|Home | Articles | Forum | Glossary | Books|
The majority of power transformers in use throughout the world are oil filled using a mineral oil, complying with IEC 60296. In the UK the relevant specification is British Standard (BS) 148, Unused and reclaimed mineral insulating oils for transformers and switchgear which, in its 1998 edition, differs in a number of respects from IEC 60296. More will be said about this later. The oil serves the dual purpose of providing insulation and as a cooling medium to conduct away the losses which are produced in the transformer in the form of heat.
Mineral oil is combustible - it has a fire point of 170ºC - and transformer fires do sometimes occur. It is usual, therefore, to locate these out of doors where a fire is more easily dealt with and consequentially the risks are less. It is necessary to consider the need for segregation from other plant and incorporate measures to restrict the spread of fire.
Because of the fire hazard associated with mineral oil, it has been the practice to use designs for smaller transformers which do not contain oil. These may be entirely dry, air insulated; or they may contain non-flammable or reduced flammable liquid; they have the advantage that they may be located inside buildings in close proximity to the associated switchgear. More will be said about this type of transformer in Section 7.
It is necessary to mention dielectrics thus far in order to distinguish between the principal types of transformers, oil filled and air insulated; this section will examine in detail the basic materials which are used to build transformers and mineral oil will be examined in some depth later. It is appropriate to start at the fundamental heart of the transformer, the steel core.
2. CORE STEEL
The purpose of a transformer core is to provide a low reluctance path for the magnetic flux linking primary and secondary windings.
In doing so, the core experiences iron losses due to hysteresis and eddy cur rents flowing within it which, in turn, show themselves as heating of the core material. In addition, the alternating fluxes generate noise, which, in the case of a large system transformer, for example, can be as invasive in the environment as a jet aircraft or an internal combustion engine at full throttle.
Core losses, though small in relation to the transformer throughput, are present whenever the transformer is energized. Thus they represent a constant and significant energy drain on any electrical system. It has been estimated that some 5 percent of all electricity generated is dissipated as iron losses in electrical equipment, and in the UK alone in the year 1987/1988 the cost of no-load core losses in transformers were estimated to be at £110 million.
At that time around 10^9 units of electricity were estimated to be wasted in core losses in distribution transformers each year, equivalent to 7 million barrels of oil to produce it and releasing 35,000 tonnes of sulphur dioxide and 4 million tonnes of carbon dioxide into the atmosphere. A similar study in the USA in 1998 made the cost some 10 times greater, partly, no doubt, due to the more extensive system but also an indication of the extent of approximation involved in making such estimates. Nevertheless, it is clear that the financial sums involved are quite vast. The cost implications identified above were, of course, particularly exacerbated by the significant increase in energy costs initiated by the oil crisis of the early 1970s.
Not surprisingly therefore, considerable research and development resource has been applied to electrical steels and to transformer core design in recent years directed mainly towards the reduction of losses but also to the reduction of noise. As a result a great deal of progress has been made and many changes have taken place since the basic principles of modern power transformer design and construction were laid down in the 1920s and 1930s.
Core loss is made up of two components: the first, the hysteresis loss, is proportional to the frequency and dependent on the area of the hysteresis loop, which, in turn, is a characteristic of the material and a function of the peak flux density; the second is the eddy current loss which is dependent on square of frequency but is also directly proportional to the square of the thickness of the material. Minimizing hysteresis loss thus depends on the development of a material having a minimum area of hysteresis loop, whilst minimizing eddy current loss is achieved by building up the core from a stack of thin laminations and increasing resistivity of the material in order to make it less easy for eddy currents to flow as will be seen by reference to Fig. 1.
The components of core loss can be represented by the expressions:
Hysteresis loss, Wh _ k1fBmax n W/kg (eqn. 1) and
Eddy current loss, We _ k2f 2 t 2 Beff 2/? W/kg (eqn. 2)
where k1 and k2 are constants for the material
f is the frequency, Hz
t is the thickness of the material, mm
? is the resistivity of the material
Bmax is the maximum flux density, Tesla
Beff is the flux density corresponding to the r.m.s. value of the applied voltage.
n is the 'Steinmetz exponent' which is a function of the material. Originally this was taken as 1.6 but with modern materials and higher flux densities n can vary from 1.6 to 2.5 or higher.
In practice the eddy current term is a complex one and can itself be considered to consist of two components; the first truly varies as the square of frequency times material thickness and flux density as indicated by the expression above.
This can be calculated in accordance with classical electromagnetic theory and is referred to as the classical eddy current loss. The second is dependent of the structure of the material such as grain size and magnetic domain movement during the magnetizing cycle and is known as anomalous loss or residual loss.
Anomalous eddy current loss can account for around half the total loss for any particular steel. It is this anomalous loss which can be greatly reduced by special processing of the core material so that this forms the basis of most of the modern approaches towards the reduction of core loss. More will be said about this later.
The first transformers manufactured in the 1880s had cores made from high-grade wrought iron and for a time Swedish iron was preferred. However in about the year 1900 it was recognized that the addition of small amounts of silicon or aluminum to the iron greatly reduced the magnetic losses. Thus began the technology of specialized electrical steel making.
The addition of silicon reduces hysteresis loss, increases permeability and also increases resistivity, thus reducing eddy current losses. The presence of silicon has the disadvantage that the steel becomes brittle and hard so that, for reasons of workability and ease of core manufacture, the quantity must be limited to about 4.5 percent. The elimination of impurities, including carbon, also has a significant effect in the reduction of losses so that although the first steels containing silicon had specific loss values of around 7 W/kg at 1.5 Tesla, 50 Hz, similar alloys produced in 1990 having high levels of purity have losses less than 2 W/kg at this condition.
As briefly mentioned above, electrical sheet steels have a crystalline structure so that the magnetic properties of the sheet are derived from the magnetic properties of the individual crystals or grains and many of these are dependent on the direction in the crystal in which they are measured.
The crystals of steel can be represented by a cube-lattice as shown in Fig. 2.
The principal axes of this lattice are designated by x, y, z co-ordinates enclosed in square brackets, thus, which represents the axis along the cube edge.
Planes intersecting the vertices of the cubes are similarly designated by co ordinates enclosed in round brackets, representing the plane intersecting diagonally opposite edges.
In the crystal lattice the direction is the easiest direction of magnetization, the direction is more difficult and the is the most difficult.
