Sunteți pe pagina 1din 52

2009

Transformers in power
distribution networks

Stefan Fassbinder
Leonardo ENERGY Initiative
8/31/2009
Transformers in power distribution networks

Contents
1 Why do we need transformers? 3
2 The design and manufacture of conventional and special-purpose transformers 6
2.1 Transformer tank and oil 6
2.2 Core 8
2.3 Windings 12
2.4 Special types of transformers 13
3 Operational behaviour 15
3.1 Short-circuit voltage 15
3.2 Resistive load 16
3.3 Inductive load 20
3.4 Capacitive load – Care required! 21
3.5 Vector groups 23
3.6 Protection 26
3.7 Operating transformers in parallel 27
4 Efficiency 30
4.1 New standard governing transformer efficiencies 30
4.2 Driving up costs by buying cheap 34
4.3 An example 35
4.4 Amorphous steel 37
4.5 Transformers used in renewable energy generation systems 39
4.6 Other countries, other customs 40
4.7 Outlook 40
5 Special solutions for special loads 41
5.1 Bad loads 41
5.2 Practical measures 45

www.leonardo-energy.org/node/4156 Issue: 31.08.2009 10:33:00 Page 2 of 52


1 Why do we need transformers?
The answer is simple: impedance. If power losses in electricity distribution networks are to be
kept within reasonable limits, then large amounts of electric power have to be transmitted
across long distances at the highest feasible voltage. In absolute terms, the higher the trans-
mission voltage, the smaller the current and hence the smaller the (resistive) power losses in
the transmission cables. But increasing the transmission voltage also reduces a given power
loss in relative terms. Whereas a voltage drop of 3 V in a motor vehicle‟s 12-volt on-board
electrical system is a significant loss, it would hardly raise an eyebrow in a low-voltage dis-
tribution network (230/400 V), and certainly wouldn‟t impair the function of any load. In a high-
voltage network, that same 3-volt drop would be almost immeasurably small. This is why cars
will be fitted in future with a more powerful 36-volt battery (generator voltage: 42 V), as many of
the secondary and auxiliary functions currently driven by pneumatic or hydraulic systems or
driven mechanically by the crankshaft will be electrically powered. The demand for these higher
voltage systems is set to grow. There will be a transitional period in which vehicles will have a
dual-voltage architecture with a DC-DC converter coupling the two voltage levels. And the
coupling of two voltages in the very restricted space of a motor vehicle links in to our main sub-
ject of discussion: the coupling of voltages in public and industrial AC and three-phase power
networks.
The question posed at the start should strictly have addressed the situation much earlier in the
transmission process by asking why the transmission distances need to be so long? Aren‟t
certain political groups demanding decentralized, local power generation, and isn‟t that anyway
the most sensible path to take from a technical and economic perspective?
Perhaps we should start by saying that transmission paths do not „need‟ to be as long as they
are. They can of course be shorter, and indeed are shorter in the case of such currently popular
developments as co-generation plants, wind power and even the creation of very small scale
hydropower stations. Nevertheless, although there is now no need to construct long trans-
mission paths, and the power losses that would have been incurred along them are saved, this
still turns out to be the more expensive option. But why?
Consider a transformer of certain dimensions. Now double each of the three dimensions: length,
width and height, while retaining the same transformer structure. Clearly, the area of any face of
the transformer will increase fourfold. This will also apply to those surfaces available for dis-
sipating heat losses, to the cross-sectional area of the conductors and to the cross-sectional
area of the iron core – all important transformer design parameters. However, if the linear di-
mensions are doubled, all volumes will increase by a factor of eight and so too will the corres-
ponding mass.
Assuming that the current densities in the conductors remain unchanged, the current carrying
capacity (or „ampacity‟) of the conductors will increase fourfold, as the cross-sectional area of
the conductors is now four times as great. The current density measured in all transformers
rated from 10 VA to 1 GVA is indeed roughly about 3 A/mm² for copper conductors and about
2 A/mm² when the conductor material is aluminium. However, doubling the dimensions of the
wire not only increases the conductor‟s cross-section by a factor of four, it also doubles its
length. The eightfold increase in the volume of the conductor material mentioned above corres-
ponds to an eightfold increase in the mass of copper or aluminium used. For a given current
density and temperature (though the effect of temperature is not critical in this simple analysis),
every kilogram of a particular conductor material will generate the same amount of heat loss. So
a transformer whose length, width and height have all been doubled will weigh eight times as
much, and the heat losses it generates will consequently also rise by a factor of eight. But this
eightfold increase in heat loss has to be dissipated by cooling surfaces whose area is „only‟ four
times as great – a fact that we ignored above. In practical applications, larger transformers

www.leonardo-energy.org/node/4156 Issue: 31.08.2009 10:33:00 Page 3 of 52


therefore needed additional cooling. The first step is to introduce liquid cooling of the trans-
former windings. Further cooling can then be achieved by increasing the area of the transformer
cooling surfaces. This type of cooling system is known as ONAN cooling („oil natural circulation,
air natural circulation‟). Forced cooling is used in transformers with ratings above about 40 MVA.
In this type of cooling, known as ONAF cooling („oil natural circulation, air forced circulation‟),
liquid cooling is augmented by cool air blown in by fans. Above about 400 MVA, it becomes
necessary to use pumps to help circulate the oil coolant. This form of cooling is abbreviated
OFAF and stands for „oil forced, air forced circulation‟. In transformers with power ratings
greater than 800 MVA, simply circulating the oil is no longer sufficient and these transformers
use ODAF cooling („oil directed, air forced cooling‟) in which a jet of cooling oil is directed into
the oil channels of the transformer windings.
Current
Example transformers S/Cu S/Cu4/3 Energy
S [kVA] Cu [kg] Density
found [kVA/kg] [kVA/kg4/3]
[A/mm²]
Efficiency
Minimum Transformer 0.001 0.014 0.070 0.291 7.000 45.00%
Small Transformer 0.100 0.500 0.200 0.252 3.000 80.00%
Industrial Transformer 40.000 48.200 0.830 0.228 3.397 96.00%
Distribution Transformer 200.000 200.000 1.000 0.171 98.50%
Bulk Supply Point Transformer 40000.000 10000.000 4.000 0.186 3.000 99.50%
Generator Transformer 600000.000 60000.000 10.000 0.255 99.75%
Geometric Mean Value --- --- --- 0.227 --- ---
Table 1: Power densities and efficiencies of a range of real transformers from a miniature transformer to a
generator transformer

1E+05kg 100%

Efficiency 
1E+04kg 90%
Copper content 

80%
1E+03kg
70%
1E+02kg
Specific copper content 60%
1E+01kg of transformers
50%
Transformer rated throughput 
1E+00kg
40%
1E-03kVA 1E-01kVA 1E+01kVA 1E+03kVA 1E+05kVA 1E+07kVA
1E-01kg Example transformers found 30%
Theoretical Deduction
1E-02kg Energy Efficiency 20%
Fig. 1: Graph showing the copper content (blue) and the efficiencies (green) of the sample transformers listed in
Table 1 as well as the theoretical copper content derived from the formula (red).

The fourfold increase in the cross-sectional area of the transformer core permits a fourfold in-
crease in the voltage. This multiplied by the fourfold higher current in the fourfold greater cross-
sectional conductor area means a sixteen-fold (24) rise in the rated output of a transformer
www.leonardo-energy.org/node/4156 Issue: 31.08.2009 10:33:00 Page 4 of 52
whose mass is eight (23) times as great. The data in Table 1 show that this theoretically de-
rivable correlation is indeed roughly confirmed in practice. If the nominal power of the trans-
former is raised by a factor of 104, the size of the transformer (i.e. its volume and mass) only
increases by a factor of 103 (as the length, width and height have each increased by one power
of ten), which in turn means that the material costs and the costs of manufacturing and installing
the transformer system also rise by a factor of 103. Consider a high-power transformer rated at
1100 MVA, which is the largest size of transformer currently being built. From an engineering
point-of-view, it is perfectly possible to build even larger units. The problem is that the only
means of transporting these devices is by rail (Fig. 2, Fig. 3) and even then a specially designed
32-axle low-loader wagon has to be used. After all a transformer of this size weighs in at around
460 tonnes, 60 tonnes of which are copper and another 60 tonnes oil (though the oil is actually
transported in a separate tank wagon to keep the axle loads within tolerable limits). If a trans-
former with 60 tonnes of cooper has a power rating of 1100 MVA, then one might naively
imagine that a small transformer containing 60 g (most nearly equivalent to an EI48 core size )
should have an output of 1100 VA. In fact, a transformer of this size only manages about 11 VA.
Similar scaling laws apply to motors and generators. It is for this reason and, of course, because
of the associated labour costs, that it is more economical to generate electric power in large
gigawatt power stations and then to distribute this power to the region lying within a radius of
100 km, rather than generating smaller quantities of electrical power locally and feeding them
into the low-voltage distribution network. And this is where the transformers come in. It is a
commonly held misconception that a fully decentralized electricity generation system would
remove the need for the interconnected pan-European grid and its transformers. Although grid
loads would fall, the presence of the grid would be more important than ever as it would have to
compensate for sporadic and strongly fluctuating local loads and it would be needed to take up
and distribute the unpredictable supply of solar and wind-generated power.

Fig. 2: A 32-axle low-loader rail wagon for transporting Fig. 3: A high-power transformer ready for transport,
high-power transformers shown here mounted on a ‘small’ 24-axle low-loader
(source: www.lokomotive- rail wagon (source: www.lokomotive-
online.de/Eingang/Sonderfahrzeuge/Uaai/uaai.html). online.de/Eingang/Sonderfahrzeuge/Uaai/uaai.html).

A kilogram of copper in a large machine causes more or less the same power losses as a
kilogram of copper in a small machine. However, each kilogram of copper in the generator of a
large power station is responsible for, say, a power output of 10 kVA, whereas a kilogram of
copper in a bicycle dynamo would yield only 100 VA. It is clear then that the efficiency of larger
units is greater than that of smaller units, as already seen in Table 1 and Fig. 1. Although trans-
formers actually cause power losses, they are – as we have seen – minimal in these large
transformers and so one could argue that large transformers actually help to save power. This
effect also makes it more expedient to deploy a few large generators rather than a greater
number of smaller ones. Generators also have high efficiencies, with larger generators sig-
nificantly more efficient than smaller generators. However, as generators also have to produce

www.leonardo-energy.org/node/4156 Issue: 31.08.2009 10:33:00 Page 5 of 52


excitation power and suffer from mechanical losses, their efficiencies are substantially lower
than a transformer of equivalent size. The reduction in power loss that comes from choosing a
large generator rather than several smaller ones is larger than the losses that are incurred be-
cause of the need to use four or five voltage transformation stages (see Fig. 4).
If more renewable energy is to be used in future, there will be no way to avoid decentralized
generation and feed-in. This will mean that fewer transformers are in use, but the renewable
resources still suffer from the fact that small-scale facilities tend to be less efficient and thus less
cost-effective. That is why they have only become established in those regions where they have
been financially subsidized. But at some point in the future, and some sources predict some
time soon, fossil fuel resources will be so scarce and prices so high that renewable energy re-
sources will become economical. And the amount of uranium available means that it will not last
much longer than the fossil fuels, unless fast breeder technology comes back into use. There
will as a result be a global demand for renewable energy resources, though not for ideological,
but simply for financial reasons. But if „green energy‟ is to be utilized, it will mean using more of
the „red metal‟ per kWh generated and per kVA of connected load. As we set out to prove.
Structure of Public Electricity Supply in Germany Structure of Railway Electricity Supply in Germany
380kV
Transmission grids
220kV 110kV
110kV 162/3 Hz
1~
50 Hz
3~

21 kV, e. g. coal
27 kV, nuke
10 kV, e. g. hydro
21 kV, e. g. coal

10 kV, e. g. hydro 15kV

0.5 kV, e. g. wind


20kV 1.5kV
10kV
Distribution networks
0.4kV
Fig. 4: Transformers in a public power supply network Fig. 5: Three transformer stages are also used in
Yellow: sub-station transformer railway traction power systems to step the voltage
Red: generator transformer down from the generator voltage to that required to
Blue: grid-coupling transformer, drive the motors
Green: distribution transformer

2 The design and manufacture of conventional and special-


purpose transformers
Transformers – never changing and a bit dull as a result? Not true! There‟s a great deal more to
these „passive‟ devices than meets the eye. While transformers may be simple in principle,
designing them and optimizing them for specific applications requires a great deal of detailed
knowledge and considerable experience. Without such knowledge and experience, it would not
be possible to create transformers with efficiencies of up to 99.75%. But even if you are not
responsible for designing or building a transforming yourself, purchasing the right transformer
for a specific application still requires a solid understanding of transformer fundamentals and
transformer characteristics.

2.1 Transformer tank and oil


In the EU, there are about 2 million oil-immersed distribution transformers in use with power
ratings up to 250 kVA and a further 1.6 million rated above 250 kVA. There are also estimated
to be about 400,000 cast-resin transformers in use. The oil-immersed transformer is therefore
the most common type of distribution transformer.
www.leonardo-energy.org/node/4156 Issue: 31.08.2009 10:33:00 Page 6 of 52
Fig. 6: Structure of a modern oil-immersed trans-
former.

Fig. 8: At the time, the yoke frames were still made of


wood. The winding taps and terminal leads are clearly
visible.

Fig. 7: The interior of the distribution transformer (here


a museum exhibit) exposed to view. (Stadtwerke
Hannover)

Fig. 9: Manufacturing a wide copper foil winding


(Wieland Werke AG, Ulm)

The most widespread design found today is the hermetically sealed transformer with flexible
corrugated walls that can deform to compensate for the thermal expansion of the oil. These
transformers do not need an expansion tank with a dehydrating breather and some of the main-
tenance procedures that need to be performed on large transformers with attached radiators
can be dispensed with. Most of today‟s distribution transformers remain maintenance-free for
the duration of their scheduled service life of 20 to 30 years, and often over the 30 to 40 years
that many of them are actually in service. With service lives that span decades rather than years
many older transformers no longer comply with current technical requirements. As a result,
transformers that are technically outdated but not actually defective (figure 6) tend to be left in
service. (Fig. 7).
The oil serves both as a cooling and electrical insulating agent. Flashover distances
(clearances) and creepage paths can be reduced to about one fifth of their values in air, and the
active part of the transformer (i.e. the pre-assembled core-and-coil unit) has a relatively small
area to be cooled. Heat transfer from the core-and-coil assembly to a liquid medium is about 20
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 7 von 52
times better than to air. The surface of the corrugated tank (Fig. 17), in contrast to that of the
active part, can be enlarged as required to ensure an adequate rate of heat transfer to the
ambient air. Oil-immersed transformers are therefore more compact than air-cooled designs.

2.2 Core
The materials used in both the core and the coils contribute significantly to cost of a power
transformer, whose manufacture is in any case a highly labour-intensive process. Selecting the
right sheet steel for the laminations, accurate stacking with frequent staggering (every two
sheets), and minimization of the residual air gap are all key parameters in reducing open-circuit
currents and no-load losses. Today, practically all core laminations are made from cold-rolled,
grain-oriented steel sheet (the thinner the laminations, the lower the eddy currents) despite the
significantly higher cost of this type of steel.
Thick- Loss at flux
Year Material
ness (50Hz) density
1895 Iron wire 6,00W/kg 1,0T
1910 Warm rolled FeSi sheet 0,35mm 2,00W/kg 1,5T
1950 Cold rolled, grain oriented 0,35mm 1,00W/kg 1,5T
1960 Cold rolled, grain oriented 0,30mm 0,90W/kg 1,5T
1965 Cold rolled, grain oriented 0,27mm 0,84W/kg 1,5T
1970 Cold rolled HiB sheet 0,30mm 0,80W/kg 1,5T
1975 Amorphous iron 0,03mm 0,20W/kg 1,3T
1980 Cold rolled, grain oriented 0,23mm 0,75W/kg 1,5T
1980 Cold rolled HiB sheet 0,23mm 0,70W/kg 1,5T
1983 Laser treated HiB sheet 0,23mm 0,60W/kg 1,5T
1985 Cold rolled, grain oriented 0,18mm 0,67W/kg 1,5T
1987 Plasma treated HiB sheet 0,23mm 0,60W/kg 1,5T
1991 Chemically etched HiB sheet 0,23mm 0,60W/kg 1,5T
Table 2: Historical development of core sheet steels

Minimizing noise levels requires application of the right amount of pressure to the yoke frame
that holds the yoke laminations in place (Fig. 10, Fig. 22). Applying the greatest possible
pressure is not necessarily the best approach. One of the key aspects in core construction is
ensuring the absence of eddy current loops. It is for this reason that even in small transformers
with ratings of above about 1 kVA (depending on the manufacturer), the clamping bolts are
electrically insulated on one side (see Fig. 10 and Fig. 11), although these benefits would also
be apparent in transformers with power ratings below 100 VA. Given the advantages that in-
sulated fastening bolts can yield in relatively small transformers, the benefits gained in much
larger distribution and high-power transformers should be obvious.
An interesting real-life case in which a transformer was earthed twice via its yoke clamping bolts
illustrates just how important it is to take these apparently innocuous elements into con-
sideration. The transformer was fitted with an earth conductor on the high-voltage side that ran
from one of the yoke clamping bolts to the earthing system, a similar earth conductor was in-
stalled on the low-voltage side. However, the technical expert examining the transformer dis-
covered a current of 8 A in each of the earth conductors. The two conductors formed a current
loop that was short-circuiting the insulation of the bolt. It was only because the engineer had a
detector for magnetic leakage fields that he was able to discover the current in the loop.

