|Home | Articles | Forum | Glossary | Books|
MAINTENANCE IN SERVICE
The subject of transformer maintenance has received only cursory attention in many earlier editions of this guide. Perhaps this is not surprising. Transformers generally have no moving parts and there is nothing to wear out, so that there is little to be maintained. Indeed, many small transformers, particularly those installed in distribution networks, once commissioned remain in service for many years with minimal attention. These are literally examples of the 'fit and forget' philosophy. However, most transformers will operate more reliably if given some attention, and the larger the transformer and the more important its role, the greater is the justification for regular and relatively frequent attention.
The following section will identify the benefits to be gained from a regular maintenance regime and describes the procedures which can be carried out in order to achieve maximum reliability. It will deal almost exclusively with oil-filled transformers. There is very little maintenance which can be carried out on a dry-type transformer with the exception of keeping it clean and dry.
The objects in maintaining any item of plant are:
• To obtain the maximum practicable operating efficiency.
• To obtain optimum life.
• To minimize the risk of premature and unexpected failure.
In the case of a power transformer, there is very little that the operator can do which will affect operating efficiency except to ensure that the cooling equipment is functioning correctly. Obtaining optimum life and minimizing risk of unexpected failure are therefore the principal objectives of transformer maintenance. Although to some extent interlinked, these are separate activities but both involve obtaining a close awareness of the transformer condition which, with the present state of the art of condition monitoring, comes principally from close monitoring of the condition of the transformer oil. Much of this section, therefore, concerns the sampling and testing of transformer oil and the information which can be obtained from this activity. It assumes an under standing of the properties of mineral oil in conjunction with cellulose insulation discussed in Section 3.5. Reference should also be made to IEC 60422 Mineral insulating oils in electrical equipment -- Supervision and maintenance guidance. In the UK the applicable document is BS 5730:2001, which bears the title Monitoring and maintenance guide for mineral insulating oils in electrical equipment. Whilst these two documents have a great deal in common, it is still the case that UK practice with regard to transformer oil is very much in a minority in relation to most of the world, primarily in its use of uninhibited oil. It should also be emphasized that the content of this section is intended as a starting point. Any suggestions that it makes are not meant as a substitute for any instructions issued by the plant manufacturer, or any special requirements of the user. The object is simply to draw attention to the factors which have a bearing on the life of the transformer and its likely service reliability.
Oil sampling procedures
Monitoring of transformer oil - the taking of oil samples, which is the essential pre-requisite to any maintenance, can be carried out with the minimum of plant shut down, or even, if necessary, whilst the plant remains on load.
Sampling of oil is a fairly common activity on the part of electrical maintenance staff. It might, therefore, appear superfluous in a volume such as this to attempt to give instruction on the taking of oil samples. Clearly, a textbook is not the best medium for the provision of such practical instruction.
However, the oil sample is such an important source of information as to the transformer condition that it is essential that extreme care is taken in obtaining it in order to ensure that the information gained is not misleading and it is therefore worthwhile emphasizing the most important aspects of the sampling procedure. Reference should also be made to IEC 60475 Method for sampling liquid dielectrics. In the UK the relevant document is BS 5263 which has the same title. The documents are not the same since there are significant differences in procedures.
In the UK the procedure is to allow the oil to drain, preferably from a valve provided solely for the purpose of taking an oil sample, into a sample bottle.
In many countries, notably a large part of Europe, the oil is drawn into a large purpose-made syringe. In the UK it is also practice to position the sampling valve about 1 m above the tank base so that the oil drawn off is truly representative of the bulk oil in the transformer. FIG. 119 shows an oil sampling valve as commonly used in the UK.
On initially opening the sampling valve a few liters of oil should be drawn and discarded in order to flush the valve and ensure that no contamination from around the valve is allowed to enter the sample bottle. It is usually convenient to attach a short length of flexible pipe to the valve to enable the oil to be more easily directed into a bucket for this purpose. Once this steady flushing flow has been established the valve should not be disturbed until taking of the sample is completed so as to avoid the risk of disturbing any debris from around the valve and allowing this to enter the sample. The sample bottle should then be filled to about half way, the bottle rinsed using this oil which should then be discarded before filling the bottle with the sample. The sample should be allowed to flow into the bottle down the side of the glass to minimize aeration of the sample and filling should continue until the bottle overflows, which helps to dispel any air bubbles. A small amount of the sample should then be poured off so as to leave some space for expansion. The bottle should then be tightly stoppered.