Silicon steel laminations of thickness around 0.35 mm used in transformers, in the USA until the 1940s and in the UK until somewhat later, were produced by a hot-rolling process in which the grains are packed together in a random way so that magnetic properties observed in a sheet have similar values independent of the direction in which they are measured. These represent an aver age of the properties for all directions within the individual crystals. These materials are known as isotropic.
It had been recognized in the early 1920s that the silicon steel crystals were themselves anisotropic but it was not until 1934 that the American N P Goss patented an industrial production process, which was chiefly developed by ARMCO in the USA, that commercial use was made of this property. The first commercial quantities were produced in 1939. The material was the first commercial grain-oriented cold-rolled silicon steel. It had a thickness of 0.32 mm with a loss of 1.5 W/kg at 1.5 Tesla, 50 Hz.
The material is cold reduced by a process set out diagrammatically in the left hand half of Fig. 3. This has formed the basis of the production of cold-rolled grain-oriented steels for many years. The initially hot-rolled strip is pickled to remove surface oxides and is then cold rolled to about 0.6 mm thickness from the initial hot band thickness of 2-2.5 mm. The material is given an anneal to re-crystallize the cold-worked structure before cold rolling again to the final gauge.
Decarburization down to less than 0.003 percent carbon is followed by coating with a thin magnesium oxide (MgO) layer. During the next anneal, at 1200ºC for 24 hours, purification and secondary re-crystallization occur and MgO reacts with the steel surface to form a thin magnesium silicate layer called the glass film or Forsterite layer. Finally, the material is given a flattening anneal, when excess MgO is removed and a thin phosphate coating is applied which reacts with the magnesium silicate to form a strong, highly insulating coating.
During hot rolling, small particles of manganese sulphide, which has been added to the melt as a grain growth inhibitor, precipitate out as the steel cools.
At the same time, some crystals with the Goss texture, that is, having the required orientation, are formed along with many other orientations. After the cold rolling, nuclei with the Goss texture re-crystallize during the decarburization anneal, as the material develops a 'structure memory.' The grain size, at this stage, is around 0.02 mm diameter, and this increases in the Goss-oriented grains at over 800ºC during the high-temperature anneal when the manganese sulphide particles retard the growth of other grains. During this secondary re-crystallization process, the Goss grains each consume 106 _107 primary grains and grow through the thickness of the sheet to diameters of 10 mm or more. All grains do not have the ideal Goss orientation but most are within 6º of the ideal
(110) shown in the diagram, Fig. 4.
Use of cold-rolled grain-oriented steel as described above continued with only steady refinement and improvement in the production process until the late 1960s. However, in 1965 the Japanese Nippon Steel Corporation announced a step-change in the quality of their electrical steel; high-permeability grain oriented silicon steel. The production process is shown in the right-hand half of Fig. 3. Production is simplified by the elimination of one of the cold-rolling stages because of the introduction of around 0.025 percent of aluminum to the melt and the resulting use of aluminum nitride as a growth inhibitor.
The final product has a better orientation than cold-rolled grain-oriented steel (in this context, generally termed 'conventional' steel), with most grains aligned within 3º of the ideal, but the grain size, on average 1 cm diameter, was very large compared to the 0.3 mm average diameter of conventional material. At flux densities of 1.7 Tesla and higher, its permeability was 3 times higher than that of the best conventional steel, and the stress sensitivity of loss and magnetostriction were lower because of the improved orientation and the presence of a high tensile stress introduced by the so-called stress coating. The stress coating imparts a tensile stress to the material which helps to reduce eddy current loss which would otherwise be high in a large grain material. The total loss is further offset by some reduction in hysteresis loss due to the improved coating. However, the low losses of high-permeability steels are mainly due to a reduction of 30-40 percent in hysteresis brought about by the improved grain orientation. The Nippon Steel Corporation product became commercially available in 1968, and it was later followed by high-permeability materials based on MnSe plus Sb (Kawasaki Steel, 1973) and on boron (Allegheny Ludlum Steel Corporation, 1975).
Domain refined steel
The continued pressure for the reduction of transformer core loss identified above led to further improvements in the production process so that in the early 1980s the Nippon Steel Corporation introduced laser-etched material with losses some 5-8 percent lower than high-permeability steel. By 1983 they were producing laser-etched steels down to 0.23 mm thick with losses as low as 0.85 W/kg at 1.7 Tesla, 50 Hz.
It has been briefly mentioned above, in defining the quantity 'anomalous eddy current loss' that this arises in part due to magnetic domain wall movement during the cycles of magnetization. Messrs Pry and Bean as early as 1958 had suggested that in a grain-oriented material anomalous eddy current loss is proportional to the domain wall spacing and inversely proportional to sheet thickness. This is illustrated in Fig. 5 which shows an idealized section of grain-oriented material in which 180º magnetic domains stretch infinitely at equal intervals of 2L. Clearly eddy current loss can be reduced by subdividing the magnetic domains to reduce L.
It had been recognized for many years that introduction of strain into sheet steels had the effect of subdividing magnetic domains and thus reducing core loss. This was the basis for the use of the stress coatings for high-permeability steels mentioned above. The coatings imparted a tensile stress into the material on cooling due to their low thermal expansion coefficient. Mechanical scribing of the sheet surface at intervals transverse to the rolling direction also serves as a means of inducing the necessary strain but this is difficult to carry out on a commercial basis and has the disadvantage that the sheet thickness at the point of the scribing is reduced, thus creating a localized increase in the flux density and causing some of the flux to transfer to the adjacent lamination with the consequent result that there is a net increase in loss.
Nippon Steel Corporation's solution to the problem was to employ a non contact domain refining process utilizing laser irradiation normally referred to as laser-etching.
FIG. 6 shows a diagrammatic arrangement of the process. When the high-power laser beam is trained onto the surface of the sheet, the outermost layer of the sheet vaporizes and scatters instantaneously. As a result, an impact pressure of several thousand atmospheres is generated to form a local elastic-plastic area in the sheet. Highly dense complex dislocations due to plastic deformation occur leaving a residual strain which produces the required domain refinement. FIG. 7 shows domain structures before and after laser irradiation. As the laser irradiation vaporizes and scatters the outermost layer of the sheet, an additional coating is necessary in order to make good the surface insulation layer.
