand strengths of both the ‘as-received’ and ‘processed’ mineral, synthetic ester and nature ester oil samples were investigated. The oil dielectric strength is increased after processing, in terms of mean breakdown voltages, highest and lowest breakdown voltages. The standard deviation is also decreased. The ester-based oils have comparable dielectric strength as mineral oil, but has larger coefficient of variations, indicating its unpredictability and probably a higher chance for ester-based oil to breakdown at lower voltages. 'Processed' mineral 50 40 30 20 10 0 0 20 40 60 Water content (%) 80 100 b) The water effect on ‘as-received’ and ‘processed’ mineral oil were also investigated to show the combined effect of particle and moisture. For ‘as-received’ mineral oil, the breakdown voltage drop sharply with the increase of water content expressed in relative humidity. For ‘processed’ mineral oil, the breakdown voltage is kept steady when water content is less than 40% RH, and decreased gradually when relative water content is increased afterwards. The coefficient of variation is increased with water content for both ‘as-received’ and ‘processed’ oil samples. Figure.6. Breakdown voltages as a function of water content in % RH for 'processed' mineral oil Table 4. Distribution characteristics for 'processed' mineral oil water content (%RH) Mean breakdown voltage (kV) Standard deviation (kV) Coefficient of variation 3% 22% 41% 67% 83% 98% 44 40 39 24 16 14 3.0 3..2 4.4 3.5 3.6 3.7 7% 8% 11% 15% 23% 27% The distribution characteristics of oil breakdown are shown in Table 4. The standard deviations for oil samples with different water content, is kept around 3.5 kV, and the coefficient of variation increased from 7% at 3% water content RH to 27% at water content of 98% RH. However, both the standard deviations and the coefficient of variations of ‘processed’ mineral oil are much less than those of ‘as-received’ mineral oil, indicating that the breakdown voltage is much consistent for clean oil, and the water effect on the dielectric strength of ‘as-received’ mineral oil is more unpredictable. Therefore, not only the mean breakdown voltage of ‘as-received’ mineral oil is lower, also is the withstand voltage in the presence of water. These results verified Chadband’s ‘weakest link’ theory that the breakdown is caused by the weak link in the oil [7], for this case the weak link is the conductive particles which is formed by particles absorbing water. When the oil is clean, there are fewer particles suspended in the oil. As the oil is moistening by water, there is little chance for the water molecule to combine with particles to form conductive ‘weak link’ in the oil. At low quantities of water, the water dissolved in the oil evenly by forming hydrogen bonds between itself and the polar part of the oil. Although the oil is generally considered to be non-polar, there will always be a small quantity of polar structures within itself. The dielectric withstands strength, therefore, will not be influenced much by the water content. Once there is more water in the oil, both the chance for the water to combine with particles and the chance for the water molecules to form water cluster are increased, leading to more ‘weak links’ suspended in the oil. Therefore the dielectric withstand strength starts to reduce. For ‘as-received’ oil, there are many particles and even at low water content large quantity of ‘weak links’ could be Acknowledgement The first author would like to thank EPSRC DHPA award to provide scholarship for his PhD study. AREVA T&D, EdF Energy, M&I Materials, Cooper Power Systems, National Grid, Scottish Power, TJ|H2b Analytical Services and Electricity North West for their financial and technical support to form the research consortium on ‘Alternative fluids for large power transformers’ at The University of Manchester. Grateful thanks are given to Dr Hongzhi Ding for his helpful suggestions. References [1] D. Martin, "Evaluation of the Dielectric Capability of Ester-based Oils for Power Transformers," in School of Electrical and Electronic Engineering. vol. PhD Manchester: Unviersity of Manchester, 2007, p. 223. [2] CIGRE, "Effect of Particles on Transformer Dielectric Strength," in Working Group 17 of Study Committe 12, June 2000. [3] D.Martin, "Evaluation of the Dielectric Capability of Ester-based Oils for Power Transformers," in School of Electrical and Electronic Engineering. vol. PhD Manchester: Unviersity of Manchester, 2007, p. 223. [4] D. Martin and Z. D. Wang, "A Comparative Study of the Impact of Moisture on the Dielectric Capability of Esters for Large Power Transformers," in Electrical Insulation and Dielectric Phenomena, 2006 IEEE Conference on, 2006, pp. 409-412. [5] M. G. Danikas, "Technical Report of Particles in Transformer Oil," IEEE Electrical Insulation Magazine, vol. 7, pp. 39-40, March/April, 1991. [6] K. Miners, "Particles and Moisture Effect on Dielectric Strength of Transformer Oil Using VDE Electrodes," IEEE Transactions on Power Apparatus and System, vol. PAS-101, pp. 751-756, March, 1982. [7] W. G. Chadband, "The Electrical Breakdown of Insulating Oil," Power Engineering Journal, vol. 6, pp. 61 - 67, Mar. 1992. 602