Immediately the sample has been taken, the sample bottle should be inverted and held up to enable the contents to be visually examined. Any air bubbles will rise. Any bubbles which fall are bubbles of free water and there is little point in carrying out any electrical tests on a sample which contains free water. The sample should therefore be discarded and a further sample taken to establish whether the water has entered the sample inadvertently or whether this is truly representative of the transformer bulk oil. Although it is possible that the transformer will have been taken out of service for the oil sampling to take place, it is desirable that the sample is taken as soon as practicable after shut down to obtain the best possible representation of the bulk of the oil to ascertain its general condition and to avoid the possibility that foreign materials and water, which tend to sink to the bottom, may settle and not be obtained in the sample. It is also important that the transformer oil temperature should be recorded at the time of taking the sample. Finally the sample should be clearly identified with the transformer details and the date of sampling.
Oil in storage
Although not strictly part of transformer maintenance procedure, it is appropriate at this time to make some comment concerning maintenance of oil in store. Oil may be stored in bulk tanks, or more usually it is stored in drums as delivered from the refinery. Bulk oil storage tanks will be fitted with a breather usually of the desiccant type. It is important that the desiccant in this should be frequently checked and maintained in a dry state. Oil stored in drums will also breathe via the bungs on the covers. To ensure that water cannot collect on the covers, to be breathed in when the contents are cooled, drums should be stored in the horizontal position with the bungs at the 3 and 9 o'clock positions.
Ideally the drums should be stored indoors where temperature cycling will be minimized. If they must be stored out of doors they must be at least protected from extremes of temperature and shielded from direct sunlight. The storage period should be minimized and a system adopted which ensures that the first received is the first to be used. Even with these precautions oil in drums will have a higher water content than oil stored in bulk and this is the reason why IEC 60296 allows new oil delivered in drums to have a higher water content than oil delivered in bulk.
It is possible that it may be required to sample oil from store to establish its water content. This is, in fact, more important if the oil is to be used in switchgear than if it is for topping up a transformer. As explained in Section 3.5, transformer insulation has a high capacity for water and even a moderately large transformer with 1 or 2 percent water in its insulation will contain several liters of water in total. If the transformer is topped up by adding, say, 50 liters of oil from a drum having a water content of 40 ppm, the maximum permitted for delivery in drums, the total amount of water added to the transformer is:
40 50 10 1000 2 6 3 _ __ cm
which is insignificant compared with the quantity of water already in the insulation and even if the oil in the drum had twice its permitted level this would still have no significant impact. This fact should not, of course, lead to slackness in storage procedures since laxity can soon lead to a serious deterioration in discipline and ultimately to accidents.
If it is required to take a sample from a bulk storage tank then the procedure is similar to that employed for a transformer. If a sample is to be taken from a drum then a glass pipette or 'thief' is used. This is of a length to enable it to reach to the bottom of the drum. FIG. 120 shows such a device being used to take a sample from a drum of oil. Before insertion into the drum the thief must be scrupulously clean, particularly on the outside since this could otherwise introduce contamination into the drum contents which would not be seen in the sample. To ensure that this is the case it may be wiped using a lint free cloth of welded polypropylene or similar material. Ordinary rag or paper toweling which will shed fibers is not permitted. The thief should initially be inserted to a level to allow it to be approximately half filled and this oil used to rinse the inside. This oil is then discarded and the procedure repeated to obtain sufficient oil to rinse the sample bottle. After this the thief may be inserted to extract the sample. Once the outside of the thief has become 'wet' with the oil even greater care is necessary to ensure that it does not pick up airborne contamination. To take the sample it should be inserted with the top end closed with the thumb until it reaches the bottom of the drum. The top is then uncovered, thus admitting the sample from the bottom of the drum where any free water or contaminants are likely to be found. The sample bottle can then be filled using the same procedure as when sampling from the transformer.
