Industrial Power Transformers-- Transformer construction (part 4)

Home | Articles | Forum | Glossary | Books



When the resistive and other losses are generated in transformer windings heat is produced. This heat must be transferred into and taken away by the transformer oil. The winding copper retains its mechanical strength up to several hundred degrees Celsius. Transformer oil does not significantly degrade below about 140ºC, but paper insulation deteriorates with greatly increasing severity if its temperature rises above about 90ºC. The cooling oil flow must, there fore, ensure that the insulation temperature is kept below this figure as far as possible.

The maximum temperature at which no degradation of paper insulation occurs is about 80ºC. It is usually neither economic nor practical, however, to limit the insulation temperature to this level at all times. Insulation life would greatly exceed transformer design life and, since ambient temperatures and applied loads vary, a maximum temperature of 80ºC would mean that on many occasions the insulation would be much cooler than this. Thus, apart from premature failure due to a fault, the critical factor in determining the life expectancy of a transformer is the working temperature of the insulation or, more precisely, the temperature of the hottest part of the insulation or hot spot. The designer's problem is to decide the temperature that the hot spot should be allowed to reach. Various researchers have considered this problem and all of them tend to agree that the rate of deterioration or ageing of paper insulation rapidly increases with increasing temperature. In 1930, Montsinger reported on some of the materials which were then in common use and concluded that the rate of ageing would be doubled for every 8ºC increase between 90ºC and 110ºC. Other investigators of the subject found that rates of doubling varied for increases between 5ºC and 10ºC for the various materials used in transformer insulation, and a value of 6ºC is now generally taken as a representative average for present day insulation materials.

It is important to recognize that there is no 'correct' temperature for operation of insulation, nor is there a great deal of agreement between transformer designers as to the precise hot spot temperature that should be accepted in normal operation. In fact it is now recognized that factors such as moisture content, acidity and oxygen content of the oil, all of which tend to be dependent upon the breathing system and its maintenance, have a very significant bearing on insulation life. Nevertheless EN 60076 and other international standards set down limits for permissible temperature rise which are dictated by considerations of service life and aim at a minimum figure of about 30 years for the transformer. These documents are based on the premise that this will be achieved with an average hot spot temperature of 98ºC.

It must also be recognized that the specified temperature rise can only be that value which can be measured, and that there will usually be, within the transformer, a hot spot which is hotter than the temperature which can be measured and which will really determine the life of the transformer.

Study of the permitted temperature rises given in EN 60076 shows that a number of different values are permitted and that these are dependent on the method of oil circulation. The reason for this is that the likely difference between the value for temperature rise, which can be measured, and the hot spot, which cannot be measured, tends to vary according to the method of oil circulation. Those methods listed in EN 60076-2, which deals with tempera ture rise, are:

• Natural

• Forced, but not directed;

• Forced and directed

Natural circulation utilizes the thermal head produced by the heating of the oil which rises through the windings as it is heated and falls as it is cooled in passing through the radiators.

With forced circulation, oil is pumped from the radiators and delivered to the bottom of the windings to pass through the vertical axial ducts formed by the strips laid 'above' and 'below' the conductors. In referring to axial ducts within the windings, the expressions 'above' and 'below' mean 'further from the core' and 'nearer to the core', respectively. Radial ducts are those which connect these. In a non-directed design, flow through the radial, and horizontal, ducts which connect the axial ducts above and below is dependent entirely on thermal and turbulence effects and the rate of flow through these is very much less than in the axial ducts ( FIG. 37(a)). With a forced and directed circulation, oil is fed to a manifold at the bottom of the windings and thence in appropriate proportions to the individual main windings. Oil flow washers are inserted at intervals in the winding which alternately close off the outer and then the inner axial ducts so that the oil in its passage through the winding must weave its way through the horizontal ducts thus ensuring a significant oil flow rate in all parts of the winding. This arrangement is illustrated in FIG. 37(b). The rate of heat transfer is very much a function of the rate of oil flow so that the directed oil flow arrangement will result in a lower winding to oil differential temperature or gradient. Typical values of gradient will be discussed shortly.

