International Colloquium Transformer Research and Asset Management

Cavtat, Croatia, November 12 – 14, 2009

Branimir , Ph.D
Kon ar - Distribution and Special Transformers Inc.
branimir.cucic@koncar-dst.hr

MAGNETIC FIELD IN THE VICINITY OF DISTRIBUTION TRANSFORMERS

SUMMARY

A quasi-static analysis of an oil-type distribution transformer was described to determine the

ambient magnetic field in the surrounding region. Windings and terminals were assumed to be primary
sources of the magnetic field. Core and tank of the transformer were modeled by secondary sources
(surface current density, surface charge density and eddy current density). In order to find these sources,
the system of integral equations was transformed into linear system. Magnetic field was calculated by
evaluating volume and surface integrals. The calculated and measured results of the magnetic induction
were compared on a 630 kVA transformer. Evaluation of the magnetic induction was made for distribution
transformers in the range 50-2500 kVA.

Key words: distribution transformer, magnetic field, magnetic induction, integral equations

1. INTRODUCTION

The demand for a regulation of electromagnetic field limits resulted from the public concern about
the electromagnetic emissions. The ICNIRP guidelines [1] for limiting exposure to time-varying, magnetic
and electromagnetic fields have been widely accepted and implemented in many countries. According to
them, magnetic induction reference level at 50 Hz is 500 T for the occupational exposure, and 100 T
for the public exposure. All values are unperturbed rms. However, some countries have prescribed lower
limit values referring to the precautionary principle. Switzerland has limit of 1 T [2], Italy of 3 T, Slovenia
of 10 T, Croatia of 40 T [3], etc. Oil-type distribution transformers are used in urban areas. As a part of
power distribution system, they emit the electromagnetic field. In this paper 1 T and 40 T curves are
determined in the surrounding region of such transformers at 50 Hz. The only source of the magnetic field
is assumed to be transformer itself.
These curves can be determined by measuring or computing [4]. The measurement of the
magnetic field is time consuming. Also, special equipment and the space with no other sources are
required. Therefore a numerical computation technique was presented to evaluate the magnetic field of
distribution transformer.

2. DISTRIBUTION TRANSFORMERS

When talking about oil-type distribution transformers in this paper, we mean on all three-phase
oil-immersed transformers from 50 kVA to 2500 kVA with the highest voltage for equipment not exceeding
36 kV and with rated secondary voltage 400-433 V [5]. The analysis of magnetic field was made for the
typical 630 kVA oil-type transformer (table I).

1
 Table I. Transformer technical data.
Type Oil-immersed, three-phase distribution transformer
Rated power 630 kVA
Rated voltages 20 kV / 420 V
Vector group Dyn5
Impedance voltage 4%
Rated currents in terminals 18.2 A / 866 A
Rated currents in windings 10.5 A / 866 A
Rated number of turns in windings 1732 / 21

Main transformer parts relevant to

the magnetic field calculation are
(figure 1):

windings
low voltage (hereinafter
referred to as "LV")
terminals
high voltage (hereinafter
referred to as "HV")
terminals + feeding
core
tank
clamping system

Figure 1 - Main parts of oil-type distribution transformer

There are two groups of parts; primary sources and ferromagnetic materials.
In the group of primary sources are windings, LV terminals, HV terminals and feeding cables.
Known currents flow through them. They are primary sources of the magnetic field (denoted by P in figure 1).
In the group of ferromagnetic materials are core, tank and clamping system. They were modeled
by secondary sources (denoted by S in figure 1). The clamping system was ignored in the magnetic field
calculation due to memory limitations and due to author's evaluation that its influence on a magnetic field
in the surrounding region of transformer is low. Core and tank were assumed to be linear isotropic
homogeneous materials with two constant parameters; magnetic permeability - and electrical
conductivity - [6]. The values of r and are given in table II.

