International Colloquium Transformer Research and Asset Management

Cavtat, Croatia, November 12 – 14, 2009

Farzad Zhalefar Majid Sanaye-Pasand

Iran-Transfo Co., Zanjan, Iran School of ECE, University of Tehran, Iran
f.zhalefar@iran-transfo.com msanaye@ut.ac.ir

CALCULATION OF NO-LOAD LOSS IN POWER TRANSFORMERS WITH FIVE-

LIMB MAGNETIC CORE; A CASE STUDY

SUMMARY

The main purpose of this paper is to calculate no-load loss in three-phase power transformers
which use five-limb core in their structure. The most important difference between these types of
transformers with conventional type ones which use three-limb cores is distribution of flux in the magnetic
core. As would be observed, non-linear behavior of the magnetic core would be reflected in the flux
distribution and therefore magnetic fluxes with distorted wave-shapes would be produced. This
phenomenon could lead to some difficulties in calculation of no-load losses in various degrees of
excitation.
In this paper a practical three-phase power transformer with a five-limb core has been chosen for
simulation. For this purpose, the principle of duality method has been chosen for modeling of this 200
MVA, 400 kV transformer. This work is performed employing EMTP.
As would be shown, obtained results are in good agreement with the predefined value of no-load
losses at various applied voltages which confirms capabilities of the used modeling method.

Key words: Five-limb, no-load loss, power transformer

1. INTRODUCTION

The power transformer is one of the most important and expensive equipment of electric power
systems. Cost-effective operation of electric power generation, transmission and distribution systems is
related to reliability and availability of their power transformers. However, most investigations have been
performed for the three-phase three-limb core, generally since the vast majority of all built transformers
have cores of this type. As an example, Z. Valkovic has investigated the effect of core geometry, core
material, joint design, and induction on core losses, primarily for three-phase three-limb cores, and to
some degree for single-phase transformers [1], [2]. Z. Valkovic and A. Rezic also investigated the effect
of the step lap joint design on core losses [3], [4].
The present day situation dictates fewer very large transformers where the transportation height
limitations would force the design to be a three-phase five-limb core or have the transformer bank split up
into three single-phase transformers. However, there are not abundant analytical works dealing with
transformer core losses which have been focused upon this type of cores. From rare research works in
this field, X.S. Chen and P. Neeudorfer had a motivation for developing a transient model for the five-limb
transformer with an interest in transformer ferroresonance, a problem which has been vexing the power
industry for over 60 years [5]. D.D. Mairs, D.L. Stuehm and B.A. Mork also studied this type of
transformers from overvoltages point of view [6]. Due to calculation of transformer losses, E.G.
teNyenhuis, G.F. Mechler and R.S. Girgis employed the 2-D finite difference method (FDM) to predict flux
distribution and magnetic loss in cores with this structure [7]. In spite of high presision, such calculation
needs considerable degrees of proccessing ability.

1
 This paper presents calculation results of flux distributions, flux density waveshapes and loss
distributions in three-phase five-limb cores for a manufactured and practical large power transformer. The
method of duality has been employed for modeling purpose. A comparison of calculated and pre-defined
values of core loss on the actual transformers is also given.

2. COMMON STUCTURES OF MAGNETIC CORES

A part of a core, which is surrounded by windings, is called a main limb or leg. There also might
be some limbs with no winding wounded in some transformers. These are called end limbs. Remaining
part of the core, which is not surrounded by windings, but is essential for completing the path of flux, is
called as yoke. These part of cores are shown in figure 1 [8].
As mentioned in introduction, there are various structures for manufacturing magnetic cores of
three-phase transformers which have their own features and applications. These stuctures could be
categorised mainly in two major groups; single-phase and three-phase cores.
Transformer core construction mainly depends on technical specifications, manufacturing
limitations and transport considerations. It is economical to have all the windings of three phases in one
core frame. A three-phase transformer is cheaper (by about 20 to 25%) than three single-phase
transformers connected in a bank [8]. But from the spare unit consideration, users find it more economical
to buy four single-phase transformers as compared to two three-phase transformers. Also, if the three-
phase rating is too large to be manufactured in transformer works (weights and dimensions exceeding the
manufacturing capability) and transported, there is no option but to manufacture and supply single-phase
units. In figure 1 various types of core construction are shown [8].

