Draft - to be published and presented with "2012 IEEE International Symposium on Electrical Insulation", 10-13 June 2012, San Juan, Puerto Rico On-site Transformer Partial Discharge Diagnosis Detlev W. Gross, Markus Soeller Power Diagnostix Systems GmbH Aachen, Germany Abstract—In the lab, partial discharge diagnosis has widely replaced the traditional RIV measurements. Additionally, partial discharge acceptance levels are being reduced due to the increasing use of composite material and a growing awareness of partial discharge phenomenon and their consequences. Adequate filtering of the supply voltage for induced voltage testing and the use of sensitive acoustic measurement has greatly improved the detection and location of partial discharge activity in power transformers. An increasing population of service aged substation equipment having reached their projected service life demands on-site repair and, hence, on-site testing to factory standards. Using inverter based three-phase mobile test sets allows on-site application of tests previously limited to a mere test room environment. Besides the unmatched portability of inverterbased sources, especially the less critical generation of reactive power simplifies on-site testing, if compared with motorgenerator sets. Additionally, the lack of high short circuit currents limits the damage in case of breakdown. However, adequately removing the switching noise spectrum becomes a demanding task in order to reach the required sensitivity of the partial discharge measurements. Thus, acceptance testing as commonly applied in the lab, will be increasingly used on-site to extend the service life of old transformers, to validate the success of on-site repair, and to ensure successful commissioning of new units. Likewise, applying on-line monitoring of various operation parameters including partial discharge can assist to extend the lifetime of service aged substation equipment. II. PARTIAL DISCHARGE TESTING Detecting high frequency signals using narrow-band receivers based on heterodyne principles has been used already very early in the history of high voltage insulation systems [1]. With early meter-type instruments diagnosis was mostly limited to the observation of magnitude and inception vs. extinction voltage, while later oscilloscope-base instruments added the phase position of the discharge activity. An in-depth understanding of the gas discharge physics and the statistics of partial discharge was supported with the introduction of instruments using the phase resolved partial discharge (PRPD) pattern, or ϕ-q-n pattern [3, 4, 5]. Keywords: partial discharge; on-site; diagnosis; inverter-based; acoustic location I. INTRODUCTION The deregulation and privatization of the energy sector, which started in the 1970ies in many areas of the world, has had a further impact on transformers and their life. Before deregulation with often government-owned or government controlled utilities, availability was the most prominent design and sourcing criterion. A side effect of this policy was “overengineering”. Deregulation shifted the emphasis to profitability. As a consequence, re-investment into the grid dropped significantly. However, in light of the typical service life of a large power transformer of about 40 years, the consequences of this change took decades to materialize, while profits went up immediately. Nowadays, we do have in many parts of the world serviceaged populations of sub-station equipment, of which a large portion has reached or already exceeded its projected service life. In Europe, for instance the majority of the 400kVtransmission-system was commissioned in the 1970ies and 1980ies. Moreover, the ongoing change from fossil fuels to renewable sources further increase the required transmission capacity of a service-aged grid. Figure 1. ϕ-q-n pattern of multiple cavities (voids) As an example, fig. 1 shows such a ϕ-q-n pattern of several voids in epoxy resin. Here, each individual sine-shaped trace belongs to an individual gas inclusion. Moreover, the welldistributed pattern is caused by a low availability of the starting electron for the discharge avalanche, as it is typical for bubbles in polymeric material such as fresh epoxy resin [6]. Fig. 2, instead, shows the activity of several gas inclusions with the casein glue of barriers and spacers on top of the static shield of large distribution transformer coils. Here, increasing the field strength due to customer demands reached the limitations of the factory's production methods. With a better understanding of the deterioration processes of materials used in power transformer, partial discharge measurements gained a more prominent role in the acceptance testing of large power transformers. The relevant standards for transformer testing, such as the IEEE C57.113 [9] have shifted the emphasis to partial discharge detection with the more recent revisions. A partial