You are on page 1of 12

21, rue d’Artois, F-75008 PARIS http : //www.cigre.

org

A2-305

CIGRE 2006

TRANSFORMER INTERNAL OVER-VOLTAGES CAUSED BY REMOTE ENERGISATION J. A. Lapworth, P. N. Jarman and T. Breckenridge Doble Powertest Ltd., National Grid and SP Power Systems Ltd. United Kingdom

SUMMARY There has been continued interest over recent years in transformer problems arising from interaction with the system, particularly the effects of fast transients associated with SF6 or vacuum breaker switching. This paper discusses several unexplained dielectric faults and failures that have occurred over the last 10 years in a large population of otherwise very reliable UK power transformers. No problems associated with fast transients were experienced but there appeared to be evidence of problems arising from interactions with the system of a more mundane form. All the problems could be attributed to a common failure mode: internal over- voltages arising from part-winding resonance initiated by remote energisation, either manually or by the action of delayed automatic re-closure schemes. A range of typical symptoms are described. Although the possibility of producing transformer over- voltages by remote energisation has been previously reported, it appears that the importance of the problem in causing internal faults has been underestimated and is not widely recognised. Possible mitigation measures are discussed but cannot be fully effective until this comparatively rare but nevertheless important phenomenon is understood better.

KEYWORDS
Transformer - Dielectric Failure - Internal Over-voltage - Part-winding Resonance - Remote Energisation - Transformer Feeder

jlapworth@doble.com

In other words the 'bathtub curve' is flat. with a failure rate which is low by international norms: less than 0. including chopped waves to simulate the flashover of coordinating gaps. modern power transformers are usually very reliable. Every transformer on the 400 kV transmission system has undergone lightning impulse tests at 1.fit these at critical locations where switching over-voltages are expected. UK transformers have traditionally been protected by screened coordinating gaps. rated voltage.line diagnostic tests have been developed to supplement routine DGA monitoring. Frequency Re sponse Analysis (FRA) which can detect winding movement caused by short circuits before final failure. Fortunately. switching impulse ‘type tests’ are also carried out. Nevertheless a few faults and failures still occur.line (continuous) monitoring systems are becoming available to allow closer monitoring where this is required and can be justified.5% p. And last but not least. A sophisticated system of monitoring transformers has been evolved to provide an early warning of problems. on average.g. This may be attributed to several factors: • Good specification • Experienced manufacturers • Robust dielectric testing in the factory before acceptance • A mature and well planned and operated network UK power transformers undergo rigorous dielectric testing in the factory before acceptance. In the UK it has been the practice to use graded insulation for HV voltages of 132 kV and above. but factory tests may not fully simulate all possible system events. before and after the peak over-potential.a.a. Sophisticated off. and retro.u. for the UK population of over 800 large transmission units. As far as possible. often initiated by unusual external events such as short circuits and lightning strikes. including partial discharge measurements at 1. being by far the most expensive asset in a substation and with long manufacturing times.7 p. with automated mercury. It is also likely that the ‘spare margins’ of dielectric and mechanical strength over expected operational stress built in by specification and design is degraded over time as a result of minor faults and ageing processes.5% p. less than 0. transformers are designed and tested to ensure that they withstand such extreme events. In service.u.1 INTRODUCTION Power transformers are critical elements in electric power networks. is also a ‘routine test’ on every unit. e.6 p. In the UK. an increasing variety of on. A short duration induced over-potential test at 2. and this has been developed to a fine art in the UK.free vacuum extraction systems allowing accurate analysis down to very low levels of acetylene.425 kV peak. but it has been normal practice for some years to fit surge arrestors to all new installations. since these are by far the most severe stresses that transformers experience in service. failure rates of large power transformers (where they have to be replaced or repaired at works) are very low. with neutrals solidly earthed. On every new design. Failures in the early years of service due to manufacturing faults have been virtually eliminated and up to 50 years of service experience to date does not show a correlation between failure rate and age. providing reliable trend analysis. 1 . Most faults in transformers can be detected at an early stage by dissolved gas analysis (DGA).