Table II - Relative permeability and electrical conductivity for core and tank
relative permeability - r electrical conductivity - [MS/m]
core 20000 0
tank 400 6.4

The core was made of classical grain oriented material - M5. Relative permeability of such
material at measured (0.07 T) and rated (1.63 T) magnetic induction is 20000 [6]. It's worth mentioning
that changes of r in the 10000 to 50000 range do not affect the results significantly [6], [7]. Electrical
conductivity of the core was assumed to be zero (no eddy currents) due to lamination of the core [6], [7],
[10]. The values for the tank (construction steel) were obtained experimentally [6], [8].

3. METHOD FOR COMPUTING THE MAGNETIC FIELD

Today's commonly used software based on differential approach (finite element method) is not
appropriate tool for computing the magnetic field of the transformer. The points of interest are relatively
far from the sources (windings, terminals). Domain of such problem is very large which results in huge

2
number of elements. Instead of differential approach, an integral approach was applied to the magnetic
field computation. The method is based on equivalent sources in magnetic empty space ( = 0). These
equivalent sources replace the ferromagnetic material and they were called secondary sources. In a
quasi-static analysis of a system of a known primary source and a linear ferromagnetic material (figure 2),
the secondary sources are the following [6], [9]:

K m - surface current density (A/m) on the surface S of ferromagnetic material

v - surface charge density (C/m2) on the surface S of ferromagnetic material

2
J v - eddy current density (A/m ) inside the volume V of ferromagnetic material

Figure 2 - Equivalent sources in a quasi-static model of linear ferromagnetic material

All time-dependent variables were assumed to be sinusoidal functions of time. Analysis was made in
phasor domain which was denoted by a point above the variable.

K m and J v are vectors, while v is a scalar.

These sources generate the following magnetic field (Biot-Savart's law):

_ _
1 R
primary source H o (r ) J o (r ' ) x dV ' (1)
4 R3
Vo

_ _
1 R
H s (r ) K m (r ' ) x dS' (2)
4 R3
secondary S
sources _
_ _
r (r ) R
(3)
H v (r ) J v (r ' ) x dV '
4 R3
V

where R is the distance between the observation and integration points;

_ _
R r r' (4)

Magnetic field at any point r of the space is

_ _ _ _
H (r ) H o (r ) H s (r ) H v (r ) (5)

Magnetic induction is (space is considered "empty")

_ _
B( r ) H (r ) (6)
0

3
4. LINEAR SYSTEM OF INTEGRAL EQUATIONS

In order to find the magnetic induction, all sources must be known. The primary sources are
known and to find the secondary ones, integral equations are written as follows [6], [9]:

_ _ _ _ _
j 0 r dV ' j 0 dS ' R
J v (r ) J v (r ' ) K m (r ' ) v (r ' ) dS ' j Ao ( r ) (7)
4 R 4 R 4 0 R3
V S S
r' r

_ _ _ _ _ _ _ _ _ _
m r R m R
K m (r ) J v (r ' ) x x n(r ) dV ' K m (r ') x x n(r ) dS ' 2 m H o (r ) x n(r ) (8)
2 R3 2 R3
V S
r' r

_ _ _ _ _ _
_ _ _ _ _ _ _
j 0 r 0 n(r ) j 0 0 n(r ) 1 R n(r )
v (r ) J v (r ' ) dV ' K m (r ' ) dS ' v (r ' ) dS ' 2j 0 Ao (r ) n(r ) (9)
2 R 2 R 2 R3
V S S
r' r

where
-1
2 f 2 50 314.16 s (10)
4 10 7 Vs
0 (11)
Am
12 As
0 8.854 10 (12)
Vm
r 1
m (13)
r 1
_ _
0 dV '
Ao ( r ) J o (r ' ) (14)
4 R
Vo

Equations (7), (8) and (9) can be written as a linear system in matrix form [6]:
A x b (15)

where
Dimensions
A Matrix of the system 3(NV + NS) x 3(NV + NS)
x Vector of unknown secondary sources 3(NV + NS) x 1
b Vector of primary sources 3(NV + NS) x 1

The ferromagnetic material was divided into NV volume elements and NS surface elements.
Secondary sources were assumed to be constant all over the elementary volume and surface.

