Industrial Power Transformers-- Transformer construction (part 1)

Home | Articles | Forum | Glossary | Books


Power transformer construction follows similar principles for units rated from a few kVA up to the largest sizes manufactured, but as the unit size increases a greater degree of sophistication becomes justified. Many manufacturers subdivide their construction activities into 'Distribution' and 'Large Power' although exactly where each one makes this division is varied. Usually the dividing line depends on the weight of the major components and the type and size of handling facilities which are required in the factory. Manufacturers of distribution transformers rated up to between 1 and 2 MVA often utilize roller conveyors and runway beams for the majority of their handling. Large power transformers require heavy lifting facilities such as large overhead cranes.

Those manufacturers who produce the largest sizes may further subdivide their operations into 'Medium power' and 'Large power' sections. Since the largest transformers require very heavy lifting facilities - up to 400 tonnes capacity including lifting beams and slings is not uncommon - it is usual to restrict the use of these very expensive facilities exclusively to the largest units so that the medium construction factory may only possess lifting facilities of up to, say, 30 tons capacity.

These subdivided construction arrangements often coincide with divisions of design departments so that design practices are frequently confined within the same boundaries.

In the descriptions of transformer design and constructional methods which follow the aim will generally be to describe the most developed 'state of the art' even though in some instances, for example for distribution transformers, more simplified arrangements might be appropriate. In Section 7, which describes specialized aspects of transformers for particular purposes, aspects in which practices might differ from the norm will be highlighted.

A note on standards

Since the writing of the previous edition of this book in the mid-1990s there have been a great many changes in the transformer industry worldwide. Due to the very high labor content involved in transformer production and the consequent high manufacturing costs in many parts of Europe and North America, factories have been closed and many mergers across international boundaries have taken place. Utilities which historically have supported their home manufacturers have been forced to procure transformers from further afield.

The process has been accelerated by privatization of utilities in many parts of Europe and this has, in turn, accelerated the process of internationalization of standards. In addition, in Europe, development and expansion of the European Union has led to the production of European Norms (EN) which are versions of the standards produced by the International Electrotechnical Commission (IEC) having specific applicability throughout the European Union. The large purchasing power of the USA has helped to preserve the independence of American Standards despite the loss of most of its indigenous transformer industry, so that throughout the world three main systems of standards for the design and manufacture of transformers are in existence. In most cases there are fairly few differences between EN and IEC Standards, but in the case of American Standards, these differ significantly from EN and IEC including differences in the standards approach. There are now formal links in place between the American National Standards Institute (ANSI) and IEC, but it will, at best, be many years before any significant convergence of transformer standards occurs.

From its inception, this book has developed in the UK, with the result that it has described transformers from a UK standpoint that are designed to British Standards (BS). It will continue to make reference to BS and EN, but where it is considered relevant or helpful to cross refer to American Standards it will aim to do so. It is hoped that the reader will find this acceptable.


Design features

Section 3 has described the almost constant developments which have taken place over the years to reduce the specific losses of core material. In parallel with these developments manufacturers have striven constantly to improve their core designs in order to better exploit the properties of the improved materials and also to further reduce or, if possible, eliminate losses arising from aspects of the core design. Superficially a core built 30 years ago might resemble one produced at the present time but, in reality, there are likely to be many subtle but significant differences.

Core laminations are built up to form a limb or leg having as near as possible a circular cross-section ( FIG. 1) in order to obtain optimum use of space within the cylindrical windings. The stepped cross-section approximates to a circular shape depending only on how many different widths of strip a manufacturer is prepared to cut and build. For smaller cores of distribution transformers this could be as few as seven. For a larger generator transformer, for example, this might be 11 or more. Theoretically, these fill from just over 93 percent to over 95 percent, respectively, of the available core circle. In reality the actual utilization is probably slightly less than this since the manufacturer aims to standardize on a range of plate widths to cover all sizes of cores, or he may buy in material already cut to width, in which case he will be restricted to the standard range of widths provided by the core steel manufacturer, usually varying in 10 mm steps. In either circumstance it will be unlikely that the widths required to give the ideal cross-section for every size of core will be available.

