JEEETRANSACTIONS ON MAG"ICS, VOL. 28, N0.

2, MARCH 1992 1355

CALCULATING STRAY LOSSES IN POWER TRANSFORMERS
USING SURFACE IMPEDANCE WITH FINITE ELEMENTS
S.A. Holland G.P. O'Connell L. Haydock
Brush Transformers Ltd Department of Electrical Engineering
PO Box 20, Loughborough, Not tin gham Polytechnic, Burton Street,
Leics., England, LE 11 1HN. Nottingham, England, NG1 4BU.

Abstract -The paper describes the results obtained

from a rarely published method of calculating eddy 11. TEACHING COMPANIES
losses in thick conducting materials. The Surface
Impedance formulation is used in a finite element A Teaching Company is a government sponsored
software package which is applied, in turn, to the partnership between industry and an academic
design of large power transformers. The institution in the pursuit of technology transfer.
transformer and model details are discussed as are Haydock et a1 131 have recently discussed these types
the excellent results obtained by employing these
techniques. The paper then discusses the further of arrangements in more detail.
application of these methods and their
implications for transformer design. 111. SKIN DEPTH OF TRANSFORMER FITTINGS AND
TANK.

I. INTRODUCTION The method of surface impedance is particularly

applicable to transformer design since a considerable
loss is due to eddies induced in steel fittings, clamps
With the increased pressure on manufacturers to and the tank wall. Equation (1) shows the generally
produce lower loss power plant for reasons of energy accepted expression for the skin depth in a
conservation and environmental considerations as conductor, 6 is the skin depth, 6.1 is the angular
well as the obvious money savings, designers are frequency, p is the permeability and G is the
having to turn to more reliable methods of calculating conductivity of the material. For non-saturated mild
the losses caused by stray field effects. steel at 50Hz the skin depth is of the order of 0.5".

Until recently many companies have relied mainly

on empirical calculations, either derived .....(1)
experimentally from older designs or from papers
dating from as far back as the early part of the Therefore, considering the large size of power
century to determine the stray losses on power transformers and the appreciable thickness of tank
transformers [l]. For the majority of standard walls it can be assumed that the loss is on the
designs these have served as good gauges of eddy surface.
induced loss but for designs of special transformers
or units with very tight specification tolerances these
have often proved to be inadequate. This has often IV. SURFACE IMPEDANCE AND CLASSICAL
led, after testing, to expensive modifications with the METHODS
changes being made without the full knowledge of
the effects of the alterations. The origins of the formulation are from the
Poynting Vector and applications of this method have
It is to this end that Brush Transformers Ltd, been used in transformer design for many years [4].
within the Hawker Siddeley group, have employed At the surface of good conductors the tangential
finite element and other modern techniques [2] in component of the electric field (ET) is approximately
order to improve their product design and efficiency. proportional to the tangential component of magnetic
The work has come out of a "Teaching Company field (HT) and thus;
Scheme" with Nottingham Polytechnic.
E ~ = z ( ~ ) [ Hx
T n] .......(2)
Manuscript received July7,1991

0018-9464/92$03.00 0 1992 IEEE

1356
where Z(m) is the complex surface impedance of the on the tank wall is found by integrating the specific
metal where it is equal to losses over the inside surface of the tank.

l
e ........(3)
where the symbols are the same as in equation (1)
V. SURFACE IMPEDANCE WITH FINITE ELEMENTS

The calculation of eddy currents in 3-D finite

elements is computationally expensive as well as
except (T'is 0-jwe which is the specific admittance, being difficult to build models where thin skin depths
and E is the dielectric permittivity. This leads to the occur. A power transformer has dimensions in
use of (4) which yields the surface loss per unit area metres and the skin depths are in millimetres and
of the conductor; using solid elements would cause errors due to the
poor aspect ratio of these elements. Consider a large
transformer of tank dimensions 8m long, 6m high
........(4) and 5m depth. To accurately calculate the fields
required up to 60,000 solid elements may have to be
where used. Then consider putting on the tank wall with
solid elements and modelling a number of skin
depths. The size of the model would increase rapidly
........( 5 ) as well as solution time.