When to take samples
IEC 60422 gives a varying frequency of testing related to the category of importance of the unit. For most transformers it is recommended that a sample be taken for electric strength and water content after filling or refilling, followed by a sample after 1 year. Subsequent to this, samples should be drawn from important transformers every 2 years and from transformers of lesser importance every 4 or 6 years.
These proposals should be viewed in a very flexible manner. Clearly, it is sensible to sample after filling or refilling. Beyond that, there is merit in sampling annually and tabulating the results, or plotting these graphically until a trend can be established, following this, the decision can be taken to sample more or less frequently as appropriate. CEGB had a policy of sampling generator transformers every 3 months. A nickel smelter in a remote part of Indonesia is known to take samples from critically important furnace transformers on a monthly basis. Obviously there will be some smaller distribution units, which, once installed, would only be sampled at a frequency to coincide with other maintenance schedules perhaps at five or 6 yearly intervals.
What are samples tested for
Whenever a sample is taken initially it should be examined for odor, appearance and color and if this is a sampling carried out after filling of refilling, then it would be sensible to carry out an electrical strength test. If this sample gives any cause for suspicion that the water content is high, that is if there is free water present or if the oil should fail the electrical strength test, then it is desirable to determine the actual water content by carrying out a Karl Fischer test. This is described in EN 60814 Method for determination of water in liquid dielectrics by automatic coulometric Karl Fischer titration.
As indicated in Section 3.5 acceptable water content depends on the age of the transformer, its voltage class and its strategic importance. Although it is always desirable that water content be maintained as low as practicable, the upper bounds of acceptability might be around 40-50 ppm at 80ºC in a 132 kV transformer after some years in service. A value above this would not necessarily mean that the oil should be processed. As explained earlier, the quantity of water absorbed in the paper is quite large and it would be a slow process to attempt to remove it by drying the oil. The main priority is to find out how the level got to be so high. Has the breather charge been left in need of renewal for some considerable time? Has a valve been wrongly set so that the transformer has been open to atmosphere with the breather bypassed? For a water-cooled transformer, is there a leak of water into oil? Clearly, more frequent sampling should be introduced to ensure that any corrective measures have been successful in dealing with the situation.
Should the oil's odor or color suggest that it may have become significantly oxidized, then an acidity check is called for. IEC 60422 considers this to be a routine test coupled with a check of inhibitor content. The same document gives recommended acidity limits according to voltage class, from 0.15 mgKOH/g for the highest voltages to 0.3 mgKOH/g for 72.5 kV class transformers. BS 5730 suggests that, except for the most important transformers, an acidity level of below 0.3 mgKOH/g is satisfactory assuming no other characteristic is unsatisfactory. Between 0.3 and 0.5 mgKOH/g it is suggested that more frequent testing should take place to ensure that the acidity does not exceed the 0.5 mgKOH/g level. Above this level the BS suggests that the oil should be replaced. IEC 60422 gives broadly similar advice.
In replacing the oil, the contamination of the new oil by residual used oil in the transformer becomes a more serious risk in respect of subsequent deterioration the higher the acidity of the old oil is allowed to become. When replacing oil which is very acid it is important to allow the drained down core and windings to stand for a few hours to allow as much as possible of the old oil to drain off, and then to flush the transformer out as thoroughly as possible with clean oil before refilling.
By far the most worthwhile test of the oil sample for all important transformers is to carry out a dissolved gas analysis. The levels of hydrocarbon gases should be recorded and compared with previous values for the transformer. Any unexplained step changes should be investigated.
It may well be the case that for an older transformer which has spent a significant proportion of its life highly loaded, the gas ratios along with the previous records simply show steadily increasing levels of the lower temperature gases, methane, hydrogen and possibly some ethane indicating possible mild overheating. If this is the case and there are no sudden changes in the general trends, and provided all other characteristics of the oil are satisfactory, then no action is necessary.
Dissolved gas analysis
Dissolved gas analysis (d.g.a.) is the most valuable and important tool avail able to the maintenance engineer concerned about the condition and life expectancy of transformers. Some authorities make excessive claims for its efficacy. It has its limitations. Its use was pioneered from the late 1960s by E. Dornenburg in Switzerland and a little later by R.R Rogers and others in the UK in co-operation with a large transformer manufacturer. At quite an early stage in its use some spectacular successes were achieved which saved thou sands of pounds in avoidance of catastrophic failures and lost generation costs.
