|Home | Articles | Forum | Glossary | Books|
MAINTENANCE IN SERVICE (cont.)
The second case-study might be considered less of a success in that it did not enable a fault to be pinpointed and repairs made. This is partly because there were a number of relatively minor faults taking place concurrently and partly because operations and maintenance staff had carried out a number of degasifications of the oil. Maintenance staff can sometimes be faced with a dilemma when transformer dissolved gas levels are increasing. The concern is that these might reach a level at which free gas will be released leading to Buchholz relay operation. To avoid this the strategy is to treat the oil in order to lower dissolved gas levels. This procedure unfortunately greatly confuses efforts at fault diagnosis, since although gas-in-oil levels are reduced, the gas in the insulation remains at a high level. This then diffuses into the oil until an equilibrium is reached, thus increasing gas-in-oil levels, but at rates which are not related to the fault. When attempting to obtain a diagnosis, therefore, it is always preferable not to treat the oil.
The transformer was a 19.5/300 kV generator transformer and the dissolved gas levels first gave cause for concern when the transformer had been in service for about 6 years and continued for a further period of some 14 years until it was finally removed from service for scrap following Buchholz relay operation. Table 12 gives the dissolved gas levels over this 14 year period and indicates the timing of oil treatments. Rogers Ratios are also included. During the final 3 years of service the generator voltage was restricted to 18.7 kV due to machine problems and it will be noted that during this period dissolved gas levels were seen to stabilize. Throughout the period in service the generator AVR was known to exhibit a control problem which resulted in genera tor voltage frequently exceeding 19.5 kV. It was considered that some of the transformer problems were due to overfluxing resulting from these periods of overvoltage and this was probably confirmed by the reduction in gas evolution following the reduction in machine operating voltage. It will be seen that there is no clear pattern to the Rogers Ratios. Because of the complex gas evolution history of this transformer and the large amount of monitoring data which had been amassed it was decided that it might well be instructive to dismantle it for as detailed an inspection as possible before totally scrapping it. The inspection revealed that there had been several faults, some of which had probably developed earlier than others and some which probably owed their origin to the overfluxing. Among the faults identified were:
• Arcing of winding clamping-pressure adjusting screws.
• Arcing of a connection to a winding stress-shield.
• Burning of core plates at their edges consistent with severe circulating currents.
• Indication of overheating of core frames and adjacent core frame insulation.
It was considered possible that these latter two faults owed some of their existence to the overfluxing incidents.
This latter case study demonstrates some of the difficulties which can be experienced on some occasions when attempting to draw meaningful conclusions from d.g.a. results. Dissolved gas analysis can, at best, serve to alert an operator to the existence of a problem. There can then often be many additional problems such as whether to take the unit out of service in an effort to locate the fault by means of an in-tank inspection and, even if this decision has been taken, should the fault be buried deep within windings it will not be located from such an inspection. The next problem is then whether to go further and take the unit completely out of commission for dismantling. Most operators will , rightly, fight shy of this decision.
Very occasionally it will be the case that a serious fault will be detected of a considerable magnitude. Often, despite the seriousness, there will be pressures to retain the transformer in service, perhaps until an approaching out age, or perhaps until a spare unit can be brought from another location. In these circumstances very frequent sampling will be called for and it is possible that it will be economic to install an on-line gas monitor. Such equipment can be connected into the oil circuit and arranged to take and analyze samples at intervals as frequently as hourly. An alarm level, either for a particular gas, or for total gas content in the sample, can be set to indicate at a remote location should this alarm level be exceeded.
An early device of this type was developed by CEGB in conjunction with Signal Instruments of Camberley, Surrey. For some years it continued to be manufactured and marketed by Signal Instruments, although being quite costly and not, therefore, justifiable except in special circumstances, there was a very limited market so that production was discontinued in the early 1990s.
A more economical on-line device was the Hydran continuous gas monitor which was also developed in conjunction with CEGB. This operates on the principle that hydrogen is produced whenever there is a fault (see FIG. 122). It is therefore designed only to detect the presence of hydrogen and can be set to alarm as soon as this is found in a continuous sample.
The disadvantage of this device is that it simply alarms at the presence of a particular gas. As should be evident from the above, it is not so much the presence of gas, or gases, which are indicative of a fault so much as a sudden change in the status quo. It is understood that the most modern versions of the Hydran can, in fact, be set to ignore a steady situation and only raise the alarm should a step-change occur.