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 8 von 52


Fig. 10: A small three-phase transformer has a very similar structure
to a distribution transformer. While these small transformers do not
generally need to be equipped with round coils … Fig. 11: … the yoke frames are similar
in shape to those used in distribution
transformers and the single-side
insulation for the yoke clamping bolts
is essential.

This would not have been a problem when yoke frames were still being manufactured from
wood (Fig. 8), were it not for the fact that the frame (whether made of wood, or steel as is the
case today) and the yoke laminations are frequently drilled (or punched) to accept the clamping
bolts. These holes have to be large enough so that an insulating bushing can be pushed over
the shaft of the bolt to ensure that the bolt does not come into contact with the burred edges of
the yoke plates and only touches one side of the yoke frame. If multiple contact points occur, it
essentially short circuits the relevant section of the yoke. In addition, cutting bolt holes effective-
ly reduces the cross-sectional area of the core, and eddy currents are also induced in the bolt,
which, for obvious reasons, cannot be manufactured from laminated sheet. Sometimes
clamping bolts made of stainless steel are chosen, because, perhaps surprisingly, stainless
steel is not in fact ferromagnetic although it consists predominantly of iron and nickel – both
ferromagnetic elements. The magnitude of the magnetic field in these stainless steel bolts is
therefore lower, thus reducing eddy current losses. In addition, stainless steel is much better at
suppressing eddy currents because its electrical conductivity is only about one seventh of that
of conventional steels. However, stainless steel bolts can in no way replace the sheet iron that
was removed when punching the bolt holes, which is to some extent possible when conven-
tional steel bolts are used. These two effects can be illustrated in the following experiment per-
formed on a small transformer (Fig. 12 and Fig. 13). Transformers of this size are typically not
fitted with insulating flanged bushings. Inserting the bolts results in a reduction in the mag-
netizing reactive power of up to 7%, as the bolts are to some extent able to replace the sheet
iron lost through the creation of the bolt holes. However, no-load losses increase by 20% partly
as a result of eddy currents in the bolts, but, primarily, because of the earth loops created when
the bolts are inserted.

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 9 von 52


Fig. 12: Unfortunately a less stringent approach is Fig. 13: The sheet metal casing and fixing screws
taken in the case of single-phase transformers. slightly reduce the magnetizing reactive power, but
the no-load active power is significantly greater.

Fig. 14: Not the most intelligent fastenings for a small Fig. 15: The no-load active power measured for the
transformer – the no-load power increases to more same transformer without the fastening clamps.
than six times that measured without the fastening
clamps in place (Fig. 15)

A better means of clamping the yoke laminations, though more costly than employing stainless
steel bolts, is to use a clamping frame that wraps around the yoke (Fig. 22). However, it is
essential to ensure that the clamping ring does not form a closed electrical circuit that would
short circuit the yoke. An experimental set-up using a small single-phase transformer, which
admittedly no one would actually build, clearly demonstrates the potential consequences of an
electrically closed clamping ring (Fig. 14 and Fig. 15).

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 10 von 52


Fig. 16: A Swiss tubular tank transformer (photo:
Rauscher & Stoecklin) 1958. This type of transformer
is still being widely built in newly industrializing
countries where labour costs are not an issue.
Fig. 17: The oil-immersed transformer has been the
standard since about 1930. The typical corrugated
tank design was introduced around 1965. (photo:
Pauwels)

But eddy currents can also be induced in


electrically conducting parts that are not
actually located within the transformer core
but are simply situated in its immediate
vicinity – this is particularly relevant in the
case of ferromagnetic materials that attract
stray magnetic fields. In larger transformers
the insides of the tank are sometimes fitted
with so-called flux traps made from core
sheet steel that attract stray magnetic fields
and through which the field flux lines will
preferentially flow rather than through solid,
non-laminated, structural steel parts. In some
Fig. 18: A typical, commercially available cast-resin dry-type transformers the clamping bolts
transformer. (Fig. 20 and Fig. 22) and other screws are
made from glass-cloth laminate. In oil-im-
mersed transformers one occasionally finds
nuts and bolts made from a moulded syn-
thetic resin/compressed wood compound,
but this material is of insufficient strength to
be used for coil clamping bolts.

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 11 von 52


2.3 Windings
In distribution transformers, the low-voltage coil is usually foil-wound because of the low number
of windings and the high cross-sectional area of the conductor required. The length of the
finished coil is approximately equal to the width of the foil (Fig. 9). For smaller sized trans-
formers or when higher voltages are involved, several strip-wound coils arranged adjacently in
the axial direction can be used.The high-voltage coil is also usually constructed in this way
though in this case round-wire windings are used in smaller transformers with shaped wire
windings used in larger devices.

Fig. 19: In small transformers, such as the 40 kVA


device shown here, the coil windings can be
approximately rectangular in section reflecting the
rectangular geometry of the core. (photo: Riedel)

Fig. 20: The upper yoke frame and the coils clamped
tightly in the axial direction (photo: Rauscher & Fig. 21: In larger transformers, the rectangular core is
Stoecklin) approximately adapted to match the circular
geometrical form of the coil

The coils in small transformers are rectangular in section (Fig. 10 and Fig. 11) and the same
type of coil geometry is sometimes found in special types of low-rating distribution transformers.
In larger transformers elliptical coils are used, and in transformers with a power rating of about
1 MVA (though some manufacturers start even earlier) circular-section coils are manufactured.
And if the coils were not circular before, they certainly are if they ever suffer a short-circuit. Such
a change in coil geometry is the result of the magnetic forces acting between the conductors.
These forces play no role at the transformer‟s nominal current density, but increase propor-
tionately with the product of the currents in the low-voltage and high-voltage windings. As the
currents in the LV and HV coils flow in opposite directions (Fig. 21), the coils repel each other. If
a short circuit does occur and the winding current is correspondingly large, the outer coil will try
to expand outward and, as the circle is the geometrical form that encloses the greatest possible
area for a given circumference, the coil will seek to adopt a circular shape as this offers the
maximum average distance from the inner coil. The inner coil, which is usually the low-voltage
winding, will be pressed against the core. As the low-voltage coil is typically a copper foil

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 12 von 52


winding, a short-circuit will often result in a core that looks as if it has been copper clad. This is
the reason why the coils are very tightly clamped in the axial direction (Fig. 20) and why any
taps in the high-voltage winding, which allow for any variation in the input voltage (typically two
steps of +2.5% above the nominal voltage, and two steps of -2.5% below), are located in the
central section of the winding (Fig. 22) and not at its upper or lower ends. This ensures that the
effective axial height of that portion of the high-voltage coil that carries current is essentially
constant as is the relative height of the HV and LV coils. Without the tight clamping a number of
windings at the upper or lower end of the coil would be lost if a short circuit caused a significant
force in the axial direction between the coils. In transformers that have been in service for a long
time, the coils may no longer be as rigidly clamped as they were at the time of manufacture and
the insulating materials may be showing signs of age. A short-circuit in such a transformer or a
breakdown of the insulation material as a result of a lightning strike often causes the device to
fail completely. At installations where short circuits or lightning strikes only occur every few
decades, a transformer can remain operational for as long as 60 years before finally having to
be replaced for economic reasons.
The rectangular core is altered to roughly match the geometry of the circular coils as shown in
Fig. 21. The yokes have exactly the same cross-sectional area. Five-leg cores are only used in
high-power transformers as this allows the cross-sectional area of the yoke to be halved, which
slightly reduces the total height of the transformer, making transport somewhat easier. Looked
at mathematically, the five-leg core only has four legs (three + two half-legs), because the two
outermost return legs only need to carry half of the flux in this type of core. We will take a look at
the special case of a five-leg core in a distribution transformer later on.
The structure of the transformer‟s active part can be seen in Fig. 22 and Fig. 23, though in these
diagrams the active part cannot be depicted large enough so that the staggering of the core
laminations is visible. This detail has therefore been shown in the magnified image on the right
in Fig. 23. Normally every two, sometimes every four, laminations are staggered by, for
example, 15 mm relative to the previous two or four core laminations. The yoke laminations are
staggered to the left and to the right, while in the legs, the laminations are displaced upward and
downward. In addition, the upper and lower asymmetrical tips of the central leg differ in that one
tip is located more to the left and the other more to the right. Offsetting the joints in this way
improves magnetic contact between the abutting surfaces.
Yoke frames

Coil clamping
bolts

Tapping points

HV coil

Wooden coil-
clamping
blocks

Yoke
lamination
retaining strap Yoke
lamination
clamping bolts

LV coil

Fig. 22: The structure of a transformer’s core-and-coil Fig. 23: The yoke frame and coil clamping bolts have
assembly (‘active part’). The design shown here is the been removed and the upper yoke lifted off to expose
more ‘elegant’ solution with unperforated yokes. the inner structure.

2.4 Special types of transformers


As the oil in a hermetically sealed transformer tank has to provide 40 or more years of service
and as it cannot be subjected to tests during that time, the quality specifications that the oil has

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 13 von 52


to fulfil are extremely high. But irrespective of its quality, these mineral oils are of course com-
bustible. It was for this reason that several decades ago oil-immersed transformers were not
allowed to be used in interior locations and in locations where there was a risk of fire. In these
locations the mineral cooling oil was replaced by PCBs (polychlorinated biphenyls), a group of
substances that are classified as non-combustible or hardly combustible. Unfortunately it was
realized, particularly in the wake of the Seveso disaster, that although these substances are not
combustible, they form dioxins when partially oxidized.
The search for alternatives led to the use of low-flammability, non-toxic, synthetic silicone oils.
Roughly speaking these are like alkane hydrocarbons in which the carbon atoms have been
replaced by silicon-oxygen atom pairs (-Si-O-). However, silicone oils never really became
established, at least not for the size of transformer being discussed here. As a result, dry-type
transformers enjoyed a revival, but the new models no longer used paper and varnish for in-
sulation, but were manufactured as cast-resin transformers. Depending on the required degree
of protection, these cast-resin transformers can be used unenclosed or with the corresponding
type of protective enclosure.
As in other kinds of transformers, the conductor materials used in distribution transformers can
be either copper or aluminium. Though more expensive, copper is usually chosen because it
enables more compact (but heavier) designs to be manufactured. Because the volume of con-
ductor material used is less if copper is chosen, the volume of the winding space is corres-
pondingly smaller, resulting in a somewhat heavier but slightly smaller device. Aluminium, how-
ever, is the preferred material in cast-resin transformers because its greater thermal expansion
coefficient is closer to the generally very high expansion coefficient exhibited by organic
materials, and this helps to reduce the thermal stresses within the rigid winding assembly.
One very special type of transformer was (re-)developed in 1987. Gas-cooled transformers had
in fact already been the subject of research some 25 years earlier. When gas cooling is in-
volved, physicists tend to think immediately of hydrogen as it has a very high heat capacity.
However, heat capacity is generally expressed relative to mass, and the density (i.e. the mass
per unit volume) of hydrogen is almost one tenth of that of air. If on the other hand the key
parameter is the speed of circulation in a cooling circuit, then heat capacity per volume is more
relevant as a resistance to flow is proportional to the square of the volume flow in any given
system. The gaseous material therefore selected was sulphur hexafluoride (SF 6), a well-known
substance that was already being used as an insulating material in switchgear and that has a
density five times that of air and has a considerably better dielectric strength. Engineers carrying
out electrical breakdown tests on a specially designed open-top test vessel were able to enjoy
the rather unusual observation of air-filled balloons apparently floating mid-air within the test
vessel. The balloons were of course resting on a bed of higher density SF 6 that had been slowly
and carefully filled into the container. By the way, breathing in this completely non-toxic and
chemically inert gas (its inertness is directly linked to its high dielectric strength) results in the
opposite of the well-known helium „squeak‟. The engineers who carried out these experiments
on themselves all lived to tell the tale, proving just how harmless the gas sulphur hexafluoride is.
Similar tests using hydrogen should, however, be avoided at all costs – especially by smokers!
Although the heat capacity of a kilogram of SF6 is only half that of a kilogram of air, its heat
capacity per litre is 2.5 times greater. That means that if SF6 is used as the coolant, it only
needs to circulate at 40% of the speed used in air-cooled devices in order to produce the same
cooling effect. As a result, the fan power can be reduced to about 32% of that needed in an
equivalent forced-air cooling system. Two prototype transformers each with a power rating of
2 MVA, a corrugated tank and internal forced cooling (i.e. cooling class GFAN – gas-forced, air
natural) were built and operated in an explosion hazard area within a chemical manufacturing
plant owned by the gas manufacturing company, which, it goes without saying, had an under-
standable interest in these trials.

DistributionTransformers.docx 31.08.2009 10:33:00 Page 14 of 52


The dielectric strength and the cooling capacity of SF6 can be increased by raising the pressure
and compressing the gas. A hand-welded tubular tank transformer, similar to the one shown in
figure 16, was therefore built to test this effect. This type of transformer design used to be
common but its construction is far too labour-intensive for it to be economical today. Neverthe-
less, the test device allowed the test engineers to demonstrate that the observed temperature
rise agreed approximately with that expected from computational analysis, which is what they
set out to show. The transformer with GNAN cooling handled 630 kVA at an overpressure of
3 bar and was of an acceptable size. Having completed these trails, the project team set about
developing a more economical method of production. Apparently, these transformers sell well in
the Far East, or at least sold well for a time, where they were used in high-rise buildings. Wide-
spread use in the domestic market failed because of the very stringent regulations governing
the construction and use of pressure vessels. The principle behind the technology had, how-
ever, been shown to work.

3 Operational behaviour
Transformers inevitably affect the power networks to which they are connected. However, to a
certain extent some of the operating parameters of a transformer can have a beneficial, in some
cases even essential influence on the operation of the supply network. In what follows, we will
be investigating how to manufacture and select transformers to optimize these parameters.
In a distribution transformer with a short-circuit voltage of 6% that is operating at its rated
current, there will be a drop of 6% in the voltage across the device‟s internal impedances. That
means that when the transformer id operating at its rated load, the voltage is 6% lower than the
open-circuit voltage. There are additional voltage drops along the wires and cables that lead
away from the transformer as well as in the upstream power supply network. In total, we should
expect voltage losses totalling around 10%. Ten percent? That sounds excessive. Luckily, the
situation isn‟t as bad as it might at first appear. To see why, let‟s take a closer look at what we
mean by the term „short-circuit voltage‟.

3.1 Short-circuit voltage


The operating behaviour of a transformer is principally characterized by its rated voltage and its
rated power output – something that does not need to be explained further here. The next most
important parameter is the short-circuit voltage. To anyone training to become an electrical
technician, the expression „short-circuit voltage‟ might appear at first to be something of a
misnomer. After all, when a short-circuit occurs, the voltage is generally defined to be zero.
However in this particular case, the short-circuit is on the output side, while the voltage is that
on the input side. If the secondary winding of the transformer is short-circuited and if the voltage
applied to the primary winding is just large enough to generate the rated current in the
secondary winding, then the value of the applied voltage is known as the short-circuit voltage
(Usc). Generally speaking, the short-circuit voltage is not expressed in volts, but as a percentage
of the rated voltage (usc). In large transformers, the short-circuit voltage can reach values of 18-
22%. In contrast, the rated short-circuit voltages in distribution transformers are typically 4% and
6%. The actual value is measured during the final testing of the device and printed on the
transformer's rating plate (Fig. 24). However, deviations from the typical values of 4% or 6% are
usually of negligible practical relevance.

DistributionTransformers.docx 31.08.2009 10:33:00 Page 15 of 52


The short-circuit voltage thus characterizes
the voltage drop within the transformer. If the
short-circuit voltage is known, the short-
circuit current Isc can be readily calculated.
For example, if all upstream impedances are
ignored, the short-circuit current in a trans-
former with usc = 6% is:
I I
I sc  N  N  16.7 I N .
usc 6%

Fig. 24: No matter whether it’s called ‘Tension court


circuit’, ‘Kortsluitspanning’ or ‘Short-circuit voltage’ –
the actual value is measured during final testing and is
printed on the transformer’s rating plate.

