Artículo de investigación A comparison of the PDIV characteristics of ester and mineral oils Carlos Gustavo Azcárraga Ramos1, Andrea Cavallini2 and Ugo Piovan3 Article originally published on the IEEE Electrical Insulation Magazine, Sept.-Oct. 2014 Abstract The voltage-withstand properties from the point of view of partial discharge inception voltage (PDIV) of ester oils are comparable or even better to those of mineral oils. Ester oils could therefore replace mineral oils as insulation in high voltage transformers. Instituto de Investigaciones Eléctricas Universidad de Bologna, Italia 3 Trafoexperts GmbH, Uster, Suiza 1 2 123 Boletín IIE julio-septiembre-2014 Artículo de investigación Introduction The growing price of mineral oils (MOs) over the last 15 years, and environmental and safety concerns, have promoted research on alternative insulating fluids. Ester oils (EOs), which were used initially in transformer insulation [1], have again attracted attention; both synthetic and natural liquids have been investigated thoroughly, in the search for suitable substitutes for MOs. EOs have interesting properties(CIGRE WG A2-35, 2010). Their most attractive properties are those concerning flammability, biodegradability and hygroscopicity, a property that favors the drying of solid insulation, thus ensuring that depolymerization of cellulose is slowed down. EOs also have drawbacks. They can be oxidized easily, so that additives are necessary, and their use is generally limited to non-breathing transformers. Their viscosities are larger than those of MOs, so that replacing MOs with EOs without adjusting the oil gaps between adjacent pressboard barriers requires caution. Also, their pour points are higher, which could limit their use in cold climates. Finally, streamer propagation under positive lightning impulses is faster than in MO, leading to lower lightning impulse breakdown voltages. Their advantages promoted the use of EOs mostly in medium voltage transformers (generally below 35 kV)(Bertrand, 2012). Their use in high voltage (HV) transformers is hampered by the limited experience of manufacturers. The higher electric stress levels in HV transformers make detailed knowledge of the dielectric properties of the insulating oil essential. Partial discharge inception voltages (PDIVs) under ac or lightning impulse, and under corona and creepage discharges, are also relevant (Sokolov et al., 1999). In this paper, we report a comparison between an MO and an EO based on soy seeds, both of commercial grade. Partial discharge tests were carried out in order to characterize fully the dielectric behavior of both oils. The measured data were treated statistically in order to deal with uncertainty. The experimental results are compared with the available body of knowledge and the results of FEM simulations. Sample preparation Before testing, virgin MO an EO oil samples were: 1. Filtered using a filter with 2 µm pores 2. Dehydrated and degassed for 24 hours at ambient temperature and a pressure 0.1 mbar. The effectiveness of the drying process was evaluated using a commercial Coulometric titrator. Following the above procedures and using a magnetic stirrer during oil drying, final water concentrations of 2-3 and 20-22 ppm were obtained for MO and EO respectively. Pressboard (PB) samples were dehydrated and degassed using the following procedure(ASTM D2413-99, 2009): 124 1. The samples were placed in an open vessel that was heated in an oven at 115 °C for 24 hours at atmospheric pressure. 2. After closing the vessel, it was connected to the vacuum pump, maintaining a pressure less than 0.5 mbar for 48 hours. 3. Without breaking the vacuum, dry insulating oil was added to the vessel until the samples were completely immersed. 4. The temperature of the vessel was hold at 60 °C for 48 hours before breaking the vacuum. 5. After removing the vessel from the oven, the samples were extracted and stored in a dessicator for 24 hours, allowing them to cool to room temperature. Test procedure The PDIV measurements were performed for corona (point/plane electrodes in oil) and creepage (point/ plane electrodes laid on a board sheet) discharges. Artículo de investigación All tests were randomized in order to avoid illusory correlation. Thus corona PDIV tests were performed following the gap sequence 20, 5, 40 and 10 mm. This procedure was followed to ensure that the effects of progressive, unnoticed changes in the components, e.g., needle tip losing its sharpness, board deteriorating as a result of burning marks, oil absorbing humidity from the atmosphere, were