Certainly there are occasions when it can be beneficial to monitor a transformer with some such device which is able to warn of a very rapidly developing fault. Whenever a manual system of sampling is instituted there is a limit to the frequency at which this can be carried out. It is impractical to operate a system of taking routine samples more frequently than, say, every one or 2 months and a lot can happen in this interval. It is understood that the National Grid Company in the UK now specifies that all new transformers supplied for the 400 and 275 kV grid systems are provided with provision for installation of on-line dissolved gas monitoring devices.
Another proposal worthy of consideration and requiring the facility for easy connection of an on-line gas monitor, is that all newly commissioned EHV transformers should be monitored on-line for at least the first 3 months of service.
Degradation of cellulose
Although by the late 1970s considerable progress and many successes had been achieved in CEGB using the dissolved gas analysis techniques described, one or two major catastrophic failures which had not been predicted had occurred and these underlined one of the weaknesses of the dissolved gas analysis approach. One of the problems is that at very low levels of overheating, which are nonetheless serious enough to result in significant shortening of life expectancy, the volumes of hydrocarbon gases produced are so low that it is difficult to measure their concentration in the oil with any accuracy.
Column C18 alkane bonded Peaks to silica 1 2-furoic acid Particle size: 10 µm 2 solvent front Length 300 mm 3 5-hydroxy methyl Mobile phase: 20% 2-furfuraldehyde Methanol in water to 100% 4 2-furfuryl alcohol methanol 5 2-furfuraldehyde Flow rate: 0.025 ml/s 6 2-acetyl furan Detector: UV216 nm 7 5-methyl-2-furfuraldehyde Sample: 15 µl methanol 8 oil compounds soluble extract in methanol
It was therefore felt that a more precise and sensitive method of detecting paper degradation was required. It was against this background that the work described below was initiated. The information is taken from a paper presented to CIGRE at its August/September meeting in 1984 by Messrs P.J Burton, J. Graham, A.C Hall, J.A Laver and A.J Oliver of CEGB [6.9]. The method developed is based on the analysis of the oil for compounds that are produced exclusively by thermal degradation of paper at temperatures as low as 110ºC. The intention was that the procedure should be used in conjunction with dissolved gas analysis rather than independently. That is, reliance should be placed on normal d.g.a. techniques to raise the alarm that a possible fault condition exists, but the new technique should then be applied to obtain more information regarding the nature and magnitude of the fault.
Oil samples are taken as for the d.g.a. procedure but are then mixed with methanol. A certain quantity of the compounds sought then become dissolved in the methanol at concentrations determined by the equilibrium levels with the oil (a similar process to the oil/water/paper equilibrium situation previously described) The methanol and its solutes are then injected into a high performance liquid chromatograph (HPLC) for separation and measurement of the individual compounds. A typical chromatogram is shown in FIG. 127.
The main reason for using the methanol extraction stage is to avoid those constituents of the oil which would crowd the chromatograph making the detection of the compounds being investigated more difficult.
The paper degradation products identified are also listed in FIG. 127 Of these compounds, 2-furfuraldehyde is the most common product detected from transformers in service.
Following the development of this method many samples of oil from transformers in service have been analyzed for furfuraldehyde. In one particular 22/400 kV generator transformer, it was noted that the dissolved ethane and methane concentrations were increasing fairly rapidly indicating that an over heating fault existed having a temperature within the range 150-200ºC. The furfuraldehyde concentration also increased over the same period of 16 weeks from about 0.7 to 1.7 mg/l suggesting that paper as well as oil was being over heated. The transformer eventually failed.
Investigations into the failure revealed that the A and B phase windings were loose, probably as a result of a fault on the transmission system and that the transformer tripped due to an interturn fault on the B phase LV winding.
Paper insulation had been overheated confirming the conclusions drawn from the furfuraldehyde measurements.
Confidence is now growing in the use of this method of detecting paper degradation. However, there are also problems similar to those identified with d.g.a. There is no such thing as a norm for furfuraldehyde level in a healthy transformer, so that it is not possible, as some authorities might have hoped, to carry out general measurements of furfuraldehyde levels throughout transformer populations to identify those for which the insulation is prematurely ageing. There is no reliable way of differentiating between a large mass of paper which is just slightly degraded and a localized area for which degradation is seriously advanced. It is also the case that if a short term overload causes overheating and significant ageing, with associated furfuraldehyde production, and then this is followed by a period of normal loading without over heating, the furfuraldehyde will be absorbed into the mass of the paper, so that the levels in the oil will appear little different from normal. Once again, as in the case of d.g.a. it is observation over a period and the detection of step changes which can be regarded as indicative of a fault condition.