So for a transformer with usc = 6%, six percent of the rated input voltage is needed to generate
the rated („nominal‟) current IN in the short-circuited secondary winding. However, only a small
part (uR) of the voltage drop in a medium-sized distribution transformer is due to the ohmic
resistances in the windings, by far the largest factor contributing to the voltage drop is the
reactive / inductive voltage drop uX. This stems from the leakage inductance caused by that
portion of the magnetic flux that bypasses the core (leakage flux) and permeates only a single
winding. This leakage flux does not flow in the primary and secondary windings simultaneously,
but in the main leakage channel between the high-voltage winding, which is generally located
on the outside, and the low-voltage winding on the inside [c.f. section 2 „Design‟]. The leakage
flux is therefore part of the magnetic flux of the outer but not the inner winding. This also has the
incidental effect that the short-circuit voltage cannot be influenced by the non-linearity of the iron
core. uX is also referred to as the leakage reactance voltage. The principal function of the main
leakage channel is cooling, its second function is insulation, and thirdly, it serves to maintain the
so-called leakage reactance, which is in effect a defined short-circuit voltage.
As shown in Fig. 28, the sum of the squares of the inductive voltage drop uX and the ohmic
voltage drop uR equals the square of the overall voltage drop usc (c.f. Pythagoras‟ theorem
concerning the sides of a right-angled triangle). Fortunately, the ohmic voltage drop uR is, as
already mentioned, the smaller part, and the larger the transformer, the smaller it is. A simple
calculation proves the point: If a transformer in the 630 kVA range has an efficiency of 98.5%
when operating at its rated load, then the total ohmic voltage drop across the two windings can
be no more than 1.5% of the rated voltage. In practice, however, the value is lower, for example
1%, because the 1.5% includes losses other than the ohmic losses in the coils. Our usc of 6% is
therefore made up of uR = 1% and uX = 5.91% of the rated voltage (6² =1² + 5.91²).

3.2 Resistive load


A transformer‟s power rating is always specified relative to its resistive load. The ohmic
resistance of the winding contributes linearly to the (rated) load, while the inductive resistance
(reactance) of the leakage inductance contributes quadratically. So the resistance of the
winding contributes only 1% to the load resistance, with the remaining 5.91% having only a
negligible effect on the total voltage drop across the transformer and the load. We now want to
determine just how small this effect really is.
The non-ideal behaviour of a transformer can be illustrated by an equivalent circuit model (Fig.
25). The model assumes that the input and output windings have the same number of turns. As

DistributionTransformers.docx 31.08.2009 10:33:00 Page 16 of 52


this is obviously not usually the case, the values associated with one side of the transformer are
„moved‟ (i.e. referred) to the other side by multiplying them by the ratio of the numbers of turns
on the two windings. The behaviour of the transformer can then be calculated for the relevant
reference side; in Fig. 25 all elements have been referenced to the load (i.e. secondary) side.
RCu1„ RCu2
X1s„ X2s

Load
RFe
Xm
RCu1„ RCu2
X1s„ X2s

Load
RFe
Xm

RCu1„ RCu2
X1s„ X2s

Load
RFe
Xm

Fig. 25: General single-phase equivalent circuits for a two-winding transformer

If we are only interested in the modulus (absolute magnitude) of the voltage drop at the rated
load, which by definition is an ohmic load, then we can adopt the simplified expression
 ( X ' X 2s ) 2 
U 2  U1 ' I Last  RCu1 ' RCu 2  1s  , where:
 RLast 
U2 secondary voltage,
n2
U1 '  U1 primary voltage referred to the secondary side,
n1
2
n 
RCu1 '   2  RCu1 resistance of primary winding referred to the secondary side,
 n1 
RCu2 resistance of secondary winding,
2
n 
X 1s '   2  X 1s leakage reactance of primary winding referred to the secondary side,
 n1 
X2s leakage reactance of secondary winding,
n1 number of turns in primary winding,
n2 number of turns in secondary winding.
ILoad and RLoad are related to one another in accordance with Ohm‟s law:
U2
Rload 
I load

and cannot therefore be changed independently of one another.

DistributionTransformers.docx 31.08.2009 10:33:00 Page 17 of 52


Fig. 26: No-load currents in a high-quality Swiss distribution transformer with a power rating of 630 kVA to which a
voltage of 400 V has been applied to the low-voltage side. The currents are relative to the rated output current of
909 A!

The parameters
RFe core resistance,
XH magnetizing reactance
have not been taken into account in this simplified model. Except in small transformers, the
magnitudes of these quantities are usually so large that the currents flowing through these
elements are insignificant, at least as far as the effect of the transformer on connected loads is
concerned. (Their relevance for the transformer‟s internal losses is far greater, as will be shown
in section 4). XM is the magnetizing reactance under open-circuit conditions (in this case, XM is
referred to the secondary voltage, as if the excitation voltage was being applied to the output
side of the transformer, which is perfectly possible, and was indeed the case when making the
measurements for Fig. 26.) The no-load current in a good-quality distribution transformer is only
around 0.5% of the rated current (Fig. 26) and more than half of the no-load current is
attributable to the magnetization current. Consequently, as the magnetizing reactance XM is the
main cause of the open-circuit current, its magnitude must be at least 200 times that of the load
impedance. RFe is a fictitious resistance that represents the iron (or core) losses and whose
magnitude, if good-quality iron is used, is generally substantially greater than XM. The shunt
impedance of these two elements that determine the open-circuit behaviour of the transformer is
therefore significantly more than 100 times greater than the load impedance. In contrast, for a
transformer with a short-circuit current of 6%, the short-circuit impedance, i.e. the effective sum
of X1σ‟, X2σ, RCu1‟ and RCu2, which are all in series with the load, is by definition only 0.06 times
as large as the load impedance. The short-circuit impedance is therefore more than 100/0.06
(i.e. almost 2000) times smaller than the shunt impedance comprising RFe and XM. As a result,
the currents flowing through RFe and XM can be neglected when describing the transformer‟s
behaviour with respect to its connected load and this is even more true when the transformer is
under short-circuit conditions. The expression for the leakage reactance can therefore be
simplified to:
X s  X1s ' X 2s  2 X 2s .

DistributionTransformers.docx 31.08.2009 10:33:00 Page 18 of 52


A similar expression applies for the resistive components of the short-circuit impedance.
RK  R1 ' R2  2R2 .
Inductive drop uX in the transformer

Total voltage Ohmic drop


drop usc in the uR in the
transformer transformer
(e.g. 6%)

Inductive drop uX in the transformer


Input voltage U1 referred to the secondary side (100%)

Ohmic drop uR in the transformer


Fig. 28: Diagram showing the voltage drops in the
Fig. 27: Phasor diagram showing the voltage drops in transformer itself, shown here at a magnification of
a transformer and its rated (ohmic) load about ten times that in figures Fig. 27, Fig. 29 and Fig.
30

This approach provides a simple means of describing the operation of a non-ideal transformer
and enables quantities of interest to be calculated with sufficient accuracy. Although Fig. 25
does not model the substantial non-linearity of the iron core and the resulting strong distortion of
the magnetizing current waveform, the fact that the magnetizing current is so small means that
this equivalent circuit model is applicable in practice. In fact, the magnetizing current can usually
be neglected when compared to the rated current, as was assumed above. In that case, the
voltage drops found in a transformer driving a load are as shown in Fig. 27. The phasor diagram
can be thought of as an instantaneous snapshot. The individual voltages are represented as
vectors that precess about the fixed origin (below), in a manner analogous to the generation of
an AC voltage in rotating machines. The input voltage (green) serves as the reference and is
always shown as its peak positive value, i.e. as a vertically aligned voltage vector. The vector
sum of all the voltage drops must equal the applied voltage. Graphically, this means that if the
voltage-drop vectors are placed in sequence starting at the origin (base of the green vector),
they must arrive at the same point as the tip of the green vector representing the applied
voltage. It then becomes apparent that in this transformer with a short-circuit voltage drop (usc)
of 6% and an ohmic voltage drop (uR) of 1%, the modulus of the output voltage at rated load is
almost 99% (and not about 94%) of the open circuit voltage. The only difference is the phase
shift on the output side relative to that of the input voltage, but that has no effect on the load.
The voltage drops within the transformer are barely discernible in the diagram. The resistive
(ohmic) component in particular is almost inconspicuous – and it is good that it is, as it re-
presents the ohmic losses. If one wants to visualize these voltage drops, they need to be
magnified and shown without the input and output voltage vectors (see Fig. 28).

DistributionTransformers.docx 31.08.2009 10:33:00 Page 19 of 52


Total voltage drop usc in the
Inductive drop uX in the transformer transformer (e.g. 6%)

Inductive
Total voltage Ohmic drop drop uX in Ohmic drop
drop usc in the uR in the the
uR in the
transformer transformer transformer
transformer
(e.g. 6%)

Input voltage U1 referred to the secondary side (100%)

Input voltage U1 referred to the secondary side (100%)

Voltage drop across load (99%)


Voltage drop across load (99%)

Fig. 30: Capacitive load


Fig. 29: Inductive load

3.3 Inductive load


The situation changes, however, if we are dealing with an inductive load. This adds linearly (i.e.
completely) to the inductive voltage drop within the transformer and quadratically to the much
smaller ohmic component. In this case, the output voltage is calculated by means of the
following equation:
 ( RCu1'  RCu 2 )2 

U 2  U1'  I load X1s '  X 2s 
 Rload 
 
This can be rewritten using the simplifications introduced earlier:
 R 2
U 2  U 20  I load  Xs  Cu 
 Rload 

where U20 is the no-load voltage on the low-voltage side. When the rated current is flowing, the
output voltage does indeed drop by almost 6% at least as far as the absolute value (magnitude)
of the voltage is concerned, because, as already mentioned, the rated load is defined to be an
ohmic load (see fig. 25). By introducing power-factor compensation in parallel to the load, we
can reduce the overall load being driven by the transformer. This not only reduces the copper
loss, it also reduces the size of the voltage drop, as this is predominantly inductive in nature. If
we are dealing with an alternating inductive load, the compensation circuit must be controllable.
The more rapidly the load changes, the faster the controller must be able to respond. These
control systems can be used to avoid the flicker that is caused by rapidly changing, strongly
inductive loads such as spot welding machines or three-phase induction motors (without a
power converter). Anyone who has lived for a time next to a building site where a tower crane
has been operating (without power-factor correction) will know exactly what we are referring to.1

DistributionTransformers.docx 31.08.2009 10:33:00 Page 20 of 52


Luckily, mobile units are now available that can be used to counteract such mobile sources of
flicker.2
The same approach can be used to determine whether the predominant cause of the voltage
drop at a domestic power socket is the leakage inductance of the distribution transformer or the
ohmic resistance of the power cables. For instance, when a fan heater with a current input of
9 A is operating, the voltage at a particular power socket is found to decrease from 230.7 V to
226.3 V. If a large inductive load that draws over four times as much current (see fig. 27) is
connected, we see that there is no discernible increase in the voltage drop. With reference to
the phasor diagrams shown in section 3.2, it can be concluded that the ohmic resistance is by
far the larger component – at least in the case of this cable and this transformer. However, this
test must be performed with great caution. As the test involves a substantial overload, the
duration of the test must be kept short. Failure to do this will trigger the 16-ampere circuit
breaker, which is normally rated for a power factor range from 1 to 0.6. A circuit breaker is not
really the best means of switching off what is essentially a purely inductive load.

Fig. 31: Large inductive load on a domestic power socket

3.4 Capacitive load – Care required!


Things get really exciting when dealing with capacitive loads. In principle, this is the situation
described earlier in which the transformer is excited from the output side, i.e. compensation of
the inductive magnetizing reactive power by the capacitive load. If the load is so small that it is
just sufficient to provide compensation, then only the transformer will be excited – a perfectly
normal situation, irrespective of the winding. However, as the capacitive load increases, the
picture changes. The passage of the load current through the leakage reactance results in a
voltage drop with a phase lead of 90°, its passage through the capacitive load is accompanied
by a voltage drop lagging by 90°. In other words: these two voltage drops have opposite signs
and therefore subtract from one another rather than add. At 100% of the rated input voltage, we
have a voltage „drop‟ of almost 6% across the leakage reactance and therefore almost 106% of
the open circuit voltage across the load. The result: overvoltage on the output side at full load
current, despite applying only the rated voltage on the input side (Fig. 30). As the load, or more
precisely its impedance, must be assumed to be constant, the current rises in accordance with
the increased voltage, creating an even greater voltage drop within the transformer and hence
an even greater voltage at the output. Ultimately, the voltage settles down at a level slightly
above rather than just below 106%. This was exactly what was observed when a medium-
voltage customer of an EUC (electric utility company) wanted to have a power factor correction
system of 1400 kVAr connected on the MV side, as that is the side where active and reactive
power are metered. It turned out, however, to be cheaper to buy a low-voltage power factor
correction unit in combination with a 1600-kVA transformer. Note that although this solution was
cheaper, it was not more cost-effective, as the method saves only the additional price that the
EUC would have charged for the reactive power, but not the costs generated by letting this
reactive current circulate around the installation and through the transformer. After all, the
reason the utility company charges for reactive energy is because it has a detrimental effect on

DistributionTransformers.docx 31.08.2009 10:33:00 Page 21 of 52


network power quality and causes losses. But reactive power also causes losses in the
customer‟s installation and this fact is not altered by installing one central power-factor cor-
rection unit rather than using lots of local, decentralized units. But this has more to do with
reactive power compensation or power factor correction rather than transformers, which is our
subject here.
A test run on the customer‟s system mentioned above resulted in the phase voltage increasing
from 230 to 255 V, because the transformer was driving a capacitive load only and was working
at pretty much full load. In this particular case, the problem was solved by reconnecting the
input side of the transformer for an input voltage of 22 kV, even though only 20 kV were actually
being applied. This enabled the output voltage to be lowered to about its rated value. As the
excitation current was coming from the output side, this proved to be the only way to prevent
overexcitation of the transformer. One of the rare cases in which the output side is the reference
side.
If overloading is severe, things will eventually escalate. Current overload creates a voltage
overload in both the load and in the transformer, overvoltage across the load then generates
further current overloading that can drive the overvoltage even higher. Taken together, the
leakage reactance and the capacitive load form an LC oscillator circuit, whose resonance
frequency f0 can be calculated as follows:
1
f0 
2π LC
At this frequency, the inductive leakage reactance and the capacitive reactance of the load are
of equal magnitude but opposite sign and thus cancel each other out. This is not, however, the
case when the transformer is passing its rated current. In the example transformer discussed
earlier (for which usc = 6%), the reactance of the load is about 16 times greater than that of the
internal reactance of the transformer. The resonance frequency is then:
f0  50Hz * 16  200Hz .
16.7*rated load (ZSC=ZLoad or YSC=YLoad, resp.) 
100 2,0
Relative load current I/IN 

Relative load current I/IN 

Current with resistive load Current with resistive load


90 1,8
Current with inductive load Current with inductive load
80 1,6
Current with capacitive load Current with capacitive load
70 1,4
60 1,2
50 1,0
40 0,8
30 Short-circuit current 0,6
ISC=16.7*IN  Rated load
20  0,4
10 0,2
Relative load admittance Y/YN 
0 0,0
0 1 2 3 Relative
4 5 6 load
7 8admittance Y/Y
9 10 11 12  15 16 17 18 19 20
13N 14 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 1,8 2,0

Rated load
Fig. 32: The output current of a transformer with Fig. 33: Detail from Fig. 32 (see box bottom left) with
uR = 1% and usc = 6% (for which Isc ≈ 16.7 * IN) can in the current values typically found in practice
theory climb to 100 * IN at a capacitive overload of
16.7 times the rated load, rather than just the 16.7 * IN,
that would arise in a short circuit.