discharge acceptance level of 500pC was commonly used. However, the increasing use of composite materials and their defect mechanism led to a reduction of the partial discharge acceptance test levels during the past decade. III. MOBILE TEST SET Generally, on-site partial discharge testing of large transformers is a demanding task. In order to allow tests at elevated voltage, as it is part of the standard short duration or the long duration test [10], the transformer needs to be energized at a higher frequency. Heavy motor-generator-sets are commonly used in the test room of a transformer factory. A motor-generator-set, the step-up transformer and the control circuits cover several freight container loaded up to their permissible weight, when testing transformers of 500MVA or more. Figure 2. ϕ-q-n pattern of gas inclusions trapped in casein glue Finally, fig. 3 shows a pattern that is caused by a delamination in transformer pressboard. Here, the Lichtenberg figure of the surface discharge does cause the steep increase of the partial discharge magnitude vs. phase. In order to overcome the operational and logistic limitations of such a conventional solution, an inverter-based three-phase source was developed [11]. The unit is built into a modified 40ft high-cube container (Fig.4) and stays within the load and size limitations of a conventional road-worthy trailer truck. Figure 4. Mobile transformer test set built into a 40ft container Figure 3. ϕ-q-n pattern of pressboard delamination Power transformer acceptance testing initially focussed on "radio interference voltage" (RIV). The original intention of this test, however, was to avoid hampering AM radio reception due to partial discharge activity. Therefore, narrow-band circuits and a weighting circuit were used with the RIV meters [2]. However, the used bandwidth does not allow to process high-repetition partial discharge, while the weighting circuit confuses the detection of low-repetition discharge with high magnitudes. The mobile test set requires a 400V three-phase supply feeding three individual 450kVA inverter units covering an output frequency range of 20-200Hz. The 2MVA step-up transformer consists of three single-phase transformers in a common tank. This allows running the unit in single-phase mode at full power on all three inverters. Both ends of the HV winding as well as two taps are accessible via four bushings in line for each of the single-phase units (fig. 5). Additionally, each low voltage coil has a tap as well. Thus, by selecting the LV and by applying jumpers equipped with multi-contact connectors, a large variation of output voltages ranging from 8.5kV to 90kV full-scale can be chosen by interconnecting the different taps in star or delta configuration (Table 1) to have a close match to the load requirements. Figure 6. Noise pattern due to (unfiltered) inverter switching action Figure 5. Step-up transformer with taps and HV filters Running the inverters with 120° phase shift offers threephase induced voltage testing, while 0° phase shift allows single phase induced voltage testing at full power. Although intended mainly for transformer testing, this also offers testing a cable of 5µF at 36kV and 50Hz, for instance. TABLE I. Finally, the mobile test set comes with a 500kV reactor for resonant applied voltage testing. The reactor sits on a frame that can be moved out of the container to provide the required spacing (fig. 7). The reactor has an inductance of 400H, which together with the coupling capacitor of 2nF results in a resonant frequency of about 178Hz. With the minimum frequency of the inverter of 15Hz, a load capacitance of up to 200nF can be covered for applied voltage testing. However, with increasing capacitance, the current limitation of the reactor limits the maximum voltage. Given the current limit of 4A, the 500kV can be reached up to about 25nF, whereas the resonance frequency is close to 50Hz, then. MOBILE TEST SET, 2MVA STEP-UP TRANSFORMER Configuration Three-Phase Output, Voltage and Current LV Input 1 LV Input 2 HV Output 1, Delta 11.8kV 97.9A 8.5kV 135.2A HV Output 1, Star 20.4kV 56.5A 14.8kV 78.1A HV Output 2, Delta 26.1kV 44.3A 18.9kV 61.2A HV Output 2, Star 45.1kV 25.6A 32.7kV 35.3A HV Output 3, Delta 52.0kV 22.2A 37.7kV 30.7A HV Output 3, Star 90.1kV 12.8A 65.2kV 17.7A The inverters offer full four-quadrant operation and, hence, can supply reactive power up to their output current limit. Additionally, the mobile test set comes with switchable inductive (3 x 180kVA) and capacitive (3 x 603kVA) compensation to minimize the inverter current. Generally, of course, the inverters produce switching noise, which strongly hamper partial discharge measurements, if not sufficiently filtered. Depending on the load situation, the inverters produce various impulse noise patterns. Fig. 6 gives an idea of such noise pattern. One set of filters is fitted on the LV side between inverter and transformer. Another set of filters is placed on the high voltage side (fig. 5), which additionally carry the fiber-optically-isolated load current measurement. Figure 7. 