There were also notable occasions when major transformer failures occurred which were not predicted. It is worthwhile therefore considering the process in some depth and examining what it can achieve and what it cannot do.
The generation of gas in oil-filled equipment by disruptive discharges (sparks and arcs), and severe overheating results from the chemical reactions which occur as a result of such faults. The resultant effect of the high thermal and disruptive discharge conditions are due to the severity of the fault and the presence of other materials such as solid insulation. Solid and liquid materials are also produced, but it is the gaseous products that are of the most concern and interest. The analytical and interpretative techniques that are used differentiate between those which are due to air contamination, oxidation and partial discharge, and those from more severe thermal and electrical faults which can destroy insulation and result in costly and severe damage to the equipment.
The identification and significance of gases in electrical equipment was first used to distinguish between combustible and non-combustible gases produced in transformers as long ago as the 1920s. This was carried out by applying a light to the gas collected from the sample or vent tap of the Buchholz relay.
Initially the procedure aimed to detect the presence of hydrogen, which meant that there was a 'real' fault within the transformer. Over the next 30 years the procedure was refined to enable hydrogen, acetylene and carbon monoxide to be detected, which enable some indication of the nature of the fault to be deduced. In particular, the presence of acetylene meant that very high temperatures existed, and carbon monoxide was taken as an indication that solid insulation was involved.
The development of chromatography, mass spectrometry and infrared analytical techniques in the period 1955-1965 led to their use for analyzing gases from Buchholz relays, and ultimately the realization that by extracting these gases from an oil sample, their presence could be detected and interpreted long before the oil was saturated and the fault had developed to the stage at which free gas could be collected in the Buchholz relay.
This development coincided with the expansions of the electricity sup ply networks and the use of higher voltages which was taking place in several countries, including the UK, leading to increased failure rates. Analysis of gases coupled with inspection of the failed equipment led to further study of the gas evolution processes and to the appreciation of certain gas ratios as being indicative of different fault temperatures.
Theory of gas evolution The composition of the gas produced in a fault is decided by many factors. In addition the gases which are seen in any sample taken for analysis are further influenced by factors other than those relating to the fault. The previous history of the transformer, the loading regime, the amount of insulation that it contains and the dryness of this insulation as well as the precise location of the fault are just some of these. Nevertheless, it is possible to relate certain patterns of gas evolution to temperatures existing at the fault and from a knowledge of these, along with a careful assessment of all other relevant factors, to obtain some appreciation of the nature and seriousness of the fault.
The immediate effect of the breakdown of the hydrocarbon molecules as a result of the energy of the fault is to create free radicals as indicated in FIG. 121. These subsequently recombine to produce the low molecular weight hydrocarbon gases. It is this recombination process which is largely determined by the temperature, but also influenced by other conditions.
The result is that the pattern of gases appearing in the oil has a form as shown in the chart of FIG. 122. For the lowest temperature faults both methane and hydrogen will be generated, with the methane being predominant. As the temperature of the fault increases ethane starts to be evolved, methane is reduced, so that the ethane/methane ratio becomes predominant. At still higher temperatures the rate of ethane evolution is reduced and ethylene production commences and soon outweighs the proportion of ethane. Finally, at very high temperatures acetylene puts in an appearance and as the temperature increases still further it becomes the most predominant gas. It will be noted that no temperature scale is indicated along the axis of FIG. 122, but the diagram has been subdivided into types of fault. The area indicated as including normal operating temperatures goes up to about 140ºC, hot spots extend to around 250ºC, and high temperature thermal faults to about 1000ºC. Peak ethylene evolution occurs at about 700ºC.
A curve which frequently appears in articles dealing with the subject of dissolved gas analysis is shown in FIG. 123. This shows the partial pressures exerted by the hydrocarbon gases in oil plotted on a logarithmic scale as the temperature increases and was initially published in the Journal of the Institute of Petroleum in a paper by W.D Halstead in 1973 . Whilst this might pro vide a more scientifically accurate statement of the composition of the gases with temperature, since this can be shown to be proportional to the partial pressure exerted by the gas, FIG. 122 conveys a more easily comprehensible picture of what is happening.