Dissolved gas analysis during works testing
It should be recognized, of course, that the value of d.g.a. as a diagnostic tool need not be restricted to transformers that are in service. D.g.a. can serve a very useful function during works testing.
Utilities are becoming increasingly conscious of the fact that a few hours in works testing is a very limited time in which to demonstrate that a large transformer will be suitable for thirty or more years satisfactory service. In addition specifications are tending to allow higher operating temperatures and, although these, in theory, still allow margins above the average values which can be measured on test for hot spots, the customer has no guarantees that there will not be hot spots which exceed this allowance. There is also a tendency for transformer manufacturers to shorten the overall times for temperature rise tests by reducing the cooling of a forced-cooled unit during the initial phase of the test, thereby reducing further the likelihood of some faults being brought to light. As a counter to this many users are specifying that the temperature rise test, or perhaps, more correctly, load-current run, should be continued for 24 hours. On this timescale it is possible to obtain meaningful d.g.a. figures from oil samples taken before and after this load-current run.
It is not normally considered practicable to set any acceptance/rejection level on d.g.a. figures but analysis of the oil samples will not only clearly show the presence of any more significant fault, but can also be expected to reveal the presence of modest overheating of the insulation which would affect the transformers overall life expectancy.
Of course, one criticism of a short-circuit temperature rise test is that the core flux-density is low and consequently leakage fluxes which could give rise to overheating in service will be very much reduced. The CEGB response to this was to specify a prolonged overvoltage run, equivalent to about 8.3 percent overfluxing for 3 hours. This was considered long enough to produce detectable gas levels in the oil should there be any significant overheating resulting from leakage fluxes.
Many manufacturers, of course, recognize the benefits of identifying incipient faults during works testing rather than having these possibly dam age their reputations by coming to light in service and so advocate the use of d.g.a. as an aid to assessing performance during works tests. The two case studies which follow were described in a paper presented to the IEEE Power Engineering Society summer meeting in July, 1981, by the Westinghouse Electric Corporation . Both units tested were of the shell type.
The first case was a three-phase transformer with on-load tapchanger. Table 13 shows d.g.a. results at the end of the factory temperature rise test. It is assumed that an oil sample would also have been tested before the temperature rise test but no figures are given. Being a new transformer newly filled and processed it must be assumed that the initial gas levels were very low. Even without taking ratios it is clear from the ethylene level that severe overheating is taking place. In addition, the presence of any acetylene in a new transformer should always be regarded as indicative of a fault. The paper reports that investigation revealed the overheating to be due to the effect of leakage fluxes. After taking corrective measures a repeat of the temperature rise test showed that the problem had been resolved.
The second reported case was that of a furnace transformer with a very high-LV rated current. The LV leads and connections were made from large section copper bar with bolted joints. Table 14 shows the d.g.a. results following the temperature rise test. Hydrocarbon gas levels are, in reality, quite modest. It is very unusual to find no hydrogen present at all, however, once again, in a newly processed transformer none of the gas levels should be expected to exceed a few parts per million. Certainly the methane and ethane figures must be taken seriously, but the very low ethylene suggests that on this occasion any overheating is quite modest. The paper reports that the tightness of all bolted joints was checked and although none were found to be loose, it proved possible to tighten some by a further quarter to half turn. A thorough inspection of the transformer revealed no other fault. That the source of the problem had indeed been found was proved by repeating the temperature rise test without the production of any hydrocarbon gases.
Establishment of norms
Most authorities experienced in the use of d.g.a. for hydrocarbon gases and for cellulose degradation products, are emphatic in the view that it is not possible to identify 'norms' for healthy transformers for the reasons given above and that it is change in the status quo which is the clearest indication of a transformer fault. However, many transformer users feel that there ought to be norms and there are authorities who have endeavored to provide these. The American IEEE Standard C57.104-1991, Guide for the interpretation of gases generated in oil-immersed transformers, infers that in operation all transformers will have norms appropriate to their age and duty in so far as detection of a fault first requires the determination that an abnormality exists. This is akin to looking for a step-change in d.g.a. levels as described above, except that the document is aiming to identify, in the absence of a d.g.a. history for a particular transformer, what dissolved gas levels represent the status quo.