500kV reactor for applied voltage testing moved out IV. ON-SITE TESTING On-site acceptance testing of transformers including partial discharge tests is either triggered by abnormal behavior of the transformer in service detected by on-line dissolved gas analysis (DGA) or in more extreme cases by tripping a Buchholz relay. Moreover, such on-site acceptance testing greatly simplifies assessing the current status in preparation for an on-site repair or validation of the results after such an on-site repair [11, 14, 16, 17]. Of course, the transformer needs to be disconnected from the grid and prepared for the test. Setting up a mobile test set as such requires few hours, only. Typically, a rented 400V diesel generator is used to supply the mobile test set. Hence, decommissioning, measurement, and commissioning the transformer again can be done within two days, if needed. However, subsequent attempts to acoustically locate a found partial discharge source may be more time consuming. The IEC60270-2000 [7] limits the frequency range to an upper corner frequency of 400kHz, whereas the highest permissible lower corner frequency is kept to 100kHz. Often, this frequency range is (partly) occupied by noise signals, when doing field tests. Thus, a tunable detection circuit is very helpful to optimize the signal-to-noise ratio (SNR). Currently, an addendum to the IEC60270-2000 is under preparation. Redefining frequency ranges is part of the discussion. Besides the three-phase induced voltage test at elevated frequency, the unit support as well single-phase induced and applied voltage tests. The available 1.3MVA active power does not pose any limitation for no-load tests on even very large transmission transformers. However, accurately measuring the load losses is limited by the available output current and by the limited capacitive compensation. Generally, the lower cut-off frequency shall exclude residual noise signals from the three-phase-supply for induced voltage testing regardless, whether it is IGBT switching noise from an inverter-based source or thyristor noise of the excitation system of a conventional motor-generator set. The upper corner frequency shall not be chosen too high to still cover larger parts of the winding. However, for proper pulse processing a bandwidth of several hundred kHz is mandatory [8]. As an example, table 2 shows the power requirements for two different transformers tested. In both cases a diesel generator well below the maximum power was used. TABLE II. MOBILE TEST SET, POWER REQUIREMENTS Given a well de-noised source and the comments above, on-site partial discharge testing of large power transformers offers similar sensitivity levels as found in an average, nonoptimized transformer test room. Power Requirements Transformer I Transformer II 112/22kV 235/20kV Nominal Power 63MVA 660MVA Apparent Power 38kVA (@1.0UN) 240kVA (@1.5UN) Power (Diesel Gen.) 24kW (@1.0UN) 190kW (@1.5UN) Nominal Voltage V. PARTIAL DISCHARGE TESTING As in the lab, typically, the test tap or potential tap of the bushings is used for coupling to the partial discharge signal. Only for smaller LV or tertiary bushings coupling capacitors are used instead. Quadrupoles and preamplifiers (fig. 8) are fitted to make the signals available to an eight-channel partial discharge detector. Besides covering the frequency range as defined with the IEC60270, this unit offers additionally a wider frequency range using a tunable heterodyne circuit covering frequencies up to 10MHz [12, 13]. VI. ACOUSTIC LOCATION Besides electromagnetic signals and light emission, partial discharge cause also acoustical signals. Hence, its acoustic emissions and the respective travel time in different materials can be used to locate partial discharge. However, the classical triangulation approach with three sensors on three faces of a cubic tank fails in most cases of real-life transformers, as two essential conditions of this method aren’t met. Firstly, the internal medium of the tank must be homogeneous and, secondly, the tank wall must have the properties of a thin membrane [15, 17]. Instead, a transformer is filled with different materials having different density and, hence, causing different travel speeds of the sound wave. Thus, for a partial discharge source deeply buried within the insulation system, the different possible travel paths, their delays and attenuation strongly hamper the location of the source. Additionally, the tank wall, instead of being a thin membrane, has its own transmission properties and does add lateral transmission within the steel as an additional path. Thus, understanding the transmission paths and their properties of a remote partial discharge source in a winding is as complex as trying to visually