Similar symptoms were exhibited in most of the cases described. metallic components at a floating potential. Many are unexplained. which usually remove the transformer from service. and sometimes repaired.g. the damage was not at the HV line end. Acetylene is the key diagnostic gas. In the following. Practically any discharge event can be detected from the trace of dissolved acetylene that is left in the oil. while acoustic emission techniques may be able to provide a location.2 ppm) of acetylene in their main tanks. Most discharge events are probably triggered by system over. Sometimes this can be due to some relatively innocuous sparking activity. some discharges seem to be ‘one-off’ events. and often without any obvious cause.2 UNEXPLAINED UK DIELECTRIC FAULTS AND FAILURES Despite advances in manufacturing and monitoring technologies. These are usually dielectric in nature. e.g. although not proved: energisation of the transformer from a remote location via a significant length of overhead line. detected by a step change in acetylene in the next routine oil sample. The remote energisation may arise from either manual switching or automatic re-energisation of the overhead line following a circuit trip initiated by an automatic Delayed Auto Re-close (DAR) scheme since it is common practice in UK mesh substation arrangements for transformers to be connected to incoming lines without an intervening circuit breaker. from just DGA step changes to a catastrophic dielectric failure. With sophisticated DGA analysis techniques it is possible to detect dielectric problems at a very early stage in their development. If the discharge is of a sufficient magnitude and active. particularly differential protection • Winding over. particularly acetylene • Buchho lz gas alarms and/or oil surge trips • Operation of electrical protection. but it is not always possible to correlate with certainty evidence of discharge activity with a particular system event. including some but not necessarily all of the following: • Step changes in DGA. and yet often no fault is found and the transformer is successfully returned to service. for almost all of these a common cause can be suggested. with a range of consequences. this invariably involved the HV or series winding. However. Many dielectric faults are detected and subsequently diagnosed. Radio Frequency Interference (RFI) techniques can be used to detect it and determine under which circumstances it occurs. Most UK transformers operate with no detectable level (< 0. Significantly. several cases of dielectric faults and failures are described. usually at some 2 . with no subsequent sign of further activity and no prospects for further investigation. judging by the level of acetylene.voltages arising from switching or lightning strikes. which tap-position.voltage damage Where winding damage was caused. some unexpected or 'random' faults and failures still occur. but elsewhere. Some of these discharge events are of immediate concern since they cause protection operations. However. e. In some cases the discharge event must have been substantial. the most common examples being loose winding clamping screws or floating stress shields in tap-changers.

routine monthly oil sampling revealed a step change in main tank acetylene to 20 ppm (Figure 1). presumably caused by some switching transient phenomenon. as expected for this particular design of transformer which has no known generic dielectric faults.discontinuity between parts of the HV winding. The tap-changer could be ruled out as the cause of the problem because. as per usual UK practice. 100 90 80 70 60 PPM 50 40 30 20 10 0 28/10/1995 11/03/1997 24/07/1998 H2 06/12/1999 CH4 C2H4 19/04/2001 C2H6 01/09/2002 C2H2 14/01/2004 28/05/2005 Small step change in main tank acetylene for generator transformer Figure 1 Subsequently. to about 750 ppm acetylene (Figure 2). it was installed in a separate oil compartment. None of the other 14 single phase units of this design have exhibited any such symptoms of a dielectric problem. so some form of internal over-voltage event arising from this is suspected as the most likely explanation.2 Case 2: Large step change in DGA This is an example of a much more substantial step change in DGA. Since the sample had been taken some time before and the transformer was still in service it did not seem likely that this was part of a rising trend before failure. Previously there had been no detectable acetylene in the main tank. In November 2000 the analysis of a routine yearly oil sample from a 40 year old 275/132 kV 120 MVA autotransformer showed what can only be described as a huge jump. 3 CASE EXAMPLES 3.1 Case 1: Small step change in DGA This case involves a 30 year old generator transformer: one phase of a 423 kV 783 MVA single phase bank. 3. such as the junction between the main and tap windings.voltage was recently observed when one of the transformers at the station was de-loaded. but earlier that month the long overhead lines to the station tripped several times during summer storms and were reclosed by DAR. In late June 2000. with dead line charge being applied from the remote end. with no significant recurrence of the discharge activity. and gas levels were much lower there. the DGA history has shown a downward trend in gas levels. since it would 3 . It may be relevant to note that an unexpected transient over. At the moment there is no definite explanation for what appears to have been a ‘one-off’ event.