An important aspect of the domain refinement process described above is that the residual strains will be removed if the material is subsequently annealed at a temperature above 500ºC thus reversing the process. It is import ant therefore that any processes carried out after laser-etching should not take the temperature above 500ºC.
In summary of the foregoing, Table 3.1 gives a simple reference guide to the methods of reducing losses in sheet steels produced by the conventional rolling process.
Amorphous steels are relatively recent in their appearance and their development stems from a totally different source than the silicon core steels described above. Originally developed by Allied Signal Inc., Metglas Products in the USA, in the early 1970s as an alternative for the steel in vehicle tire reinforcement, it was not until the mid-1970s that the importance of their magnetic properties was recognized. Although still restricted in their application some 35 years later due to difficulties in production and handling, they offer considerable reduction in losses compared to even the best conventional steels.
Amorphous metals have a non-crystalline atomic structure, there are no axes of symmetry and the constituent atoms are randomly distributed within the bulk of the material. They rely for their structure on a very rapid cooling rate of the molten alloy and the presence of a glass-forming element such as boron.
Typically they might contain 80 percent iron with the remaining 20 percent boron and silicon.
Various production methods exist but the most popular involve spraying a stream of molten metal alloy onto a high-speed rotating copper drum. The molten metal is cooled at a rate of about 106 ºC/second and solidifies to form a continuous thin ribbon. The quenching technique sets up high internal stresses so these must be reduced by annealing between 200ºC and 280ºC to develop good magnetic properties. Earliest quantities of the material were only 2 mm wide and about 0.025-0.05 mm thick. By the mid-1990s a number of organizations had been successful in producing strip up to 200 mm wide.
Table 1 Summary of loss reduction processes of conventionally rolled core steels
POWERCORE [POWERCORE strip is a registered trademark of Allied Signal Inc., Metglas Products.]
The original developers of the material, Metglas Products, had towards the end of the 1980s produced a consolidated strip amorphous material named strip, designed to be used in laminated cores. The material is produced in the thickness range 0.125-0.25 mm, by bonding several sheets of as-cast ribbon to form a more easily handleable strip. The ribbons are effectively bonded over 15-75 percent of their surface area by a local plastic action combined with a chemical bond of diffused silicon oxide. The weak bond does not allow significant eddy current flow between layers of the composite and the bulk properties are similar to those of single ribbon.
The need for a glass-forming element, which happens to be non-magnetic, gives rise to another of the limitations of amorphous steels, that of low saturation flux density. POWERCORE strip has a saturation level of around 1.56 Tesla. Specific loss at 1.35 Tesla, 50 Hz, is just 0.12 W/kg. At 1.5 Tesla, 50 Hz, this is 0.28 W/kg.
Another important property is the magnetizing VA. At 1.3 Tesla this is 0.25 VA/kg compared with 0.69 VA/kg for 3 percent silicon steel. An indication of the effect of the low saturation flux density can be gained from comparing these again at 1.5 Tesla. In the case of POWERCORE strip this has risen to 1.3 VA/kg whilst for conventional silicon steel it is typically only 0.94 VA/kg.
Whilst the sizes of strip available as POWERCORE are still unsuitable for the manufacture of large power transformer cores; in the USA in particular, many hundreds of thousands of distribution transformer cores with an average rating of around 50 kVA have been built using amorphous material. In Europe use of the material has been on a far more limited scale, the main impetus being in Holland, Sweden, Switzerland, Germany and Hungary. One possible reason for the slower progress in Europe is that the thin strip material does not lend itself to the European preferred form of core construction, whereas the wound cores, which are the norm for distribution transformers in the USA, are far more suitable for this material. In the UK its use has been almost exclusively by one manufacturer who have built several hundred small distribution transformers. All were manufactured from plain unlaminated ribbon material.
This manufacturer has also built a small number of experimental units using amorphous steel, see Fig. 8, but report that the difficulties of cutting and building this into a conventional core can tend to outweigh any benefits gained.
Another of the practical problems associated with amorphous steel is its poor stacking factor which results from a combination of the very large number of layers of ribbon needed to build up the total required iron section and also the relatively poor flatness associated with this very thin ribbon. Plain ribbon 0.03 mm thick has a stacking factor of only 0.8. POWERCORE strip 0.13 mm thick can give a figure of 0.9, but both of these are poor compared to the 0.95-0.98 attainable with conventional silicon steel.
Another approach towards the optimization of the magnetic and mechanical performance of silicon steel, which has received much attention in Japan, is the production of high silicon and aluminum-iron alloys by rapid solidification in much the same manner as for amorphous steels. No glass-forming additives are included so a ductile microcrystalline material is produced, often referred to as semi-crystalline strip. Six percent silicon iron strip has been produced which has proved to be ductile and to have losses less than those of commercial grain-oriented 3 percent silicon-iron. A figure of 0.56 W/kg at 1.7 Tesla, 50 Hz, is a typically quoted loss value.
Rapidly quenched microcrystalline materials have the advantage of far higher field permeability than that of amorphous materials so far developed for power applications. FIG. 9 indicates typical loss values attainable for the whole range of modern core materials and shows how the non-oriented micro crystalline ribbon fits between amorphous ribbon and grain-oriented steel.
Adoption of improved steels
The cold-rolled grain-oriented steels introduced in the 1940s and 1950s almost completely replaced the earlier hot-rolled steels in transformer manufacture over a relatively short timescale and called for some new thinking in the area of core design. The introduction of high-permeability grain-oriented steels some 30 years later was more gradual and, because of its higher cost, its early use tended to be restricted to applications where the capitalized cost of no-load loss (see Section 8) was high. A gradual development in core design and manufacture to optimize on the properties of the new material took place but some of these improvements were also beneficial for designs using conventional materials. In 1981 some 12 percent of the world-wide production of grain-oriented steel was high-permeability grade. By 1995 high-permeability material was the norm. A similar situation occurred with the introduction of laser-etched steel, which for reasons of both availability and cost, remains very much a 'special' material, to be used only where the cost of no-load losses is very high, more than 15 years after its announcement.
The ways in which core design and construction developed to reflect the properties of the available material will be discussed in the next section.
Designation of core steels
Specification of magnetic materials including core steels is covered inter nationally by IEC 60404. This is a multipart document, Part 1 of which, Magnetic materials, classification provides the general framework for all the other documents in the series. Part 8 is the one dealing with individual materials of which Section 8.7, Specification for grain-oriented magnetic steel sheet and strip, covers the steels used in power transformers.