FIG. 37 Directed and non-directed oil flow

The designer generally aims to achieve a 'balanced' design, in which both top oil temperature-rise and temperature-rise by resistance for LV and HV windings approach reasonably close to the specified maxima by control of the winding gradient. If the gradient is 'too high' it will be necessary to limit the top oil temperature-rise to ensure that the permitted temperature-rise by resistance is not exceeded. Given that the oil flow arrangement used will itself be dictated by some other factors, the designer's main method of doing this will be by adjustment of the number of horizontal cooling ducts which he employs in the winding design.

The average temperature-rise of the winding is measured by its change in resistance compared with that measured at a known ambient temperature.

There are many reasons why the temperature-rise in some parts of the winding will differ significantly from this average however and, whilst some of the differences can be accurately estimated, there are others which are less easily predicted. For example some of the winding at the bottom of the leg is in cool oil and that at the top of the leg will be surrounded by the hottest oil.

It is a relatively simple matter to measure these two values by placing a thermometer in the oil at the top of the tank near to the outlet to the coolers and another at the bottom of the tank. The average oil temperature will be halfway between these two values and the average gradient of the windings is the difference between average oil temperature-rise and average winding tempera ture-rise, that is, the temperature-rise determined from the change of winding resistance. The temperature of the hottest part of the winding is thus the sum of the following:

• Ambient temperature.

• Top oil temperature-rise.

• Average gradient (calculable as indicated above).

• A temperature equal to the difference between maximum and average gradient of the windings (hot spot factor).

It will be seen that this is the same as the sum of:

• Ambient temperature.

• Temperature-rise by resistance.

• Half the temperature difference between inlet oil from cooler and outlet oil to cooler.

• Difference between maximum and average gradient of the windings, as above.

This latter sum is, on occasions, a more convenient expression for hot spot temperature. In both cases it is the last of these quantities which cannot be accurately determined. One of reasons why there will be a difference between maximum gradient and average gradient will be appreciated by reference to FIG. 38 which represents a group of conductors surrounded by vertical and horizontal cooling ducts. The four conductors at the corners are cooled directly on two faces, whilst the remainder are cooled on a single face only. Furthermore, unless the oil flow is forced and directed, not only will the heat transfer be poorer on the horizontal surfaces, due to the poorer oil flow rate, but this oil could well be hotter than the general mass of oil in the vertical ducts. In addition, due to the varying pattern of leakage flux, eddy current losses can vary in different parts of the winding. Fortunately copper is as good a conductor of heat as it is of electricity so that these differences can be to a large extent evened out. However in estimating the hot spot temperature this difference between average and maximum winding gradient cannot be neglected. For many years this was taken to be approximately 10 percent of the average gradient, that is, the maximum gradient was considered to be 1.1 times the average. It is now suggested that this might have been somewhat optimistic and the 1991 issue of IEC 60354, Guide to loading of oil-immersed power transformers, concludes that a value of 1.1 is reason able for small transformers but that a figure of up to 1.3 is more appropriate for medium and large transformers. More will be said on this aspect in Section 6.8.

FIG. 38 Winding hot spots

BS EN 60076-2 deals with temperature-rise. In this document the type of cooling for a particular transformer is identified by means of a code of up to four letters. These are as follows.

The first letter refers to the type of internal cooling medium in contact with the windings. This may be:

O mineral oil or synthetic insulating liquid with a fire point 300ºC;

K insulating liquid with fire point 300ºC;

L insulating liquid with no measurable fire point.

The second letter refers to the circulation mechanism for the internal cooling medium from the options described above:

N natural thermosiphon flow through cooling equipment and windings;

F forced circulation through cooling equipment, thermosiphon through windings;

D forced circulation through cooling equipment, directed from the cooling equipment into at least the main windings.

Frequently, tapping windings which might contain only 20 percent of the total ampere-turns and thus have far less losses to dissipate than the main windings, will be excluded from the directed flow arrangements and cooled only by natural circulation.

The third letter refers to the external cooling medium, thus:

A air;

W water.