not attributed to unidirectional changes in gap length. Measurement data were fitted to the 2-parameter Weibull distribution. Alternative distributions, i.e., the Gaussian and the 3-parameter Weibull distribution, were also investigated. However, the Gaussian distribution did not fit the data convincingly, while the 3-parameter Weibull distribution, which is much more difficult to implement, did not yield significantly better fits than the 2-parameter Weibull distribution. The scale parameter a (63.2% probability percentile), and the 10th percentile, B10, with their 95% confidence intervals, are shown in the figures next. B10 is reported since, for design purposes, it is better to compare oils at low unreliability levels (Wang et al., 2011). The Weibull shape parameter b is not reported, since an indirect measure of data dispersion can be obtained from the relationship between a and B10, i.e., the closer a and B(10), the higher b. PDIV tests are not at present completely specified in the relevant standards (IEC 61294, 1993). The existing specifications should be reviewed considering that: a) new detector types have become available and b) tests in the laboratory can be carried out with much greater sensitivities than tests in the field. As an example, IEC 61294 (IEC 61294, 1993) dating back to 1993, specifies that PDIV is attained when the PD magnitude exceeds 100 pC while raising the voltage at a rate of 1 kV/s. This prescription is reasonable in the field, but seems rather conservative in the laboratory, where sensitivity can be much better than 100 pC. Indeed, Wang and coworkers (Wang et al., 2011) found that (IEC 61294, 1993) leads to PDIV values which are much larger than those obtained using more accurate procedures, e.g., better sensitivity, lower rate of voltage increase. Considering the above, the PDIV measurements reported here were made following a procedure which departs significantly from that specified in (IEC 61294, 1993). PDs were detected using a commercial UWB (40 MHz) detector with the capability of displaying PD pulse waveforms. PDs were coupled through a 50 resistor in series with the test sample and a 100 pF capacitor. The sensitivity of the detection circuit was better than 5 pC. The following procedure was used to measure PDIV: 1. Increase the voltage at 1 kV/min till PD inception (a clear PD pattern or a recognizable PD pulse should be recorded). 2. Reduce the voltage to extinguish PD activity, i.e., to the partial discharge extinction voltage (PDEV). 3. Wait 5 minutes at the PDEV to allow positive and negative ions remaining after the previous PD activity to recombine. 4. Start increasing the voltage from 2kV below the PDEV in steps of 100 V or 250 V (depending on the gap size), waiting 300 s at each step. 5. Record PD pulses and patterns during each voltage step. 6. While testing at the next voltage level, check for the presence of PD pulses in the record for the previous voltage level (corona PD can occur but is extremely rare). 7. If PD pulses are found, stop the test and record the PDIV. In all cases, ten PDIV measurements were carried out at each test gap. The results presented below show that the variance of the experimental data is low, suggesting that the procedure minimizes measurement randomness. a. Corona in oil A hermetic Teflon test cell equipped with a micrometric adjustment device was used to measure PDIV at gap lengths of 5, 10, 20 and 40 mm gaps, with point-to-plane geometry electrodes immersed in MO or EO. The high voltage electrodes of the test cell were tungsten needles 0.5 mm in diameter and 5 mm long, supplied by Fine Science Tools Inc. The tip radius was less than 5µm. 125 Boletín IIE julio-septiembre-2014 Artículo de investigación b. Creeping discharges The test setup consisted of a point/plane electrode configuration laid on a board sheet. The vessel was insulated from the ground to minimize the effect of neighboring objects at ground potential. In order to control exactly the distance between the tip and the ground electrode, and be able to change the contact point between tip and board if burning marks appeared, a movable sample holder was constructed. The adjustable arm of the HV electrode was positioned in such a way that the needle was at an angle of 30° with respect to the horizontal plane (pressboard sheet). In order to maintain the same contact forces between the needle and the pressboard in all experiments, the body of the HV electrode was hinged to the frame of the test cell. In this way, the weight of the electrode became the contact force and was therefore