However, if a ripple-control signal of similar frequency is present, the ripple-control signals


through the series resonant circuit, comprising the transformer‟s leakage reactance and the
capacitive load, will be shorted and lost, and therefore fail to reach the low-voltage level. In a
transformer with a short circuit voltage usc = 4%, the critical point is shifted to 224 Hz. If we were
dealing with capacitive load that could be varied (e.g. by means of a VAr controller), the critical
point would also vary in a rather uncontrollable manner.
DistributionTransformers.docx 31.08.2009 10:33:00 Page 22 of 52
If the transformer is subjected to a capacitive overload of 16 times the transformers rated load,
which can be achieved, say, by connecting a 1600 kVAr power-factor correction unit to a
100 kVA transformer, we find that the resonant frequency drops to 50 Hz. In this case, the
inductive leakage reactance would be compensated by the capacitive reactance of the load,
and as the current would be limited only by the resistance of the winding, it would rise to almost
six times the short-circuit current or 100 times the rated current (see Fig. 32). In Fig. 32, the
actual current is plotted against the ratio of the magnitude of the conductance G to the rated
conductance GN:
IN
GN  ,
UN
depending on the phase angle of the ohmic, inductive or capacitive load. For
GN
G  16.7G N
6%
the impedance of the load is equal to the internal impedance of the transformer. But that should
not be interpreted to mean that the output voltage always corresponds to half the open-circuit
voltage. If that were the case, we wouldn‟t have needed to include the phase angle in the
calculations. Only at „infinite conductance‟ (which corresponds to short-circuit conditions) will the
curves for the ohmic, inductive and capacitive loads converge.
Luckily, capacitive loads of that magnitude do not occur in practice. Nevertheless, the thought
experiment shows just how rapidly things change when the capacitive load starts to increase. A
more detailed look at that part of Fig. 32 that covers realistic operating currents (see Fig. 33),
confirms remarks made earlier: for the ohmic rated load, the transformer‟s output voltage is 99%
of the open-circuit voltage. If the transformer drives an inductive load of equal size, the output
voltage drops to 94%, while if the load is capacitive in nature (and of equal magnitude), the
voltage is about 107%. Care is therefore paramount when dealing with capacitive loads.
However, the transient overloading of a transformer with a capacitive load does occur and is in
fact purposely used on occasions to eliminate voltage dips and the resulting flicker in ohmic and
ohmic-inductive impulse loads. For this to be achieved, the required correction capacitance has
to be connected to the transformer at the same time as the critical load. The resulting voltage
rise compensates for the voltage dip that would have been caused by the critical load alone.
However, considerable care has to be taken when dimensioning such systems. Although
temporarily connecting the capacitive load to an ohmic load (e.g. a spot welding machine)
causes the voltage dip to disappear, the total load current increases and places the transformer
under greater stress than would have been the case with the critical ohmic load on its own. It is
only when the flicker is generated by an inductive load, such as the tower crane referred to
above, that this type of compensation actually reduces the load on the transformer. In both
cases, it is recommended that the load and the compensation unit are connected jointly to the
transformer so that compensation is immediate and proactive rather than delayed and reactive.

3.5 Vector groups


Distribution transformers are normally designed with the Dyn5 vector group. That means that
the transformer has a delta-connected HV winding, a star-connected LV winding and the star
point is brought out as a neutral terminal. The input and output voltages have a relative phase-
shift of 150° (5 x 30°). Phase shifts are restricted to steps of 30°, hence the index 5 to signify
150°. At least that is the case if one limits oneself to zigzag connections in which the delta-
connected and the star-connected parts of the relevant winding have the same voltage. This
transformer classification scheme simply reflects the fact that in the past the use of other ratios
would have made little sense. Today, special-purpose transformers are available that offer a 1:3
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 23 von 52
voltage ratio and therefore phase-shift steps of 15° – though, perhaps understandably, no one
has attempted to market them as, say, transformers with „vector group Dyn4½‟. This type of
transformer is used to generate „greater than 12-pulse‟ rectifiers such as those used to reduce
harmonic emissions in electrolysis plants and in very large converter drives.
A transformer that is star-connected on both sides, whether with neutral points (YNyn) or
without (Yy), can only have the vector group Y(N)y(n)0 or Y(N)y(n)6 as the sole variation
possible in star-star transformers: either the three „start of winding‟ points are joined to form a
neutral point and the „end of windings‟ are brought out as phase conductor connections or vice
versa i.e. reversing the polarity of each of the coils.
The sole possible variation? Strictly speaking, no. One could re-label the phase conductors and
define, say, conductor L1 on the input side as L2 or L3 on the output side. This would lead to
something that one could designate as a Y(N)y(n)4 or a Y(N)y(n)8 vector group – or one could
simply call it a mistake. Those in the electrical industry would probably tend to favour the
second interpretation – assuming of course that the „mistake‟ was ever detected. Only if the
output sides of a correctly wired and an incorrectly wired transformer were to be connected in
parallel would the mistake become apparent and very clearly apparent as it would in all
probability destroy the installation – not the sort of failure detection method that one would use
intentionally!
Some in the electrical industry refer to the brought-out neutral point as a PEN connection. That
is incorrect. Despite what is sometimes assumed, the carefully insulated neutral point brought
out from an oil-immersed transformer is never actually connected to the inside of the tank. While
conceivable in theory it is unlikely that a customer would ever want a transformer configured in
that manner. If a transformer was customized in this way, it would no longer be possible to
operate it in the modern, standardized3 multi-feed power distribution networks in use today.
In principle, transformers with other vector groups could conceivably be used to feed into public
low-voltage networks. However, these transformers cost more and have no advantage over the
standard vector group Dyn5 – at least that has been the case up until now. Recent develop-
ments have begun to reshape the power engineering landscape and we will be looking at these
later on.
As medium-voltage networks do not generally have a neutral conductor, only a delta winding is
feasible on the HV side of the transformer. If the windings were star-connected, this would
reduce the voltage across a coil by a factor of 3 making it easier to control. This is exactly the
reason why all grid coupling transformers (i.e. transformers that interconnect EHV and HV
power networks) are configured with the YNyn0 vector group, but these networks always have a
neutral conductor even if this just means that the neutral points on each side of the transformer
have been earthed. As a rule a medium-voltage power network does not have a neutral con-
ductor. If you need a loadable neutral on the output side so as to be able to tap two different AC
output voltages, then under single-phase supply conditions current will flow in only one LV coil
generating magnetic flux only in that one limb. The operating principle of a transformer dictates
that this flux has to be compensated by a corresponding counter flux generated by the HV coil
in that same limb. But his can only happen if the input voltage is applied to both ends of the
input winding directly, which itself is only possible in the case of a delta winding or a connected
neutral point. The delta connection could be approximated by a parallel circuit, and the star
connection without a neutral terminal could be modelled by an equivalent series circuit (as
shown in the modified equivalent circuit diagram in Fig. 34). In the current case of an open
neutral point on the input side and the single-phase load on the output side, we would have
something like a loaded coil in series with an unloaded coil, and the unloaded coil represents a
reactor with a very high magnetizing reactance XM that reduces the load current. As a result,
almost all of the voltage across the quasi-series circuit drops across this unloaded coil, which
excites the core and drives the core into magnetic saturation, with the consequence that iron
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 24 von 52
losses and magnetic leakage losses rise dramatically. Although the magnetizing reactance falls,
it does not fall far enough for it not to cause a substantial shift in the voltages. The voltage
across the load collapses to a fraction of the rated voltage, while a substantial overvoltage is
present on the unloaded output winding. If the winding is in fact not completely unloaded, but
actually feeds a relatively small load, then this load will have to cope with a continuous 3-fold
overvoltage. A situation that brings with it a real risk of damaging or destroying the load and of
fire damage. This is why Yyn0 and Yyn6 vector groups are generally not used if the only
brought-out neutral point, the one on the output side, is loaded. It is for this reason that the input
windings of distribution transformers are usually delta-connected.

L1 RCu1„ RCu2
X1s„ X2s

RLoad
Xm

RFe
N
RCu1„ RCu2
X1s„ X2s
Xm

RFe

L2

Fig. 34: Equivalent circuit representing single-phase loading of a Yyn transformer: As the load impedance is
considerably less than XM and RFe, the total impedance of the upper circuit is much smaller than that of the lower
circuit

Unless, of course, you happen to be dealing with a TT system of a type still frequently found in
Belgium. These TT systems are fed by a transformer with the usual Dyn5 vector group but for
which the output voltage is only 133/230 V. In this case, we could equally well have used Yyn0
or the Yyn6 vector groups as the neutral point on the low-voltage side is brought out but is not
connected and therefore not loaded. Its only use is for measurement and testing procedures,
such as monitoring earth faults. The voltage at the AC power socket is that between two phase
conductors, and so the current from the power socket, which for example in Germany would be
a single-phase current, flows as a two-phase current through two low-voltage windings and ex-
pects a corresponding current through the two high-voltage coils on the relevant limbs. If the
high-voltage side is star-connected without a neutral terminal, there is nothing to prevent this
current from flowing. But a single-phase load on the output side would mean that the current in
the HV winding would have to flow first through a loaded coil and then through an unloaded
one, the latter acting, as already described, as a reactor that attenuates the current. This sort of
behaviour needs to be taken into account if the LV neutral point is earthed in the hope of re-
ducing the earth-fault loop impedance, which is anyway much higher in TT systems than in TN
systems because of the resistance of the earth path in the impedance loop. Clearly the high
load-imbalance impedance of the transformer will prevent any noticeable reduction in the im-
pedance of the earth-fault loop.
A voltage tester will light up (though only weakly) whenever it touches any active conductor in
an AC power socket, because the voltage on each active conductor is only about 133 V relative

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 25 von 52


to earth. In an IT earthing system, the circuit for the voltage tester is closed via the stray
coupling capacitances. An ordinary tungsten-filament lamp would no longer be able to light up.
However in a TT system, this current – although sufficient to illuminate the lamp and to cause a
dangerous electric shock – would not be enough to trigger an overcurrent protection device.
This is the well-known problem of the TT system and one that is exacerbated somewhat by this
type of „Belgian‟ supply. Although it has to be said that depending on the extent and the age of
the supply network and the possible capacitive leakage currents, especially those in modern
loads, dangerous shock currents can also arise in IT earthing systems – making RCD protection
necessary in both TT and IT systems. In addition, neither variant allows for the possibility of
driving single-phase and three-phase AC loads designed for a 400 V supply. For example, the
popular German flow heater requiring a three-phase 400 VAC connection would not function in
parts of Belgium (unless of course someone was prepared to accept only a third of the device‟s
rated output). In contrast, an electric cooker designed for a three-phase supply works without
difficulty. Cookers are usually designed to be able to cope with these supply networks by having
a means of reconnecting the terminals on the terminal board so that 230 V is always supplied to
each of the three load groups. The cooker does not actually need 400 V to operate. However,
this approach will not work in the case of the flow heater, as the individual heating elements are
usually dimensioned for 400 V, i.e. for a delta connection. While it would be possible in principle
to design a flow heater to run on a 230 V / 400 V delta / star supply, there is a big difference
between providing 7.5 kW for an electric cooker or 27 kW for a flow heater. Providing the latter
level of power becomes complicated if no 400 V supply is available.

3.6 Protection
The distribution transformers used in public power supply networks are generally not protected
on the output side. The input side, in contrast, is equipped with HV HRC (high-voltage high-
rupturing capacity) fuse links. However, if such a fuse is subjected to an overcurrent in the
range between one and three-times the fuse‟s current rating, it will tend to overheat but won‟t
interrupt the fault current. But before a cynic turns round and says that “HRC” is obviously an
abbreviation of „Hopelessly Redundant Component‟, we need to set the record straight.
This type of fuse provides protection against a short circuit fault, but not against overloading.
Protection of this kind is usually perfectly adequate, because by properly planning the network
based on parameters gained from years of experience, and by designing-in sufficient levels of
reserve power, it is possible to prevent overloading in, say, Germany. It simply doesn‟t occur. In
other regions of the world, however, overloading is the normal state. Transformers are pretty
tough characters who can put up with a lot – whether this makes economic sense is something
that we will be examining a little later. The crucial areas in which protection is needed are short-
circuiting or arcing on the output side and the albeit rare event of an internal fault within the
transformer. Particularly turn-to-turn faults in the LV foil windings can result in some spectacular
damage. Arcing can cause some of the oil to vaporize or can cause it to decompose into
gaseous components. The resulting pressure wave swells the sides of the transformer tank.
Just like the coils react when a short-circuit occurs, the tank tries to attain a spherical form that
offers greater volume per surface area. If such a damaged transformer is subsequently dis-
assembled, the conductor material found in the base of the tank has the form of egg-sized, egg-
shaped globules of red copper or pale silver aluminium speckled with soot. The coils, and in
many cases the tank as well, are now simply scrap, with only the core and transformer ac-
cessories still usable. If the fault had persisted for even a few seconds, the tank would in fact
have burst, leaking copious amounts of oil that would have ignited and acted as a fire ac-
celerant. It is therefore quite clear, that short-circuit protection is essential, but overload pro-
tection is not. The development of electronic control systems means that it is now common to
have remote monitoring of the oil temperature. This not only helps to prevent hazardous
situations from arising, but it also helps to optimize the operation of the supply network, as the
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 26 von 52
difference between the temperature of the oil and the ambient temperature provides an in-
dication of the degree of transformer loading.
The issue of short-circuit protection underscores once again the problems described earlier
concerning the use of inappropriate vector groups. If, for instance, there is a single-pole-to-earth
fault on the output side and an unconnected neutral point on the input side, and the magnetizing
reactance of a limb not involved in the short circuit is „in the way‟ of the short circuit, then the
short-circuit current that flows will be too small to trigger the fuse. The extreme shift in the
voltage system of almost a factor of √3, causes the limb in question to become overexcited and
magnetically saturated. As a result, that part of the voltage that exceeds the rated voltage is
practically only affected by the leakage reactance of the corresponding limb of the core. The
amount of current flowing can therefore be well above the rated current for the transformer itself
and for the fuse, but still too small to trigger the fuse. The question then is which of the two
blows first. It is worth repeating here our call to spend time carefully choosing the right vector
group so that if a short circuit does occur, it is definitely big enough to be identified as such by
the fuse system, which can then react as it should and interrupt the fault current. If that is not
possible, then other protective and monitoring systems need to be put in place.

3.7 Operating transformers in parallel


Connecting transformers in parallel is of course possible in principle. This obviously means that
the voltage ratings of the two coils to be operated in parallel must be identical. Any off-load tap
changers or strap panels must also have the same settings. As an example, we will assume that
we want to operate two transformers, which have the same output power rating, with their
primary windings in parallel and the secondary windings also connected in parallel. Each trans-
former has a voltage changer with a range of ±5% of the rated voltage, but one is set to +5%,
while the other is set to -5%. One transformer has a short-circuit voltage of usc = 4%; for the
other usc = 6%. In this case, the two short-circuit impedances are in series and both trans-
formers are coupled via an impedance of 4% + 6% = 10% of the load impedance relative to the
rated load of one of the transformers. Similarly, the difference between the parallel voltages is
also 10% (of the rated voltage). As a result, the transformer that exhibited the 10% higher open-
circuit voltage will drive the rated current through the other transformer, which will in turn trans-
form this current back onto the input side. All windings on both transformers would therefore be
carrying the rated current, the one forwards, the other backwards, and without supplying any
electrical power. If a load was then connected, the voltage across the parallel output terminals
would decrease slightly. This would reduce the load on the backward-feeding transformer, but
the forward-feeding transformer would be overloaded.
Even if the asymmetric input voltage settings were to be corrected, the assumption made above
should never have been made. Even when all the voltages are identical, one should never
operate transformers that have different short-circuit voltages with their output sides connected
in parallel, as the load will still be distributed unequally. We now assume that our two example
transformers are both connected to the same voltage source and have the same tap changer
positions or that some other measures have been taken to ensure that the open-circuit terminal
voltages on the LV sides of both transformers are of the same magnitude and have the same
phase relationship. The two secondary windings are now connected in parallel and drive a load
that corresponds to the sum rated output of the two transformers. In this situation, each device
would be working at full capacity (but no more), if, that is, each played its part. But they don‟t!
The one with a short-circuit voltage of 6% will only be loaded to 4/5 of its rated load, the one with
usc = 4% has to deal with 6/5 of its rated load, i.e. with a 20% overload.
But once again, that‟s not quite the whole truth. The size of the transformers (i.e. the ratio of
their rated outputs) also plays a role. Experts generally state that transformers that differ in size
by more than a factor of three should not have their secondary windings connected in parallel.
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 27 von 52
In fact, since 1997, the recommendation is that the size difference should not exceed a factor of
two4. As was already briefly mentioned, the ohmic voltage drop across the winding decreases
as the size of the transformer increases. In a small transformer, the short-circuit voltage usc
contains substantially more uR and slightly less uX (Fig. 35). As the size of the transformer
increases, uR gradually becomes more and more negligible, at least in terms of what we are
considering here. But as Fig. 27, Fig. 29 and Fig. 30 show, the size of the voltage drop in the
transformer depends on whether the device is driving an ohmic or an inductive load. If the
voltage drop in the transformer has only a small ohmic component, it will be reduced more if the
transformer is attached to an inductive load than when it is attached to an ohmic load and vice
versa. So two transformers with different uR/uX ratios may well have exactly the same open-
circuit voltage, but when operating under load there will be a slight difference in their voltages.
Depending on the phase angle of the load, either the reactive voltage or the ohmic voltage will
drop more strongly. So if the transformer is operating under load (i.e. not under open-circuit
conditions) then connecting the secondary windings in parallel will generate a circulating
current. The magnitude and direction of the circulating current will depend on the phase angle of
the load – something that is hard to predict. So connecting a large and a small transformer in
parallel requires the introduction of a safety factor, though this in turn makes the whole
argument somewhat circular as the following example illustrates. There is, it has to be said, little
point in providing a large 630 kVA transformer with a small „assistant transformer‟ with a rated
power of say 63 kVA or 100 kVA. The large transformer should anyway have been dimensioned
so that it itself offers at least that amount of reserve capacity. If an „assistant‟ is required, the
safety factor mentioned above means that the actual capacity required is more like 250 kVA.
But that would satisfy the 3 : 1 ratio rule, making the safety factor superfluous to requirements –
and we‟re back to where we started.
uX = 3.91%

uX = 2.95%

uR = 0.9% uR = 2.7%

Fig. 35: Differing ratios of the active to the reactive voltage drop in a large transformer with a 630 kVA Class C
rating according to HD 428 (left) and a smaller transformer with a 50 kVA Class B rating (right)