Figure 1 – Various Types of Single-Phase Core Construction

2.1. STRUCTURES OF SINGLE-PHASE CORES

In a single-phase three-limb core (figure 1-a), windings are placed around the central limb, called
as the main limb. Flux in the main limb gets equally divided between two yokes and it returns via the end
limbs. The yoke and end limb area, whih is oval, should be only 50% of the main limb area for the same
operating flux density. This type of construction can be alternately called as single-phase shell-type
transformer. Zero-sequence impedance is equal to positive-sequence impedance for this construction (in
a bank of single-phase transformers).
Sometimes in a single-phase transformer windings are split into two parts and placed around two
limbs as shown in figure 1-b. This construction is sometimes adopted for very large ratings. Magnitude of
short-circuit forces are lower because of the fact that ampere-turns height are reduced. The area of limbs
and yokes is the same. Similar to the single-phase three-limb transformer, one can have additional two
end limbs and two end yokes as shown in figure 1-c to get a single-phase four-limb transformer to reduce
the height for the transport purpose.

2.2. STRUCTURES OF THREE-PHASE CORES

The most commonly used construction, for small and medium rating transformers, is three-phase
three-limb construction as shown in figure 2-a [8]. For each phase, the limb flux returns through yokes

2
and other two limbs (the same amount of peak flux flows in limbs and yokes). In this construction, limbs
and yokes usually have the same area. Sometimes the yokes are provided with a 5% additional area as
compared to the limbs for reducing no-load losses. It is to be noted that the increase in yoke area of 5%
reduces flux density in the yoke by 5% and reduces watts/kg by more than 5% (due to non-linear
characteristics) but the yoke weight increases by 5%. Also, there may be additional loss due to cross-
fluxing since there may not be perfect matching between lamination steps of limb and yoke at the joint.
Hence, the reduction in losses may not be very significant. The provision of extra yoke area may improve
the performance under over-excitation conditions. Eddy losses in structural parts, due to flux leaking out
of core due to its saturation under over-excitation condition, are reduced to some extent [9], [10].

Figure 2 – Various Types of Three-Phase Core Construction

The three-phase three-limb construction has inherent three-phase asymmetry resulting in

unequal no-load currents and losses in three phases. One can get symmetrical core by connecting it in
star or delta so that the windings of three phases are electrically as well as physically displaced by 120
degrees. This construction results into minimum core weight and tank size, but it is seldom used because
of complexities in manufacturing [10], [11].
In large power transformers, in order to reduce the height for transportability, three-phase five-
limb construction depicted in figure 2-b is used. The magnetic length represented by the end yoke and
end limb has a higher reluctance as compared to that represented by the main yoke. Hence, as the flux
starts rising, it first takes the path of low reluctance of the main yoke. Since the main yoke is not large
enough to carry all the flux from the limb, it saturates and forces the remaining flux into the end limb.
Since the spilling over of flux to the end limb occurs near the flux peak and also due to the fact that the
ratio of reluctances of these two paths varies due to non-linear properties of the core, fluxes in both main
yoke and end yoke / end limb paths are non-sinusoidal even though the main limb flux is varying
sinusoidally [2,4]. Extra losses occur in the yokes and end limbs due to the flux harmonics [11].
The zero-sequence impedance is much higher for the three-phase five-limb core than the three-
limb core due to low reluctance path (of yokes and end limbs) available to the in-phase zero-sequence
fluxes, and its value is close to but less than the positive sequence impedance value. This is true if the
applied voltage during the zero sequence test is small enough so that the yokes and end limbs are not
saturated.

3. NO-LOAD TEST IN POWER TRANSFORMERS

A transformer dissipates a constant no-load loss as long as it is energized at constant voltage, 24

hours a day, for all conditions of loading. This power loss represents a cost to the user during the lifetime
of the transformer. Maximum values of the no-load loss of transformers are specified and guaranteed by
the manufacturer. No-load-loss measurements are made to verify that the no-load loss does not exceed
the specified or guaranteed value [10], [11], [12].