which is termed the surface resistance rind is It is in these types of cases that surface impedance
simplified since displacement current is neglected modelling has been developed [5]. Using the simple
compared with conduction current. formulation of reduced scalar potential as an energy
function from [6] as in (6) can be used with the
The problem is now one of calculating the simple addition of the surface impedance terms as in
tangential component of magnetic field at the tank (7). The letters t i and t2 are the tangential directions
wall of a transformer. There are many ways that this to the sur.f;iceelements.
is done analytically, some methods use a basic
formula to calculate it at a point half way up the wall
and use geometric factors for the rest of the surface
[4]. For clamps and other internal fittings this is a
very difficult calculation to perform because of the
complicated geometries involved. Most methods are and
based on curved tanks as in Fig (1)

Three Phase Windings

Vi. EXAMPLE TRANSFORMER

Work has been continuing at Brush Transformers

Tank Wall \ Ltd for some time now, analysing a whole range of
With Curvcd Sidcs Yokc transformer problems with the application of finite
element computations . The first example presented
demonstrates the advantages of using surface
Fig. 1. Plan View of a Power Transformer with impedance with finite elements.
Curved Tank Walls.
The following example is a 90MVA, 3-phase,
and difficulties arise when straight tank walls are 132/33kV transformer mounted in a straight sided
used because some of the symmetry is lost. The loss mild steel tank. In this unit tank wall mounted
magnetic shunts were employed in order to reduce
 1357
the stray losses. These shunts made of plates of Pcffcctive = CL
transformer core steel, stacked to around 1Omm S.F. + p(1-S.F.) ....(8)
thick, and mounted close to the transformer wall.
They are arranged so that they shield the wall from Non-linearity in the tank wall and fittings is
the leakage field produced by the unit when on load. accounted for by the application of Aggarwal's
The shunts have a much lower loss characteristic than approximation, which is detailed in [5]. The B-H
the mild steel of the tank as well as a higher curve of the mild steel is not used in any way,
permeability and so reduce the stray loss. Table I although the application of this type of modelling
shows results from tests performed on the would probably improve the accuracy of the analysis.
transformer.
TABLE I
TEST DETAILS (@NORMAL TAP) FOR YOMVA VII. RESULTS FROM ANALYSIS
TRANSFORMER
The following details the results of the tank wall
loss and impedance calculations.
TABLE I1
In Tank Out of Tank Difference RESULTS OF FINITE ELEMENT ANALYSIS FOR
No Shunts 51.68 14.38 37.30 YOMVA TRANSFORMER
With Shunts 29.78 15.40
Tank Wall Loss (no shunts) 35.5kW

