IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 18, NO. 3, JULY 2003821
Reducing Losses in Distribution Transformers
Juan Carlos Olivares, Member, IEEE, Yilu Liu, Senior Member, IEEE, Jose M. Cañedo, Member, IEEE,
Rafael Escarela-Pérez, Member, IEEE, Johan Driesen, Member, IEEE, and Pablo Moreno, Member, IEEE
Abstract—This paper examines three methods of reducing
distribution transformer losses. The first method analyzes the
effects of using aluminum electromagnetic shields in a distribution
transformer. The goal of placing electromagnetic shields in the
distribution-transformer tank walls is to reduce the stray losses.
A 500-kVA shell-type transformer was used in the experiments.
The overall results presented indicate that stray losses can be
considerably reduced when electromagnetic shielding is applied
in the transformer tank. In the experiment, the tank walls were
lined with aluminum foil. The possibility of reducing the dielectric
losses was shown through experiments in the second method.
And the third method of this work analyzes the behavior of
wound-cores losses in distribution transformers, as a function of
joint configuration design parameters. The joint configuration
used in this paper is called step-lap joint.
Index Terms—Dielectric losses, loss measurements, shielding,
transformer, transformer cores.
on transformer efficiency . Although the efficiency of a
modern transformer lies above 99%, the loss cost is still
significant . The present study is part of such effort to further
increase the efficiency. The loss topic continues to be a topic
of huge interest. In 1990, only about 92.5% of the energy
generated at U.S. power plants was actually distributed to the
consumer. The other 7.5% of the energy (approximately 229
billion kilowatt-hours annually) was dissipated as losses in the
transmission and distribution system .
The distribution transformer efficiencies steadily increased
with the introduction of improved materials and manufacturing
methods . Even so, 26.6% of the average transmission and
distribution losses are still associated with distribution trans-
formers . The previous figures of losses are a consequence of
the large number of transformers installed among other factors.
It is estimated that there are 50million distribution transformers
 in use in the U.S.
The objective of this research is to find ways to reduce the
distribution transformer losses. In order to reduce these losses,
UE to environmental considerations and rising energy
costs, customers have been putting high requirements
Manuscript received July 18, 2001; revised May 9, 2002. This work was sup-
ported in part by CONACYT.
versity, Bradley Department of Electrical and Computer Engineering, Blacks-
burg, VA 24061 USA (e-mail: firstname.lastname@example.org).
J. M. Cañedo and P. Moreno are with CINVESTAV, Jalisco 45090, México
(e-mail: email@example.com; firstname.lastname@example.org).
R. Escarela-Pérez is with Universidad Autónoma Metropolitana-Az-
capotzalco, México, D.F. (e-mail: email@example.com).
J. Driesen is with Katholieke Universiteit Leuven, Research Group Electrical
Energy ESAT-ELEN Kasteelpark Arenber, Heverlee 10B-3001, Belgium
Digital Object Identifier 10.1109/TPWRD.2003.813851
COMPARATIVE ANALYSIS OF TRANSFORMER COSTS
we are using three methods. The first method investigates the
effects of electromagnetic shields upon distribution transformer
tank losses. In today’s competitive market, accurate estimation
and subsequent reduction of the stray loss by shielding tech-
niques could give a competitive advantage .
measurements of four transformers were made in order to show
how the transformer losses could vary during its manufacturing
process. The dielectric losses can be reduced if the transformer
manufacturer carries out an adequate drying process.
The third method deals with the core joints. Core joints play
an important role in the performance of transformer cores. Due
to the importance of improved electrical core performance,
transformer manufacturers – and research institutions
– are very active in the development of better electrical
steel and the optimization of the core design parameters.
Of the various materials required to build a transformer, the
electrical steel comprises the largest investment. Table I shows
a comparative analysis of transformer costs by component .
Table I was made taking into account six transformers, from
15 to 1000 kVA and the voltage considered was 13200 V–220
In this study, the effect of the following two factors is consid-
ered: (a) overlap length and (b) number of laminations per step
or group. The experimental data presented in this work will be
helpful for a practicing engineer in the transformer industry.
II. EXPERIMENTAL STUDY TO REDUCE
DISTRIBUTION-TRANSFORMERS STRAY LOSSES USING
A 500-kVA experimental transformer on loan from a factory
was used to investigate in detail the stray losses and to examine
how these losses can be described physically. This transformer
(see Fig. 1) is a shell-type, which means the windings are sur-
rounded by the core.
In the transformer, the leakage flux is high in the tank walls,
which causes high-power losses . The main effort to reduce
0885-8977/03$17.00 © 2003 IEEE
822IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 18, NO. 3, JULY 2003
side core is not assembled yet.
Shell-type transformer picture taken during manufacturing; the right
load losses has concentrated in the area of stray losses . A re-
duction of the magnetic flux is required to reduce these losses.
Placing a physical barrier, called a shield, between the elec-
tromagnetic field source and the region of interest can accom-
plish this purpose. Shielding materials include magnetic and
electric conducting materials . Magnetic materials are high
permeability material and shield by a mechanism called “flux
shunting.” In this case, the flux from a source is diverted into
the magnetic material and away from the region to be shielded.