The fourth letter refers to the circulation mechanism for the external cooling medium:

N natural convection;

F forced circulation (fans, pumps).

A transformer may be specified to have alternative cooling methods, for example, ONAN/ODAF, which is a popular dual rating arrangement in the UK. The transformer has a totally self-cooled or ONAN rating, usually to cover base load conditions and a forced cooled ODAF rating achieved by means of pumps and fans to provide for the condition of peak load. A ratio of one to two between the ONAN and ODAF ratings is common.

For normal ambient conditions, which are defined in EN 60076-2, as air never below _25ºC and never hotter than _40ºC, not exceeding _30ºC aver age during the hottest month and not exceeding _20ºC yearly average, or water never exceeding 25ºC at the inlet to oil/water coolers, permitted temperature rises are as follows:

Temperature rise of top oil 60 K

Average winding temperature rise by resistance

- for transformers identified as ON.. or OF.. 65 K

- or transformers identified as OD.. 70 K

No tolerances are permitted on the above values.

In all except the smallest transformers cooling of the oil will be by some external means, tubes or radiators mounted on the side of the tank, external banks of separate radiators or even oil/water heat exchangers. If the oil is required to circulate through these coolers by natural thermosiphon, that is, ON.. type cooling is employed, then a fairly large thermal head will be required to provide the required circulation, possibly of the order of 25 K. If the oil is pumped through the coolers, that is, OF.. or OD.. type cooling is employed, then the difference between inlet and outlet oil temperatures might be, typically, 10-15 K. Thus temperatures within designs of each type of transformers, using the second of the two alternative derivations identified above, might typically be:

Type of cooling ODAF ONAN

(a) Ambient (BS EN 60076) 30 30

(b) Temperature-rise by resistance (BS EN 60076) 70 65

(c) Half (outlet-inlet) oil 8 12

(d) Maximum gradient-average gradient, typical value 4 5

Hot spot temperature 112 112

The differences between maximum and average gradient are estimates simply for the purpose of illustration. The value has been taken to be less for the ODAF design on the basis that there are likely to be fewer inequalities in oil flow rates.

The fact that the hot spot temperature is the same in both cases is coincidence.

For each of the above arrangements the permitted top-oil rise according to EN 60076-2 is 60ºC, so the mean oil rises could be (60 - 8) _ 52ºC and (60 - 12) _ 48ºC, respectively, for the ODAF and ONAN designs. Since temperature-rise by resistance is mean oil temperature-rise plus gradient, it would thus be acceptable for the winding gradient for the ODAF design to be up to 18ºC and for the ONAN design this could be up to 17ºC. This is, of course, assuming 'balanced' designs as defined above. It should be remembered that, if one of the windings is tapped, the transformer is required to deliver full rating on the maximum minus tapping and that the EN 60076-2 temperature rise limits must be met on this tapping.

It must be stressed that in the examples given above, items (c) and (d) can not be covered by specification, they are typical values only and actual values will differ between manufacturers and so, therefore, will the value of hot spot temperature. It will be noted also that the hot spot temperatures derived significantly exceed the figure of 98ºC quoted above as being the temperature which corresponds to normal ageing. It will also be seen that the figure used for ambient temperature is not the maximum permitted by EN 60076-2, which allows for this to reach 40ºC, giving a hot spot temperature of 122ºC in this case. Such temperatures are permissible because the maximum ambient temperature occurs only occasionally and for a short time.

When a transformer is operated at a hot spot temperature above that which produces normal ageing due to increase in either ambient temperature or loading, then insulation life is used up at an increased rate. This must then be offset by a period with a hot spot temperature below that for normal ageing, so that the total use of life over this period equates to the norm. This is best illustrated by an example; if 2 hours are spent at a temperature which produces twice the normal rate of ageing then 4 hours of life are used in this period. For the balance of those 4 hours (i.e. 4 - 2 _ 2) the hot spot must be such as to use up no life, that is, below 80ºC, so that in total 4 hours life are used up. This principle forms the basis of IEC 60354. This subject will be discussed at greater length in Section 6.8. The system works well in practice since very few transformers are operated continuously at rated load. Most transformers associated with the public electricity supply network are subjected to cyclic daily loading patterns having peaks in the morning and afternoon. Many industrial units have periods of light loading during the night and at weekends, and ambient temperatures are subject to wide seasonal variations. In addition, in many temperate countries such as the UK a significant portion of the system load is heating load which is greater in the winter months when ambient temperatures are lower, thus reducing the tendency for actual hot spot temperatures to reach the highest theoretical levels.

