Industrial Power Transformers-- Transformer construction (part 3)

Home | Articles | Forum | Glossary | Books



Mention has already been made of the fact that the LV winding is placed next to the core because it has the lower insulation level. It is now necessary to look in further detail at the subject of insulation and insulation levels and to examine the effects of these on the disposition of the windings.

Transformer windings may either be fully insulated or they can have graded insulation. In EN 60076-3 these are termed uniform insulation and non uniform insulation, respectively. In a fully or uniform insulated winding, the entire winding is insulated to the same level, dictated by the voltage to which the entire winding is to be raised on test.

Graded or non-uniform insulation allows a more economical approach to be made to the design of the insulation structure of a very high-voltage (EHV) winding. With this system, recognition is made of the fact that such windings will be star connected and that the star point will be solidly earthed. The insulation of the earthy end of the winding thus need only be designed for a nominal level.

Before the adoption of EN 60076-3 and its predecessor IEC 76 in the UK, BS 171 required that all windings up to 66 kV working level should he uniformly insulated. Above this, which in the UK means 132, 275 and 400 kV, non-uniform insulation was the norm. Although EN 60076-3 allows for either system to be used at all voltage levels, the UK practice has been continued partly for reasons of custom and practice and also because in many instances new equipment being procured must operate in parallel with equipment designed to earlier standards. In addition the systems themselves have been designed for this standard of equipment. Since most transformers having two EHV windings, that is, each winding at 132 kV or higher, tend to be auto transformers, this means that most double wound transformers will have, at most, only one winding with graded insulation and many will have both windings fully insulated.

FIG. 26(a) shows the arrangement of a transformer in which both windings are fully insulated. This might be a primary substation transformer, 33/11 kV and perhaps around 20 MVA. The LV winding must withstand an applied voltage test which will raise the entire winding to 28 kV above earth.

The winding insulation must therefore withstand this voltage between all parts and earthed metalwork, including the core. Along the length of the winding this test voltage appears across the dovetail strips plus the thickness of the s.r.b.p. tube. At the ends, these strips and the tube are subjected to surface creepage stress, so that the end-insulation distance to the top and bottom yokes must be somewhat greater.

FIG. 26 Arrangements of windings and leads for transformer having uniform insulation

The 33 kV winding is tested at 70 kV above earth. The radial separation between LV and HV must be large enough to withstand this with, say, a single pressboard wrap and spacing strips inside and outside ( FIG. 26). The end insulation will be subjected to creepage stress and so the distance to the yoke must be somewhat greater than the HV/LV distance. Between the transformer limbs, the HV windings of adjacent phases come into close proximity.

To withstand the 70 kV test voltage between phases, it is necessary to have a clearance similar to that between HV and LV windings with, say, a single pressboard barrier in the middle of this distance, as shown in FIG. 26(b).

The LV winding leads are taken out at the top and bottom of the leg, which means that they must of necessity pass close to the core framework. Since they are at relatively low voltage, it is probable that the necessary clearance can be obtained by bending these away from the core as close to the winding as possible and by suitably shaping the core frame ( FIG. 26(c)).

The HV winding leads also emerge from the top and bottom of the leg but these are taken on the opposite side of the coils from the LV leads. Being at a greater distance from the core frame than those of the LV winding, as well as having the relatively modest test voltage of 70 kV, these require little more insulation than those of the LV winding.

It is usually convenient to group the tapping sections in the center of the HV windings. This means that when all the taps are not in circuit, any effective 'gap' in the winding is at the center, so that the winding remains electromagnetically balanced. More will be said about this aspect below. The tapping leads are thus taken from the face of the HV winding, usually on the same side of the transformer as the LV leads.