which failed to find any indication of a dielectric fault. 800 700 600 500 PPM 400 300 200 100 0 24/07/1998 09/02/1999 28/08/1999 15/03/2000 01/10/2000 H2 CH4 19/04/2001 C2H4 05/11/2001 C2H6 24/05/2002 C2H2 10/12/2002 28/06/2003 14/01/2004 Large step change in main tank acetylene for transmission transformer Figure 2 Since the design.e. the hydrogen concentration was much lower than the acetylene. which were re-closed with dead line charge being applied from the remote ends. Moreover. which was located to the bottom of the windings by acoustic emission techniques. 3. A decision was made to accelerate the planned replacement of both transformers.e. suggesting an ‘old’ discharge event. both due to internal defects.most likely already have failed. The RFI survey at site had detected intermittent discharge coming from this transformer. However. i.3 Case 3: ‘Spurious ’ Protection Operation In the summer of 2001 a relatively new 275/132 kV 240 MVA auto-transformer installed at a remote part of the UK grid system was tripped out of service by differential protection and Buchholz oil surge relays immediately after being energised from a remote location.line diagnostic tests failed to find any dielectric fault. this was the obvious explanation for the step change in dissolved gases. so the transformer was re-energised off the system from the tertiary to working volts for several hours using a diesel generator set without any problem being found. It seems too much of a coincidence that two transformers at the same substation should exhibit similar substantial step changes in DGA in the same time period. Because of lingering concerns. i. apparently a winding dielectric fault rather than sparking at a loose winding clamp. but of a lesser extent. A more likely explanation for the apparently coincident ‘one-off’ discharge events is considered to be over-voltages arising after tripping of the lines to the substation. As soon as possible the transformer was switched out of service for off. Oil samples indicated that a discharge had taken place in the main tank. an internal inspection was also carried out but 4 . with the second one being the priority. Off. RFI checks failed to detect any discharge and further oil samples showed a downward trend.line diagnostic tests. had a history of sparking activity at loose winding clamping screws. and this transformer. no evidence of an active fault. analysis of oil samples taken from a sister transformer at the same site had shown a similar step change in the same period.

2 p. so the possibility of an internal flashover caused by a switching transient must also be considered. which failed to find any definite evidence of damage.line diagnostic tests suggested the possibility of a dielectric fault that had been observed previously in other transformers of the same design. In early 1998 another transformer of the same 30 year old design had failed catastrophically at another site as a result of a tracking fault along an inter-phase barrier board.again no fault was found. A decision was made to take the transformer to a manufacturer’s works for detailed examination. a ‘one-off’ overvoltage incident caused by this phenomenon was originally suspected. so the closest barrier board had been able to come into contact with the paper conductor insulation of the series windings after becoming warped with age. so the transformer was removed from service. Off. The transformer was subsequently the subject of a comprehensive discharge detection experiment in which the transformer was energised from the tertiary using a diesel generator set at various voltages up to 1. which appeared to be coming from within the winding assembly. No conclusive evidence of a problem was found. Discharge was detected. Presumably the internal discharge was not at a critical location since there has apparently been no subsequent deterioration. However. but a precise and unambiguous location for the fault could not be agreed. Because ferro-resonance alarms were activated during the initial incident. With this particular design there are no external wraps on the winding assemblies. Unfortunately. Figure 3 5 . Eventually the transformer was switched back into service and has operated without any further problem since. 3. ferroresonance would normally only produce core overheating and possibly a Buchholz alarm.000 MVA auto-transformer at a substation that was teed off a long 400 kV line started to show signs of dielectric distress after being energised from the other end of the 400 kV line instead of the normal practice of being energised locally from the 275 kV side.4 Case 4: Dielectric Faults of Auto-Transformers of a Particular Design In 2001 a 400/275 kV 1. between the middle (400 kV) and bottom (275 kV) of the series winding (Figure 3). Some insulation external to the main windings was replaced and dielectric testing was carried out before the transformer was returned to service at another substation.u. after a short time further evidence of a dielectric fault appeared. so an internal inspection was carried out.