Until the late 1980s core steels used in power transformers were specified in BS 601 in the UK. BS 601:1973, was a five part document, Part 2 of which specifically referred to grain-oriented steel greater than 0.25 mm thick.
Most cold-rolled grain-oriented steels used up to this time complied with this document which identified particular materials by means of a code, for example, 28M4 or 30M5, which were 0.28 and 0.30 mm thick, respectively. The final digit referred to the maximum specific loss value. With the introduction of high-permeability steels this code was arbitrarily extended to cover these materials giving designations as, typically, 30M2H. This is a high-permeability grade 0.30 mm thick with specific loss in the '2' band. Although they continue to be used, designations such as 30M2H no longer have any status in the current Standard.
Readers seeking more detailed information relating to core steels may consult an IEE review paper by A.J. Moses which contains many references and provides an excellent starting point for any more extensive investigations.
3. WINDING CONDUCTORS
Transformer windings are made almost exclusively of copper, or to be precise, high-conductivity copper. Copper has made possible much of the electrical industry as we know it today because, in addition to its excellent mechanical properties, it has the highest conductivity of the commercial metals. Its value in transformers is particularly significant because of the benefits which result from the saving of space and the minimizing of load losses.
The load loss of a transformer is that proportion of the losses generated by the flow of load current and which varies as the square of the load current. This falls into three categories:
(1) Resistive loss within the winding conductors and leads.
(2) Eddy current loss in the winding conductors.
(3) Eddy current loss in the tanks and structural steelwork.
Resistive loss can be reduced by reducing the number of winding turns, by increasing the cross-sectional area of the turn conductor, or by a combination of both. Reducing the number of turns requires an increase in Fm, that is, an increase in the core cross-section (frame size - see Section 4.2), which increases the iron weight and iron loss. So load loss can be traded against iron loss and vice versa. Increased frame size requires reduced axial length of winding to compensate) and thus retain the same impedance, although as already explained there will be a reduction in the number of turns (which was the object of the exercise) by way of partial compensation. Reduction of the winding axial length means that the core leg length is reduced, which also offsets the increase in core weight resulting from the increased frame size to some extent. There is thus a band of one or two frame sizes for which loss variation is not too great, so that optimum frame size can be chosen to satisfy other factors, such as ratio of fixed to load losses or transport height.
The paths of eddy currents in winding conductors are complex. The effect of leakage flux within the transformer windings results in the presence of radial and axial flux changes at any given point in space and any moment in time. These induce voltages which cause currents to flow at right angles to the changing fluxes. The magnitude of these currents can be reduced by increasing the resistance of the path through which they flow, and this can be effected by reducing the total cross-sectional area of the winding conductor or by subdividing this conductor into a large number of strands insulated from each other.
(In the same way as laminating the core steel reduces eddy current losses in the core.) The former alternative increases the overall winding resistance and thereby the resistive losses. Conversely, if the overall conductor cross-section is increased with the object of reducing resistive losses, one of the results is to increase the eddy current losses. This can only be offset by a reduction in strand cross-section and an increase in the total number of strands. It is costly to wind a large number of conductors in parallel and so a manufacturer will wish to limit the total number of strands in parallel. Also, the extra insulation resulting from the increased number of strands results in a poorer winding space-factor.
Compact size is important for any item of electrical plant. In transformer windings this is particularly so. The size of the windings is the determining factor in the size of the transformer. As explained above they must have a sufficiently large cross-section to limit the load losses to an acceptable level, not only because of the cost of these losses to the user but also because the heat generated must be removed by the provision of cooling ducts. If the losses are increased more space must be provided for ducts. This leads to yet larger windings and thus a larger core is needed to enclose them. Increasing the size of the core increases the no-load loss but, along with the increase in the size of the windings, also means that a very much larger tank is required, which, in turn results in an increased oil quantity and so the whole process escalates.
Conversely, any savings in the size of windings are repaid many times over by reductions in the size of the transformer and resultant further savings else where. As the material which most economically meets the above criteria and which is universally commercially available, high-conductivity copper is the automatic choice for transformer windings.
Eddy current losses in tanks and internal structural steelwork such as core frames only constitute a small proportion of the total load losses. These are also a result of leakage flux and their control is mainly a matter of controlling the leakage flux. More will be said about this in Section 4.
During the 1960s, at a time when copper prices rose sharply, attempts were made to explore the possibilities offered by the use of the then very much cheaper aluminum in many types of electrical equipment. Indeed, about this time, the use of aluminum in cables became widespread and has tended to remain that way ever since. However, although some quite large transformers were built using aluminum for windings, mainly at the instigation of the aluminum producers, the exercise largely served to demonstrate many of the disadvantages of this material in large power transformers.
Aluminum has some advantages for certain transformer applications, not ably for foil windings which are intended to be resin encapsulated, where its coefficient of thermal expansion matches much more closely the expansion of the resin than does that of copper. This leads to less of a tendency for resin cracking to occur under load cycling. These properties of aluminum will be discussed when these applications are described.
Copper is in plentiful supply, being mined in many places throughout the world, but it also has the great advantage of being readily recycled. It is easily separated from other scrap and can be reused and re-refined economically, thus preventing unnecessary depletion of the earth's natural resources.
There are an enormous range of electrical applications for which high conductivity copper is used, and there are a number of different coppers which may be specified, but for the majority of applications the choice will be either Electrolytic Tough Pitch Copper, (Cu-ETP-2) or the higher-grade Cu-ETP-1.
The former is tough pitch (oxygen bearing) high-conductivity copper which has been electrolytically refined to reduce the impurity levels to total less than 0.03 percent. In the UK it is designated Cu-ETP-2, as cast, and C101 for the wrought material. This copper is readily available in a variety of forms and can be worked both hot and cold. It is not liable to cracking during hot working because the levels of lead and bismuth which cause such cracking are subject to defined limits.
The latter, Cu-ETP-1, as cast, and C100 when wrought, is now available for use by manufacturers with advantage in modern high-speed rod breakdown and wire drawing machines with in-line annealing. It makes excellent feed stock for many wire enameling processes where copper with a consistently low annealing temperature is needed to ensure a good reproducible quality of wire.