Core, leads and internal structural steelwork

Although the cooling of the transformer windings represents the most important thermal aspect of the transformer design, it must not be overlooked that considerable quantities of heat are generated in other parts. The core is the most significant of these. There is no specified maximum for the tempera ture rise of the core in any of the international standards. One of the reasons for this is, of course the practical aspect of enforcement. The hottest part of the core is not likely to be in a particularly accessible location. In a three phase three-limb core, for example, it is probably somewhere in the middle of the leg to yoke joint of the centre limb. Its temperature could only be measured by means of thermocouple or resistance thermometer, even this exercise would be difficult and the accuracy of the result would be greatly dependent on the manufacturer placing the measuring device in exactly the right location. EN 60076 resolves this difficulty by stating that the temperature rise of the core or of electrical connections or structural parts shall not reach temperatures which will cause damage to adjacent parts or undue ageing of the oil. This approach is logical since, in the case of all of these items, temperatures are unlikely to reach such a value as to damage core steel or structural metalwork or even the copper of leads. It is principally the material in contact with them, insulation, or oil, which is most at risk of damage. Hence 'dam age to adjacent parts' usually means overheating of insulation and this can be detected during a temperature rise test if oil samples are taken for dissolved gas analysis. More will be said about this in Section 5, which deals with testing.

Cooling of the core will usually be by natural circulation even in transformers having forced cooling of the windings. The heat to be removed will depend on grade of iron and flux density but direct heat transfer from the core surface to the surrounding oil is usually all that is necessary up to leg widths (frame sizes) of about 600 mm. Since the ratio of surface area to volume is inversely proportional to the diameter of the core, at frame sizes above this the need to provide cooling becomes an increasingly important consideration.

Because the concern is primarily that of overheating of insulation, some users do specify that the maximum temperature rise for the surface of the core should not exceed the maximum temperature permitted for windings. Some users might also agree to a localized hot spot of 130ºC on the surface of very large cores in an area well removed from insulation, on the basis that oil will not be significantly degraded on coming into contact with this temperature provided the area of contact is not too extensive and recognizing that cooling of these large cores is particularly problematical. Enforcement of such restrictions, of course, remains difficult.

Cooling of the oil

In discussion of the typical internal temperatures identified above, little has been said about the cooling of the oil, which having taken the heat from the windings and other internal parts, must be provided with means of dissipating this to the atmosphere. In a small transformer, say up to a few kVA, this can be accomplished at the tank surface. As a transformer gets larger, the tank surface will increase as the square of the linear dimension whereas the volume, which is related to rating and thus its capacity for generating losses, will increase in proportion to the cube of this, so the point is soon reached at which the available tank surface is inadequate and other provision must be made to increase the dissipation, either tubes or fins attached to the tank, or radiators consisting of a series of pressed steel 'passes.' Whilst the transformer remains small enough for fins or tubes to be used, heat loss is by both radiation and convection. The radiation loss is dependent on the size of the envelope enclosing the transformer, convection loss is related to the total surface area. The effectiveness of a surface in radiating energy is also dependent on its emissivity, which is a function of its finish. Highly polished light colored surfaces being less effective than dull black surfaces. In practical terms, however, investigators soon established that most painted surfaces have emissivities near to unity regardless of the colour of the paint.