5 Case 5: Dielectric failure after remote switching A new protection system was being commissioned on a circuit comprising a 92 km overhead line route incorporating a 2 km long cable 9 km from the substation.100 90 80 70 60 50 40 30 20 10 0 18/2/82 18/2/83 18/2/84 17/2/85 17/2/86 17/2/87 17/2/88 16/2/89 16/2/90 16/2/91 16/2/92 15/2/93 15/2/94 15/2/95 15/2/96 14/2/97 14/2/98 H2 CH4 C2H4 C2H6 C2H2 DGA history in years before failure Figure 4 The transformer was being monitored by monthly oil samples because the DGA signature was considered to be of concern. Some years previously another transformer of the same design at another substation had been switched out of service after developing a gassing fault that resulted in persistent Buchholz alarms. There had been a step change in the DGA signature 6 years prior to the failure. There was no obvious evidence from the DGA history (Figure 4) of any imminent failure: in fact all dissolved gases had been showing a consistently falling trend up to and including the last sample. but it is believed that an over. presumably due to some over. Obviously. The particular winding design used (simple disc) may also be a factor in allowing a higher than expected internal overvoltage at a particular point. taken some time before. dielectric damage was observed in the major insulation between common and tertiary windings. the last sample having been taken 14 days before the failure. most likely a lightning strike on the attached line. The subsequent investigation revealed the same inter phase barrier board tracking fault. In addition. It was probably relevant that this transformer was attached to long lines over a range of mountains which are known to suffer from a higher than average incidence of lightning activity. of any dielectric fault at that stage. so it would appear that a significant over-voltage event had been experienced. As part of the protection commissioning the transformer was energised 6 6 . 3.voltage event is a necessary pre-requisite to initiate the generic failure mode.voltage event. so this particular fault had apparently developed to failure within a year. There was no evidence from the previous routine oil sample. At the substation a 34 year old 400/132 kV 240 MVA autotransformer was connected directly (without a circuit breaker) to the line. possibly at the bottom of the series winding (junction of common and series windings). particular features of this design allow this type of tracking failure to occur and ageing processes obviously also contributed.

thereby causing an explosion.7 Case 7: Dielectric failure after lightning strike In late 2003 a 30 year old 275/33 kV 100 MVA transformer tripped out of service after a lightning strike at the other end of the attached 275 kV overhead line. On the occasion in question. 3. Oil samples confirmed 7 . the transformer had to be energised from another substation 70 km away via an SF6 breaker.times from the remote end of the line. On the sixth energisation there was a violent explosion within the tank that operated the Buchholz oil surge trip and caused a minor rupture in a tank weld. Subsequent inspection revealed that one phase of the centre-entry series winding had flashed over from about half way down the lower half of the winding. The initial impulse failure left a small mark on the outer copper conductor and a puncture through the conductor insulation and the first winding wrap. because of maintenance work. Almost immediately the transformer was tripped out of service by protection and the pressure relief device operated. A flashover on the main tank side of the tap-changer barrier board had taken place between tap leads at the ‘top’ of the tap winding (closest to the main winding) (Figure 5). The stress ring then broke down to the core taking the full 132 kV to earth fault current.6 Case 6: Dielectric failure after remote switching This case involved a 40 year old 132/33 kV 60 MVA distribution transformer that was normally energised from another substation 30 km away via a bulk oil breaker. along an insulating wrap to the stress ring at the bottom 132 kV end of the winding. which had been protected with screened co-ordinating gaps. There was no previous history of problems or bad DGA results on this transformer. Flashover damage between tap leads inside main tank Figure 5 3. Note that this transformer had surge arrestors fitted to the HV side.