The above IEEE Standard provides values for norms in relation to the total dissolved combustible gas content (TDCG) for the transformers and these are set out in Table 15. As will be seen from the table the values of combustible gases enable the transformer to be placed into conditions 1-4 which are set out in Table 16. The document states that the values are consensus values based on the experiences of many companies. Only condition 1 is regarded as satisfactory, but it will be recognized that even a condition 1 transformer could, according to the table, be expected to contain up to 35 p.p.m. of acetylene. The document goes on to make the point that in a fairly new transformer, the presence of any acetylene would give rise to concern, but in a 20 year old transformer the gas levels quoted in the table, including acetylene, would not be considered extraordinary.
Condition 1 TDCG below this level indicates the transformer is operating satisfactorily.
Any individual combustible gas exceeding specified levels should prompt additional investigation.
Condition 2 TDCG within this range indicates greater than normal combustible gas level.
Any individual combustible gas exceeding specified levels should prompt additional investigation. Action should be taken to establish a trend. Fault(s) may be present.
Condition 3 TDCG within this range indicates a high level of decomposition. Any individual combustible gas exceeding specified levels should prompt additional investigation.
Immediate action should be taken to establish a trend. Fault(s) are probably present.
Condition 4 TDCG within this range indicates excessive decomposition. Continued operation could result in failure of the transformer.
In the UK the organization EA Technology's Dr M.K Domun studied and collated oil analysis data from around 500 transformers, mainly of 132 kV, for many years and as a result of this work published figures in a paper presented to an IEE Conference on Dielectric Materials, Measurements and Applications in September, 1992, [6.11] as 'optimal values' for transformers which have been on load for a lengthy period and which are considered to be in a 'healthy' condition. These are listed in Table 17.
Hydrogen 20 ppm Methane 10 ppm Ethane 10 ppm Ethylene 10 ppm Acetylene 1 ppm Carbon dioxide 5 000 ppm Carbon monoxide 100 ppm Acidity 0.08 mgKOH/g Moisture 25 ppm (no temperature quoted) Electric strength 27 kV Furfuraldehyde 2 mg/l
It will be noted that Dr Domun's norms are somewhat less than those suggested by IEEE. Dr Domun stresses that there is a wide variation between individual units and says that the above figures were chosen on the basis of the 50 percent rule, i.e. at least 50 percent of the samples conform to the values of these parameters. CEGB experience is of large generator transformers which are operated at high loadings for long periods, unlike the network transformers studied by Dr Domun, and many of these continued to operate satisfactorily with hydrocarbon gas levels considerably higher than the values given Table 17.
The above examples illustrate both the disadvantage of aiming to identify norms as well as the benefit. It is with this note of caution that the norms in Tables 15 and 17 are provided.
It should be noted that, at the time of writing this Thirteenth edition, IEEE C57.104 is under revision.
Other monitoring systems
Put into the simplest terms it can be said that transformers have three basic failure modes:
• They can suffer insulation failure leading to electrical breakdown between internal parts.
• They can fail due to severe internal overheating.
• They can suffer mechanical failure due to their inability to withstand the effects of a close-up external fault.
It is the first two of these modes which are truly 'faults' for which dissolved gas analysis can be of assistance in providing indication of incipient break down before this has reached the catastrophic stage. But it is the third which represents 'end of life' failure. When paper insulation is severely degraded it loses its mechanical strength but nevertheless much of the dielectric strength of the paper/oil combination is retained so that in a transformer of which the insulation has aged to the extent of approaching the end of its useful life, there is no immediate failure and the transformer will continue to operate satisfactorily until it receives a 'shock' mechanical loading due to some external factor such as a system fault relatively close to its terminals. Ideally a user would like to replace his transformer just before it is due to fail in this way. If he replaces it several years before it is due to fail he has not obtained maximum use and there will be an economic 'cost.' If he defers replacement until failure has actually occurred he is involved in the high costs of an unscheduled outage and the need to find a replacement on an urgent basis, and possibly even some consequential damage costs.