constant. The LV electrode was moved to obtain the required test gaps (5, 10, 20 and 40 mm). Between tests the pressboard was moved in the transverse direction in order to minimize surface damage (tracking burns). However, no tracking damage was observed during the short time required to complete the tests. A note on needle electrodes As stated above, the needle electrodes were tungsten steel needles supplied by Fine Science Tools Inc. The field at the needle tip is given by 2 Etip ~ V• ln (4 • d/r + 1) • r Where r is the tip radius, d is the gap length, and V is the applied voltage (Mason, 1955). As the current associated with PD activity can blunt the needle tip, a 1 k resistor was inserted in series with the tip. The tips were regularly inspected using an optical microscope, and the needles were replaced if signs of degradation were observed. In addition, the accumulated data were examined for evidence of trends, possibly due to electrode wear after every ten PDIV measurements. Inspection of needles from different production batches showed that the variation in tip radius was small within the same batch, but significantly larger between batches. So, in order to achieve stable and consistent results, only comparative tests should be carried out, using needles from the same production batch. Experimental results Corona in oil Corona in oil tests showed that EO has larger PDIVs than MO (figure 1). The confidence intervals (confidence g=95%) show that the EO data are characterized by a larger variance than the MO data. Creeping discharges PDIV a and B10 and their confidence intervals (g=95%) are presented in figure 2. In this case the EO outperformed MO for short gaps, but not for longer gaps. The variance of the data is small for both oils. Figure 3 shows a comparison of the PDIV results for corona and creepage discharges. Discussion Figure 1. Statistics of partial discharge inception voltage (PDIVs) of EO and MO measured at different gap lengths and using point-plane electrodes. 126 As stated above, the PDIVs of EO are higher than those of MO under corona and creeping discharges. In order to explain this observation, it is instructive to focus on inception fields, rather than inception voltages, since the former provides more insight into the inherent insulating properties of the oils and the board/oil interfaces. The expression for the electric field at the needle tip (1) was Artículo de investigación derived under reasonable assumptions, e.g., needle assumed to be a perfectly conducting hyperboloid, space charge effects neglected (Mason, 1955). However, it is not applicable when the needle is laid on a board sheet. We therefore carried out FEM simulations using COMSOL Multiphysics. In order to make the problem tractable, the radius of the needle tip was taken as 5 mm, and the electrical properties of the materials shown in table 1 were adopted. These properties were measured in the laboratory and compared with some relevant references (Schultz and Küchler, 1998). Charge injection from the needle tip into the oil was not included in the simulations. The calculated PD inception fields under corona discharge are shown in table 2. Since the permittivity of the oil does not influence the solution, the fields obtained by applying 1 V are the same for MO and EO. It follows that the relationship between the measured PDIVs for MO and EO will be the same as the relationship between the corresponding calculated PD inception fields. Thus EO performs better than MO at all gap lengths. This result might be explained in terms of the difference between the ionization energies of the chemical species formed in the two oils. However, the distribution of chemical species and their ionization energies are not known precisely. Another factor that should be taken into account is the larger electrical conductivity of EO with respect to MO. Since space charge injected from the needle tip into the oil during the negative half-cycle of the supply voltage may have played an important role in PD inception, it is possible that the greater conductivity of the EO prevented the buildup of space charge, thus reducing the field at the needle tip when the tip voltage became positive. The results of similar simulations with the board present are presented in table 3. It will be seen that the electric fields for 1 kV applied are larger than those obtained without the board. The MO electric fields are about twice those of the EO, which partly explains why EO tends to have higher PDIV values. The difference can be attributed to the fact that EO and the board have similar permittivities. The inception field values are higher than those obtained in the absence of pressboard, showing that the presence of the board increases the voltage withstand capability of both oils. Figure 2. Statistics of creeping partial discharge inception voltage (PDIV) for EO and MO measured at different gap lengths, and using point (needle)-plane electrodes with the needles lying on a board sheet. Figure 3. PDIVs for MO and EO measured for corona PD (without board) and for creepage PD (with board). 