In contrast to an electric motor, the most economical operating point for a transformer is well
below its rated load, so it makes sense to design in plenty of reserve capacity during the
planning phase. In the example discussed above, it is worth budgeting for a transformer with a
rating of 1000 or even 1250 kVA – or, better still, introducing system redundancy by including a
pair of 630 kVA transformers, each of which is capable in an emergency of handling the load on
its own. Having reserve capacity also helps to settle the nerves. First, conversion or retrofitting
costs are far lower if there is a need to handle greater loads at some later date. Secondly,
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 28 von 52
power losses are reduced and, thirdly, the voltage drop in the transformer is lower as a result of
using either two 630 kVA devices in parallel or from using a single large device. Indeed this can
be the optimum solution for flicker problems as it gets things right at the start rather than
attempting to deal with the problem by grafting on a solution later.5 Of course, the price for this
improved resistance to flicker in the supply system is a higher short-circuit power. The short-
circuit power rises linearly with the size of the transformer (provided the rated short-circuit
voltage is constant) and is the sum of the short-circuit powers of the individual transformers if
they are connected in parallel. This needs to be considered when configuring the downstream
distribution network – and the term „short-circuit power‟ needs to be used with care. According
to the definition, the short-circuit power is calculated by multiplying the open-circuit (i.e. no-load)
voltage by the short-circuit current. The operating states „no-load‟ and „short-circuit‟ are however
mutually exclusive. Short-circuit power is a purely fictional computational parameter, but never-
theless one that is useful in estimating what could happen in the event of a short circuit.
There is one further and very obvious condition for operating transformers in parallel: the vector
group codes of the units to be connected in parallel must be the same. The important aspect
here is that the digits following the letter codes are identical. If they were not the same, one
would end up connecting windings with different phase relationships – a situation that is ob-
viously unacceptable. For instance, two transformers with the vector group codes Dd0 and
YNyn0 can certainly be operated in parallel. But if the neutral point is loaded, the single-phase
or non-linear load will be borne by only one of the transformers, as the other does not have a
neutral point. That needs to be taken into consideration. Additionally, the short-circuit voltages
on the rating plates refer to symmetric, linear, three-phase loads. If another type of load is con-
nected, quite different values will apply with the size of the deviation depending strongly on the
vector groups involved.6 By this point, things can have begun to get quite confusing. Connecting
transformers with different vector groups in parallel is certainly not to be recommended, and
should only really be seen as an emergency measure.
When we talk here about operating transformers in parallel, we normally mean that the output
sides are connected in parallel, as the input sides are usually connected either directly or in-
directly in parallel. It is possible, however, that a distribution transformer is fed from different MV
systems, which themselves are fed from the same HV system but via HV transformers with dif-
fering vector group codes so that they have different phase relationships. In this case, the sum
of the vector group code digits for the HV transformer and the downstream distribution trans-
former must be the same for each HV transformer so that the voltages at the secondary
windings have the same phase.
But as so often, that‟s not quite the whole truth. Vector group codes are not the only significant
elements. We also need to take the power transmission networks into account. The case
capacitance (i.e. the capacitance per unit length)7 of underground cable is very large, while its
series inductance (i.e. inductance per unit length) is rather small. In overhead power lines, on
the other hand, the capacitance is smaller and the inductance is greater. Phase relationships
will therefore vary depending on the specific load conditions. Let‟s assume that we have two
distribution transformers that are connected in parallel on their output sides. They have the
same rated outputs, the same vector groups, the same short-circuit voltages, the same output
voltages and even approximately the same copper losses – in other words, they are ideally
suited to be operated in parallel. One of the transformers is fed via a relatively long underground
MV cable, the other via a relatively long overhead MV line. As we assume that the voltages
supplied at the start of the two cables have the same phase and the modulus of the impedance
is similar for the two cables, we do not expect there to be any appreciable imbalance in the
distribution of the common load. However, the phases of the voltages arriving at the input sides
of the two parallel distribution transformers (i.e. at the ends of the MV supply cables) can indeed
be different. The phase relationships measured at the LV bushings are therefore also different
and if the bushings are connected in parallel a circulating current will begin to flow and the two
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 29 von 52
transformers will appear to heat up even under open circuit (no-load) conditions. But only ap-
parently, as the transformers are not actually under no-load conditions. In fact, they drive the
circulating current. The power losses are almost purely reactive in nature, with the exception of
the ohmic losses in the two transformers as well as in their connection cables on the high-
voltage and low-voltage sides as far as the coupling points.
The underground cable also represents a substantial capacitive load for the HV feeder trans-
former and, as already described, this can cause an increase in the output voltage (Fig. 33). In
contrast, the capacitance of the overhead line is probably small enough to be compensated (at
least at full load) by the leakage inductance of the distribution transformer that is being fed. The
voltages can therefore differ not only in terms of their absolute magnitudes but also with respect
to phase. The question of whether this difference could become critical is something that has to
be calculated for a wide range of load cases during the planning phase. That itself is not as
simple as it perhaps sounds, given the large number of transformer and supply system para-
meters that need to be taken into account.
Summary:
Conditions for operating transformers in parallel:
 Same voltage across the windings to be connected in parallel
 Same rated short-circuit voltages
 Same vector group codes
 Ensure supply networks have the same phase relations
 If the transformers are not connected in parallel on the input side, ensure that the supply
networks have approximately the same short-circuit power levels
 Maximum size ratio of transformers operated in parallel: 3 : 1.

4 Efficiency
In 1999, the Swiss journal „Bulletin SEV/VSE‟8 carried a cover story entitled „Replacing old
transformers pays off‟9. The article showed that as a result of the significant improvements in
the efficiency of modern transformers, there are now sound economic reasons (in addition to
important environmental arguments) why older transformers should be decommissioned even
when they are still functioning properly. In this section, we explain how these efficiency improve-
ments have been achieved and their current and future significance for those responsible for
purchasing and deploying transformers.

4.1 New standard governing transformer efficiencies


„If only energy saving transformers were used in the EU, around €1.2 billion a year (based on
1999 electricity prices) could be saved from reduced energy consumption.‟10 Unfortunately,
when new products are purchased, preference is not always given to the most economical
technology (viewed over the lifetime of the product), but to the technology with the lowest price
tag. This is all the more unfortunate when the products concerned are transformers, as these
cheaper „suboptimal‟ devices will be used for an average of thirty years and will over the course
of their lifetime continually lose ground to the best available technology as further efficiency im-
provements are made. A study conducted for the EU and supported by the European Copper
Institute, KEMA11, a Dutch electricity utility company, a Belgian transformer manufacturer and a
British energy agency estimated that about 22 TWh per year could be saved in the European
Union through the use of energy-efficient distribution transformers. This corresponds to an
annual reduction in carbon dioxide emissions of 9 million tonnes, equivalent to 4% of the
European Union‟s Kyoto targets.12
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 30 von 52
Load losses
rel. Cast
Oil-immersed transformer
Power short- resin
rating circuit List DK List CK List BK List AK HD538
volt. ≤24kV ≤24kV ≤36kV ≤24kV ≤36kV ≤24kV ≤36kV ≤12kV
SN uk PK PK PK PK PK PK PK PK
50kVA 4% 1350W 1100W 1450W 875W 1250W 750W 1050W
100kVA 4% 2150W 1750W 2350W 1475W 1950W 1250W 1650W 2000W
160kVA 4% 3100W 2350W 3350W 2000W 2550W 1700W 2150W 2700W
250kVA 4% 4200W 3250W 4250W 2750W 3500W 2350W 3000W 3500W
315kVA 4% 5000W 3900W 3250W 2800W
400kVA 4% 6000W 4600W 6200W 3850W 4900W 3250W 4150W 4900W
500kVA 4% 7200W 5500W 4600W 3900W
630kVA 4% 8400W 6500W 8800W 5400W 6500W 4600W 5500W 7300W
630kVA 6% 8700W 6750W 5600W 4800W 7600W
800kVA 6% 10500W 8400W 10500W 7000W 8400W 6000W 7000W
1000kVA 6% 13000W 10500W 13000W 9000W 10500W 7600W 8900W 10000W
1250kVA 6% 16000W 13500W 16000W 11000W 13500W 9500W 11500W
1600kVA 6% 20000W 17000W 19200W 14000W 17000W 12000W 14500W 14000W
2000kVA 6% 26000W 21000W 24000W 18000W 21000W 15000W 18000W
2500kVA 6% 32000W 26500W 29400W 22000W 26500W 18500W 22500W 21000W
Table 3: Load losses in standardized distribution transformers

No-load losses

Oil-immersed transformer Oil

List E0 List D0 List C0 List B0 List A0 List C036


≤24kV ≤24kV ≤24kV ≤24kV ≤24kV ≤36kV
P0 Noise P0 Noise P0 Noise P0 Noise P0 Noise
190W 55dB(A) 145W 50dB(A) 125W 47dB(A) 110W 42dB(A) 90W 39dB(A)
320W 59dB(A) 260W 54dB(A) 210W 49dB(A) 180W 44dB(A) 145W 41dB(A)
460W 62dB(A) 375W 57dB(A) 300W 52dB(A) 260W 47dB(A) 210W 44dB(A)
650W 65dB(A) 530W 60dB(A) 425W 55dB(A) 360W 50dB(A) 300W 47dB(A)
770W 67dB(A) 630W 61dB(A) 520W 57dB(A) 440W 52dB(A) 360W 49dB(A)
930W 68dB(A) 750W 63dB(A) 610W 58dB(A) 520W 53dB(A) 430W 50dB(A)
1100W 69dB(A) 880W 64dB(A) 720W 59dB(A) 610W 54dB(A) 510W 51dB(A)
1300W 70dB(A) 1030W 65dB(A) 860W 60dB(A) 730W 55dB(A) 600W 52dB(A)
1200W 70dB(A) 940W 65dB(A) 800W 60dB(A) 680W 55dB(A) 560W 52dB(A)
1400W 71dB(A) 1150W 66dB(A) 930W 61dB(A) 800W 56dB(A) 650W 53dB(A)
1700W 73dB(A) 1400W 68dB(A) 1100W 63dB(A) 940W 58dB(A) 770W 55dB(A)
2100W 74dB(A) 1750W 69dB(A) 1350W 64dB(A) 1150W 59dB(A) 950W 56dB(A)
2600W 76dB(A) 2200W 71dB(A) 1700W 66dB(A) 1450W 61dB(A) 1200W 58dB(A)
3100W 78dB(A) 2700W 73dB(A) 2100W 68dB(A) 1800W 63dB(A) 1450W 60dB(A)
3500W 81dB(A) 3200W 76dB(A) 2500W 71dB(A) 2150W 66dB(A) 1750W 63dB(A)
Table 4: No-load losses and noise levels in standardized distribution transformers up to 24 kV

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 31 von 52


No-load losses
Cast
Oil-immersed transformer
resin
List C036 List B036 List A036 HD538
≤36kV ≤36kV ≤36kV ≤12kV
P0 Noise P0 Noise P0 Noise P0
230W 52dB(A) 190W 52dB(A) 160W 50dB(A)
380W 56dB(A) 320W 56dB(A) 270W 54dB(A) 440W
520W 59dB(A) 460W 59dB(A) 390W 57dB(A) 610W
780W 62dB(A) 650W 62dB(A) 550W 60dB(A) 820W

1120W 65dB(A) 930W 65dB(A) 790W 63dB(A) 1150W

1450W 67dB(A) 1300W 67dB(A) 1100W 65dB(A) 1500W


1370W
1700W 68dB(A) 1450W 68dB(A) 1300W 66dB(A)
2000W 68dB(A) 1700W 68dB(A) 1450W 67dB(A) 2000W
2400W 70dB(A) 2100W 70dB(A) 1750W 68dB(A)
2800W 71dB(A) 2600W 71dB(A) 2200W 69dB(A) 2800W
3400W 73dB(A) 3150W 73dB(A) 2700W 71dB(A)
4100W 76dB(A) 3800W 76dB(A) 3200W 73dB(A) 4300W
Table 5: No-load losses and noise levels in standardized distribution transformers in the 24 kV to 36 kV range and
in cast-resin transformers up to 12 kV

99,4% Efficiency with max. P(Fe), min. P(Cu)


η 

Efficiency with max. P(Cu), min. P(Fe)


99,3%
99,2%
99,1%
99,0%
98,9%
98,8%
98,7%
98,6%
Degree of loading 
98,5%
0% 25% 50% 75% 100% 125%

Fig. 37: The operational characteristics of the trans-


former depend on whether one minimises the no-load
(iron, Fe) losses or the load (copper, Cu) losses, as
shown here in a comparison of the A0DK and E0AK
classes for a 1000 kVA transformer using data from
Table 3 and Table 4

Fig. 36: Reduced noise levels and improved efficiency


go hand in hand

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 32 von 52


Looked at superficially, the data above might seem to suggest that there is no real need to do
anything, as the efficiencies of distribution transformers are not only very high, they are also
standardized. Choosing the most economical transformer would seem to involve little more than
selecting a transformer with a class CC' rating in the European Harmonization Document 428.
But has the most economical transformer really been selected? In many cases it hasn‟t. HD 428
specifies three energy efficiency classes, from the least efficient AA' class to the (at that time)
most efficient class CC'. Today, however, it is certainly possible to manufacture transformers of
even greater efficiency (which we will refer to here as „DD' ‟ transformers ) and whose use would
certainly make economic sense. In fact, transformers of this type have been manufactured and
used in Switzerland for some time. In another example, a Belgian company has designed and is
marketing a low-noise transformer suitable for use in residential areas. This transformer also
happens to exhibit particularly low loss levels (see fig. 36). In the light of these new develop-
ments, the HD 428 document is soon to be replaced by a European standard, though the
number of the new standard has yet to be decided. What has been decided, however, is the
classification scheme to be adopted in the new standard. The new standard will introduce four
main changes compared to the HD 428 efficiency classification scheme:
 While most of the existing classification limits are retained, the new standard will introduce
one higher efficiency class with regard to load (copper) losses and two new higher efficiency
classes for no-load (iron) losses.
 Limits for sound pressure levels will be assigned to the no-load losses.
 Load losses will be specified for two groups: transformers up to 24 kV HV and transformers
not exceeding 36 kV HV. At higher voltages, the wire used for the coil windings is thinner,
while the wire insulation, interlayer insulation and flashover distances (clearances) are all
greater. There is therefore less room for copper.
 The order of the energy efficiency classes is to be reversed, so that class A will in future
represent the most efficient design. However, this change in nomenclature has the drawback
that the series can no longer be extended to accommodate future improvements in trans-
former efficiency. With the letter A already representing the currently most efficient class of
transformers, there is now no obvious means of designating transformers of even higher
efficiency that may well be developed in future. And it seems highly unlikely that there will be
a need to extend the A, B, C, … classification series to accommodate lower efficiency trans-
formers. This change does though bring the classification system used for transformers into
line with numerous other energy efficiency classification schemes, such as those used for
electric motors, domestic appliances, all of which use A as the upper (i.e. most energy
efficient) end of the scale. So in order to retain this classification sequence while at the same
time accommodating future technological developments, the values that define each energy
efficiency class will need to be revised accordingly so that class A always represents the best
available technology at any specific time.
As was the case with the lists in the harmonization documents, each load loss list can in
principle be combined with every no-load loss list. The relative weighting given to load losses
and no-load losses in the design of a transformer can therefore determine whether the trans-
former has more conductor material in the coil windings and less core material or vice versa.
The design choices made will also affect the transformer‟s operational behaviour, particularly its
losses. For instance, optimum efficiency can be achieved at a load factor of 24% or at 47% de-
pending on the design (see fig. 37). When compared at constant current density, a transformer
with more conductor material will exhibit greater load losses, or „copper losses‟ as they are also
known. Strictly speaking, a more accurate trivial name for these losses would be „aluminium
losses‟, as the losses in an aluminium conductor are 35% greater than those in a copper con-
ductor of identical cross-section. But the term „copper losses‟ is unlikely to change, as it reflects
the fact that copper is, historically, the standard conductor material used in transformers.