FIG. 27 shows the arrangement of a transformer in which the LV winding is fully insulated and the HV winding has non-uniform (graded) insulation. This could be a bulk supply point transformer, say, 132/33 kV, star/delta connected, possibly 60 MVA, belonging to a Distribution Network Operator (DNO). Some DNOs take some of their bulk supplies at 11 kV, in which case the transformer could be 132/11 kV, star/star connected, and might well have a tertiary winding. This too could be 11 kV although it is possible that it might be 415 V in order to fulfill the dual purpose of acting as a stabilizing winding and providing local auxiliary supplies for the substation. Whichever the voltage class, it would be placed nearest to the core. If 11 kV, the test levels would be the same as the 11 kV LV winding and that of the LV winding of the 33/11 kV transformer described above. If 415 V, the test levels would be very modest and the insulation provided would probably be dictated by physical considerations rather than electrical. In either case the tertiary and LV insulations would be similar to that of the 33/11 kV transformer. The LV winding would be placed over the tertiary and the tertiary to LV gap would require radial and end insulation similar to that between LV and core for the star/delta design. The 132 kV HV winding is placed outside the LV winding and it is here that advantage is taken of the non-uniform insulation.

FIG. 27

For 132 kV class non-uniform insulation, when it is intended that the neutral shall be solidly connected to earth, the applied voltage test may be as low as 38 kV above earth. (More will be said about the subject of dielectric test levels in Section 5.) When the overpotential test is carried out, around 230 kV is induced between the line terminal and earth. Consequently the neutral end needs insulating only to a level similar to that of the LV winding, but the line end must be insulated for a very much higher voltage. It is logical, therefore, to locate the line end as far as possible from the core and for this reason it is arranged to emerge from a point halfway up the leg. The HV thus has two half-windings in parallel, with the neutrals at the top and bottom and the line ends brought together at the center. If, with such an arrangement, the HV taps are at the starred neutral end of the winding, the neutral point can thus be conveniently made within the tapchanger and the voltage for which the tap changer must be insulated is as low as possible. Unfortunately it is not possible to locate these tapping coils in the body of the HV winding since, being at the neutral end, when these were not in circuit there would be a large difference in length between the HV and LV windings. This would greatly increase leakage flux, stray losses and variation of impedance with tap position as well as creating large unbalanced forces on short-circuit. It is therefore necessary to locate the taps in a separate winding placed outside the HV winding. This winding is shorter than the HV and LV windings and split into upper and lower halves, with an unwound area in the middle through which the HV line lead can emerge.

The center of the HV winding must be insulated from the LV winding by an amount capable of withstanding the full HV overpotential test voltage. This requires a radial distance somewhat greater than that in the 33/11 kV transformer and the distance is taken up by a series of pressboard wraps interspersed by strips to allow oil circulation and penetration. Alternatively, it is possible that the innermost wrap could be replaced by an s.r.b.p. tube which would then provide the base on which to wind the HV winding. The disadvantage of this alternative is that the HV to LV gap is a highly stressed area for which s.r.b.p. insulation is not favored (see Section 3) so that, whilst it might be convenient to wind the HV onto a hard tube, the use of such an arrangement would require a reduced high to low design stress and a greater high to low gap. High to low gap, a, appears in the numerator of the expression for percent reactance (Eq. (2.1) of Section 2) multiplied by a factor three. If this is increased then winding axial length, l, must be increased in order to avoid an increase in reactance, thus making the transformer larger. The designer's objective is normally, therefore, to use as low a high to low as possible and it is probably more economic to wind the HV over a removable mandrel so that it can be assembled onto the LV on completion thus avoiding the use of a hard tube.

FIG. 28 Arrangement of HV line lead with outer HV tapping winding and non-uniform insulation

The voltage appearing on test between the line end of the HV winding and the neutral-end taps is similar to that between HV and LV winding so it is necessary to place a similar series of wraps between the HV and tapping windings. These wraps must be broken to allow the central HV line lead to emerge; an arrangement of petalling (see Section 3) or formed collars may be used to allow this to take place without reducing the insulation strength ( FIG. 28).

Although the system of non-uniform insulation lends itself well to the form of construction described above, which is widely used in the UK for transformers of 132 kV and over, there are disadvantages and there are also circumstances when this cannot be used. The main disadvantage is seen when the transformer rating is such that the HV current is small. An example will make this clear.