Production of high-conductivity copper
As indicated above, copper is extracted and refined in many places throughout the world. FIG. 10 illustrates these. The output from a refinery is in a variety of forms depending on the type of semi-finished wrought material to be made. Cathodes are the product of electrolytic refining of copper. They must be re-melted before being usable and may then be cast into different 'refinery shapes.' The shapes are billets for extrusions, cakes for rolling into flat plate, wire bars for rolling into rod and wire rod for wire drawing. Sizes of cathodes vary depending on the refinery. Typically they may be plates of 1200 _ 900 mm in size weighing 100-300 kg.
Billets are usually about 200 mm in diameter and no more than 750 mm in length to fit the extrusion chamber. Extrusions are usually subsequently drawn to the required finished sizes by one or more passes through the mill drawblocks.
Cakes (or slabs) are used when flat plate, sheet, strip and foil are required.
They are nowadays mostly cast continuously. Copper is commonly hot rolled from 150 mm down to about 9 mm and then cold rolled thereafter.
Wirebars were previously the usual starting point for hot rolling of rod. They were generally cast horizontally and therefore had a concentration of oxide at and near the upper surface. It is now possible to continuously cast them vertically with a flying saw to cut them to length but they are now almost obsolete however.
Wire rod is the term used to describe coils of copper of 6-35 mm diameter (typically 9 mm) which provide the starting stock for wire drawing. At one time these were limited to about 100 kg in weight, the weight of the wirebars from which they were rolled. Flash-butt welding end to end was then necessary before they could be fed into continuous wire drawing machines.
It is now general practice to melt cathodes continuously in a shaft furnace and feed the molten copper at a carefully controlled oxygen content into a continuously formed mould which produces a feedstock led directly into a multistand hot-rolling mill. The output from this may be in coils of several tonnes weight each. For subsequent wire drawing these go to high-speed rod breakdown machines which carry out interstage anneals by resistance heating the wire at speed in-line. This has superseded the previous batch annealing techniques and shows considerable economies but does require a consistently high quality of copper.
Table 2 Properties of high-conductivity copper alloys of interest to transformer designers (Copper Development Association)
Electrical and thermal properties
Besides being a good conductor of electricity, copper is, of course, an excellent conductor of heat. The standard by which other conductors are judged is the International Annealed Copper Standard on which scale copper was given the arbitrary value of 100 percent in 1913. A list of some of its more import ant properties, particularly to transformer designers, is given in Table 2.
High-conductivity copper alloys
There are many alloys of copper and high-conductivity copper all of which have their specific uses in different types of electrical equipment. The approximate effect of impurities and some added elements on conductivity is shown in Fig. 11. Most of the elements shown have some solubility in copper. Those which are insoluble tend to have little effect on conductivity and are often added to improve properties such as machinability of high-conductivity copper.
By far the most important alloy of copper to transformer designers is silver bearing copper. The addition of silver to pure copper raises its softening temperature considerably with very little effect on electrical conductivity.
A minimum of 0.01 percent of silver is normally used, which also improves the other mechanical properties, especially creep resistance, to provide transformer winding copper with the necessary mechanical strength to withstand the forces arising in service due to external faults and short circuits. The disadvantage of this material can be the added difficulty introduced into the winding pro cess due to its increased hardness, so its use tends to be restricted to those very large units for which high mechanical strength is demanded. More will be said about this aspect in the following section.
Further information concerning copper for electrical purposes, including its refining as well as its electrical and mechanical properties, can be obtained from the booklet 'High Conductivity Coppers' published in the UK by the Copper Development Association, on which much of the foregoing description is based.
Copper winding wires
Almost all of the copper used in transformer windings is in the form of rectangular-section wire or strip complying with EN 13601:2002 Copper and copper alloys. Copper rod bar and wire for general electrical purposes. In addition to specifying the required characteristics of the copper including degree of purity, edge radii, resistivity and dimensional tolerances, the standard gives in Appendix B a table of recommended dimensions. Wire of circular cross-section cannot be wounded into windings having as good a space-factor as can rectangular-section wire, nor does it produce a winding with as high a mechanical stability. Circular-section wire is therefore generally restricted to small medium-voltage distribution transformer sizes for which it is used in plain enamel covered form. Special types of winding must be used for these circular conductors and these are described in Section 7.8.
Much of the foregoing data relating to copper, including that in Table 3.2, is taken from the booklet 'High Conductivity Coppers' published by the Copper Development Association to which the reader is referred for further information.
It is hardly necessary to emphasize the importance of a reliable insulation system to the modern power transformer. Internal insulation failures are invariably the most serious and costly of transformer problems. High short-circuit power levels on today’s electrical networks ensure that the breakdown of transformer insulation will almost always result in major damage to the transformer.
However, consequential losses, such as the non-availability of a large generating unit, can often be far more costly and wide reaching than the damage to the transformer itself.
The ever growing demands placed on electricity supplies have led to increasing unit sizes and ever higher transmission voltages. Transformer ratings and voltages have been required to increase consistently to keep pace with this so that they have been nudging the physical limits of size and transport weight since the 1950s. That transformer rated voltages and MVA throughputs have continued to increase since this time without exceeding these physical limits has largely been due to better use being made of the intrinsic value of the insulation. (Assisted by better analysis of dielectric stress distribution in insulation using finite element analysis; see Appendix 7.) A vital aspect is the transformer life, and this is almost wholly dependent upon the design and condition of the insulation. It must be adequate for a lifespan of 40 years or more and this probably explains the increasingly demanding testing regime of impulse testing, switching surges and partial discharge measurement. At the other end of the scale, distribution transformers have become more compact and manufacturers' prices ever more competitive. Many of the savings achieved have been as a result of improvements and innovation in insulating materials and the production of special insulation components. FIG. 12 shows some of the insulation items which have been developed specifically for the distribution transformer industry in recent years.
As indicated at the start of this section, today’s transformers are almost entirely oil filled, but early transformers used asbestos, cotton and low-grade pressboard in air. The introduction of shellac insulated paper at the turn of the twentieth century represented a tremendous step forward. It soon became the case, however, that air and shellac-impregnated paper could not match the thermal capabilities of the newly developed oil-filled transformers. These utilized kraft paper and pressboard insulation systems supplemented from about 1915 by insulating cylinders formed from phenol-formaldehyde resin impregnated kraft paper, or Bakelized paper, to give it its proprietary name. Usually referred to as s.r.b.p. (synthetic resin bonded paper), this material continued to be widely used in most transformers until the 1960s and still finds many uses in transformers, usually in locations having lower electrical stress but where high mechanical strength is important.