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 33 von 52


If the magnetic flux density, frequency and iron quality are held constant, the no-load losses in a
transformer (also known as „core losses‟ or „iron losses‟) depend only on the amount of iron
used in the core. Similarly, if the current density is held constant, then, roughly speaking, the
copper losses will depend only on the amount of copper used. On the other hand, iron losses
can be reduced by increasing the number of windings on the core and thus reducing the
magnetic flux density (induction). In contrast, copper losses can be reduced by operating at a
higher flux density and using fewer windings on the core – but this can only be realised within
strict limits, as high-quality magnetic materials have quite sharp magnetic saturation points and
most conventionally designed transformers operate close to this limit. The primary means of
reducing copper losses is to lower the current density, while maintaining the number of turns
and the core cross-section and modifying the core in such a way that the winding window is
larger and thicker wire can then be used for the windings.
A transformer that spends most its life operating under no load or minimal load conditions
should therefore be designed to minimise the no-load losses, i.e. less iron and more copper. It
would however be wrong to conclude from this that any transformer designed for permanent full-
load operation (something that only really occurs in generator transformers in power stations
and in certain industrial applications) should contain as little copper as possible. In this case, the
preferred approach is to maximise the cross-section of the iron core in order to minimise the
number of turns. The cross-section of the conducting wires should also be as large as possible
in transformers running under continuous full load.
Unfortunately, splitting a 1250-kVA transformer into two units each with a power rating of
630 kVA, as suggested earlier in the section on operational characteristics, results in a slight
rise in all losses irrespective of the size of the load. Nevertheless, the beneficial redundancy
achieved means that this sort of splitting is frequently performed in practice. With two smaller
transformers there is also the option of switching off one of the transformers during light-load
periods and thus reducing the losses during these periods to below the level that would be in-
curred if a single larger transformer were used. Of course, both sides of the transformer have to
be disconnected from the power supply. If only one side is disconnected, the transformer re-
mains excited and no-load losses continue to be incurred.

4.2 Driving up costs by buying cheap


If the power required from a conventional small transformer is a little below what is in principle
physically obtainable from a transformer of that size, one often finds that the coil formers are not
fully wound – almost as if achieving a minimum temperature rise is an essential part of the de-
sign. In fact, this problem is not restricted to conventional small transformers, it is also common
in the type and size of transformers considered here. And this is completely consistent with the
view, widely held by both transformer manufacturers and users alike, that a transformer be-
longing to a higher thermal insulation class is better than one in a lower insulation class. Trans-
formers in a higher thermal insulation class are by definition allowed to get hotter and, as a rule,
they generally do heat up more than those with a lower thermal insulation rating. And because
they get hotter, the manufacturers employ (expensive) insulation materials. The temperature is
higher, the losses are higher and the transformer generates correspondingly more heat in the
electrical system to which it is connected. The best that can be said is that the transformer is a
little smaller in size. But that is its sole advantage and even that benefit can be eaten up be-
cause the higher temperatures generated mean that other components have to be kept further
away from it. Furthermore, the higher losses often require more effort and expense to be spent
on managing heat dissipation. The voltage drop is also larger. It is essential that transformer
professionals rid themselves of the ludicrous notion, which, when expressed provocatively,
might be phrased: “my transformer is better than yours, because it‟s hotter.” If „progress‟ means
stepping up from class H (with a continuous duty temperature rating of 180°C) to class C
(220°C) then the heat really will be on.
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 34 von 52
But how did this mindset arise? Price pressure is the usual reason cited, with „price‟ meaning
only the purchase price and all follow-up costs conveniently ignored. But that would imply that
market forces should have eliminated any manufacturer offering more expensive devices. There
are however a number of transformer manufacturers who have decided to put quality first, and
who are thriving as a result, despite the fact that the first thing the customer is interested in is
the price. The key here is to explain carefully to the customer why the coil former is always fully
wound, why these transformers are usually somewhat more expensive and occasionally
significantly more expensive than apparently equivalent competitor products, and why using the
cheapest transformer usually leads to the most expensive system overall. The commercial
success of these companies13 is clear proof that they are indeed managing to convince their
clients that the purchase price of a transformer is not its most important feature.
One manufacturer uses only grain-oriented sheet steel in the cores of all its transformers with a
power rating above about 1 kVA. This may well be due to the fact that the company uses some
of these (auto)transformers in one of its other business units to manufacture energy manage-
ment units14 as well as special DC link converters. And an article on the latter reiterates what
was said above about the danger of focusing only on the price: „Over the last few years, cost
and space considerations have led to an expansion of the thermal insulation classes up to class
H (180°C) – a development that has brought with it numerous disadvantages…. 15‟. Apart from
the fact that – strictly speaking – it should have read „…price and space considerations…‟ (as
„price‟ and „cost‟ are not the same), the article underscores the company‟s praiseworthy attitude
to the question of transformer efficiency.

4.3 An example

Fig. 39: …similar to the one considered in the analysis


Fig. 38: The coil of a single-phase transformer, both
below.
the primary and secondary sides have been con-
figured as multilayer windings…

In the following illustrative example, we study a 40 kVA single-phase dry-type industrial trans-
former in order to demonstrate just how strongly transformer losses depend on the specific
transformer design, and just how quickly the extra investment in a higher-quality device can be
recovered. The reason we focus here on an industrial application is because there is a greater
need for action in the industrial sector than in the public electricity supply network. Eight dif-
ferent transformer variants were computed and quoted to the customer (Table 6). Version 0 was
the cheapest and most basic variant. Note that we have avoided the use of such popular but
often misleading euphemisms like „most economical‟ or „most cost-effective‟ to describe this
basic version of the transformer. It is neither; it is simply the version with the lowest purchase
price. The rectangular windings were designed in such a way that cooling ducts were required
between all layers and on all sides. Starting from this basic version, loss-reduction measures
were then progressively designed into the following seven variants, of which the first six simply

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 35 von 52


involved using conductor wire of successively larger cross-section. One might think that this
would result in a larger sized transformer, but in fact the converse is true. Although the trans-
former gets heavier on moving from version 0 to 7, it also gets smaller rather than bigger. While
the thicker wire obviously takes up some of the space occupied by the air cooling ducts, as the
wire becomes thicker, the need for a cooling duct disappears. The first ducts that can be dis-
pensed with are those between the layers of a winding on the long sides of the transformer
(similar to the coil shown in Fig. 38 and Fig. 39). Then, as the diameter of the wire used in-
creases, it becomes possible to do without the cooling ducts located on the long sides between
the two coils and finally to eliminate the ducts on the end faces. In version 7, the winding data
(wire diameters and number of turns) are identical to those in version 6, but the core uses grain-
oriented rather than hot-rolled sheet steel. Although moving from version 6 to 7 only involves a
change in the core material, the copper losses also decrease due to the higher magnetizability
of the grain oriented steel. The stack height of the core in version 7 is lower and the average
length of a turn is therefore shorter.
Channels / Design Measures & Weights Calculated Electrical Values Pay-
Price back
Ver- betw. core & betw. LV & stack
in LV winding in HV winding width length mFe mCu mtot Pv Fe Pv Cu Pvtot U time
sion LV winding HV winding height

front long front long front long front long


[mm] [mm] [mm] [kg] [kg] [kg] [W] [W] [W] [V] [%] [€] [a]
[mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm]

0 10 10 10 10 10 10 10 10 100 450 360 202 30.7 232.7 417 1634 2051 13 95.1% 877 ---
1 10 10 10 0 10 10 10 0 100 415 365 196 42.4 238.4 406 1343 1749 11 95.8% 932 0.944
2 10 10 0 0 10 10 10 0 100 417 342 196 46.6 242.6 406 1217 1623 10 96.1% 946 0.839
3 10 10 0 0 10 0 10 0 100 400 342 196 48.2 244.2 406 1090 1496 9 96.4% 955 0.723
4 10 10 0 0 10 10 0 0 100 406 340 196 59.9 255.9 406 874 1280 6 96.9% 1027 1.004
5 10 10 0 0 0 0 0 0 100 408 335 196 65.9 261.9 406 753 1159 5 97.2% 1062 1.072
6 As in 5, but with even thicker wire 100 412 341 196 71.3 267.3 406 626 1032 4 97.5% 1100 1.133
7 As in 5, but with grain-oriented steel, lower stack height 80 412 311 155 64.7 219.7 223 580 803 4 98.0% 1249 1.541

Table 6: Improving a cheap transformer (version 0) in seven steps

The effects of the stepwise introduction of loss-reducing measures are not perhaps immediately
apparent in the data in Table 6 and for this reason have been presented graphically in Fig. 40.
Two results stand out straight away:
 Viewed across the series of improvements, the losses decrease significantly faster than the
price rises.
 The payback period is nearly always under 1.5 years, only the version with the higher quality
sheet steel core needs a little longer. The calculations assumed an electricity price of 10
cents per kWh, 242 working days per year and a single eight-hour shift per day. If there are
two shifts a day, the payback periods are halved.
There is one other beneficial technical „side effect‟ from the loss reduction measures: The
voltage drop in the transformer decreases as one moves from version 0 to 7. This is not always
advantageous, especially in transformers larger than the one considered here, where a defined
voltage drop is highly desirable. In large transformers, cooling and electrical insulation require-
ments mean that it is not possible to do without the cooling ducts. But the example transformer
discussed here is relatively small, and both the input and output sides are in the low-voltage
range. Furthermore, a small voltage drop (both ohmic and inductive) was advantageous from
the point of view of the particular industrial process under consideration.
Asked one year later about which of the eight variants the customer finally chose, the manu-
facturer becomes a little embarrassed: „I really ought not to say. The customer went for the
cheapest product. But not only that, he also got the loads wrong. So one by one, the trans-
formers are now burning out.‟ Any one of the improved transformer variants 5, 6 or 7 would

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 36 von 52


have had sufficient reserve capacity to cope with the erroneous load specifications and to
prevent transformer failure.
2500W 98.5% 1,400 € 1.8a
98.0% 1,200 € 1.6a
2000W 1.4a
97.5%
1,000 €
1.2a
1500W 97.0%
800 € 1.0a
96.5%
600 € 0.8a
1000W 96.0%
0.6a
400 €
Pvtot 95.5% Price 0.4a
500W
Efficiency 95.0% 200 € Payback 0.2a
0W 94.5% 0€ 0.0a
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7

14V 500mm 300kg


Voltage drop no load / rated load Weights and measures
12V 250kg
400mm
10V
200kg
8V 300mm
width length 150kg
6V 200mm mtot mCu
100kg
4V
100mm
2V 50kg

0V 0mm 0kg
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Fig. 40: Graphical presentation of the data in Table 6

But there is light at the end of the tunnel. A number of manufacturers of resin-encapsulated
distributor transformers still aim to save their customers a few euros by doing without a couple
of extra kilograms of copper (in the high-voltage winding) and aluminium (in the lower-voltage
winding) and by employing forced ventilation to cool the windings. Other manufacturers, how-
ever, are wary of the risk of failure associated with these mechanical ventilation systems and
have consciously decided to avoid this approach wherever possible. Naturally air-cooled cast-
resin transformers are now available with power ratings of up to 6.3 MVA. At these sorts of
powers, highly efficient insulating materials are essential – unlike their use in smaller trans-
formers, as discussed earlier, where they serve to mask avoidable energy wastage. Other
manufacturers use fans when the transformer has a power rating of around 1 MVA, though they
are only activated in emergencies when the transformer is overloaded; when the transformer
operates at its rated load and at normal ambient temperatures, the fans are not needed. This is
a sensible approach, as deploying a large transformer in order to cope with a few hours of
emergency loading makes neither economic nor environmental sense. In fact, averaged over
the year, losses can be higher if an oversized transformer is running for long periods under
capacity (see Fig. 37).

4.4 Amorphous steel


In terms of energy conservation, any transformer operating at low loads for large periods of its
service life should ideally have minimal iron losses. Whereas load losses depend on the square
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 37 von 52
of the current and will drop to a quarter of their nominal value when the transformer operates at
half load, the no-load (i.e. „core‟ or „iron‟) losses depend on the voltage and the frequency. As
both the voltage and the frequency normally remain constant, the iron losses also remain at
their maximum value for as long as the transformer is operating, irrespective of whether it is
connected to a large load, a small load or no load. Hence the term „no-load losses‟. Despite the
fact that no-load losses are almost an order of magnitude smaller than the load losses (compare
Table 4 to Table 5), the former are far more important from a practical point of view.
We should even perhaps revise our earlier statement that no-load losses are particularly
significant at low loads, as they remain (almost) unchanged and certainly do not disappear
when the load and therefore the associated load losses increase. Once a transformer has been
installed, the load and thus the load losses can be controlled by demand side management
methods, but there is no way to influence the no-load losses. Nevertheless, if a new transformer
is to be chosen or configured for a particular application, it makes good sense to look at the
potential loss ratios (Fig. 37).

Fig. 41: An amorphous steel core (photo: Pauwels), Fig. 42: Coil assembly with amorphous core (photo:
shown here in a five-leged design normally only seen Pauwels)
in high-power transformers

No-load losses can be reduced by lowering the magnetic flux density and by using special core
steels. The thinner the sheet steel is, the smaller the extent of eddy current formation. Eddy
currents are completely absent in core materials that do not conduct electricity (so-called
ferrites), but these are reserved for radio-frequency applications as their magnetizability is too
low for transformers operating at grid frequencies. Amorphous steel is a new type of core
material that offers a compromise between sufficiently high magnetizability and significantly
reduced core losses. Amorphous steel is made by atomizing the liquid metal and spraying it
onto a rotating roller where it is quenched extremely quickly, so rapidly in fact that it cannot
crystallise and remains in a disordered amorphous state, hence the name. While the resulting
core material has a saturation magnetization of at most 1.3 T compared to the 1.75 T exhibited
by modern cold-rolled grain-oriented steels, the no-load losses in a transformer with an
amorphous steel core are around 60% lower (see Fig. 41 and Fig. 42). As the saturation flux
density of the core material is lower, these transformers tend to be larger and heavier and
correspondingly more expensive. The transformer with an amorphous steel core is also about
12 dB louder. Despite these disadvantages, there are factors in favour of amorphous core trans-
formers. Studies in Belgium, Great Britain and Ireland have shown that the expected payback
periods can be as short as three to five years. With an expected service life of 30 years, these
transformers would therefore pay for themselves six to ten times over. But the market does not
appear fully ready for amorphous core transformers. One company in Germany tried its luck.16
Extensive studies were also conducted in Belgium and Ireland and a number of amorphous core
transformers were sold.17 But then came the liberalization of the European electricity markets

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 38 von 52


and electricity prices began to fall. In spite of what economic logic dictates, it seems that once
again only prices and not costs play a role in the free market.