127 Boletín IIE julio-septiembre-2014 Artículo de investigación Relative permittivity Material MO 2.2 MO impregnated board 3.5 EO 3.2 EO impregnated board 4.1 Permittivity mismatch Conductivity (S/m) 3.8×10-13 59% 12.4×10-15 3.3×10-12 28% 17.1×10-15 Table 1. Material properties for fem simulations. Gap (mm) Electric Field (kV/mm) at 1kV PDIV (kV) Inception field (kV/mm) MO and EO MO EO MO EO 5 51.9 8.6 9.2 450 480 10 44.5 9.4 11.0 420 490 20 36.9 11.0 12.8 400 470 40 29.4 14.2 16.7 420 490 Table 2. Estimated partial discharge inception fields - without board. Gap (mm) Electric Field (kV/mm) at 1kV PDIV (kV) Inception field (kV/mm) MO EO MO EO MO EO 5 154.4 82.4 4.4 12.3 680 1010 10 134.0 71.4 9.9 13.7 1330 980 20 113.2 60.4 14.6 16.4 1650 990 40 90.9 48.7 20.0 19.4 1820 950 Table 3. Estimated partial discharge inception fields - with board. 128 However, the results are somehow contradictory, since the inception field for MO increases by a factor of almost three in going from short to long gaps, while the EO inception field is almost constant. Considering the inception fields and their confidence intervals, one might ask whether the changes shown in Tables 2 and 3 in going from short to long gaps are significant. Figure 4 shows that the inception fields are approximately constant, except for the case of MO under creepage PD. The observation that corona inception fields, i.e., measured using point-plane electrodes and without board, are generally lower than those measured when the needle is placed in contact with pressboard could be explained in terms of the structure of the pressboard/oil interface. It is assumed that the transition between oil and pressboard occurs through a mixture of pressboard fibers projecting into the oil. Mitchinson et al (Mitchinson et al., 2010) presented a microscope image of the pressboard/oil interface, confirming the presence of the board fibers protruding into the oil. These authors suggested that the interface could be more resistant to PD inception than pure oil, due to Van der Waals forces which would prevent oil molecules directly in contact with the pressboard from drifting freely, thereby forming a zone within which a gradual transition from bound to free oil molecules occurs (no-slip layer). In order to explain the difference between MO and EO inception fields as functions of gap length for creepage discharges, it is instructive to study the electric vector field near the needle tip. Figure 5 shows that, for a 5 mm gap, there are significant differences between the field vectors, but not for a 40 mm gap. In particular, at 5 mm the field in EO tends to be parallel to or directed into the board, whereas in MO the field tends to be directed from the needle into the oil. At 40 mm the electric field tends to be parallel to or directed into the board in both oils. This difference suggests different PD inception mechanisms between EO and MO at 5 mm in at short gaps, e.g., in EO PDs might be initiated at weak points of the interface, but in MO in the oil. At 40 mm PDs might be initiated at weak points of the interface in both oils. This suggestion is consistent with the PD pattern and PD pulse sequences at short gaps reported in fure 6; MO shows a pattern typical of bubble discharges with PD pulse bursts (Pompili et al., 2008; Pompili et Artículo de investigación Figure 4. Electric field at inception for MO and EO measured for corona PD (without board) and for creepage PD (with board). Confidence intervals with 95% probability are also shown). Figure 5. Electric field distribution near the needle tip for MO (upper) and EO (lower). Figure 6. PD pattern and discharge behavior close to PDIV for a 5 mm gap. al., 2009), whereas EO shows a pattern typical of surface discharges (Cavallini et al., 2010). It should be noted that, because of the structure of the interface (board fibers separated by oil), and the smaller permittivity mismatch between the board fibers and EO compared to that between the board fibers and MO, the electric field in the EO will be smaller than the electric field in