Consideration of this problem has engaged researchers for some years; to find a system of knowing just when insulation has reached the point when it no longer has sufficient strength to meet the mechanical demands placed upon it. It was hoped at the early stages of developing furfuraldehyde assessment that this might be linked to the absolute level of paper degradation and thus provide the means that were sought, but there are problems in attempting to derive absolute indications from furfuraldehyde in exactly the same way as there are from the hydrocarbon gases. Transformers vary so considerably in their relative insulation volumes, oil content, water content and acidity as well as loading patterns, and all these factors influence furfuraldehyde levels. In addition, the degradation of a fairly small localized area of insulation in the vicinity of a hot spot can be just as terminal as degradation of far greater extent in a design which has seen extensive service but which does not have significant hot spots The former, how ever, will generate a far smaller quantity of furfuraldehyde making it much more likely to go undetected. As indicated in Section 3, the properties of paper insulation depend on those of the long chain cellulose molecules of which it is made up. Deterioration of its mechanical properties is brought about by decomposition of these long chain molecules, and early researchers used tensile strength as a measure of remanent life. Current practice is to measure degree of polymerization (DP) which is an indication of the number 'links' in the long chain cellulose molecules. This starts at about 1100-1200 for new material but drops rapidly during the drying and processing stage of the transformer to around 850-900 which might be taken as a typical starting point for a new transformer.
End of life is reached when DP has dropped to about 250 and the paper loses its remaining strength suddenly at about half of its original value.
There have been suggestions that by entering the transformer and taking samples of the insulation for measurement of DP, the remanent life of the insulation could be estimated. The problem, of course, is that any insulation which is sufficiently accessible to sample will not be representative of the more critical insulation in the vicinity of the hot spot. One way of overcoming this would be to place an insulation sample in the hottest oil at the time of commissioning the transformer and to further heat this by means of a heater coil supplied from a current transformer placed in one of the winding leads in the same way as for a thermal image winding temperature indicator. The difficulty is that the hot spot temperature cannot be determined with sufficient accuracy to make this exercise worthwhile. The problem remains, therefore, that determination of imminent end of life must be based on little more than guesswork.
Another approach thought by a number of researchers to have promise is based on the detection of movement within the windings in response to impressed low-voltage impulses. As insulation ages, shrinkage occurs so that, whilst windings are initially in a state of axial compression due to the manufacturing clamping forces, as end of life is approached the effect of shrinkage will create a degree of slackness. The slackness, of itself, can accelerate the onset of failure of a transformer already weakened by the low mechanical strength of its insulation, since it will permit winding displacement and, as explained in Section 4.7, the axial forces on the transformer windings under high through fault currents are increased if there is already some initial displacement. Most of the methods employed require the transformer to be taken entirely off-line so as to avoid the presence of an external circuit making the difficult task of detection of the impulse currents and the small changes in them even more difficult. Other systems have attempted to detect winding vibration produced by the impulses, using acoustic sensors. Another technique is based on the fact that the slight change in winding inductance and capacitance values will result in changes to natural resonance frequencies. The difficulty with all of these efforts is in relating the measured parameters to transformer condition and the risk of failure. Accurate measurement of the selected parameter is itself difficult enough but making this final step is many times more so and it is unlikely that such methods will achieve meaningful results in the foreseeable future so that meanwhile the guesswork must continue.
Failures and their causes
In the foregoing paragraphs there has been a general discussion of the mechanisms of transformer failure. Earlier editions of this volume have included a more specific catalogue of the ways in which transformers have failed in service. Such an approach was reasonable in the earliest editions, since transformer design and manufacture was developing rapidly and those involved in the process were going through a phase of gaining a large amount of experience with regard to what could be done and what could not. Hopefully, more than 80 years after the publication of the first edition, this experience has been fully assimilated, failure rates have been reduced significantly and to simply include a list of failures which have occurred over the past 20 years is likely to teach very little. Designs have changed and a transformer built today will have many different features from one made 20 years ago, although it might appear superficially the same. For example, in earlier editions failure of core bolt insulation was identified as a common fault. For a purchaser to use this knowledge at the present time to specify the quality of core-bolt insulation, or even to insist that bolted cores should be avoided, would be superfluous, since small and medium sized transformers have used boltless cores for many years and core-bolts are now avoided in even the largest cores.
This is not to say that failures will cease to happen. Relatively recently, in the early 1970s, CEGB noted disturbingly high failure rates in large generator transformers. (This has also been discussed in Section 5.3). There were a number of reasons for this but significant amongst these was the large step increase in unit sizes as generator ratings were rapidly increased in the UK from around 120 MW to 500 and 660 MW. Failure rates were reduced once more in the 1980s by a combination of more extensive testing, improved QA during manufacture, moving to single-phase units rather than three phase, which had the effect of removing the severe limitations which the latter had imposed on transport weights, thus reducing the loadings imposed on the basic materials, and also by adopting a procedure whereby designs which had been proven by service were repeated, rather than accepting a process of almost continual innovation.