3.1. NATURE OF NO-LOAD LOSS

Transformer no-load loss, often called core loss or iron loss, is the power loss in a transformer
excited at rated voltage and frequency but not supplying load. The no-load loss comprises three
components: 1) core loss in the core material, 2) dielectric loss in the insulation system and 3) copper
(I2R) loss due to excitation current in the energized winding. The no-load loss of a transformer is primarily

3
caused by losses in the core steel (item 1). The remaining two sources are sometimes ignored. As a
result, the terms no-load loss, core loss, and iron loss are often used interchangeably. Core loss and iron
loss, strictly speaking, refer only to the power loss that appears within the core material [11].
The magnitude of no-load loss is a function of the magnitude, frequency, and waveform of the
impressed voltage. These variables affect the magnitude and shape of the core magnetic flux waveform
and hence affect the value of the core loss. It has been verified through measurements on power and
distribution transformers that core loss also depends, to some extent, upon the temperature of the core.
According to the IEEE test code [2], the approximate rate of change of no-load loss with core temperature
is 0.00065 per-unit core loss increase for each 1 C reduction in the core temperature [11].
The two main components of the core loss are hysteresis loss and eddy-current loss. The change
in eddy-current loss, due to a change in the resistivity of the core steel as temperature changes, appears
to be one factor that contributes to the observed core-loss temperature effect. The hysteresis loss
magnitude is a function of the peak flux density in the core-flux waveform. When the impressed voltage
waveform is distorted (not a pure sine wave), the resulting peak flux density in the flux waveform depends
on the average absolute value of the impressed voltage wave. Eddy-current loss is a function of the
frequency of the power source and the thickness of the core-steel laminations. Eddy loss is strongly
influenced by harmonics in the impressed voltage. The IEEE transformer test code [2] recommends the
average voltage voltmeter method, to be described below, for measuring no-load loss [10], [11].

3.2. MEASUREMENT OF NO-LOAD LOSS

The measurement of no-load loss, according to the average-voltage voltmeter method, is

illustrated in figure 3. Voltage and current transformers are required to scale the inputs for voltmeters,
ammeters, and wattmeters. Three-phase no-load-loss measurements are carried out the same way,
except that three sets of instruments and instrument transformers are utilized. The test involves raising
voltage on one winding, usually the low-voltage winding, to its rated voltage while the other windings are
in open circuit [10], [11].
Two voltmeters connected in parallel are employed. The voltmeter labeled Va in figure 3
represents an average-responding, rms-calibrated voltmeter. The voltmeter labeled Vr represents a true
rms-responding voltmeter. Harmonics in the impressed voltage will cause the rms value of the waveform
to be different from the average-absolute (rms-scaled) value, and the two voltmeter readings will differ.
When the voltage reading, as measured by the average-responding voltmeter, reaches a value
corresponding to the rated voltage of the excited winding, readings are taken of the rms current, the rms
voltage, and the no-load power. The ratio of the measured rms current to the rated load current of the
excited winding, expressed in percent, is commonly referred to as the percent excitation current. The
measured no-load loss is corrected to a sine-wave basis by a formula given in the IEEE test code [2],
[11], using the readings of the two voltmeters. The correction is shown in figure 3. The corrected value is
reported as the no-load loss of the transformer.

Figure 3 - Measurement of No-Load Loss According to the Average-Voltage Voltmeter Method

4. MODELING METHOD

This model was introduced by Cherry and generalized by Slemon [12], [13]. In 1981, Dick and
Watson employed duality approach and obtained a per-phase equivalent circuit for a three-phase, three-
legged transformer. German and Davies in 1981 and Arturi in 1991 employed this approach in the

4
modeling of two-winding, multi-legged transformers [14], [15], [16]. In this approach, the equivalent circuit
of transformer is made using the principle of duality between electric and magnetic environments. A
model based on this approach properly reflects the flux balance and thus the nonlinear iron core can be
represented accurately [17], [18].
As would be pointed out, fundamental component of flux densities will be used to calculate no-
load losses among various core pathes. However, phenomena like eddy and hysteresis have no
important effect on flux distribution between yokes and end limbs. Therefore, in the performed research
study, these factors have been neglected, due to simplicity. Furthure information about pointed modeling
method is available in [19].

5. SYSTEM STUDIED

For modeling purpose, a three-phase five-limb transformer has been chosen. Vector group of this
three-winding transformer is YNyn0d11 with voltage ratios of 400/66/20 kV. Rated power related to this
transformer is 120/160/200 MVA associated to ONAN/ONAF1/ONAF2 operating modes. During no-load
test, sinusoidal rated voltage has been applied to the tertiary side of this transformer.