FIG. 1 Core sections. Seven step, taped (left) and 14 step, banded (right)

FIG. 2 Typical core forms for single- and three-phase transformers

Transformer manufacturers will normally produce a standard range of core cross-sections - they often refer to these as frame sizes - with each identified by the width in millimeters of the widest plate. These might start at 200 mm for cores of small auxiliary transformers and progress in 25 mm steps up to about 1 m, the full width of the available roll, for the largest generator transformers. This cylindrical wound limb forms the common feature of all transformer cores. The form of the complete core will however vary according to the type of transformer. Alternative arrangements are shown in FIG. 2; of these, by far the most common arrangement is the three-phase, three-limb core. Since, at all times the phasor sum of the three fluxes produced by a balanced three-phase system of voltages is zero, no return limb is necessary in a three-phase core and both the limbs and yokes can have equal cross-section. This is only true for three-phase cores and for single-phase transformers return limbs must be provided. Various options are available for these return limbs, some of which are shown in FIG. 2, all have advantages and disadvantages and some of these will be discussed in greater depth in Section 7.1, which deals with generator transformers. Generator transformers represent the only occasion where single-phase units are used on three-phase systems although in some countries they are used for large interbus transformers or autotransformers. The main reason for the use of single-phase units is from transport considerations, since the largest generator transformers can be too large to ship as three-phase units.

The use of single-phase units also has advantages where very high reliability is required as, for example in the case of large generator transformers. This aspect will be considered in greater depth in Section 7.1 which deals with generator transformers. FIG. 2 also shows a three-phase, five-limb core which is another arrangement used mainly for large three-phase generator transformers and interbus transformers in order to reduce transport height. This configuration enables the yoke depth to be reduced by providing a return flux path external to the wound limbs. In the limit the yokes could be half that which would be required for a three-phase, three-limb arrangement so the saving in height can be considerable. The 'cost' is in the provision of the return limbs which add significantly to the size of the core and to the iron losses. Of course, if transport height considerations permit, the yoke depth need not be reduced to half the limb width. If the yokes are provided with a cross-section greater than half that of the limbs the flux density in the yokes will be reduced. This will result in a reduction in specific core loss in the yokes which is greater than the proportional increase in yoke weight compared to that of a half-section yoke, hence a reduction in total core loss is obtained. This will be economic if the capitalized cost of the iron loss saved is greater than the cost of the extra material. The only other occasion on which a three-phase, five-limb core might be necessary is when it is required to provide a value of zero-sequence impedance of similar magnitude to the positive-sequence impedance as explained in Section 2.

The first requirement for core manufacture is the production of the individual laminations. Most manufacturers now buy in the core material already cut to standard widths by the steel producer so it is necessary only for them to cut this to length. Production of the laminations is one of the areas in which core manufacture has changed significantly in recent years. As explained in Section 3.2, the specific loss of core steel is very dependent on the nature and level of stress within the material. It is therefore necessary to minimize the degree of working and handling during manufacture. Cutting of the laminations is, of course, unavoidable but this operation inevitably produces edge burrs. Edge burrs lead to electrical contact between plates and the creation of eddy current paths. Until the end of the 1980s BS 601 Steel Sheet and Strip for Magnetic Circuits of Electrical Apparatus laid down acceptable limits for these burrs which generally meant that they had to be removed by a burr-grinding pro cess. Burr-grinding tends to damage the plate insulation and this damage needs to be made good by an additional insulation application. Each of these operations involve handling and burr-grinding in particular raises stress levels, so an additional anneal is required. Modern cutting tools can enable the operation to be carried out with the production of the very minimum of edge burr.

This is to some extent also assisted by the properties of the modern material itself. Typically burrs produced by 'traditional' tools of high-quality tool-steel on cold-rolled grain-oriented material of the 1970s might be up to 0.05 mm in height as permitted by BS 601. These could be reduced by a burr-grinding operation to 0.025 mm. With HiB steel and carbide-steel tools, burrs less than 0.02 mm are produced so that all of the burr-grinding, additional insulating and annealing processes can now be omitted.