which could be a power transformer or a wound VT.a dielectric fault in the main tank. Therefore a decision was made to re-energise the transformer from the LV side using a diesel generator set to confirm and locate any discharge fault. Flashover damage on lead linking two HV windings Figure 6 4 POSSIBLE MECHANISMS The faults and failures described above were all apparently due to internal over-voltages. and this was located to the B phase after the star connection of the HV winding was broken in the tapchanger to separate the phases. Could these have been responsible for the faults described ? Ferro-resonance is an oscillatory phenomenon caused by the interaction of system capacitance with the non. The resonance is driven by capacitive coupling from an energised parallel circuit. most likely arising as a result of resonance and initiated by switching transients. The design of the transformer in question features a split double concentric HV winding and the subsequent investigation showed that the dielectric fault was a flashover to the earthed core from the lead connecting the two halves of the HV winding.line diagnostic tests suggested a dielectric fault affecting the HV winding. Two well known but relatively rare resonance phenomena affecting transformers are ferro-resonance and part winding resonance. There had been no evidence of any prior dielectric problem with this transformer or any others of the same design. the lightning strike was not to this phase.linear inductance of a transformer. but when the overhead line was re-energised by DAR after the lightning strike. Off. A discharge fault was detected at the top of the B phase winding assembly. The voltages involved depend on the 8 . However. or possibly through the grading capacitors of a circuit breaker and can only occur when the ferro-resonating circuit is switched out but not earthed. which passed over the top of the winding assembly (Figure 6). It would appear that this failure was due to an internal overvoltage of significant magnitude that arose not as a direct consequence of the lightning strike.

9 . but earlier work [4] provided documented examples of energising over-voltages on transformer feeders and explained how reflections of the initially impressed voltage step within the attached line can provide the necessary harmonic excitation at the transformer. Any direct damage from the phenomenon is expected to be due to core saturation and unexpected stray flux. Whereas it may be possible in some circumstances.u. In view of the evidence that remote switching can generate damaging internal over. 3] remote energisation was not considered a major problem. so the first consideration should be whether this can be avoided. That work was concerned mainly with over-voltages on the secondary side of the transformer and did not consider the possibility of internal resonances. ferro-resonance is unlikely to be a direct cause of dielectric failure. In recent reviews of transformer-system interactions [2. Because of the low voltages involved. they are usually less than 1 p. The experience described here suggests the seriousness of this particular form of switching in terms of producing internal transformer damage has been underestimated. to avoid this possibility when carrying out manual switching of the network.voltages. particularly when the possible consequences are realised. or as a result of routine chopped lightning impulses which would be expected to excite any resonant frequencies and so adequately test the winding. Power transformers are not tested to withstand ferro-resonance: it is considered to be more a nuisance than a serious threat. However. Presumably such tests do not adequately simulate the ability of the system to resonate with the transformer.u. allowing the possibility of internal resonances when excited by appropriate impulses – the part winding resonance phenomenon. to avoid or minimise the likelihood and consequences of similar events. 5 MITIGATION MEASURES Even if the causes of the faults and failures described are not fully understood. since the resonance is usually sub-harmonic. each of which may have a different and possibly variable surge impedance due to the distributed nature of winding capacitances and inductances. it does not seem likely to be able to avoid the remote energisations that occur when lines are automatically switched back in after lightning strikes. but it could leave the core in an unusual state of residual magnetism. since many lines have transformers at both ends. Remote energisation appears to be a common factor in the observed faults. The junction between the main and tap windings of an HV winding is a recognised discontinuity.resonant frequency and the saturation limit of the transformer. system frequency resonance can occur in which case voltages up to about 1. Power transformers have more than one physically separate winding. it is worth questioning why such problems have not been identified by factory Recurrent Surge Oscillograph (RSO) measurements on transformer tap windings during switching impulse tests. could be experienced. one could still consider possible mitigation measures. or perhaps cause enhanced transient voltages if re-energisation occurs during ferro-resonance.3 p.ignitions after breaker operations or reflections in the attached network can provide the appropriate harmonic content (kHz) to excite internal winding resonances [1]. Part winding resonance is the recognised explanation for how damaging over-voltages can arise within a winding while not at the terminals and it is known that oscillatory switching transients arising from repetitive re.