Paper is among the cheapest and best electrical insulation materials known.
Electrical papers must meet certain physical and chemical standards; in addition they must meet specifications for electrical properties. Electrical properties are, in general, dependent on the physical and chemical properties of the paper. The important electrical properties are:
• high dielectric strength;
• dielectric constant - in oil-filled transformers as close as possible a match to that of oil;
• low power factor (dielectric loss);
• freedom from conducting particles.
The dielectric constant for kraft paper is about 4.4 and for transformer oil the figure is approximately 2.2. In a system of insulation consisting of different materials in series, these share the stress in inverse proportion to their dielectric constants, so that, for example, in the high to low barrier system of a transformer, the stress in the oil will be twice that in the paper (or pressboard). The transformer designer would like to see the dielectric constant of the paper nearer to that of the oil so the paper and oil more nearly share the stress.
Kraft paper is, by definition, made entirely from unbleached softwood pulp manufactured by the sulphate process; unbleached because residual bleaching agents might hazard its electrical properties. This process is essentially one which results in a slightly alkaline residue, pH 7-9, as distinct from the less costly sulfite process commonly used for production of newsprint, for example, which produces an acid pulp. Acidic content leads to rapid degradation of the long-chain cellulose molecules and consequent loss of mechanical strength which would be unacceptable for electrical purposes - more on this aspect shortly. The timber is initially ground to a fine shredded texture at the location of its production in Scandinavia, Russia or Canada using carborundum or similar abrasive grinding wheels. The chemical sulphate process then removes most of the other constituents of the wood, for example, lignin, carbohydrates, waxes, etc., to leave only the cellulose fibers. The fibers are dispersed in water which is drained to leave a wood-pulp mat. At this stage the dried mat may be transported to the mill of the specialist paper manufacturer.
The processes used by the manufacturer of the insulation material may differ one from another and even within the mill of a particular manufacturer treatments will vary according to the particular properties required from the finished product. The following outline of the type of processes used by one UK producer of specialist high-quality presspaper gives some indication of what might be involved. Presspaper by definition undergoes some compression during manufacture which increases its density, improves surface finish and increases mechanical strength. Presspaper production is a continuous process in which the paper is formed on a rotating fine mesh drum and involves building of the paper sheet from a number of individual layers. Other simpler processes may produce discrete sheets of paper on horizontal screen beds without any subsequent forming or rolling processes, but, as would be expected, the more sophisticated the manufacturing process, the more reliable and consistent the properties of the resulting product.
The process commences by re-pulping of the bales of dry mat using copious quantities of water, one purpose of which is to remove all residual traces of the chemicals used in the pulp extraction stage. The individual fibers are crushed and refined in the wet state in order to expose as much surface area as possible. Paper or pressboard strength is primarily determined by bonding forces between fibers, whereas the fibers themselves are stressed far below their breaking point. These physiochemical bonding forces which are known as 'hydrogen bonding' occur between the cellulose molecules themselves and are influenced primarily by the type and extent of this refining.
Fibers thus refined are then mixed with more water and subjected to intensive cleaning in multistage centrifugal separators which remove any which may not have been totally broken down or which may have formed into small knots.
These can be returned to pass through the refining cycle once more. The centrifuges also remove any foreign matter such as metallic particles which could have been introduced by the refining process. The cellulose/water mixture is then routed to a wide rotating cylindrical screen. While the water flows through the screen, the cellulose fibers are filtered out and form a paper layer. An end less band of felt removes the paper web from the screen and conveys it to the forming rolls. The felt layer permits further water removal and allows up to five or six other paper plies to be amalgamated with the first before passing through the forming rolls. These then continue to extract water and form the paper to the required thickness, density and moisture content by means of heat and pressure as it progresses through the rolls. Options are available at this stage of the process to impart various special properties, for example the CLUPAK [CLUPAK Inc.'s trademark for its extensible paper manufacturing process.] process which enhances the extensibility of the paper, or impregnation with 'stabilizers' such as nitrogen containing chemicals like dicyandiamide which provide improved thermal performance. More will be said about both of these later. Final finish and density may be achieved by means of a calendering process in which the paper, at a controlled high moisture content, is passed through heavily loaded steel rollers followed by drying by means of heat in the absence of pressure.
The cohesion of the fibers to one another when the mat is dried is almost exclusively a property of cellulose fibers. Cellulose is a high-polymer carbohydrate chain consisting of glucose units with a polymerization level of approximately 2000. FIG. 13 shows its chemical structure.
Hemi-cellulose molecules are the second major components of the purified wood pulp. These are carbohydrates with a polymerization level of less than 200. In a limited quantity, they facilitate the hydrogen bonding process, but the mechanical strength is reduced if their quantity exceeds about 10 percent.
Hemi-cellulose molecules also have the disadvantage that they 'hold on' to water and make the paper more difficult to dry out.
Softwood cellulose is the most suitable for electrical insulation because its fiber length of 1-4 mm gives it the highest mechanical strength. Nevertheless small quantities of pulp from harder woods may be added and, as in the case of alloying metals, the properties of the resulting blend are usually superior to those of either of the individual constituents.
Cotton fibers are an alternative source of very pure cellulose which was widely used in the UK for many years to produce the so-called 'rag' papers with the aim of combining superior electrical strength and mechanical properties to those of pure kraft paper. Cotton has longer fibers than those of wood pulp but the intrinsic bond strength is not so good. Cotton is a 'smoother' fiber than wood so that it is necessary to put in more work in the crushing and refining stage to produce the side branches which will provide the necessary bonding sites to give the required mechanical strength. This alone would make the material more expensive even without the additional cost of the raw material is itself.
When first used in the manufacture of electrical paper in the 1930s the source of cotton fibers was the waste and off-cuts from cotton cloth which went into the manufacture of clothing and this to an extent kept the cost competitive with pure kraft paper. In recent years this source has ceased to be an acceptable one since such cloths will often contain a proportion of synthetic fibers and other materials so that the constitution of off cuts cannot be relied upon as being pure and uncontaminated. Alternative sources therefore had to be found. Cotton linters are those cuts taken from the cotton plant after the long staple fibers have been cut and taken for spinning into yarn for the manufacture of cloth. First grade linters are those taken immediately after the staple. These are of a length and quality which still renders them suitable for high-quality insulation material. They may provide the 'furnish' or feedstock for a paper making process of the type described, either alone or in conjunction with new cotton waste threads.