It is perhaps appropriate at this stage to look a little further into the subject of plate insulation. The quality of this insulation was defined in BS 601: Part 2, which stated that 80 percent of a specified number of insulation resistance measurements made on a sample of the core plate should be greater than 2 Ohm and 50 percent should be greater than 5 Ohm. As indicated in Section 3.2 the purpose of this insulation is to prevent the circulation of eddy currents within the core. Preventing these currents from flowing does not, however, prevent the induced voltages from being developed. The induced voltage is proportional to the plate width and it was generally considered that plate insulation complying with the requirements of BS 601 was acceptable for plates of up to about 640 mm wide. For cores of a size which would require a plate width greater than this there are the options of subdividing the cross-section so that each part individually meets the 640 mm maximum requirement or, alternatively, additional insulation could be provided. It is often necessary to subdivide large cores anyway in order to provide cooling ducts, so that this option could normally be selected without economic penalty. It should be noted that some manufacturers had long considered that the BS 601 requirement to achieve 2 Ohm was a rather modest one. When they intended to apply additional insulation anyway there was no pressing need for change to the BS and the issue only came to the fore when this additional coating was dropped. At about this time BS 601: Part 2 was superseded by BS 6404: Section 8.7: 1988, Specification for grain-oriented magnetic steel sheet and strip, which stated that the insulation resistance of the coating should be agreed between the supplier and the purchaser. Manufacturers were thus able to take the opportunity to apply their own specifications for the material and these generally called for a higher resistance value. There also remained the question as to what was required of the remaining 20 percent of the readings. These could, in theory, be zero and dependent on the coating process control they could be located in a single area of the steel strip. Reputable transformer manufacturers in this situation issued their own individual specifications usually stipulating that the physical location of the 20 percent low resistance value readings occurred randomly throughout the samples, that is it was not acceptable that all of these should be located in the same area of the sample. As indicated in Section 3.2 many of the modern steels are provided with a high-quality insulation coating which is part of the means of reducing the specific loss. With these steels it is not normally necessary to provide additional coating regardless of the size of the core and the resistance measurements obtained are invariably considerably better than the minimum requirements of the old BS 601.

One of the disadvantages of grain-oriented core steels is that any factor which requires the flux to deviate from the grain direction will increase the core loss and this becomes increasingly so in the case of the HiB range of core steels.

Such factors include any holes through the core as shown in FIG. 3 as well as the turning of the flux which is necessary at the top and bottom corners of the core limbs. This latter effect is noticeable in that a tall slim core will have a lower loss than a short squat core of the same weight and flux density since the former arrangement requires less deviation of the flux as illustrated in FIG. 4. The relationship between the core loss of a fully assembled core and the product of core weight multiplied by specific loss is known as the building factor for the core. The building factor is generally about 1.15 for a well-designed core of grain-oriented steel. Expressed in terms of building factor the tall core discussed above has a better (i.e. lower) building factor than the squat core. In order to limit the extent to which the flux path cuts across of the grain direction at the intersection of limbs and yokes corners of laminations are cut on a 45º miter. The core plates at these mitered corners must be overlapped so that the flux can transfer to the adjacent face rather than cross the air gap which is directly in its path ( FIG. 5). These mitered corners were, of course, not necessary for cores of hot-rolled (i.e. non-oriented) steel. It was also normally accepted practice for cores of hot-rolled steel for the laminations to be clamped together to form the complete core by means of steel bolts passing through both limbs and yokes.

With the advent of grain-oriented steel it was recognized that distortion of the flux by bolt holes through the limbs was undesirable and that the loss of effective cross-section was also leading to an unnecessary increase in the diameter of the core limb. Designers therefore moved towards elimination of core bolts replacing these on the limbs by bands of either steel (with an insulated break) or glass fiber. In the former case the insulated break was inserted in the steel band to prevent current flow in the band itself and additionally it was insulated from the core to prevent shorting out individual laminations at their edges. Core bolts had always needed to be effectively insulated where they passed through the core limbs and yokes for the same reasons. The top and bottom yokes of cores continued to be bolted, however, since the main structural strength of the transformer is provided by the yokes together with their heavy steel yoke frames. FIG. 6 shows a three-phase core of cold-rolled grain-oriented steel with banded limbs and bolted yokes.