the MO. The observation that EO performs better than MO in corona inception measurements then leads to the conclusion that, for creepage discharges, larger inception fields are to be expected for EO. However, the experimental results show that MO performs better at long gaps. Any proposed explanation of this discrepancy must take into account the microscopic structure of the oil/board interface. In particular, it may be that the Van der Waals forces in the no-slip layer are stronger within the MO/board interface than within the EO/board interface. Our present description of the oil/pressboard interactions is still incomplete. Some authors (Mitchinson et al., 2010; Dai et al., 2010) have emphasized that initiation of creeping PD on a fully dried pressboard surface is very unlikely; wet pressboard sheet usually has a lower PDIV. Thus a reduction of PDIV 129 Boletín IIE julio-septiembre-2014 Artículo de investigación Figure 7. Oil humidity before and after PDIV tests, without and with board. from 38 kV to 28 kV was observed when the relative humidity of pressboard was increased from less than 0.5% to 2.5%, for a gap of 40 mm (Dai et al., 2010). In (Zainuddin et al., 2012) it is speculated that the leakage current flowing in pressboard can (a) heat the oil in the pores, causing gassing and, eventually, inception of PD in the pores, and (b) give rise to phenomena similar to dry band arcing when PD within the pores is sufficiently energetic to dry the pressboard further and release gases. Since ester oils are highly hygroscopic they contribute to drying of pressboard during the impregnation phase and beyond, thereby preventing high current densities in the board. These effects were confirmed experimentally by measuring moisture content of the oil at the beginning and end of the PDIV tests for corona (without board) and creepage (with board) discharges. A period of approximately 24 hr was required to complete measurements at one gap length for each oil. These data are presented in figure 7. It will be seen that the moisture content of both oils did not increase appreciably for tests performed without pressboard sheets (the observed increase is due to diffusion from the atmosphere into the oil). However, the increase is much more marked for EO for tests performed using a pressboard sheet. These observations show that moisture extraction continues after drying and impregnation, and can help to ensure reliable longterm behavior of the insulation. Conclusions EO has proved also to be a viable substitute for MO from the point of view of PDIV at least when new, dried and highly purified fluids are used for the comparison. The smaller permittivity mismatch between pressboard and EO, compared to that between pressboard and MO, leads to a more uniform field distribution at the oil/pressboard interface. We have observed that EO has higher PDIVs than MO for corona and creepage discharges. However, our understanding of the oil/pressboard interface is incomplete. There is a lack of standardization in PDIV measurements, particularly in the case of creepage discharges. Since industry is striving to provide new insulating oils for use at increased transmission voltages, improving electrical testing techniques and the understanding of the oil/pressboard interface is important. 130 Artículo de investigación Acknowledgments Carlos Azcarraga gratefully acknowledges the Instituto de Investigaciones Eléctricas for study leave and the Mexican Science Council (CONACYT) for financial support of his graduate studies. References CIGRE WG A2-35. Experiences in Service with New Insulating Liquids. Final Report; Brochure 436, 2010. Bertrand Y. Development of a low viscosity insulating fluid based on vegetable oil. In 2012 IEEE ISEI, Puerto Rico, USA, June 2012, pp. 413-418. M. Pompili and C., Bartnikas, R. Mazzetti. Comparative PD pulse burst characteristics of transformer type natural and synthetic ester fluids and mineral oils. IEEE Trans. Dielectr. Electr. Insul., vol. 16, no. 6, pp. 1511–1518, 2009. A. Cavallini, X. Chen, G.C. Montanari, and F. Ciani. Diagnosis of EHV and HV Transformers Through an Innovative Partial-Discharge-Based Technique. IEEE Trans. on Power Delivery, vol. 25, no. 2, pp. 814-824, April 2010. J. Dai, Z.D. Wang, and P. Jarman. Creepage Discharge on Insulation Barriers in Aged Power Transformers. IEEE Transactions on Dielectrics and Electrical Insulation, vol. 17, no. 4, pp. 1327-1335, August 2010. H. Zainuddin, P.L. Lewin, and P. M. Mitchinson. Characteristics of Leakage Current During Surface Discharge at the Oil-Pressboard Interface. In