It is possible to apply the laws of thermodynamics and heat transfer to the tank and radiators so as to relate the temperature-rise to the radiating and convecting surfaces and, indeed, in the 1920s and 1930s when much of the basic ground work on transformer cooling was carried out, this was done by a combination of experiment and theory. Nowadays manufacturers have refined their databases empirically so as to closely relate the cooling surface required to the watts to be dissipated for a given mean oil rise. For the larger sizes of transformer, say, above a few MVA, the amount of convection surface required becomes so large that the radiating surface is negligible by proportion and can thus be neglected. Then it is simply a matter of dividing the total heat to be dissipated by the total cooling surface to give a value of watts per square centimeter, which can then be tabulated against mean oil rise for a given ambient.

As an approximate indication of the order of total convection surface required when heat is lost mainly by convection, for a mean oil rise of 50 K in an ambient of 20ºC, about 0.03 watts/cm2 can be dissipated.


FIG. 39 Arrangements of cooling radiators

Center line of radiators Centre line of transformer tank h h

(a) Height, 'h' of radiator centre line above tank center line is a measure of the thermal head available to provide circulation of oil. The use of 'swan-necked' connecting pipes enables radiators to be raised and longer radiators to be used

Conservator Buchholz relay

Oil vent pipe

To header Bottom header Space for fan(s) if required Pump may be installed if required h

(b) Provision of separate bank of radiators allows 'h' to be increased considerably


An example can be used to translate this figure into practical terms. Consider a 10 MVA ONAN transformer having total losses on minimum tapping of 70 kW. Let us assume it has a tank 3.5 m long _ 3.5 m high _ 1.5 m wide.

Total cooling surface required at watts 003 . / /cm m

Tank surface (sides pl 2 2 70000

003 233.

u us cover) __

_ 235 35 215 35 15 35 40 (. .) (. .)

..25 233 4 2 m Hence, net surface of radiators __ 0 025 193 2

_ m Suppose pressed-steel radiators are used 3 m long _ 0.25 m wide, these will have a convection surface of approximately 1.5 m2 per pass, hence 193/1.5 _ 129 passes will be required or, say, 10 radiators of 13 passes per radiator.

It will be noted that in the above example, the tank is contributing about one-sixth of the total convection surface required. If the transformer were a 30/60 MVA ONAN/ODAF, having total losses at its 30 MVA ONAN rating of 100 kW, then for the same mean oil temperature-rise the total convection surface required is about 333 m^2 . The tank may have only increased to 4 m long x 3.6 m high x 1.7 m wide, so that it will contribute only 47.8 m^2 or about one-seventh of the area required and, clearly, as unit size increases the contribution from the tank is steadily reduced. At the ODAF rating when fans are brought into service, these will blow the radiator surface much more effectively than they will the tank, even if the radiator banks are tank mounted. Hence, it becomes less worthwhile including the tank surface in the cooling calculations. Additionally, there may be other reasons for discounting the tank, for example it may be necessary to provide an acoustic enclosure to reduce external noise. There can then be advantages in mounting the radiators in a separate bank. Some of these can be seen by reference to FIG. 39.

An important parameter in an ONAN cooling arrangement is the mounting height of the radiators. The greater the height of the horizontal centreline of the radiators in relation to that of the tank, the greater will be the thermosiphon effect creating the circulation of the oil and the better this circulation, the less will be the difference between inlet and outlet oil temperature. The net effect is to reduce the hot spot temperature-rise for the same heat output and effective cooling surface area. To fully appreciate this it is necessary to refer back to the derivation of the hot spot temperature given above. This is related to the top oil temperature plus maximum gradient. The area of cooling surface determines the mean oil temperature, which is less than top oil by half the difference between inlet and outlet oil. Thus, the smaller this difference, the less will be the amount added to the mean oil temperature to arrive at top oil temperature and the lower will be the hot spot temperature.

When the radiators are attached to the tank, there is a limit to the mounting height of these, although some degree of swan-neck connection is possible as shown in FIG. 39(a). If the radiators are separately mounted the height limitation is dictated solely by any restrictions which might be imposed by the location. In addition the tank height ceases to impose a limitation to the length of radiator which can be used and by the use of longer radiators fewer of them may be necessary.


Top of Page PREV.   NEXT   Guide Index HOME