Increasing economic pressures in the latter part of the 1970s to reduce losses, and in particular the core loss since it is present whenever the transformer is energized, led designers and manufacturers towards the adoption of totally boltless cores. The punching of holes through core plates has the additional disadvantage that it conflicts with the requirement to minimize the working of the core steel, mentioned above, thus increasing the loss in the material.

Both these factors together with the marginal reduction in core weight afforded by a boltless core were all factors favoring the elimination of bolt holes.

With modern steels having a very high degree of grain orientation the loss penalty for deviation of the flux from the grain direction is even more sign cant so that manufacturers are at even greater pains to design cores entirely without bolts through either limbs or yokes. On a large core this calls for a high degree of design sophistication to ensure that the necessary structural strength is not sacrificed. FIG. 7 shows a large modern core having totally boltless construction.

FIG. 3 Effect of holes and corners on core flux

FIG. 4 Cross flux at corners forms greater portion of total flux path in short squat core than in tall slim core

FIG. 5 45 deg. miter overlap construction

FIG. 6 Three-phase mitered core of a 150 MVA, 132/66 kV and 50 Hz transformer showing the banding of the core limb laminations (Bonar Long Ltd)

FIG. 7 Three-phase three-limb boltless core. Three flitch plates (tie bars) are used each side of each limb and are visible at the top of each limb below the upper frame. The temporary steel bands clamping the limbs will be cut off as the winding assemblies are lowered onto the limbs (Areva T&D)

Core building

The core is built horizontally by stacking laminations, usually two or three per lay, on a jig or stillage. The lay-down sequence must take account of the need to alternate the lengths of plates to provide the necessary overlaps at the mitered corners as shown in FIG. 5. FIG. 8 shows a large core being built in the manufacturer's works. The clamping frames for top and bottom yokes will be incorporated into the stillage but this must also provide support and rigidity for the limbs until the core has been lifted into the vertical position for the fit ting of the windings. Without clamping bolts the limbs have little rigidity until the windings have been fitted so the stillage must incorporate means of providing this. The windings when assembled onto the limbs will not only provide this rigidity, in some designs the hard synthetic-resin-bonded paper (s.r.b.p.) tube onto which the inner winding is wound provides the clamping for the leg laminations. With this form of construction the leg is clamped with temporary steel bands which are stripped away progressively as the winding is lowered onto the leg at the assembly stage. Fitting of the windings requires that the top yoke be removed and the question can be asked as to why it is necessary to build it in place initially. The answer is that some manufacturers tried the process of core building without the top yokes and found that the disadvantages outweighed the saving in time and cost of assembly. If the finished core is to have the lowest possible loss then the joints between limbs and yokes must be fitted within very close tolerances. Building the core to the accuracy necessary to achieve this without the top yoke in place is very difficult. The higher labor costs become, of course, the greater is the incentive to eliminate a step in the manufacturing sequence, so that by the late 1990s the number of manufacturers prepared to invest in the necessary tools and equipment to build cores without the top yoke has become progressively greater. Once the windings have been fitted the top yoke can be fitted, or replaced if it has been removed, suitably interlaced into the projecting ends of the leg laminations, followed by the top core frames. Once these have been fitted, together with any tie bars linking top and bottom yokes, axial clamping can be applied to the windings to compress them down to their correct length. These principles will apply to the cores of all the core-type transformers shown in FIG. 2.

FIG. 8 Four-limb (single phase with two limbs wound) core with 60/40 percent yokes and return limbs in course of building (Areva T&D)

FIG. 9 Five-step step-lapped mitered core joint.