6. SIMULATION RESULTS

Using the mentioned modeling method, some simulations have been performed that are
presented in this section. Due to presence of a symmetrical three-phase electric system, applied voltages
and produced fluxes would be symmetrical with 120 degrees phase differences (fluxes shown in figure 4).
This is shown in figure 5 for flux densities of three main limbs.
The main difficulty related to three-phase five-limb transformers is presence of fluxes with
distorted wave-shapes in the end yokes/limbs. Although for calculation of losses produced by such
distorted fluxes there are some practical relations available, however, the first step is to produce the flux
waveforms.

Figure 4 – Physical Structure of the Modeled Transformer, Main Limb Fluxes

5
 Flux Density of Main Limbs of Magnetic core
2.0
[T]
1.5

1.0

0.5

0.0

-0.5

-1.0

-1.5

-2.0
0.070 0.075 0.080 0.085 0.090 0.095 [s] 0.100
Left side: Middle: Right side:

Figure 5 – Flux Density of Main Limbs of the Modeled Transformer

6.1 FLUX WAVEFORMS

As pointed out, main limbs would have sinusoidal flux densities, while fluxes assosiated with the
yokes and end limbs would be distorted (fluxes shown in figure 6).

Figure 6 – Physical Structure of Modeled Transformer, Main Limb, Yoke and End Limb Fluxes

Figure 7 shows waveforms of flux densities related to the left side main limb in addition to the left
side yokes and end limbs. This figure provides very interesting observations. The yoke region between
the main limbs has a highly distorted and uniform flux density across it which is mainly due to nonlinearity
of the core characteristics and also the lower cross sectional area used in this regin of the core. This yoke
region (between wound limbs) is part of a magnetic circuit where two adjacent phases, 120 apart, are
supplying mmf’s in the opposite directions. The resultant mmf is, hence equals two times the phase value.
The strong nonlinearity of the core material and the greater cross-section of the yoke keep the magnitude
of the flux density in the yoke even lower than the operating induction of the core.
In the region of the yoke leading to the outside limbs (end limbs), figure 7 shows that the flux
density varies between about plus and minuse 1.5 T, for the case of nominal excitation. The generally
lower magnitudes of flux densities in this region is attributed to the fact that this part of the core magnetic
circuit is subjected to an mmf corresponding to the phase value of the exciting current only. Also, the
significantly lower permeability of grain oriented steels, in directions other than the direction of grain
orientation, forces the magnetic flux to concentrate in the inside perimeters of the magnetic circuit to

6
avoid the high reluctance of the outside perimeter of the circuit. It is interesting to see in this figure that
the flux quickly redistributes again before it travels to the outside limbs resulting in a nearly uniform
distribution across the punching width. This low flux loading and uniform flux density distribution explains
why the outside limbs do not necessarily need to have the same cross-sectional area as the yoke [10],
[16].

Flux Density of Various Locations of Magnetic Core

2.0
[T]
1.5

1.0

0.5

0.0

-0.5

-1.0

-1.5

-2.0
50 59 68 77 86 [ms] 95
Left side Main Limb: Yoke: End Limb:

(a)

Flux Density of Various Locations of Magnetic Core

2.0
[T]
1.5

1.0

0.5

0.0

-0.5

-1.0

-1.5

-2.0
0.055 0.064 0.073 0.082 0.091 [s] 0.100
Left side Main Limb: Yoke: End Limb:

(b)

7
 Flux Density of Various Locations of Magnetic Core
2.0
[T]
1.5

1.0

0.5

0.0

-0.5

-1.0

-1.5

-2.0
55 60 65 70 75 80 85 90 [ms] 95
Left side Main Limb: Yoke: End Limb:

(c)

Figure 7 - Flux Density of Left Side Main Limb, Yoke and End Limb of The Modeled Transformer
at a) 90% , b) 100% and c) 110% of excitation