4.5 Transformers used in renewable energy generation systems


Every wind turbine contains a transformer that steps-up the generator voltage from, typically,
690 V so that the power generated can be fed into the medium-voltage collection grid.
When transformers are manufactured or purchased for a specific application it is usual that a
so-called „loss evaluation‟ is carried out in accordance with established formulae used by both
transformer manufacturers and electricity companies. The loss evaluation analysis determines
the specific level of capital investment that is economically justified in yielding a unit reduction in
the transformer‟s no-load or load losses. In the case of large transformers, the loss evaluation
formulae are sometimes used to compute either contractual penalties, in which the purchase
price is reduced if the contractually agreed loss levels are exceeded in practice, or contractual
bonuses, in which an additional payment is made if the losses turn out to be lower than
contractually agreed. In the case of a transformer used in a wind turbine, the no-load loss
evaluation factor was calculated to have the relatively (but justifiably) high value of 9.28 €/W.
The load loss evaluation factor was computed using standard methods to be 0.79 €/W. That
value would be low even for a conventional transformer and in the present case it led to the
conclusion that the current CC' transformer (to be known in future as a C0BK transformer)
would still be economical, and that that the lifecycle costs of the (more efficient) Swiss „DD' ‟
transformer would be higher. However, a recently developed probabilistic method18 that
computes the expected losses for around 4000 different operational states of the transformer
yielded a load loss factor of 1.43 €/W. At that level, the Swiss transformer discussed earlier (to
which we gave the hypothetical „DD' ‟ rating and that will in future belong to class B0AK) or even
an A0AK transformer would be the device of choice. But how did this discrepancy in the load
loss factors arise? The usual method of calculation assumes an average load level, and this is
low as windless conditions are not uncommon, and when the wind does blow it is rarely so
strong that the turbine operates at full load. However, using the average load introduces errors
into the calculation as load losses vary quadratically and not linearly with the load. For instance,
if the rated copper losses are 8 kW, they are only 2 kW if the transformer is running at half load.
One hour at full load and one hour at standstill result in load losses of 8 kWh; two hours running
at half load generate load losses of only 4 kWh. The difference is analogous to that between the
real root-mean-square value and the rectified mean value of alternating currents, and the
difference is greater the more discontinuous the load profile is – and renewable sources of
energy are typically significantly more discontinuous than conventional sources.
In the case of a wind turbine, the blades are frequently motionless or rotating at slow speed
without generating electrical power. Theoretically, the turbine is then generating negative power,
as it needs to consume power to keep the control and monitoring systems active. Although the
energy required by these systems is minimal compared the situation in a coal-fired power
station, where about 7% of the gross generator output is consumed for the power station‟s own
use, it still needs to be drawn from the grid. This means that the transformer has to stay
operational even though it is essentially under no-load conditions. No-load losses thus remain
relevant for 8760 hours per year, irrespective of how long the wind turbine is operating at full,
partial or no load.
In Germany, a further issue has to be taken into consideration. The German Renewable Energy
Sources Act (EEG) specifies that owners of sources of renewable energy shall receive
payments for power fed in to the grid, and these payments are between twice and three times
the market price. This doubles or triples the value – or at least the price – of any energy losses.
If no-load losses and load losses were to be economically reassessed in the light of this, wind
farms would need very different types of transformers than those typically used in the public
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 39 von 52
electricity supply network – in fact, this would be an ideal application for an amorphous core
transformer.

4.6 Other countries, other customs


The reason why Ireland became the test bed for amorphous core transformers was not simply
because the manufacturer had a production facility there, but also because grid losses in 1980
were almost 12% – an embarrassingly high level of loss by European standards. Although the
reason losses were so high had more to do with Ireland‟s sparse population (and lower grid
density) than with poor grid management, the perceived need to act was perhaps greater than
elsewhere. Today losses are below 10%, although this improvement cannot be ascribed purely
to a couple of pilot projects involving amorphous core transformers.
Most European countries have losses of between 6 and 10% in their electricity distribution
networks. In Germany, grid losses are a very creditable 4.6%. The transformers used in the
German grid are selected, irrespective of the voltage level, so that they operate at between 30
and 60% of capacity during a typical daily load profile. That is the optimal region to operate in
(Fig. 36) – but even then it still leaves a theoretical energy saving potential of 4.6%. The only
country with lower grid losses is Luxembourg with 2%, but in such a small, densely populated
country, in which no transmission line is longer than 20 km, achieving low losses is perhaps not
such a challenge as it is elsewhere. Generally speaking, half of the energy saving potential is to
be found in the transformers, predominantly in the distribution transformers.
The situation in other more distant countries is considerably worse. Reports from India claim
that distribution transformers are regularly operating at 50 to 100% overload levels. As a result,
the failure rate is an incredible 25% a year. Quoting the February 2002 issue of the „Bulletin on
Energy Efficiency‟, the official journal of the IREDA (Indian Renewable Energy Developmental
Agency), the Indian Copper Development Centre ICDC reports that less than 50% of the
electricity consumed in India is actually billed. The rest is lost through transmission and
distribution losses (about 18%), or from „illegal use‟ or poor management, because the
electricity companies simply do not install meters.
Back in Europe, there are those who believe that the introduction of higher voltage levels would
also help to reduce the amount of electrical energy consumed illegally – referred to
euphemistically as „non-technical losses‟. Transformers as anti-theft devices? It has even been
suggested that the introduction of the lower frequency of 16.7 Hz for the railway traction power
grids used in Germany, Switzerland, Austria and a number of Scandinavian countries also
served the same purpose. The truth, however, is that the lower frequency helps to reduce
commutation problems affecting the ac commutating series motors (universal motors) that
power the traction units.

4.7 Outlook
Whereas switched-mode power supplies have replaced small transformers in numerous
applications, there is currently no sign that conventional grid transformers will be replaced by
any other technology in the near future. There are suggestions that the extra high voltage level
220 kV will disappear over the long term and that at some even later date the 380 kV level will
be replaced by a DC network.19 Such developments would dispense with the need for at least
some of the transformers required today. From the point of view of energy loss, however, these
changes would be essentially neutral, as the inverters and the requisite interference
suppression filters would also generate losses of a similar magnitude.
Attempts have also been made to develop low-loss transformers with superconducting coils.20
Unfortunately, these conductors are only really loss-free when conducting direct current, and the
iron losses can even rise if the core is also cooled. A further problem is that cooling power has
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 40 von 52
to be supplied continuously at its maximum required level, thus increasing the no-load losses,
despite the fact that in practice, the transformer hardly ever runs at full load, and when it does,
then only for a short time. When all these factors are taken into account, overall loss reductions
turn out to be minimal. The one place where superconducting transformers can be used
effectively is in railway vehicles. Once these transformers go into industrial production they will
save not only weight (and therefore extra energy), but also space. Weight and space limitations
in railway vehicles also mean that the transformers currently in use in railway vehicles are
working at their design limits and are thus significantly less efficient than comparable grid trans-
formers.

5 Special solutions for special loads


Transformer losses, specifically no-load and load losses, have already been mentioned on
several occasions. In what follows we will be focusing particularly on those load losses that
arise when a transformer is subjected to loads other than those for which it was designed – a
situation that is not uncommon in today‟s networks. Such loads require careful analysis.

5.1 Bad loads


While overloading a transformer can obviously be problematic, „bad loads‟ can be worse. But
what do mean by „bad load‟. If a transformer is fed – as envisaged – with a sinusoidal voltage
and subjected to a sinusoidal current, it is relatively straightforward to analyse the losses that
occur. The primary-side voltage generates a slight magnetization or no-load current and thus a
certain degree of no-load loss („iron loss‟) due to the eddy currents that cannot be completely
suppressed in the transformer‟s core. The load current causes ohmic loss („copper loss‟) in each
winding. The Joule heating due to the no-load current in the primary winding is negligible. As
the voltage and frequency are constant, so too is the iron loss. Additional power is lost when
stray magnetic fields induce eddy currents in electrically conducting structural components, this
is particularly the case with ferromagnetic materials that attract the stray magnetic fields. This
loss is also constant under no-load conditions and is in fact treated as part of the transformer‟s
no-load loss. The capacitive load, which we looked at earlier, is the only traditional type of „bad
load‟ as it generates a „negative voltage drop‟ in the transformer and reinforces the no-load loss
and the load voltage.
However, under no-load conditions these stray magnetic fields are not particularly strong. A
second type of stray magnetic field appears when the transformer operates under load. This
field stems from the main leakage channel, i.e. the gap between the coils, and permeates the
outermost coil, but not by the inner coil. As was described in section 3.1, the gap between the
coils is needed in distribution and larger transformers for:
1. insulation,
2. cooling,
3. limiting the short-circuit current.
The intensity of the leakage field is directly proportional to the magnitude of the load current and
induces (in proportion to its strength) what one could call an „eddy voltage‟ in those conducting
components permeated by the field. It is this voltage that drives the eddy currents in such
components. In the conductors, whose conductivity is almost an order of magnitude higher than
that of the structural steel components, an additional circulating current flows in a plane vertical
to the direction of the main current and whose magnitude is greater the thicker the conductor.
Assuming that the temperature is constant, Ohm‟s law applies and this transverse current is
directly proportional to the „eddy voltage‟. This part of the eddy current loss therefore also varies
as the square of the load current and is normally treated as a 5 - 10% supplement that is added
to the load loss calculated from the currents and winding resistances. Normally meaning here „at
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 41 von 52
the nominal frequency stated on the transformer‟s rating plate‟. Now as is known, an induced
voltage is proportional to the rate of change of the magnetic flux density, i.e. to the peak value
and the frequency of the excitation field. The eddy current loss therefore grows as the square of
the current and the square of the frequency.
If several currents of different frequencies (including direct currents) share a common con-
ductor, the total root mean-square (rms) current in the conductor is calculated by summing the
squares of the individual values and then taking the square root of the result. Let‟s first take a
look at the following simplified example:
A transformer with an eddy-current loss of 10% of its copper loss is subjected to the following
load (and no other): 70.7% of its rated current at the 50 Hz fundamental and 70.7% of its rated
current at the 150 Hz third harmonic. The total current is then:

I  ( 0.707I N )2 ( 0.707I N )2  I N .
The transformer is therefore operating with its rated current. The contribution of the fundamental
frequency to the eddy-current losses (also known as „supplementary load losses‟ PSupp) is given
by:
PSupp  0.7072 PSuppN  0.5PSuppN .

The fundamental with an amplitude of 70.7% of the rated current therefore creates eddy-current
losses that are 50% of the eddy-current power when the device is operating at its rated load,
that is, 5% of the nominal copper losses. The third harmonic current (150 Hz) appearing on the
output side, which is fed in as a third-harmonic on the input side, causes an eddy current power
loss of
150 Hz 2
PSupp  0.707 2PSuppN( )  4.5PSuppN .
50 Hz
So to recap. The total rms current is exactly equal to the rated current. One might conclude from
this that the transformer will not become overloaded. However, the eddy-current losses
generated jointly by both these components of the total current are some five times greater than
the corresponding losses the transformer would generate at its rated load (50 Hz sinusoidal) –
losses that would normally be regarded simply as a minor contribution to the transformer‟s
copper losses. This effect has in the past led to overheating whose cause was not immediately
apparent.
Converter transformers have been around for a long time and these „additional‟ supplementary
load losses resulting from harmonic distortion effects are taken into account when dimensioning
the device. The devices are also designed to reduce the size of these losses. The size of the
eddy currents can, where necessary, be reduced by splitting up the thick conductors into
numerous mutually insulated individual wires (similar to method used in high-frequency coils
wound with litz wire), or by increasing the distances between certain mechanical components
and the magnetic leakage field, or by using components made of magnetically or electrically
non-conducting materials. One such material is, perhaps surprisingly, stainless steel which is
not ferromagnetic – in contrast to conventional structural steel. Furthermore, the conductivity of
stainless steel, and therefore its ability to conduct eddy currents, is about half that of con-
ventional steel.
Now the sole use of this type of converter transformer is to feed a single power converter,
whose harmonic spectrum is known beforehand. In order to avoid overheating in a commercially
available transformer that will operate in the presence of harmonics, say in a high rise office
block, there are a number of factors that can be used to calculate the loading capacity of the
transformer relative to its rated load. In North America, the so-called K-factor was introduced for

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 42 von 52


this purpose. As just explained, the K-factor expresses the magnitude of the eddy current loss
relative to the losses associated with a purely sinusoidal load:
n
K   I n2 n 2 ,
1

where n is the harmonic order and In is the associated current expressed as a fraction of the
rated current, as in the numerical example discussed above. But this number is only of any use
if the terms of reference are known, i.e. if we know what fraction of the transformer‟s losses are
made up by the eddy-current losses PSupp. Normally, though, the only losses (if any) stated on
the rating plate are the iron losses P0 and the copper losses PCu.
In Europe, it is not the „K factor‟, but „factor K‟ that is computed, in accordance with
harmonization document HD 538.3.S1:
0.5
 2 
e  I h  n  N  q  I n  
2
K  1      n    ,
 1  e  I  n2 
  1  
I

where
0.5
 n N 
0.5 n  N  I  2 
I    I n 2   I1    n   .
 n 1   n 1  I1  
 

Fig. 43: Current and harmonic spectrum of an 11-watt compact fluorescent lamp (CFL). Top: CFL connected to a
transformer driving an average load in a residential area. Bottom: Simulation of a transformer operating at its rated
current and driving only CFLs

That‟s probably enough to frighten anyone off. If an electrical engineer or technician working on
real practical problems puts in the hard work and succeeds in correctly applying these exact
formulae, he or she is still left with the question of just what these results are actually saying and
how exact they really are. As the transformer system being planned does not actually exist, the
output values fed into the equations above can only be based on assumptions, guesswork or
experience. Clearly neither prior measurement on the system nor prior questioning of the sub-
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 43 von 52
sequent user is possible. Many buildings today are planned and built and only then does the
owner seek tenants or buyers. Consequently, the harmonic profile of the power supply system is
unknown at the time of planning. It seems that we should adopt a more practical approach to
the problem.
If a conventional transformer is to be used, a safe conservative estimate would put the supp-
lementary load losses at 10% of the load loss specified for 50 Hz operation. Having made this
assumption, let us now conduct a thought experiment in which we fully load a transformer with a
load made up purely of energy saver lamps (Fig. 43).
The rms values of the individual harmonic components (from 1st to 51st order) of the supply
voltage and the lamp current were read off the display of the measurement instrument (Table 7).
The bottom row shows the total rms values, which were computed as the square root of the sum
of the squares of the individual values. For this 11-watt CFL, the apparent power input is
230.7 V * 64.8 mA  15.0 VA .

For the sake of simplicity, let us now assume Analysis of harmonics in an 11-W
that the three-phase supply transformer has Osram Dulux CFL with serial
a rated power of 15 kVA and is driving a load
made up solely of 1000 of these 11-watt impedance R =29.1W & X L=113W
CFLs, distributed symmetrically across the U U² IL I L² P ad /P Cu
three phase conductors. Strictly, of course, if n V V² mA mA²
we want to meet this requirement, we would 1 230.2 52992.0 48.5 2352.3 5.6%
have to assume a load of 999 lamps and a 3 8.3 68.9 37.1 1376.4 29.5%
rated transformer power of 14.985 kVA – but 5 10.7 114.5 20.3 412.1 24.5%
this is only an illustrative example and we 7 4.3 18.5 5.3 28.1 3.3%
can tolerate this minor inaccuracy. As it is not 9 1.1 1.2 3.0 9.0 1.7%
so easy to actually find a transformer of the 11 2.3 5.3 3.8 14.4 4.2%
13 1.0 1.0 1.5 2.3 0.9%
right type and load it with 1000 energy saver
15 0.6 0.4 1.5 2.3 1.2%
lamps, we chose to load it with a single lamp
17 1.1 1.2 1.5 2.3 1.5%
and then connect 1000 times the short-circuit 19 0.5 0.3 0.9 0.8 0.7%
resistance Rsc and the short-circuit im- 21 0.5 0.3 1.3 1.7 1.8%
pedance Xsc, as shown in Fig. 43. It would 23 0.6 0.4 0.8 0.6 0.8%
seem that with this load the transformer is 25 0.4 0.2 0.6 0.4 0.5%
operating at its maximum capacity, without 27 0.6 0.4 0.8 0.6 1.1%
being overloaded. But that is only how things 29 0.4 0.2 0.5 0.3 0.5%
seem, if you ignore the eddy-current losses, 31 0.3 0.1 0.5 0.3 0.6%
which as already stated grow with the square 33 0.3 0.1 0.5 0.3 0.6%
of the current and the square of the 35 0.3 0.1 0.4 0.2 0.5%
frequency. The third harmonic, for instance, 37 0.3 0.1 0.4 0.2 0.5%
39 0.3 0.1 0.3 0.1 0.3%
generates an eddy-current loss that is 29.5%
41 0.1 0.0 0.3 0.1 0.4%
of the copper loss quoted for the transformer
43 0.2 0.0 0.2 0.0 0.2%
at its rated load. The fundamental, in 45 0.1 0.0 0.2 0.0 0.2%
contrast, generates an eddy-current loss that 47 0.1 0.0 0.2 0.0 0.2%
is only 5.6% of PCu. The fundamental is 49 0.1 0.0 0.1 0.0 0.1%
significantly smaller than the total rms value 51 0.1 0.0 0.1 0.0 0.1%
(itself equal to the rated current) and is the Supplementary losss: P Supp /P Cu = 81.4%
reason for our earlier assumption that the
Table 7: Values measured for a typical compact
eddy-current loss is about 10% of PCu.
fluorescent lamp and the effect on the power supply
transformer, on the assumption that the transformer is
operating at its rated current and these CFLs are its
only load

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 44 von 52


If one now adds all the contributions from the fundamental to the 51st order, it is apparent that
in such cases of high harmonic distortion, the eddy-current losses in the transformer are not the
10% of the copper loss assumed on the basis of a sinusoidal rated current, but are a massive
81.4%!