Step-lapped joints

The arrangement of the limb to yoke mitered joint shown in FIG. 5 uses a simple overlap arrangement consisting of only two plate configurations. Because much of the loss associated with a modern transformer core arises from the yoke to limb joints manufacturers have given considerable thought to the best method of making these joints. One arrangement which has been used extensively, particularly in distribution transformers, is the step-lapped joint. In a step-lapped joint perhaps as many as five different plate lengths are used so that the miter can have a five-step overlap as shown in FIG. 9 rather than the simple overlap shown in FIG. 5. This arrangement which allows the flux transfer to be gradual through the joint ensures a smoother transfer of the flux and thus provide a lower corner loss. The disadvantages are that more lengths of plate must be cut, which will increase costs, and the replacing of the top yoke after installation of the windings becomes a more complex process requiring greater care and thus further increased labor costs. On a distribution transformer core the smaller stiffer laminations are probably easier to replace than would be the case on a larger core, which is possibly the reason why this form of construction initially found wider application in distribution transformers. It is also the case that the corner joints represent a larger proportion of the total core in the case of a small distribution transformer than they do in a larger power transformer core, making such an improvement more worth while. (Of course, the other side of the coin is that it must be easier to cut and build a small core, having a yoke length of, say, 1 m, to a degree of tolerance which results in joint gaps of, say, 0.5 mm, than it is for a large core having a yoke length of, say, 4 m.) An additional factor is that the very competitive state of the world distribution transformer market probably means that any savings which can be made, however small, will be keenly sought after.

Core earthing

Before concluding the description of core construction, mention should be made of the subject of core earthing. Any conducting metal parts of a transformer, unless solidly bonded to earth, will acquire a potential in operation which depends on their location relative to the electric field within which they lie. In theory, the designer could insulate them from earthed metal but, in practice, it is easier and more convenient to bond them to earth. However, in adopting this alternative, there are two important requirements:

• The bonding must ensure good electrical contact and remain secure through out the transformer life.

• No conducting loops must be formed, otherwise circulating currents will result, creating increased losses and/or localized overheating.

Metalwork which becomes inadequately bonded, possibly due to shrinkage or vibration, creates arcing which will cause break down of insulation and oil and will produce gases which may lead to Buchholz relay operation, where fitted, or cause confusion of routine gas-in-oil monitoring results by masking other more serious internal faults, and can thus be very troublesome in service.

The core and its framework represent the largest bulk of metalwork requiring to be bonded to earth. On large important transformers, connections to core and frames can be individually brought outside the tank via 3.3 kV bushings and then connected to earth externally. This enables the earth connection to be readily accessed at the time of initial installation on site and during subsequent maintenance without lowering the oil level for removal of inspection covers so that core insulation resistance checks can be carried out.

In order to comply with the above requirement to avoid circulating currents, the core and frames will need to be effectively insulated from the tank and from each other, nevertheless it is necessary for the core to be very positively located within the tank particularly so as to avoid movement and possible damage during transport. It is usual to incorporate location brackets within the base of the tank in order to meet this requirement. Because of the large weight of the core and windings these locating devices and the insulation between them and the core and frames will need to be physically very substantial, although the relevant test voltage may be modest. More will be said about this in Section 5 which deals with testing.

Leakage flux and magnetic shielding

The purpose of the transformer core is to provide a low reluctance path for the flux linking primary and secondary windings. It is the case however that a proportion of the flux produced by the primary ampere-turns will not be con strained to the core thus linking the secondary winding and vice versa. It is this leakage flux, of course, which gives rise to the transformer leakage reactance. As explained in the previous section leakage flux also has the effect of creating eddy current losses within the windings. Control of winding eddy current losses will be discussed more fully in the section relating to winding design, however if the leakage flux can be diverted so as to avoid its passing through the winding conductors and also made to run along the axis of the winding rather than have a large radial component as indicated in FIG. 10, this will contribute considerably to the reduction of winding eddy current losses. The flux shunts will, themselves, experience losses, of course, but if these are arranged to operate at modest flux density and made of similar laminations as used for the core, then the magnitude of the losses in the shunts will be very much less than those saved in the windings. Requirements regarding earthing and prevention of circulating currents will, of course, be the same as for the core and frames. On very high-current transformers, say where the cur rent is greater than about 1000 A, it is also the case that fluxes generated by the main leads can give rise to eddy current losses in the tank adjacent to these. In this situation a reduction in the magnitude of the losses can be obtained by the provision of flux shunts, or shields, to prevent their flowing in the tank. This arrangement, shown in FIG. 11, will also prevent an excessive temperature rise in the tank which could occur if it were allowed to carry the stray flux.

Top of Page

PREV.   NEXT   Guide Index HOME