Cotton fiber may also be combined with kraft wood pulp to produce a material which optimizes the advantages of both constituents giving a paper which has good electrical and mechanical properties as well as maximum oil absorption capability. This latter requirement can be of great importance in paper used for high to low wraps or wraps between layers of round wire distribution transformer HV windings where total penetration of impregnating oil may be difficult even under high vacuum.
Other fibers such as manila hemp and jute may also be used to provide papers with specific properties developed to meet particular electrical purposes, for example in capacitors and cable insulation. BS 5626, Cellulosic papers for electrical purposes, which is a multipart document closely related to IEC 60554, lists the principal paper types and properties. Presspapers are covered by EN 60641, Pressboard and presspaper for electrical purposes. This is also a multi part document and will be mentioned further later in relation to pressboard.
Papers for special applications
The foregoing paragraphs should have conveyed the message that there are many different types of electrical papers all of which have particular proper ties which have been specifically developed to meet certain requirements of particular applications. Before leaving the subject of paper insulation it is worthwhile looking a little more closely at four special types of paper whose properties have been developed to meet particular needs of the transformer industry. These are:
• Crêped paper
• Highly extensible paper
• Thermally upgraded paper
• 'Diamond dotted' presspaper
Crêped paper was probably the earliest of the special paper types. It is made with an irregular close 'gathering' or crimp which increases its thickness and greatly increases its extensibility in the machine direction. It is normally produced cut into strips around 25 mm wide and is ideal where hand applied covering is required on connections in leads or on electrostatic stress control rings which are to be placed between end sections within windings. Its extensibility enables it to be shaped to conform to irregular contours or to form bends which may be necessary, for example, in joining and forming tapping leads.
FIG. 14 shows an arrangement of leads to an on-load tap changer which makes extensive use of crêped paper for this purpose.
A disadvantage of crêped paper is its tendency to lose elasticity with time so that after some years in service taping of joints may not be as tight as when it was first applied. A better alternative in many situations is highly extensible paper. CLUPAK extensible presspaper is one such material. Manufacture of the basic presspaper is as described above and the elastic property is added at a stage in the roll-forming process in which the action of the rolls in conjunction with heat and moisture is to axially compress the fibers in the machine direction. As a result the paper retains its smooth finish but attains greatly enhanced burst, stretch and cross machine tear properties whilst retaining its tensile strength and electrical performance. The high mechanical strength and resilience of the paper make it ideally suited to machine application for such items as electrostatic stress control rings identified above or as an overall wrapping on continuously transposed conductors (CTC) (see Section 4). CTC used in large power transformer windings often has a large cross-section making it stiff and exceedingly difficult to bend to the required radius of the winding. As a result the conductor can be subjected to very severe handling at the winding stage.
In addition, the actual process of winding this large section conductor around the winding mandrel imposes severe stress on the paper covering, creating wrinkling and distortion which can intrude into radial cooling ducts. The toughness and resilience of the extensible presspaper make it better able than conventional paper to withstand the rough use which it receives during the winding process and the elasticity ensures that any tendency to wrinkling is minimized.
As explained above, thermally upgraded paper is treated by the addition of stabilizers during manufacture to provide better temperature stability and reduced thermal degradation. The subject of ageing of insulation will be dealt with at some length later (Section 4.5). At present it is sufficient to say that degradation is temperature dependent and is brought about by the breakdown of the long-chain cellulose molecules. The permitted temperature rise for power transformers is based on reaching an average hot spot temperature in operation which will ensure an acceptable life for the insulation. This is usually between about 110ºC and 120ºC. However, within this range of temperatures insulation degradation is greatly increased by the presence of oxygen and moisture, both of which are present to some extent in most oil-filled transformers and particularly in distribution transformers whose breathing arrangements are often basic and for which maintenance can frequently be minimal. It is in these situations that thermally upgraded paper can be beneficial in retarding the ageing of paper insulation; not by permitting higher operating temperatures, but by reducing the rate of degradation at the operating temperatures normally reached.
FIG. 15 shows a length of diamond dotted presspaper being used in the construction of a distribution transformer. Mention has already been made of the fact that s.r.b.p. tubes were widely used in transformers for their good insulation properties combined with high mechanical strength. These are made by winding kraft paper which has been coated with thermosetting resin on one side onto a mandrel and then curing the resin to produce a hard tube. The reason that their use has become more selective is that the large ratio of resin to paper which is necessary to obtain the required mechanical strength makes these very difficult to impregnate with transformer oil. In the presence of electrical stress in service any voids resulting from less than perfect impregnation can become a source of partial discharge which can result ultimately in electrical breakdown. Kraft papers are used for s.r.b.p. cylinders because thin papers of the necessary width are only available from the large flat wire machines in Sweden and these machines cannot handle cotton fibers. If they could a cot ton paper would improve the drying and impregnation properties of s.r.b.p. cylinders. The diamond dotted presspaper shown in the figure represents a more acceptable method of achieving high mechanical strength without the associated difficulty of impregnation. The presspaper is pre-coated with two-stage resin in a diamond pattern which can be allowed to dry following the coating stage. The resin dots create a large bonding surface whilst ensuring that the paper can be effectively dried and oil impregnation efficiently carried out.
When the winding is heated for drying purposes, the adhesive dots melt and cure so creating permanent bonding sites which will be unaffected by subsequent heating cycles in service but which give the structure its high mechanical strength. Although the diamond adhesive pattern can be applied to any type of paper, in practice it is still desirable to use a base paper which has good drying and impregnation properties such as the wood/cotton fiber blend identified above, particularly if used in foil type LV windings (see Section 7.9) which can be notoriously difficult to dry out and oil impregnate.
At its most simple, pressboard represents nothing more than thick insulation paper made by laying up a number of layers of paper at the wet stage of manufacture. FIG. 16 shows a diagrammatic arrangement of the manufacturing process. Of necessity this must become a batch process rather than the continuous one used for paper, otherwise the process is very similar to that used for paper. As many thin layers as are necessary to provide the required thickness are wet laminated without a bonding agent. Pressboard can however be split into two basic categories:
(1) That built up purely from paper layers in the wet state without any bonding agent, as described above.
(2) That built up, usually to a greater thickness, by bonding individual boards using a suitable adhesive.