5.2 Practical measures


So what should a technician or engineer facing a practical problem do, if, as is likely, the
formulae discussed above were of no real value? Start by memorizing, writing down or copying
the following values and make sure they are available when you need them.
Leakage loss makes up no more than 10% at most of a transformer‟s load losses. The figure for
modern transformers, and converter transformers in particular, is more likely to be 5% or less.
This reduces the „additional‟ supplementary load losses caused by harmonic distortion to about
half, i.e. a good 40% of the copper loss. It is very rare that a transformer has to drive a load
made up only of compact fluorescent lamps or similar devices, such as switched-mode power
supplies, though such a situation is conceivable for the power supply system of a computer
centre. But let‟s stay with this extreme case in which the supplementary load losses make up a
good 80% rather than the 10% of copper losses typically assumed. That is, at the rated
operating point we have 70% more load losses than originally calculated. In order to reduce the
copper losses (including the eddy-current and supplementary load losses) and thus the as-
sociated joule heating effect back to a level that corresponds to 100% at 50 Hz, the rms load
current used to select the transformer must be multiplied by the following factor in order to take
account of the quadratic dependence between the current and the heat generated:
170%
K  1.3
100%
This means that the power of the transformer should be approximately 30% greater than that
calculated on the basis of the apparent power requirement Urms * Irms, as the remainder of the
power loss, the no-load loss, is essentially unaffected by the presence of harmonic currents.
The no-load loss only increases when the excitation voltage is non-sinusoidal, but in most
cases, the voltage harmonics are far smaller than the current harmonics. Later on, however, we
will take a look at a counter example.
Safety factors are not affected by any of this and they must remain part of the planning process.
If the planning process has taken into account all relevant factors except harmonics (re-
dundancy, reserve capacity to accommodate future load growth, etc.) and if it identifies a
requirement of 1000 kVA, then the planner should (based on our arguments above) choose a
transformer with a rated capacity of 1250 kVA. Over the long-term this also provides benefits for
the customer, as a distribution transformer is at its most economical when operating in the
range 24 - 47% of its rated power depending on the exact configuration of the transformer (see
section 4).
In addition, the neutral point must also be capable of handling 173% of the phase conductor
current. In most cases, this condition represents the most stringent criterion to be fulfilled. In
distribution transformers we normally assume that the neutral loading capacity is 100% and only
then, if we have a single or double-phase load, i.e. one of the limbs remains unloaded. If we
actually do have a full, three-phase load, it is assumed that the return currents in the neutral
conductor will mutually annihilate each other. This, however, is actually only true for harmonics
of order 3n (so-called triple-n harmonics) and for the fundamental. Because they all have the
same phase, triple-n harmonics sum linearly in the neutral conductor. The squares of their
amplitudes add to the squares of the amplitudes of the other frequencies flowing in the neutral
conductor to generate the total rms return current, as shown in Table 7.

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 45 von 52


Under certain conditions, power supply faults can mutually cancel each other out. It is
characteristic of the sort of single-phase loads shown in Fig. 43 that they only draw current from
the power supply when the voltage is close to its peak value, with the result that the current
flows in short bursts of large magnitude. We should mention that in a three-phase system, it is
the phase-to-neutral voltage that we are talking about. The voltage between L2 and L3, for
example, is 90° out-of-phase with the voltage between L1 and N. As a result the voltage
between L2 and L3 passes through zero when the voltage between L1 and N is reaching its
peak value (which is higher as a result of the „bad load‟). While the voltage that drives this
current collapses near its peak value as a result of the network impedances, the distortion of the
voltage between the phase conductors shows the opposite picture (Fig. 44). By selecting the
transformer vector group, a phase-to-neutral voltage (coil voltage) on the output side can be
generated either by a phase-to-neutral voltage or a phase-to-phase voltage on the input side,
depending on the particular vector group chosen. That is, in principle, the same as if one were
to use a single-phase transformer to transform the voltage shown in the lower part of Fig. 44
from 400 V down to 230 V and then use this voltage for part (ideally half) of those loads that
originally caused the voltage distortion evident in the upper part of Fig. 44. Such an arrange-
ment would almost completely eliminate the voltage distortion. In other words: transformers with
different vector groups could make a substantial contribution to clearing up distortion in power
supply networks, provided they were deployed in the right mixture.

Fig. 44: Measured and simulated phase-to-neutral voltage and phase-to-phase voltage in a residential area during
the final of the Football World Cup (Germany against Brazil)

Such a method has in fact been suggested by Professor Fender and others (Fig. 45) – and was
realized in the early years of electrical engineering at the Catholic University of Leuven
(Louvain) in Belgium as described earlier in the section on vector groups. In addition to the
advantage mentioned there, we can in the case of non-liner loads being considered here, add
the following benefit: If every second one of the usual Dyn5 transformers were to be replaced by
a corresponding transformer with a Dzn6 or Yzn6 vector group, then half of the single-phase
rectifier loads would (from the perspective of the MV supply) draw their current „hump‟ with a
30° phase lag relative to the other half. Or put another way: some of the harmonics released
into the power supply network by single-phase loads would arrive at the MV network with a
phase difference of 180° (relative to their higher frequencies) and would cancel each other out.
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 46 von 52
The simulation shown on the right in Fig. 44 makes this point clear. The middle section of Fig.
44 also shows the phase angle of the fifth harmonic and one can see that the fifth harmonic is
responsible for most of the distortion in both the upper and the lower cases illustrated. But as
these two fifth-order harmonics have opposite polarity, if one can put it that way, they would
cancel each other out if half of each voltage were to arise in the same circuit (Fig. 45).
So why doesn‟t anyone make use of this method? Once again, it‟s because people confuse
„costs‟ with „price‟ and a transformer with a zigzag winding has a purchase price that is about
5% or perhaps even 10% higher. However, the costs that arise do not appear explicitly on any
one sheet of paper and are anyway paid by another cost centre. That is the real reason. How-
ever, the reason usually put forward is that it would not be possible to connect the transformers
in parallel. But this would anyway be impossible in the sort of installations where this type of
configuration makes sense, namely in high-rise office buildings with their own LV supply net-
work. The one vector group in one building, the other vector group in the next or next-but-one
building, that‟s all that would be needed. No one is suggesting that transformers with different
vector group codes should be installed side-by-side in the same building.

Fig. 45: Using transformers with different vector groups leads to a phase shift in the current ‘hump’ on the MV level
(Source: Prof. Fender, Wiesbaden University of Applied Science)

But the usual Dyn5 transformers also play their part in cleaning up the power supply system, by
splitting the one current „hump‟ per half-cycle and per phase conductor as generated by the
single-phase rectifier loads into two „humps‟ on two phase conductors on the input side (one
leading by 60°, the other lagging by 60°) and located either side of the peak voltage (see Fig.
45b). A further effect is that due to the triple-n harmonics, which are practically short-circuited in
Dyn5 transformers. These harmonics flow in-phase from the consumer side towards the trans-
former returning via the neutral point as if all three phase conductors were connected in parallel.
These three currents demand an in-phase reverse flow of charge in the high-voltage winding,
which they get by inducing a circulating current in the delta-connected HV winding. The only im-
pedances they will come across are the leakage reactance and the rather small winding re-
sistances. But this only works if the current on the low-voltage side really does arrive from all
three phase conductors at the same time. The reason is essentially the same as that given in
section 3.5 in the discussion of neutral loading capacity. It is for this reason alone that the usual
power supply voltages contain only small fractions of the third-order harmonic, even though this
is the predominant order in the current. If it ever proved possible to gather all the distorting loads

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 47 von 52


on one phase conductor and all the linear loads on the other two, then the effect would dis-
appear. To demonstrate the effect (Fig. 46) one can connect a two-kilowatt electric heater via a
rectifier and a smoothing capacitor to a conventional domestic power outlet (caution: the power
drawn increases to over 3 kW!).
So by combining this intrinsic ability of the Dyn5 transformer to clean up the power supply with
the 30° phase shift from the transformer with another vector group, we can almost restore the
beauty of the original sinusoidal curve (sinus: lat. bosom). However, effectively eliminating the
voltage harmonics from the supply network comes at a price. Restoring the quality of the
voltage waveform results in a stronger flow of current harmonics within and through the trans-
formers. The circulating current induced in the delta winding in particular generates additional
heat loss. Such losses also contribute to the „additional‟ supplementary load losses. An ex-
periment carried out with a small transformer (Fig. 47) illustrates just how quickly this additional
loss channel can attain a significant size. The series resistance of the delta-connected
secondary winding is 0.1 ohm. A THD of only 3.2% in the primary voltage results in a circulating
current of 2.3 A (Fig. 48). The resulting I² * R loss is therefore about 0.5 W. Half a watt is about
1% of the total copper losses and doesn‟t actually sound that bad. If the voltage THD rises to
6.4%, which can occur in practice, the joule heating loss would increase to 4% of the total
copper losses or 4.6 A, which in this case would correspond to 28% of the rated current. The
transformer load would therefore have to be reduced by 28% solely in order to prevent over-
heating of the secondary winding by the 150-Hz circulating current. The 30% overdimensioning
introduced earlier to cope with the supplementary load losses caused by eddy currents would
still need to be taken into account separately.

Fig. 46: The asymmetric distribution of rectifier loads – in this case 2.5 kW on a single phase conductor – results in
a third harmonic of the voltage waveform that is larger than the fifth harmonic (see lower middle panel)

The zero-sequence system, that is, the homopolar components of the input currents in the
three-phase transformer, can be identified by means of the three no-load currents (Fig. 49). All
three currents show simultaneously a pronounced peak of the same polarity and this occurs six
times per cycle, despite the fact that the voltages driving these currents are each phase shifted
by 120° relative to each other. If one disconnects the delta-connected high-voltage winding and

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 48 von 52


thus interrupts the current in Fig. 48, the phenomenon described above disappears and the
three currents become essentially independent of one another.
One final remark on operating transformers in parallel. If you run two transformers with differing
short-circuit voltages in parallel, you will find that the harmonic currents are distributed highly
asymmetrically, in fact more asymmetrically than the current‟s fundamental component. The
reason is that the leakage reactance is greater for higher frequencies. So another good reason
not to connect transformers in parallel. A transformer with a rated short-circuit voltage of 4% has
a short-circuit voltage of almost 12% at 150 Hz, at 250 Hz it will have increased further to nearly
20% because the inner voltage drop is predominantly inductive. Obviously, if the rated short-
circuit voltage is 6%, these figures will be correspondingly higher. The only case where this
does not apply is for Dyn-type transformers and triple-n harmonics due to the formation of
circulating currents.
But let‟s take one more look at the voltage. There are a few certain situations in which the
voltage can be so distorted that it has a detrimental effect on the performance of a transformer
that it is driving. For example, it is an „inherent characteristic‟ of small UPS systems (in other
words: the alternatives are too expensive) that when power loss occurs, they generate square-
wave rather than a sinusoidal voltage. However, a non-stepped square-wave will have a form
factor that is 11% smaller than that of a sinusoidal waveform. The 11% is the factor linking the
mean value and the rms value. The quoted value is always the rms value or at least it should
be. But the degree of magnetization depends on the mean value. The right rms value at the
output side of a small UPS can cause significant overexcitation of the transformer to which it is
connected. In addition, the harmonic distortion of a square wave is so high that very substantial
no-load losses must be expected.

Fig. 47: Experimental set-up for Fig. 46

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 49 von 52


Fig. 48: A THD of only 3.2% in the primary voltage drives a circulating current in the delta-connected secondary
winding whose size is about 14% of the rated current

Fig. 49: No-load currents in a small three-phase transformer with a YNd11 vector group

As we have seen, transformers don‟t just sit around humming all day long, they can be very
beneficial in helping to clean up low-quality power supply networks. That is not to say, of course,
that they are not sometimes the source of interference themselves: computer monitors in a
ground-floor office may well flicker because of noise emissions or alternating magnetic fields
from a transformer installed in the basement immediately below. But the transformer is often not
the guilty party. The stray fields generated by transformers are typically not as large as they are
often assumed to be. In many cases, the faults stem from cable runs in an ill-conceived network
configuration that permits currents originally planned for the neutral conductor to escape to all
other conducting structural components. If this problem has been rectified, the only approach
left is to increase the distance between the source of the disturbance and any potentially sus-
ceptible equipment and to minimize the distance between the outward and return current paths.
It may also prove helpful in such situations to deploy a special transformer design in which the
bushings are arranged in a rectangular pattern near the base of the unit rather than in a row on
DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 50 von 52
the lid as is the case in conventional designs (Fig. 50 and Fig. 51). This sort of engineering
solution will not be found described in any industrial standard. Indeed the manufacturer claims
in an advert that: „When we build transformers, the first thing we focus on is complying with our
customers‟ requirements and only then on complying with standards. Why? Because it is
customer needs and not standards that offer real scope for product innovation.‟ This is
something that we have already seen in connection with transformer efficiency – an area in
which standardization is hardly outstanding. It has to be said, however, that much has been
done in the mean time to improve the standards. But the processes involved still take far too
long as the standards are increasingly subjected to international harmonization – a trend that in
itself is to be welcomed. But until the standards have finally caught up, industrial companies will
need to focus on technical creativity and on communication with their customers.

Fig. 51: Magnetic flux density as a function of the


horizontal distance from the centre of the transformer
measured two metres above the transformer’s lid
(Source: Rauscher & Stoecklin)

Fig. 50: EMC transformer (Source: Rauscher &


Stoecklin)

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 51 von 52


1
Fassbinder, Stefan: Netzstörungen durch passive und aktive Bauelemente, VDE Verlag, Berlin / Offenbach 2002,
p. 188
2
Bolliger, R.: Wenn die Lichter flackern [‘When the lights flicker’]. ET Schweizer Zeitschrift für angewandte
Elektrotechnik 11/2003, p. 26
3
Cf. amendment of EN 50174-2 (VDE 0800 Part 174-2) from January 2002 für the September 2001 edition
4
www.a-eberle.de/pdf/info_12.pdf
5
See [1], p. 51
6
Fender, Manfred: Vergleichende Untersuchungen der Netzrückwirkungen von Umrichtern mit Zwischenkreis bei
Beachtung realer industrieller Anschlußstrukturen [‘Comparative studies of the effects of converters with
intermediate circuits on power quality in real industrial installations], Ph.D. thesis, Wiesbaden 1997
7
Fassbinder, Stefan: „Erdkabel kontra Freileitung‟ [‘Underground vs overhead power transmission cables’], in de,
vol. 9/2001, p. gig9, appears in DKI reprint s180 „Drehstrom, Gleichstrom, Supraleitung – Energie-Übertragung
heute und morgen‟ [‘Three-phase AC, DC and superconducting systems – Power transmission now and in the
future’] from the German Copper Institute (DKI), Düsseldorf
8
SEV/Electrosuisse: Swiss Association for Electrical Engineering, Power and Information Technologies
VSE: Association of Swiss Electricity Utility Companies
9
Borer Edi: „Ersatz von Transformatoren-Veteranen macht sich bezahlt‟ [‘Replacing old transformers does pay’], in
Bulletin SEV/VSE, vol. 4/1999, p. 31
10
European Copper Institute: „The scope for energy saving in the EU through the use of energy-efficient electricity
distribution transformers‟, Brussels 1999, www.eurocopper.org
11
www.kema.nl
12
„Verlustminimierte Trafos können EU helfen‟ [‘Loss-minimized transformers can help the EU’], in etz, vol. 9/2000
13
www.riedel-trafobau.de, www.buerkle-schoeck.de
14
Decker, Christiane: „Energie sparen mit EMU‟ [‘Saving energy with EMUs’], in de, vol. 15-16/2000, p. 34
15
Bürkle, Thomas: „Wassergekühlte Zweipunkt-Zwischenkreisdrosseln‟ [‘Water-cooled two-point DC link reactors’],
in etz, vol. 22/2000, p. 18
16
www.marxtrafo.de
17
www.power-technology.com/contractors/switchgear/pauwels
18
www.efficient-transformers.org
19
Fassbinder, Stefan: „Hochspannungs-Gleichstrom-Übertragung (HGÜ)‟ [‘High-voltage DC power transmission’],
in de, vol. 11/2001, p. gig9, appears in DKI reprint s180 „Drehstrom, Gleichstrom, Supraleitung – Energie-
Übertragung heute und morgen‟ [‘Three-phase AC, DC and superconducting systems – Power transmission now
and in the future’] from the German Copper Institute (DKI), Düsseldorf
20
Fassbinder, Stefan: „Supraleitung – ein Teil zukünftiger Energieversorgung?‟ [‘Superconductivity – What part will
it play in energy supplies of the future?], in de, vol. 9/2001, p. 38, appears in DKI reprint s180 „Drehstrom,
Gleichstrom, Supraleitung – Energie-Übertragung heute und morgen‟ [‘Three-phase AC, DC and superconducting
systems – Power transmission now and in the future’] from the German Copper Institute (DKI), Düsseldorf

DistributionTransformers.docx Stand: 31.08.2009 10:33:00 Seite 52 von 52

S-ar putea să vă placă și