Although it is rare to require to transform down from 400 kV at ratings as low as 60 MVA, it has on occasions occurred, for example to provide station sup plies for a power station connected to the 400 kV system where there is no 132 kV available. In this case the HV line current is 86.6 A. With two half HV windings in parallel the current in each half winding is 43.3 A. A typical cur rent density for such a winding might be, say, 3 A/mm2 so that at this current density the required conductor cross-section is 14.4 mm2. This could be pro vided by a conductor of, say, 3 x 5 mm which is very small indeed and could not easily be built into a stable winding, particularly when it is recognized that possibly 1 mm radial thickness of paper covering might be applied to this for this voltage class. It would therefore be necessary to use a much lower-current density than would normally be economic in order to meet the physical constraints of the winding.

This problem can be eased by utilizing a single HV winding instead of two half-windings in parallel as indicated in FIG. 29. This would immediately result in a doubling of the conductor strand size so that this might typically become 3 x 10 mm which is a much more practicable proposition. Of course, the benefit of the central line lead is now no longer available and the winding end must be insulated for the full test voltage for the line end.

FIG. 29 Typical arrangement of windings and HV line lead for uniform insulation

Non-uniform insulation cannot be used if the neutral is to be earthed via an impedance, as is often the case outside the UK, nor is it acceptable for a delta connected HV which would be unlikely to be used in the UK at a voltage of 132 kV and above, but is used occasionally in other countries, so again there is no merit in having the line lead at the center of the leg. Hence the arrangement shown in FIG. 29 would be necessary. If a delta connection is used, any HV tappings must be in the middle of the winding and in order to meet the uniform insulation requirement, the tapchanger must be insulated to the full HV test level. Such a configuration will clearly be more costly than one with non-uniform insulation, but this simply demonstrates the benefit of a solidly earthed neutral as far as the transformer is concerned. No doubt proponents of systems having impedance earthing of the neutral would wish to identify benefits to the system as a whole of using this arrangement.


The previous section dealt with the disposition of the windings as determined by the need to meet the power frequency tests, or electromagnetic voltage distributions which are applied to the windings, but it also briefly mentioned the need to withstand the effect of steep-fronted waves. When testing a power transformer such waves are simulated by an impulse test, which is applied to the HV line terminals, and in the case of fully insulated transformers, to the HV neutral terminal, in addition to the dielectric testing at 50 Hz. Impulse testing arose out of the need to demonstrate the ability to withstand such waves, generated by lightning strikes, usually to the HV system to which the transformer is connected. These waves have a much greater magnitude than the power frequency test voltage but a very much shorter duration.

FIG. 30 Equivalent circuit of transformer for simplified uniform winding

Whilst considering the construction of transformer windings it is necessary to understand something of the different effect which these steep-fronted waves have on them compared with power frequency voltages and to examine the influence which this has on winding design. Section 6.5 which deals with all types of transients in transformers will go more deeply into the theory and examine the response of windings to lightning impulses in greater detail.

For simulation purposes a standard impulse wave is defined in EN 60076-3 as having a wave front time of 1.2 µs and a time to decay to half peak of 50 µs. (More accurate definition of these times will be found in Section 5 which deals with transformer testing.) When struck by such a steep wave front, a transformer does not behave as an electromagnetic impedance, as it would to power frequency voltages, but as a string of capacitors as shown in FIG. 30. When the front of the impulse wave initially impinges on the winding, the capacitances Cs to the succeeding turn and the capacitance of each turn to earth Cg predominate, so that the reactance and resistance values can be ignored. It will be shown in Section 6.5 that when a high voltage is applied to such a string, the distribution of this voltage is given by the expression:

e ExL x

_ sinh( / ) sinh a a

where E _ magnitude of the incident wavefront

L _ winding length

and a _ () CC gs/

… which represents a curve of the form shown in FIG. 31. The initial slope of this curve, which represents the voltage gradient at the point of application, is proportional to Cg/Cs. In a winding in which no special measures have been taken to reduce this voltage gradient, this would be many times that which would appear under power frequency conditions. If additional insulation were placed between the winding turns, this would increase the spacing between them and thus reduce the series capacitance Cs. Cg would be effectively unchanged, so the ratio Cg/Cs would increase and the voltage gradient become greater still. The most effective method of controlling the increased stress at the line end is clearly to increase the series capacitance of the winding, since reducing the capacitance to earth, which can be partially achieved by the use of electrostatic shields, is nevertheless not very practicable. [A short squat winding tends to have a lower capacitance to ground than a tall slim winding, so such an arrangement would have a better intrinsic impulse strength. There are, however, so many other constraints tending to dictate winding geometry that designers are seldom able to use this as a practical means of obtaining the required impulse strength.]