4
5. MODELS OF TRANSFORMER FOR MAGNETIC FIELD CALCULATION

Three transformer models for magnetic field calculation are shown in figures 4 - 6. Model 1 (figure
4) is the simplest and consists only of primary sources (windings, terminals). Model 2 (figure 5)
additionally takes into account the transformer core. Model 3 (figure 6) is the most complex and consists
of the primary sources, core and the tank.
Division of ferromagnetic materials (core and tank) depends on computer memory size. In this
paper computations were made on a computer with 1.8 GHz CPU and 3.3 GB RAM. As stated before,
eddy currents in the core were disregarded, and the core was divided only on surface elements (1892
elements). The tank was modeled with six plates. Due to memory limitations it was divided very rough
(756 surface and 294 volume elements). Rectangle and cuboid were chosen as the elementary surface
and volume elements, respectively.

Figure 3 - Oil-type distribution transformer Figure 4 - Model of transformer with primary

sources (model 1)

Figure 5 - Model of transformer with primary Figure 6 - Model of transformer with primary
sources and core (model 2) sources, core and tank (model 3)

5
6. COMPARISON BETWEEN CALCULATIONS AND MEASUREMENTS

Magnetic induction was measured and computed on 630 kVA transformer defined in table I. The
transformer was energised from HV side and LV side was in short circuit (figure 3). Rated currents flew
through windings and terminals. According to figures 3 and 7, the origin of coordinate system was set at
the top of the tank cover (hereinafter referred to as "origin of transformer"). The magnetic induction was
measured and computed in three planes (figure 7). Figures 8 - 10 show the comparison between
measured and computed rms values of the magnetic induction at 50 Hz.

Figure 7 - Planes of measurements and Figure 8 - Comparison between calculations and

calculations measurements in XY plane (z=-0.355 m)

Figure 9 - Comparison between calculations and Figure 10 - Comparison between calculations and
measurements in XZ plane (y=0 m) measurements in YZ plane (x=0 m)

Maximum deviations between the calculations and measurements are given in table III.

Table III - Maximum deviations between the measured and computed rms values of magnetic induction
model1 model 2 model 3
(primary sources) (primary sources+core) (primary sources+core+tank)
1 T 50 % 17 % -
40 T 100 % 56 % 18

6
 Rather than maximum deviations, more important information is how far is the most distant point
of the curves. The distances between the most distant points of the curves and the origin of transformer
are given in table IV.

Table IV – Distances between the most distant points of the curves and the origin of 630 kVA oil-type
transformer
XY XZ YZ Max. Deviation between computed
plane plane plane distance and measured max. distances
measured 2.66 m 2.46 m 2.67 m 2.67 m
1 T computed (model 1) 3.20 m 3.22 m 2.90 m 3.22 m 20.6 %
computed (model 2) 2.56 m 2.51 m 2.58 m 2.58 m -3.37 %
measured 0.83 m 0.90 m 0.79 m 0.90 m
computed (model 1) 1.31 m 1.37 m 0.96 m 1.37 m 52.2 %
40 T
computed (model 2) 1.08 m 1.14 m 0.84 m 1.14 m 26.7 %
computed (model 3) 0.90 m 0.94 m 0.87 m 0.94 m 4.44 %

Despite of high deviations for model 1, it can be useful when finding the most distant point of the
magnetic induction curve. Computation on model 1 showed the most distant points of 1 T and 40 T
curves to be 3.22 m and 1.37 m far from the origin of transformer, respectively. Maximum measured
distances were 2.67 m and 0.90 m. That gives 21 % and 52 % deviation between measurements and
computations. Furthermore, the analysis showed the influence of the terminals to be low. The
computation on simplified model 1 (with windings only) resulted in distances 3.19 m and 1.33 m, which
are almost the same as for the model with both windings and terminals. Curves of magnetic induction
computed on model 1 are always farther from transformer than measured curves (figures 8-10).
Therefore we are always on the safe side. The model 1 is recommended for quick magnetic field
computation especially when the points of interest are several meters from the transformer.
Model 2 gives much better results than model 1, but deviation is still high for the points in the
close proximity of transformer. This model is also used for magnetic field computation of dry-type
transformer [7] and cast-resin transformer [10].
To compute the magnetic field in the close proximity of oil-type distribution transformer, model 3
is needed. This model is memory demanding and time consuming. Also, fine division of ferromagnetic
material is necessary. The induced eddy currents distribute in a very thin layer on its surface. Therefore
the tank should be divided into layers which significantly increases the number of elements. In mentioned
hardware configuration (3.3 GB RAM), number of elements was to small to compute the magnetic field
adequately in all regions, except in regions where the field is generated dominantly by inner sources
(windings).