Each category is covered by a European Standard: the former by EN 60641:1996 Pressboard and presspaper for electrical purposes, and the latter by EN 60763:1996 Laminated pressboard for electrical purposes. As in the case of paper insulation, there are a number of variants around the theme and all the main types of material are listed in the European Standards. Raw materials may be the same as for presspaper, that is all wood pulp, all cotton, or a blend of wood and cotton fibers.
Pressboard in the first of the above categories is available in thicknesses up to 8 mm and is generally used at thicknesses of around 2-3 mm for interwinding wraps and end insulation and 4.5-6 mm for strips. The material is usually produced in three sub-categories:
(1) The first is known as calendered pressboard and undergoes an initial pressing operation at about 55 percent water content. Drying by means of heat without pressure then follows to take the moisture level to about 5 percent. The pressboard thus produced has a density of about 0.90 to 1.00. Further compression is then applied under heavy calenders to take the density to between 1.15 and 1.30.
(2) The second category is moldable pressboard which receives little or no pressing after the forming process. This is dried using heat only to a moisture content of about 5 percent and has a density of about 0.90. The result is a soft pressboard with good oil absorption capabilities which is capable of being shaped to some degree to meet the physical requirements of particular applications.
(3) The third material is precompressed pressboard. Dehydration, compression and drying are performed in hot presses direct from the wet stage. This has the effect of bonding the fibers to produce a strong, stable, stress-free material of density about 1.25 which will retain its shape and dimensions throughout the stages of transformer manufacture and the thermal cycling in oil under service conditions to a far better degree than the two boards previously described. Because of this high stability precompressed material is now the preferred pressboard of most transformer manufacturers for most applications.
Laminated pressboard starts at around 10 mm thickness and is available in thicknesses up to 50 mm or more. The material before lamination may be of any of the categories of unlaminated material described above but generally precompressed pressboard is preferred. This board is used for winding support platforms, winding end support blocks and distance pieces as well as cleats for securing and supporting leads.
The electrical stress between co-axial cylindrical windings of a high-voltage transformer is purely radial and the insulation in this region can simply consist of a series of cylindrical pressboard barriers and annular oil spaces as shown in Fig. 17(a). Pressboard can be rolled to form the cylinders and the axial joint in these may be in the form of an overlap or a scarfed arrangement as shown in Fig. 17(b). The winding ends create much more of a problem since the pressboard barriers cannot be extended far enough beyond the winding ends to provide adequate tracking distance without unduly increasing the length of the core limb. The interwinding insulation must thus be bent around the end of the winding as shown in Fig. 18. For many years the way of achieving this was to make the interwinding wraps of paper or alternatively provide a tube with soft unbonded ends. The ends of the tubes or paper wraps were then 'petalled' by tearing them axially at intervals of about 80 mm and folding over the 'petals.' The tears on successive layers were carefully arranged to be staggered so as to avoid the formation of direct breakdown paths through the petalling.
This process had a number of disadvantages. Firstly, it was very laborious and added greatly to manufacturing costs. Secondly, when axial compression was applied to the windings to take up shrinkage, the profile of the petalling could become displaced so as to less accurately assume the required shape and also in some circumstances create partial blockage of oil ducts. The solution is to produce shaped end rings using the process for moldable pressboard as described above. Since this requires little pressure at the forming stage it is not necessary to manufacture elaborate and expensive moulds and the resulting shapes being fairly low density and soft in character are easily oil impregnated.
A variety of molded shapes are possible, for example shaped insulation to protect high-voltage leads. Some of the possibilities are shown in Fig. 19 and a typical HV winding end insulation arrangement based on the use of shaped end rings is shown in Fig. 20. As winding end insulation, molded end rings have the added advantage over petalling that they can be formed to a profile which will more closely follow the lines of equipotential in the area, thus eliminating tracking stress and more closely approximating to an ideal insulation structure as can be seen from Fig. 20.
Other insulation materials
Before leaving this section dealing with insulation it is necessary to briefly mention other insulation materials. Paper and pressboard must account for by far the greatest part of insulation material used in power transformers, how ever there are small quantities of other materials used on certain occasions.
The most common material after paper and presspaper is wood. This is almost exclusively beech for its high density, strength and stability. It must be kiln dried to a moisture content of about 10 percent for forming, to be further dried at the time that the transformer is dried out. In small distribution transformers the use of wood for core frames can eliminate problems of electrical clearances to leads. For large transformers wood can be used economically for lead support frames and cleats. Also in large transformers wood can provide an alternative to pressboard for winding end support slabs. In this case in order to provide the necessary strength in all directions the wood must be built up from laminations with the grain rotated in a series of steps throughout 90º several times throughout its thickness.
Paper and pressboard are excellent insulation materials when used in transformer oil. If, in order to eliminate any perceived fire hazard it is required to install a transformer that does not contain oil, one possible option is to revert to the early systems in which air is the main dielectric. Paper and pressboard are not good dielectrics in the absence of oil. Without the very efficient cooling qualities of oil, transformers must run hotter in order to be economic and paper and pressboard cannot withstand the higher temperatures involved. One material which can is an organic polymer or aromatic polyamide produced by DuPont of Switzerland and known by their trade name of NOMEX [Dupont's registered trademark for its aramid paper.]
This material can be made into a range of papers and boards in a similar way to cellulose fibers but which remain stable at operating temperatures of up to 220ºC. In addition, although the material will absorb some moisture dependent upon the relative humidity of its environment, moisture does not detract from its dielectric strength to anything like the extent as is the case with cellulose based insulation. Until the mid 1970s polychlorobiphenyl (PCB) based liquid dielectrics were strongly favored where a high degree of fire resistance was required (see Section 3.5). As PCBs became unacceptable around this time due to their adverse environmental effects, the search for alternatives was strongly pursued in a number of directions. In many quarters the benefits of transformers without any liquid dielectric were clearly recognized and this led to the manufacture and installation of significant numbers of the so-called 'dry-type' transformers complying with the requirements of the former Class C of BS EN 60085 Method for determining the thermal classification of electrical insulation which permits a temperature rise of up to 210ºC. By the 1990s this class of transformers has been largely eclipsed by cast-resin insulated types so that the use of NOMEX based insulation has become less widespread. Dry-type transformers and those containing alternative dielectrics will be described in greater detail in Section 7.8.