This latter strategy was controversial, since limiting innovation can be construed as preventing manufacturers using their skills to increase their competitiveness. However the same accusation can be leveled at the practice referred to earlier of listing previous failures and their causes, since there is an inference that if such a way of doing something has caused a failure in the past, then this should always be avoided in the future regardless of the availability of improved materials and better methods of performing design calculations, and this can lead to very restrictive thinking. It can also result in purchasers' specifications becoming very proscriptive. This is a criticism which was often made of the UK Electricity Supply Industries' Specification BEBS T2 (1966), and indeed, this document identified many design and construction features which it considered unacceptable, usually because they had caused failure at some time in the past.
Now the move is towards specifications which allow manufacturers to utilize their own design skills. But if they are to be given this freedom, specifications must also call for adequate testing and they must also tell manufacturers exactly how the transformer is to be operated. Section 8.1 deals with the specification of technical requirements and it is hoped that this will enable prospective purchasers to identify all those operational features which have a bearing on how a transformer should be designed and manufactured. The reasons why these particular features are relevant should, of course, be apparent from else where within the pages of this guide.
Clearly, not all transformer failures are the fault of the designer or manufacturer. Operation and maintenance must also have an impact as it is hoped will be appreciated from a study of the earlier part of this section. Although maintenance requirements are few, users must regularly monitor the condition of their transformer and, if they seek high reliability, they must ensure that three fundamental requirements are observed:
• Breather systems must be adequately maintained so that water content is kept at the lowest practicable level.
• The transformer must be adequately cooled at all times, any overloading maintained within permitted limits and action taken on any indications of possible overheating.
• The transformer must not be subjected to excessive overvoltages.
One of the most detailed international studies of transformer failures was carried out by a CIGRE Working Group in 1978. It was reported in Electra number 88, dated May, 1983 [6.12]. Input was received from 13 countries relating to all types of transformers having HV voltages from 72 to 765 kV. The analysis took account of more than 1000 failures occurring between 1968 and 1978, relating to a total population of more than 47 000 unit-years. The transformers varied from 'just entered service' to 'aged 20 years.' Nearly half were aged between 10 and 20 years. They were categorized into power station transformers, substation transformers and autotransformers and were further subdivided into those with on-load tapchangers (OLTC) and those without. The main purpose of the survey was to establish reliability figures, but those responding to the questionnaires were also asked to categorize the reasons for failure. The overall failure rate was concluded to be 2 percent and the breakdown into voltage classes showed that the figure was higher, up to 3 percent, for the 300-700 kV voltage class. Figures for the greater than 700 kV class were left out of the main report since too few statistics were reported. These were covered in an appendix in which it was reported that the failure rate was about 7 percent.
FIG. 128, reproduced from the report, shows the causes of failure for the largest group, covering over 31 000 unit-years. These are substation transformers with on-load tapchangers, and the pattern is similar for other groups.
The histogram gives the presumed cause of failure, the component involved and the type of failure. The quantities also indicate the outage time involved, classified as either less than one day, 1-30 days or greater than 30 days. The most significant features to emerge are that the on-load tapchanger was the component most frequently involved, perhaps not too surprising since this is the only component of the transformer which has moving parts; with winding faults less than half as frequent but still the second most likely component to fail. For all groups the magnetic circuit has the lowest reported failure rate, this despite the fact that the statistics refer to transformers in service between 1968 and 1978 when the use of bolted-cores was standard practice. Although design and manufacturing errors were reported as by far the most likely cause of failure, it should be noted that incorrect maintenance figures quite prominently in this list of causes. (It must be remembered that this information has been provided by the transformer operators.)
FIG. 129 gives the reported data for the equivalent group of power station transformers, namely those with on-load tapchangers. This is a very much smaller group, covering only just over 2300 unit-years so the figures must be correspondingly less reliable. The largely similar pattern is nevertheless evident, except that terminals now become the most likely component to fail, equally with windings. This probably reflects the fact that power station transformers tend to run more fully loaded so that components such as terminals will be more highly stressed. What is surprising is that terminal failures almost invariably resulted in outages exceeding one day and a significant proportion of outages exceeded 30 days so these must have been more serious than simply requiring changing of a bushing. This does emphasize the importance of paying proper attention to the selection and installation of these components and of ensuring that connections are correctly made and checked during maintenance. The very much lower ranking of tapchanger related faults probably reflects the fact that on generator transformers these operate on manual control and perform far fewer operations than those of substation transformers which are under automatic control.