6.2 CALCULATION OF NO-LOAD LOSS

For the case of regions with sinusoidal flux waveforms, calculation of loss would not be so
complicated. However, for the encountered case, there are two distinct regions, main limbs and remaining
parts. In the main limbs due to presence of magnetic flux with sinusoidal waveform, final result is obvious.
However, the problem is calculation of no-load loss in yoke and end-limbs with non-sinusoidal shaped
fluxes. The main way to calculate this value is to solve complex non-linear differntial equations, while, in
this paper, some practical ways are used.
In the technical datasheets of Iran-Transfo Co. there are some practical relations and curves
available that could be used to calculate no-load losses produced by non-sinusoidal shaped fluxes. In this
method, one effective value for flux density (B'), which is affected by main harmonic of flux density of
yokes and end limbs, is chosen. Then using practical curves, a value for loss density of magnetic system
is introduced (W/kg). Using this value in addition to the total mass of magnetic core, total no-load loss of
the power transformer could be obtained. This method has been verified as it has been used for no-load
loss calculation of various power transformers with five-limb core. The mentioned effective flux density
(B') could be obtained using (1) to (3) [20].

B1J = B * (0.648 − 0.030 * B ) * 3 (1)

B1R = B * (0.450 + 0.052 * B ) * 3 (2)

1 ⎡ ⎤
3 3
⎛B ⎞ ⎛B ⎞
B′ = B * 3 * ⎢GS + ⎜ 1J ⎟ * GJ + ⎜ 1R ⎟ * GR ⎥ (3)
GFe ⎣⎢ ⎝ B ⎠ ⎝ B ⎠ ⎦⎥

where,

8
 B' : effective density of magnetic flux,
B : peak value of magnetic flux density associated to main limbs,
B1J : peak value of main harmonic of magnetic flux density associated to yokes,
B1R : peak value of main harmonic of magnetic flux density associated to end limbs,
GFe : total weight of magnetic core,
GS : total weight of main limbs,
GJ : total weight of yokes,
GR : total weight of end limbs.

The mentioned method has been performed for the studied transformer using both Iran-Transfo
practical equations (equations (1) to (3)) and simulation results using EMTP. Fianl results of produced
losses using two methods for three degree of excitation could be obtained in Tables I to III. These
excitaion degrees, which are 10% under-excitation, nominal excitation and 10% over-excitation, are the
most important cases which are usually emphasised by purchasers due to high probability of transformer
operation in these conditions.

Table I - Fianl Flux Densities and No-Load Losses of Studied Power Transformer, 90% Excitation

Iran-Transfo Design Method Results EMTP Results

B (T) 1.521 1.521

B1J (T) 1.587 1.700
B1R (T) 1.394 1.263
B’ (T) 1.499 1.504
P’ (W/kg) 0.78 0.81
PLoss (kW) 85.1 88.37

Table II - Fianl Flux Densities and No-Load Losses of Studied Power Transformer, 100% Excitation

Iran-Transfo Design Method Results EMTP Results

B (T) 1.683 1.683

B1J (T) 1.742 1.807
B1R (T) 1.567 1.456
B’ (T) 1.663 1.652
P’ (W/kg) 0.98 1.0
PLoss (kW) 108.8 106.914

Table III - Fianl Flux Densities and No-Load Losses of Studied Power Transformer, 110% Excitation

Iran-Transfo Design Method Results EMTP Results

B (T) 1.847 1.847

B1J (T) 1.896 1.934
B1R (T) 1.747 1.694
B’ (T) 1.828 1.841
P’ (W/kg) 1.37 1.38
PLoss (kW) 149.46 150.55

As observed, there is a small difference between results of two methods. This difference could be
resulted from the fact that during modeling process using the principle of duality, some phenomena like
hysteresis and eddy losses were neglected. These phenomena affect no-load losses directly and
indirectly. In other words, additionally to increment of no-load loss by hysteresis and eddy, they could
change waveform and distribution fluxes among yokes and end limbs, which the former was important for
simulations of this paper. Although, according to high quality of modern core steels such phenomena
could be neglected effectively. Therefore, a little difference between results of the two mentioned
methods is completely satisfying and acceptable.

9
7. CONCLUSION

In this paper, simulation results of modeling a 200 MVA, 400 kV transformer has been used to
estimate the capability of the principle of duality method. Construction of this three-phase transformer was
based on using five-limb magnetic core; a structure which is used in special cases that physical
dimensions of designed transformer with three-limb core reach some pre-defined limitations.
Obtained simulation results showed a good agreement with design values which confirms ability
of the used modeling method. As mentioned, in transformers using five-limb cores distorted fluxes exist.
Such fluxes normally could increase no-load loss value of transformers. This feature was obtained using
this modeling method. Although, in this modeling method some practical phenomena like hysteresis and
eddy had been ignored, obtained results were satisfying. This was due to the high quality of used core
steels that decreased effect of such phenomena.