The load loss test, a reduced voltage short circuit Tank Wall Loss (with shunts) 14.8kW
test, is performed by short circuiting the secondary
winding and applying impedance volts, ie; the Impedance 25.5%
voltage necessary to cause full load current to flow.
The impedance voltage establishes the leakage flux The loss is calculated for each element and so the
which in turn causes these stray losses. In this total loss is simply the integral of the elemental
transformer the impedance voltage was 25.46% losses. A designer may examine contour plots of the
loss on the surface of the tank wall to check for hot
spots which may cause oil failure due to overheating.
VII. MODEL DETAILS
These results correlate very well with those
The transformer was modelled using 28,962 obtained from actual tests performed in the factory.
surface and tetrahedral brick elements, the surface The impedance calculation is the application of;
elements being used for the tank wall. The 3-D
solver [7] uses only the reduced scalar potential L=NO/I .............(9)
formulation since the excitation of the coils is
performed using magnetic shells. The solver may where N is the number of turns in a winding, I is the
also model the non-linear characteristics of the core exciting current and 0 is the average flux in the
steel, but since there was no mutual flux then the winding.
permeability of the steel was very high and so a linear
study was used where the relative permeability was
chosen to be 28,000 which is typical of non-saturated IX. FURTHER WORK WITH STRAY LOSSES AND
transformer core steel, which was in this case 23M3 STEEL FITTINGS
grade. One important facility that the solver does
have is that the permeability of the steel across the An appreciable amount of loss is experienced by
laminations is taken into account by a stacking factor. the clamps and other fittings in the transformer. The
Equation (8) shows the calculation of the clamps generally take two forms. The yoke/core
permeability of the core steel in the direction across clamps hold the core together and are .often made of
the laminations. In this case the stacking f.dctor was mild steel plates that are placed along the length of the
chosen to be 0.95. yoke of the transformer core. Bolts and tie rods hold
these clamps together and so can significantly
contribute to stray loss production since they form a
1358
circuit with many parallel paths. Insulators are often SOOMVA unit. These units are more difficult to
used to break up these paths. The other type of model since they often employ shaped core steel
clamps are the winding clamps which are often used packets at the end of the windings to redirect the
in smaller units. These often take the form of senii- leakage flux straight back into the transformer core
circular plates of steel at the top and bottom of the thus reducing the tank wall loss.
windings, the top clamps having jacks to create the
compression required. Below are listed two more The procedure used to model the whole of a
units, having these types of devices, analysed using transfoImer is to model it as a solid model and then to
surface impedance with finite elements. cover the steel components with surface elements and
delete.the solid elements inside. This decreases the
TABLE I11 size of the model and hence reduces computation
COMPARISON OF TESTED AND CALCULATED time. There are limitations to the method.
RESULTS FOR TWO FURTHER UNITS Aluminium components cannot be modelled using
surfaces since the skin depth is too thick.
l'ransformer
Details (Loss) Tested Calculated
X. CONCLUSION
T1. (16MVA,33/1lkV) 40.33kW '38.2OkW
T2. (20MVA,66/1IkV) 39.9lkW 40.88kW The results obtained suggest that the surface
elernent/inipedancemethod is an excellent one for the
applications described. These results were also
appreciably closer to the test results than other more
traditional calculations. The transformer tank wall and
fittings c m be modelled accurately with surface
elements removing the need for complex layers of
brick elements to account for skin effects and this, in
turn, reduces the complexity and size of models.
This method thus yields advantages of reduced
complesity especially important in the analysis of
very large devices such as 3-Phase transformers.

REFERENCES

11 I S. hlamota, "Stray Load Losses in Transformers", Joum.

J;ip;iricsc Inst. Elect. Engnrs., May, 1926, No.454, pp.
5 os - 5 80
121 L. Haydock, S. Holland, W. Hudson, "A Hybrid Finite
Elcnicnt ;mi Distributed Magnetic Equivalent Circuit
h'lodclling Mcthodology for Electromagnetic Devices.",
Atlvanccs in Elcctrical Engineering Software Electrosoft
Conference 1990, pp 43-53.
Fig. 2. Contour Plot of Stray Loss o n a Cut-vcd 131 L. Haydock, S. Holland, G. P. O'Connell, "Teaching
Companies - A Case Study", proc. The 12th E.C.O.O. and
Tank Wall of a SOOMVA Unit 8th I.C.T.E. Confcrcnce, Toronto, 1990.
141 A. A. Rcrczovski et al., "Calculation of Stray Losses in
Accurate results, such as these, require nearly all Transrormcr Tanks", Elcktrichestvo, No.9, pp. 1-7.1966.
steel components to be modelled hence the niodels 151 T. W. Preston: "3-Dimensional Eddy Currents in Turbine-
C;cncrators.", I'hd Thesis, University of London, 1984.
become very complex and large but the surface A. Coulson, M.A., Preston, T.W., and Recce A.B.J., "3-
161
impedance method helps to keep the model size to a Dimcnsionnl Finite Elcmcnt Solvcrs For The Design Of
manageable one. Elcclrical Equipment", IEEE Trans. Mag. V01.2, No.6,
24762479. 198.5.
Other transformers that have been an;ilysed are 17) G.E.C. - Alsthom - S.L.I.M. Software - G.E.C. - Alsthom
large transformers, as in Fig. 2, up to 7SOMVA. Engineering Research Centre, Litchfield Road, Stafford,
U.K.
Fig. 2 shows the contour plot of stray loss for a