Annual Report Conference on Electrical Insulation and Dielectric Phenomena (CEIDP), Montreal, CA., 2012, pp. 483-486. V. Sokolov and Berler Z. and Rashkes V. Effective methods of assessment of insulation conditions in power transformers: A view based on practical experience. In Proc. of Electrical Insulation Conference (EIC), Cincinnati, OH, Oct. 1999, pp. 659–667. ASTM D2413 - 99(2009). Standard Practice for Preparation of Insulating Paper and Board Impregnated with a Liquid Dielectric. Wang Z.D. and X.Wang, P.Jarman, G. Wilson Q. Liu. Discussion on Possible Additions to IEC 60897 and IEC 61294 for Insulating Liquid Tests. IET Electric Power Applications, vol. 5, no. 6, pp. 486-493, 2011. IEC 61294 (Ed.1) – 1993. Insulating Liquids - Determination of the Partial Discharge Inception Voltage (PDIV) - Test Procedure. J.H. Mason. Breakdown of Solid Dielectrics in Divergent Fields. Proceedings of the IEE - Part C: Monographs, vol. 102, no. 2, pp. 254-263, September 1955. Schultz K., Küchler A. Dahinden V. Function of Solid Insulation in Transformers. Transform 98 - Germany, pp. 41-54, April 1998. P. M. Mitchinson, P.L. Lewin, B. D. Strawbridge, and P. Jarman. Tracking and Surface Discharge at the Oil-Pressboard Interface. IEEE Electrical Insulation Magazine, vol. 26, no. 2, pp. 35-41, March-April 2010. M. Pompili and C., Bartnikas, R. Mazzetti. PD pulse burst behavior of a transformer type synthetic organic ester fluid. IEEE Trans. Dielectr. Electr. Insul., vol. 15, no. 6, pp. 1498–1506, 2008. 131 Boletín IIE julio-septiembre-2014 Artículo de investigación CARLOS GUSTAVO AZCÁRRAGA RAMOS [carlos.azcarraga@iie.org.mx] PhD in Electrical Engineering from the University of Bologna (2011- 2014). MSc degree in Electrical Engineering at the Mexican National Polytechnic Institute (2002-2004, Cum Laude). BSc degree in Electromechanical Engineering from the Zacatepec Institute of Technology (1991-1995, honors). After pursuing the AIT program in 1997, Carlos Azcárraga was signed by the Electrical Equipment Department of the IIE to work in the development of state of the art diagnostic techniques for power transformers and substation equipment. He has represented the IIE in several technical forums, including Cigré SC A3 High Voltage Equipment. He is an author and coauthor of more than 30 technical papers related to high voltage insulation assessment. He has been a Professor at the Universidad Cuauhnahuac, Universidad Morelos and at the Postgraduate Center of the IIE. His current research interests include diagnostic techniques of high voltage systems, electromagnetic transients, partial discharge physics and numerical electromagnetics. ANDREA CAVALLINI Andrea Cavallini received the MSc and PhD degrees in electrical engineering from the University of Bologna. From 1995 to 1998 he was a researcher at Ferrara University. Since 1998 he has been associate professor at the University of Bologna. His research work relates to endurance modeling and diagnostics of insulation systems. He is interested in the physics of partial discharge (PD) phenomena, e.g., physical/stochastic modeling of PD phenomena, the effects of the waveshape and frequency of the applied voltage, and the position of defects, on PD phenomena. He is also involved in more applied research concerning PD identification in different types of apparatus, e.g., cables, rotating machines, transformers, GIS with different types of insulation, e.g., polymeric, epoxy/mica, paper/oil, gas. From 2004 to 2011 he was the Italian representative in the SC D1 (Materials and Emerging Technologies) of Cigrè, convener of Cigrè WG D1.43 (Rotating machine insulation voltage endurance under fast, repetitive voltage transients) and Administrative committee member of the IEEE Dielectric and Electric Insulation Society (DEIS). He is currently chair of the IEEE DEIS Educational Committee. UGO PIOVAN Ugo Piovan is the director and owner of Trafoexperts GmbH, Switzerland, a consulting company which he founded in 2011. He was previously Technical Leader of Weidmann Electrical AG, and prior to that, Technical Manager of TAMINI Transformers. He is an expert in ultra high voltage (up to 1200 kV AC) and ultra high current (up to 120 kA) transformers. He is active in international bodies such as CIGRE and IEC, with a particular interest in standards for converter transformers (both industrial and HVDC). In 2012 he was honored with the IEC 1906 award for his contribution to the development of IEC standards on converter transformers. Ugo holds a MSc in Electrical Power Engineering from the Polytechnic of Milan. 132