locate a light source within a complex structure of glass of different refraction indices, for example. Figure 8. Quadrupoles and preamplifiers fitted to the test tap Typically, before attempting an acoustic location, an indepth partial discharge diagnosis is performed in order to understand the rough location of the source in terms of dominant phase, phase-to-phase, or phase-to-ground discharge, for instance. Based on these results, the suspected area of the transformer is scanned placing several acoustic sensors on the tank wall (fig. 9) to find acoustic signals correlated to the electrical partial discharge signal. Figure 9. Acoustic sensors fitted to the tank wall using magnets Therefore, an oscilloscope or equivalent equipment is triggered to the dominant electrical signal, while displaying the averaged acoustic signals of the piezo sensors placed. Fig. 10 shows such typical signals with different travel time for the sensor positions. The ideal situation is having a clear oil path. Here, the typical transmission speed of about 1.4m/ms can be used for calculating the distance. Of course, knowing the internal structure of the transformer under test and the impact of different materials is essential and detailed drawings should be at hand. Practical experience of acoustic location of about one hundred large power transformers led to a comparably simple method. After having found a correlating acoustical signal, three sensors are placed in a row, whereas the positions are optimized in order to have the center sensor showing the shortest distance. Placing the sensors in a row reduces the triangulation to a two-dimensional, i.e. “flat problem”. Having started with a horizontal line, for instance, the sensors are then placed in a vertical line at the horizontal position found in the first step. This process is assisted by a software tool to create then the three-dimensional position results. Fig. 11 shows the oscilloscope screen of this software, while the location screen is found with fig. 12. Figure 10. Acoustic travel time with respect to the electrical PD signal Given the complexity of a transformer’s insulation system, the key strategy, when optimizing the sensor positions, is to have a signal path as simple as possible. Figure 12. Viewing the resulting position based on the triangulation The two-dimensional system is intentionally over determined in order to address the effects of signals traveling in the tank wall. The transmission in steel is about four times faster than in transformer oil and, hence, can produce confusing results in case of a lateral location of the discharge source. Especially with mounting turrets of high voltage bushings, the high speed in steel can produce misleading results, when signals via the epoxy resin of the bushing and the steel of the turret appear earlier at the sensor than the signals of the direct oil path. With a clear oil path, the acoustic signal has a sharp inception and placing a cursor either manually or automatically on the graph is easy. In such cases an accuracy of about +/-2cm can be achieved. If several independent transmission paths contribute to the signal placing the cursor and, hence, determining the distance becomes more and more difficult. However, even here in most cases a location precision of +/20cm is possible. Often, the precision can then be improved with the analysis of the partial discharge pattern, its properties, and the cross-coupling matrix with respect to phases and coils. Figure 11. Analyzing the vertical and horizontal results Based on this analysis, areas are identified, where the observed partial discharge activity is physically possible and compared with location result of poorer precision. VII. MONITORING Generally, partial discharge monitoring suffers from electromagnetic noise interference as found in a substation environment. Thus, the local noise situation needs to be understood before setting up a monitoring system. Figure 13. Bushing coupling unit and adapter connected to a test tap The smoothest application can be expected, if the transformer is connected to cables and/or gas insulated switchgear (GIS). Here, almost lab-type frequencies in the range of the IEC60270 can be used due to the low interference level. Partial discharge coupling is made using the measurement taps of the bushings. Care must be taken to have this connection made durable in order to maintain a safe operation of the bushing. Fig. 13 shows such coupling to a bushing tap – the measurement impedance that separates the power frequency synchronization and the high frequency signal follows a mere mechanical adapter with surge protection. Figure 14. Fitting an UHF sensor to a transformer's oil drain valve Alternatively, higher frequencies can be covered with builtin UHF antennas mounted on a spare flange or fitted to a drain valve of the transformer under test (fig. 14). Sadly, often transformer design does not provide