7. MAGNETIC FIELD EVALUATION OF DISTRIBUTION TRANSFORMERS

Simplified model 1
was used to compute
the magnetic field of
oil-type distribution
transformers in the
range 50-2500 kVA.
The distances between
the most distant point of
1 T and 40 T curve
and the origin of
distribution transformers
are shown in figure 11.
Measured values are
also shown for 630 and
1000 kVA transformers.
All transformers
were SI 24. Figure 11 – Maximum distances between curves of magnetic induction and
the origin of oil-type distribution transformers

7
 As stated before, such computed distances should be higher than measured ones, especially for
40 T curve. In relation to exposure to the magnetic fields, even if these computed values are taken into
account, the 40 T limit should not be the problem due to low distances from the transformer. But, 1 T
limit might be the problem.
In this analysis the only source of the magnetic field was assumed to be the transformer itself. But
when the transformer is in operation (in network), dominant source of the magnetic field could be LV
conductors. The magnetic field can be significantly decreased by using LV twisted cables. If it is not
possible, the individual LV phase conductors should be as close as it’s possible. LV bus-bars are not
recommended. If they are used, the magnetic field can be decreased by shielding of bus-bars.

8. CONCLUSION

Although the satisfying results for magnetic induction are achievable for oil-type distribution
transformer, the lack of computer memory is still the major limitation.
The advantage of proposed method based on integral approach is possibility to compute the
magnetic field very simply and quickly, especially when the computed values are several meters from the
transformer.

9. REFERENCES

[1] "ICNIRP Guidelines for Limiting Exposure to Time-Varying Electric, Magnetic and Electromagnetic
Fields (up to 300 GHz)", ICNIRP, 1998.
[2] "Ordinance relating to Protection from Non-Ionising Radiation (ONIR)", The Swiss Federal Council,
1999.
[3] "Pravilnik o zaštiti od elektromagnetskih polja", Ministarstvo zdrastva RH, 2003.
[4] "Directive 2004/40/EC of the European parliament and of the council on the minimum health and
safety requirements regarding the exposure of workers to the risks arising from physical agents
(electromagnetic fields)", Official Journal of the European Union L 159 of 30 April 2004.
[5] EN 50464-1 Three-phase oil-immersed distribution transformers 50 Hz, from 50 kVA to 2500 kVA
with highest voltage for equipment not exceeding 36 kV – Part 1: General requirements, April 2007.
[6] B. , "Magnetsko polje u okolini distribucijskih transformatora (Magnetic field in the vicinity of
distribution transformers)", doktorska disertacija, FER, Zagreb, 2009.
[7] B. , Z. Haznadar, Ž. Štih, "The 3D magnetic field calculation of a dry type transformer by
integral equations", The 25th International Review of Progress in Applied Computational
Electromagnetics (ACES 2009), Monterey, California, 2009.
[8] B. , "The 3D calculation of the quasistatic magnetic field around a current carrying conductor
and ferromagnetic plate by means of integral equations", Energija, Vol. 56, No. 6, pp. 712-729,
2007.
[9] J. Matjan, "Rješavanje zada a elektromagnetskog polja integralnim jednadžbama", doktorska
disertacija, ETF, Zagreb, 1977.
[10] M. Breschi, A. Christofolini, "Experimental and numerical analysis of stray field from transformers",
IEEE transactions on magnetics, Vol. 43, No. 11, November 2007.