FIG. 31 Distribution of impulse voltage within winding

FIG. 32 shows several methods by which series capacitance can be increased. The first, FIG. 32(a), uses an electrostatic shield connected to the line end and inserted between the two HV discs nearest to the line end. The second, FIG. 32(b), winds in a dummy strand connected to the line lead but terminating in the first disc. Both of these arrangements effectively bring more of the winding turns nearer to the line end. The electrostatic shield was probably the first such device to be used and is possibly still the most widely favored. The shield itself is usually made by wrapping a pressboard ring of the appropriate diameter with thin metal foil (thin to ensure minimum stray loss - see description of shield for shielded-layer winding earlier in this section) and then covering this with paper insulation of about the same radial thickness as the winding conductor. It is necessary to make a connection to the foil in order to tie this to the line lead and this represents the greatest weakness of this device, since, as indicated in the description of the shielded-layer winding, making a high integrity connection to a thin foil is not a simple matter.

FIG. 32 Types of winding stress control

As the travelling wave progresses further into the winding, the original volt age distribution is modified due to the progressive effect of individual winding elements and their capacitances, self and mutual inductances and resistance.

The voltage is also transferred to the other windings by capacitance and inductive coupling. FIG. 33(a) shows a series of voltage distributions typical of a conventional disc winding having an HV line lead connection at one end.

It can be seen that, as time elapses, the voltage distribution changes progressively -- the travelling wave being reflected from the opposite end of the winding back towards the line end, and so on. These reflections interact with the incoming wave and a complex series of oscillations occur and reoccur until the surge energy is dissipated by progressive attenuation and the final distribution ( FIG. 33(b)) is reached.

Thus it is that, although the highest voltage gradients usually occur at or near to the line-end connection coincident with the initial arrival of the impulse wave, these progress along the winding successively stressing other parts and, whilst these stresses might be a little less than those occur ring at the line end they are still likely to be considerably greater than those present under normal steady-state conditions. In many instances, therefore, stress control measures limited to the line end will be insufficient to provide the necessary dielectric strength and some form of interleaving is required ( FIG. 32(c)). This usually involves winding two or more strands in parallel and then reconnecting the ends of every second or fourth disc after winding to give the interleaving arrangement required. It has the advantage over the first two methods that it does not waste any space, since every turn remains active. However, the cost of winding is greatly increased because of the large number of joints. It is possible by adjustment of the degree of interleaving, to achieve a nearly linear distribution of impulse voltage throughout the winding, although because of the high cost of interleaving, the designer aims to minimize its use and, where possible to restrict it to the end sections of the winding.

After the line end sections, the next most critical area will usually be at the neutral end of the winding, since the oscillations resulting from interactions between the incident wave and the reflection from the neutral will lead to the greatest voltage swings in this area ( FIG. 33). If some of the tapping winding is not in circuit, which happens whenever the transformer is on other than maximum tap, the tapping winding will then have an overhang which can experience a high voltage at its remote end. The magnitude of the impulse voltage appearing both across the neutral end sections and within the tapping winding overhang will be similar and will be at a minimum when the initial distribution is linear, as can be seen from FIG. 34. It is often necessary, there fore, to use a section of interleaving at the neutral end to match that of the line end sections. The magnitude of impulse voltage seen by the tapping winding due to overhang effects is likely to be dependent on the size of tapping range (although it will also be influenced by type of tapping arrangement, for example, buck/boost or linear, and physical disposition of this with respect to the HV winding and earth), so this must be borne in mind when deciding the size of tapping range required.