such UHF access via the drain valve. In more than 50% of the cases the drain valve does not offer a fully open cross-section, continues internally with an elbow, faces a 45° stiffener plate, or runs via pipe directly to the conservator. Fig. 15 shows an example of such a partial discharge monitoring instrument (upper left corner) as part of an overall transformer monitoring system. Here, the instrument talks to the local monitoring system covering temperatures, voltages, and other parameters, which then reports to the SCADA system. Alternatively, the PD monitoring device itself is already equipped with a TCP/IP interface acting according to the IEC61850. Besides the values for partial discharge monitoring, which are already included with the relevant section of this standard, further parameters can be added into the model description, including the ϕ-q-n pattern or more in-depth trending information, for instance. Figure 15. PD acquisition system as part of an overall monitoring system VIII. SUMMARY Besides wanted effects, deregulation has caused a reduced investment in the grid and, hence, has caused a larger population of service-aged substation equipment. [4] [5] Mobile, inverter-based transformer test sets for three-phase induced voltage tests allow assessing the condition of transformers on site with partial discharge acceptance tests to factory standards. [6] Lower weight, non-critical supply of reactive power, and smooth breakdown handling makes inverter-based sources by far superior to conventional motor-generator sets. Additionally, a large reactor offers applied voltage testing in resonance mode. [8] Combining the partial discharge measurements with acoustic location of the partial discharge activity assist to prepare for an on-site repair or to decide for factory repair of a faulty unit. During the past decade, the described method of acoustic location has been successfully applied to about one hundred large power transformers having had critical partial discharge activity. Continuous on-line partial discharge monitoring allows keeping transformers in service, which have already reached their projected service life. REFERENCES [1] [2] [3] Peek, F.W., Dielectric Phenomena in High Voltage Engineering. McGraw-Hill Book Company, 1915, pp. 216-217. NEMA 107 "Methods of measurement of radio influence voltage (RIV) of High-Voltage Aparatus", NEMA Pub. No. 107-1987. Niemeyer, L. "A generalized approach to partial discharge modeling", IEEE Trans DEI, Vol. 2, No. 4, August 1995, pp. 519-528. [7] [9] [10] [11] [12] [13] [14] [15] [16] [17] Heitz, C. "A generalized model for partial discharge processes based on a stochastic process approach", J. Phys. D, September 1999, pp. 10121023. Fruth, B., Gross, D. "Phase resolving partial discharge pattern acquisition and spectrum analysis", Proc. of the ICPDAM, July 1994, Brisbane NSW, Australia, 94CH3311-8, pp. 578-581. Gross, D.W., Fruth, B.A, "Distortion of phase resolved partial discharge pattern due to harmonics and saturation," CEIDP 1998, Atlanta, GA, USA, October 25-28, 1998, pp. 416-419. IEC60270, High-voltage test techniques - Partial discharge measurements, CEI/IEC 60270:2000. Gross, D., "Signal transmission and calibration of on-line partial discharge measurements," ICPADM, Nagoya, Japan, June 1-5, 2003, 03CH37417, pp. 335-338. IEEE C57.113-2010 "IEEE Recommended Practice for Partial Discharge Measurement in Liquid-Filled Power Transformers and Shunt Reactors". IEEE C57.12.90-2006 "IEEE Standard Test Code for Liquid-Immersed Distribution, Power, and Regulating Transformers" Werle, P.; Kouzmine, O. "New methods for condition assessment and onsite testing," Transformer Life Management 2010, Schering Institut, Leibnitz Universität Hannover, Halle, Germany, June 21-22, 2010. Gross, D., Soeller, M. "Partial discharge diagnosis on large power transformers," ISEI 2004 Conference, Indianapolis, IN, September 1922, 2004, 04CH37561C, pp. 186-191. Gross, D., Soeller, M., "Partial discharge acceptance testing and monitoring on power transformers," ETG Fachtagung, Diagnostik el. Betriebsmittel, Berlin, Germany, February 26-27, 2002, ISBN 3-80072671-8, pp. 213-216. Bräunlich, R., Hässig, M., Fuhr, J., and Aschwanden, T. "Assessment of insulation condition of large power transformers by on-site electrical diagnostic methods", ISEI International Symposium on Electrical Insulation, Anaheim, CA, U.S.A., April 2-5, 2000. Bengtsson, T, Kols, H., Jönsson, B., "Transformer PD diagnosis using acoustic emission technique," Conf. Proc. of ISH, August 25-29, 1997, Montréal, Canada, Vol 4, pp. 115-119. Carlson, Å., Fuhr, J., Schemel, G., and Wegscheider, F. Testing Power Transformers. ABB Business Area Power Transformers, 2003, ISBN 300-010400-3. Gross, D. "Locating partial discharge using acoustic sensors, " Highvolt Kolloquium, Prüfen und Messen an elektrischen Betriebsmitteln der Hochspannungstechnik, Dresden, Germany, May 19-20, 2011.