The installation of surge arrestors to limit over-voltages is another obvious measure. It would appear that remote energisation can result in damaging internal over-voltages in transformers. A better understanding of the phenomenon is required before practical mitigation measures can be recommended. is critical. but the risk of causing internal damage appears to have been underestimated.voltages are not produced every time a remote energisation occurs. Previous workers have shown that energising a line terminated by a transformer can generate significant resonant over-voltages. Lastly. If the relevant parameters were understood it would be an option to relocate susceptible designs away from critical locations and avoid such designs in the future by specification and test. further consideration should be given to avoiding these situations whenever possible. it is not possible to make much progress in implementing mitigation measures.voltages. the precise point on wave of switching or the degree of phase imbalance.g. in the absence of a sufficient understanding of the problem. particularly the resonant frequency characteristics of the windings. 10 . but it is not clear from experience so far that these will be effective for this particular phenomenon. Since there is increasing evidence that serious internal dielectric damage can be caused to transformers by such events. will probably have an important influence on whether damaging internal overvoltages are produced. and is known to have caused internal damage in other switching situations. the transformer design itself.g. so it is probably the case that some aspect of the way in which the switching occurs. even though they occur relatively rarely. Since another key aspect of the problem is the interaction of system capacitance with the transfo rmer. In particular there is a need to record examples of transient terminal waveforms so as to determine the voltage amplitudes and critical frequencies involved. 6 CONCLUSIONS It would appear that several unexplained dielectric faults and failures on UK transformers have been initiated by a common event: energising from a remote location. e. e. Therefore it may be possible to avoid the problem by controlled switching. either intentionally or by the operation of an auto-reclose scheme following a line trip. It is likely that the winding design of the transformer is an important aspect in determining its susceptibility to damage from such events. Unfortunately. one should consider if helpful changes to system or terminal capacitances can be made. It is suspected that damaging over. These appear to occur particularly at discontinuities in the HV winding. particularly if the over-voltages at the terminals are not high enough to trigger the surge arrestors. Part winding resonance is suspected as providing the mechanism for generating damaging internal over. the junction of the main and tap winding. Two of the cases discussed involved transformers with simple disc winding arrangements which are known to have less smooth frequency responses from FRA measurements.

2005 L. al.. Preininger et. pages 83-104) 11 . “Electrical Environment of Transformers – Impact of Fast Transients”. Glinkowski et.7 ACKNOWLEDGEMENTS The authors would like to acknowledge the contributions of several colleagues from various UK utilities in providing information for this paper and agreeing to publication. Foreman and H.07 “A Guide to Describe the Occurrence and Mitigation of Switching Transients Induced by Transformer and Breaker Interaction”. Glavitsch. BIBLIOGRAPHY [1] [2] [3] [4] G.6. “Resonance Behaviour of High-Voltage Transformers” CIGRE 1984 Session paper 12-14 presented by Working Group 12. F. al. M. April 2004.142/D1..21. 18. Csuros. IEEE Draft Standard PC57. Summary paper of CIGRE JWG 12/13/23. K. CIGRE Study Committee 33 paper “Energising Overvoltages on Transformer Feeders” (Electra No. particularly Simon White of British Energy and Duncan Shepherd of Scottish & Southern Energy. July 1971.