FIG. 33 Voltage distribution through windings

FIG. 34 Impulse voltage distribution in tapping winding overhang - tapchanger selected on minimum tap The need for an interleaved HV winding arrangement, as opposed to, say, a simpler line end shield, is often determined by the rating of the transformer as well as the voltage class and impulse test level required. The lower the MVA rating of the unit the smaller the core frame size, which in turn leads to a lower volts per turn for the transformer and a greater total number of turns. A disc winding with a high total number of turns must have a large number of turns per disc, perhaps as many as fifteen or sixteen, in order to accommodate these within the available winding length, compared with a more normal figure of, say, eight or nine. As a result, the maximum volts appearing between adjacent discs might be as high as 32 times the volts per turn compared with say, only 18 times in a more 'normal' winding. This large difference in power frequency voltage between adjacent sections, that is discs, can become even more marked for the impulse voltage distribution, thus necessitating the more elaborate stress-control arrangement.

For very HV windings the impulse voltage stress can be too high to be satisfactorily controlled even when using an interleaved arrangement. It is in this situation that it may be necessary to use a shielded-layer winding as described earlier in the section. When the steep-fronted impulse wave, impinges on the line end of this type of winding, the inner and outer shields behave as line and earth plates of a capacitor charged to the peak magnitude of the impulse voltage. The winding layers between these plates then act as a succession of intermediate capacitors leading to a nearly linear voltage distribution between the shields. (This is similar to the action of the intermediate foils in a con denser bushing which is described in Section 4.8). With such a near linear distribution, the passage of the impulse wave through the winding is not oscillatory and the insulation structure required to meet the impulse voltage is the same as that required to withstand the power-frequency stress. Electrically, therefore, the arrangement is ideal. The disadvantage, as explained earlier, is the winding's poor mechanical strength so that a disc winding is used when ever the designer is confident that the impulse stress can be satisfactorily con trolled by static shield, dummy strand or by interleaving.

FIG. 35 Arrangement of rod gap on 11 kV bushing

FIG. 36 Chopped-wave impulse test record for 132/33 kV transformer

Chopped waves

For many years it has been the practice to protect transformers of all voltages connected to overhead lines and therefore exposed to lightning overvoltages, by means of surge diverters or co-ordinating gaps. More will be said about the devices themselves in Section 6.6. Although such devices undoubtedly protect the windings by limiting the magnitude of the wavefronts and the energy transferred to them, operation of these does itself impose a very steep rate of change of voltage onto the line terminal which can result in severe inter-turn and inter-section stress within the windings. The most modern surge arresters are designed to attenuate steep-fronted waves in a 'softer' manner than the majority of those used hitherto, but the cost of protecting every transformer connected to an overhead line in this way would be prohibitive. By far the most practicable and universal form of protection used in the UK is the rod gap or co-ordinating gap. FIG. 35 shows a simple arrangement as used on the 11 kV HV bushings of a 11/0.415 kV rural distribution transformer.

The co-ordinating gap is designed to trigger at a voltage just below that to which the winding may be safely exposed. If it is set too low it will operate too frequently. Set too high it will fail to provide the protection required.

Because of the severe dV/dt imposed on the transformer windings by the triggering of a rod gap it has been practice to test for this condition by means of chopped wave tests when carrying out impulse tests in the works.

FIG. 36 shows a typical chopped impulse wave as applied during these tests. For many years the chopping was carried out by installing a rod gap across the impulse generator output. In order to ensure that this gap flashed over as close as possible to the nominal impulse test level, it was practice in the UK electricity supply industry to specify that the impulse voltage for the chopped-wave test should be increased by a further 15 percent over the nor mal full-wave test level. Specification requires that the gap should flash over between 2 and 6 µs from the start of the wave and since the nominal time to peak is 1.2 µs, this means that the peak has normally passed before flashover and the winding has been exposed to 115 percent of the nominal test voltage.

Designers were thus required to design the windings to withstand this 115 percent as a full-wave withstand. It is now possible to use triggered gaps whose instant of flashover can be very precisely set, so the need to specify that the test be carried out at 115 percent volts no longer arises and BS EN 60076-3, which deals with dielectric testing of transformers, now specifies that the chopped-wave tests should be carried out at 110 percent volts. As far as with standing the rapidly collapsing voltage wave is concerned, this will, of course, be better dispersed through the winding with a high series capacitance, so that the winding design will follow the same principles as for the full-wave withstand.


Top of Page PREV.   NEXT   Guide Index HOME