It is evident from FIG. 122 that the ratios of the evolved gases change at various points along the temperature scale. The ratios mentioned above were:
1 Methane Hydrogen 2 Ethane Methane 3 Ethylene Ethane 4 Acetylene Ethylene
These are the ones which were proposed by Messrs I. Davies and P. Burton in the UK in 1972. For each of the ratios, if they have a value of less than unity they are given the code zero. If they are greater than unity they are given the code one. It is then possible to compile a table, shown as Table 7, which relates each likely combination of codes to a position along the temperature scale.
In 1974, after a detailed study of dissolved gas data and associated transformer faults, Mr R.R Rogers proposed some refinement of the ratios into bands according to their magnitudes. These are given in Tables 8 and 9.
These have become known as Rogers ratios and are still widely used as a means of attempting to identify fault conditions from dissolved gas analysis.
Other authorities use other ratios, for example in the late 1970s the compilers of IEC 599, Guide for the Interpretation of the analysis of gases in transformers and other oil-filled electrical equipment in service believed that the diagnostic process could be simplified by the use of only three ratios, omit ting that for ethane to methane. That document also aimed to indicate the temperatures reached, rather than categorizing the faults. It then gave what it called typical faults which would give rise to these temperatures. This is reproduced this as Table 10. In its current guise as IEC 60599 this continues to advocate the use of the three ratios methane/hydrogen, acetylene/ethane, and ethylene/ethane.
Also in the 1970s Michel Duval, a research chemist working at the Hydro Québec's Institute of Research (IREQ) in Canada, developed a graphical method of diagnosis that does not rely on ratios of pairs of gases, but rather the relative concentrations of three principle gases. This has become known as the Duval Triangle and is an approach which still finds favor at the present time.
Duval's method is to plot concentrations, in p.p.m. of methane, ethylene, and acetylene expressed as percentages of the total content of those three gases in the sample, against a set of triangular co-ordinates having the percentage of each gas along each of the axes as shown in FIG. 124. The triangular field is divided into seven zones as shown and the diagnosis of the fault is indicated according as to the zone within which the plot of the three percentage gas concentrations falls. The zones have the interpretations as shown in FIG. 124.
A criticism that has been made of the Duval approach is that it must always identify a fault because there is no zone within the triangle that corresponds to 'no fault.' Duval would accept this by saying that his approach only recommends that a diagnosis should be made when there is a change in the status quo and that the gas percentage concentrations to be used to plot a point within his triangle must represent changes in dissolved gas levels from the levels that existed when the transformer was considered to be in a healthy state. More will be said on this aspect a little later.
In none of the foregoing has mention been made of carbon monoxide and carbon dioxide levels. There is now a view that attempting to draw conclusions from the level of carbon monoxide and carbon dioxide can be very misleading. It was once considered that the only source of carbon monoxide was from overheating of cellulose (i.e. paper) insulation. However it is now recognized that both carbon dioxide and monoxide can also arise from normal oxidation of the oil, in relative proportions which differ widely in different transformers, so unless there is a very marked change in a long established pattern of carbon monoxide and carbon dioxide evolution it is considered that it is more reliable to ignore these gases. It is further argued that serious insulation degradation only occurs in conjunction with significant overheating of metal surfaces, so such conditions will be detected by the presence of the other gases.
In addition to the above, more recent research has achieved some success in detecting and measuring other products of cellulose degradation so that investigation of low temperature overheating has been directed in this quarter. More details of this technique are given below.
Diagnosis in practice
There are those persons and organizations who see dissolved gas analysis as the answer to all transformer operating problems. This is not the case. There are occasions when it can create as many problems as it resolves.
The first word of caution is to avoid drawing any conclusions on the basis of a single sample, and by single sample is meant a sample taken at a particular point in time, because the first thing an operator must do on carrying out an analysis which suggests that a fault exists, is to take a second sample and repeat the analysis. On the assumption that the repeat sample confirms the initial diagnosis, the next step is to look at the previous history of the transformer. When was the last sample taken and what was the result? Which gases have changed since that sample and to what extent?