8. REFERENCES

[1] Z. Valkovic, “recent problems of transformer core design”, Physica Scripta, vol. T24, Mar. 1988, pp.
71–74.
[2] ____, “influence of transformer core design on power losses”, IEEE Trans. Magnetics, vol. MAG-
18, no. 2, Mar. 1982, pp. 801–804.
[3] Z. Valkovic and A. Rezic, “flux and loss distribution in three-limb core with staggered T-joint”, in
Symposium on Electromagnetic Fields in Electrical Engineering, Sept. 1988, pp. 85–88.
[4] ____, “improvement of transformer core magnetic properties using the step-lap design”, Journal of
Magnetism and Magnetic Materials, vol. 112, 1992, pp. 413–415.
[5] X.S. Chen and P. Neudorfer, “digital model for transient studies of a three-phase five-legged
transformer”, IEE Proceedings-C, vol. 139, no. 4, July 1992, pp.351-358.
[6] D.D. Mairs, D.L. Stuehm and B.A. Mork, “overvoltages on five-legged core transformers on rural
electric systems“, IEEE Trans. Ind. Appl., vol. 25, no. 2, March/April 1989.
[7] E.G. teNyenhuis, G.F. Mechler and R.S. Girgis, “flux distribution and core loss calculation for single
phase and five limb three phase transformer core design”, IEEE Trans. Power Del., vol. 15, no. 1,
Jan. 2000, pp. 204-209.
[8] S.V. Kulkarni and S.A. Kharparde, “transformer engineering, design and practice”, Marcel Dekker
Inc., 2004
[9] W.N. Harry, R. Hasegawa, A.C. Lee and L.A. Lowdermilk, “amorphous alloy core distribution
transformers”, proc. IEEE, vol. 79, no. 11, Nov. 1991, pp. 1608-1623.
[10] R.S. Girgis, C. J. Baldwin, W.N. Kennedy, S. Monaco, S.R. Lindgren, R.D. Holm, and R.E.
Kothmann, “design and performance improvements in power transformers using ribbon cable”,
IEEE Transactions on Power Delivery, vol. 10, no. 2, April 1995, pp. 869-877.
[11] J.H. Harlow, et al, “electric power transformer engineering”, CRC Press LCC, 2004.
[12] E. C. Cherry, “the duality between interlinked electric and magnetic circuits and the formation of
transformer equivalent circuits," Proc. of the Physical Society, vol. (B) 62, pp. 101-111, Feb. 1949
[13] G.R. Slemon, "equivalent circuits for transformers and machines including nonlinear effects," Proc.
IEE, Part IV, vol. 100, 1953, pp. 129-143
[14] C.M. Arturi, "transient simulation and analysis of a five-limb step-up transformer following an out-of-
phase synchronization", IEEE Trans. Power Delivery, vol. 6, no. 1, Jan. 1991, pp. 196-207
[15] D.M German, and A.E. Davies, “the simulation of transformer feeders following switching
operations,” IEEE Trans. Power Apparatus and Systems, vol. PAS-100, no. 11, Nov. 1981, pp.
4510-4514
[16] X. Chen and S.S. Venkata, "a three-phase three-winding core-type transformer model for low-
frequency transient studies," IEEE Trans. Power Delivery, vol. 12, no. 2, April 1997, pp. 775-782
[17] F. De Leon and A. Semlyen, "reduced order model for transformer transients," IEEE Trans. Power
Delivery, vol. 7, no. 1, Jan. 1992, pp. 361-368
[18] F. De Leon and A. Semlyen, "complete transformer model for electromagnetic transients," IEEE
Trans. Power Delivery, vol. 9, no. 1, Jan. 1994, pp. 231-239
[19] F. Zhalefar and M. Sanaye-Pasand, "Transformer core modeling as applied to slow transients and
protective studies", IEEE Power India Conference, New Delhi, India, March 2006
[20] Iran-Transfo Co. Datasheets, vol. 1, transformer cores

10