If the transformer has been in service for a long time with no significant change in loading pattern and there is a long history from past sampling of only steadily changing gas levels, then a sudden step change should be taken seriously. Even then it is necessary to proceed with caution. Could the gases have diffused from the diverter-switch compartment of an on-load tapchanger? Has someone topped up the transformer recently using contaminated oil? Only when all these questions have been asked and appropriate answers obtained should it be accepted that a fault exists. Following the receipt of such confirmation the next step is to consider an increased sampling frequency. If the procedure was, as identified above, of taking a second confirmatory sample as soon as the initial suspicion was raised, then it is likely that by the time definite confirmation of a fault has been obtained, some indication of the rate of gas evolution will have been gained, since there is likely to have been at least a day or two time lag between these two samples. The increased sampling frequency will clearly depend on both the rate of evolution and the type of fault. If the fault indicated is only modest overheating, then clearly it is not so important to achieve rapid response than if a very high temperature, perhaps indicated by the presence of acetylene, is indicated.
When the presence of a fault has been definitely confirmed, and its development perhaps been monitored for several weeks, there comes a time when a decision must be made as to how to proceed further. Perhaps it will be decided to drain the oil and enter the tank for a visual examination. Perhaps it will be considered that dissolved gas levels are approaching saturation and danger of free gas production is a possibility. In this situation, processing of the oil, either on-line or by briefly taking the transformer out of service, is an option which may be considered.
The other important aspect to be recognized is that some hydrocarbon gases are present in the oil of most power transformers. Those which have been in service for some years and have been operated for considerable periods at or near to rated load can have levels of many tens, maybe even a hundred or more, parts per million of some gases and still be healthy. Because of this, and because of the many other variable factors involved, as identified above, it is not generally possible to obtain an indication of the condition of a particular transformer, or of its life expectancy, simply by carrying out a dissolved gas analysis. The most reliable indications are those obtained when a d.g.a. history has been maintained and a step-change in an established pattern is suddenly observed. Typical is the case of a large 400 kV generator transformer which had operated satisfactorily for 10 years or so. A routine three-monthly sample then revealed a sudden increase, by a factor of two or three, in ethane and ethylene levels with acetylene, which had previously been showing only 1 or 2 ppm, increased to around 10 ppm. After a repeat sample had confirmed the figures it was decided to take the transformer out of service and drain down the oil in order to gain access to the tank. Following an internal inspection it was found that a flexible connection on a main 400 kV line lead had worked loose. Since this connection was covered by about 15 mm radial thickness of paper insulation, finding the source of the problem was not easy. However the flexible joint was repaired, the lead re-insulated and the transformer returned to service without further problems.
It is appropriate at this stage to consider some case histories in more detail.
The first of these is an example of one of the early successes achieved by CEGB in the UK. After some weeks of site commissioning runs on a large generator, its 22/400 kV generator transformer dissolved gas levels were found to be excessively high for a new transformer. These are shown in the graph, FIG. 125. The actual dissolved gas figures are set out in Table 11.
Table 11 also gives the gas ratios as they were calculated at the time and the Rogers, 1974, Ratios for comparison. It will be seen that the original ratios vary between 1010 and 0010, with the former being the most prevalent.
Interpretation using Table 7 would give the diagnosis 'circulating currents and/or overheated joints' or possibly 'general conductor overheating.' Rogers Ratios are predominantly 1020 and occasionally 0020. These would be interpreted from Table 9 as 'core and tank circulating currents.' Rogers did not envisage a 0020 pattern.
The transformer was de-tanked on site using the turbine hall crane and from a visual examination it soon became clear that the problem was associated with core frame and core frame to tank insulation. The insulation had become damaged during shipment so that arcing was taking place between sections of the frame and also between frame and tank, permitting circulating currents to flow through the frames and the tank. This was a relatively common problem with very large transformers at one time because of the very modest test requirements for the core insulation. The problem was eliminated by specification of higher test levels for this insulation. Additional insulation was inserted at the locations where arcing had been taking place, the active parts were replaced within the tank and the transformer returned to service. FIG. 126 shows a graph of d.g.a. figures following the return to service and it can be seen that these are virtually nil, confirming that the repair has been successful.