Electric materials are high-conductivity materials and shield by
a phenomenon known as “eddy-current cancellation.” In this
case, currents are induced in the conductor, which create mag-
netic fields that partially cancel those from the source .
The main load-losses are the
arise from eddy currents induced in metallic parts of the trans-
former; for instance, in clamps and in tank walls. Therefore,
good understanding of stray losses and their reduction mecha-
nisms are necessary for improving the transformer design ,
. The stray losses are a function of many factors including
physical geometry of the cores and windings, voltage class, and
the material used in the tank construction .
According to Fig. 2, the stray losses increase with the
growing of transformer rating –. Hence, the application
of shielding in very small transformers is not attractive to
reduce stray losses. Using the least square fitting method, the
stray losses that could be reduced by placing shielding are
expressed as a function of the transformer rating as follows
losses. The stray losses
The ratio of stray losses to load losses
using regression analysis as given below 
Fig. 3 shows the behavior of
A magnetic shield comprises a large number of packets of
aluminum laminations mounted on the vertical sides of the steel
tank. The height of the aluminum shield is equal to the height of
the steel tank and the separation between the laminations is of
the order of 0.3 mm. The process of lining the steel tank wall of
the transformer with aluminum foil is rather time-consuming.
versus transformer rating.
Fig. 2. Stray losses versus the transformer rating at temperature 85 C.
Stray loss load loss ratio versus the transformer rating (kilovolt
The total aluminum cost was about U.S.$ 311.70, which repre-
sents 10% of the total transformer material cost. See Table II.
The load loss and stray loss are measured under three condi-
tions: (a) without shield, (b) with aluminum shield of 1.2 mm of
thickness, (c) with aluminum shield thickness of 10 mm.
Table III shows the measurement values of stray losses and
the efficiency for cases (b) and (c) of the previous paragraph.
The load losses without shield were 504.12 W at ambient tem-
perature. This base-case efficiency was of 99.09%.
It is observed that stray losses are increased by 20.9% when
unshielded case, since the depth of penetration is larger than the
aluminum shield thickness so the magnetic flux density reaches
etration is less than the steel-wall thickness, the half thickness
of the plate, the inner part of the plate remains unmagnetized,
and the magnetic flux as well as the eddy currents are confined
to a layer of depth
on the plate surface .
Placing aluminum shielding in the internal tank wall reduces
tude in the shielding produce a magnetic field that partially can-
cels the incident field. In other words, the magnetic flux density
OLIVARES et al.: REDUCING LOSSES IN DISTRIBUTION TRANSFORMERS823
MASS OF THE ALUMINUM SHIELD, THICKNESS OF 10 MM
MEASUREMENT VALUES OF STRAY LOSSES
induced is opposed to the magnetic flux density incident. The
superposition of induced field and incident field gives a total
field, which is repelled from the tank superficies. It is important
to recognize that the previous phenomenon occurs regardless of
the application of aluminum shielding. Then, the difference is
is in the order of 2 10 , while for the aluminum shield
this value is 500 times less, that is, 3.8
in the losses.
10 , which is reflected
III. IMPORTANCE OF DIELECTRIC LOSSES IN
THE NO-LOAD TEST
The no-load losses
, the hysteresis losses
. Since the no-load current will be very small compared
to the full load value, the
negligible. Thus , 
include the eddy-current
, and the dielectric losses
losses in the windings will be
process. Column 1 shows the no-load losses when a test coil of
12 turns is used (see Fig. 4). Column 2 indicates the no-load
losses when the cores are assembled with their design windings
but without the tank (The high-voltage winding has 1066 turns
and the low-voltage winding has 16 turns). Column 3 indicates
the no-load losses when the transformer is completed, that is,
when the transformer has a tank and it is filled with oil.
In Table IV, two main observations can be made. First, the
no-load losses of the active-element (set core winding) are
higher than the no-load losses of the completed transformer
because when the active-element is tested, its insulation con-
tains a high content of moisture, which causes high dielectric
losses. The dielectric losses are determined by the expression
Fig. 4. No-load test with a test coil of 12 turns.
NO-LOAD LOSSES DURING THREE STAGES OF TRANSFORMER
Second, column 2 losses are higher than the column 1 losses
because undesirable stresses are created in the electric steel
for manufacturing operations introduced during the core-coil
assembly. The stresses due to the slitting of the core steel
as well as the stresses due to the core winding and forming
operations are relieved by the heat treatment process (or stress
relief annealing). Normally, stresses create a harmful effect by
causing a degradation of magnetic properties. These changes
occur because the metal crystals are distorted. All of the mea-
surements in this paper were carried out on a new transformer
and before the impulse test. This is important because when the
test transformers are old  or have received the impulse test,
the no-load losses can be higher. This difference exists because
there are local breakdowns between individual laminations,
which would result in higher loss .
is voltage (V),is angular frequency (rad/s),
is capacitance of the configuration.
IV. IMPACT OF THE JOINTS DESIGN PARAMETERS OF THE
WOUND CORE IN DISTRIBUTION TRANSFORMER LOSSES
The following core manufacturing-parameter definitions
are related to Fig. 5. These definitions are exclusive for the
wound-core distribution-transformer family.
• Step or book.Set of laminations,which can varybetween
four and 25 and this set of laminations form a cycle. In
Fig. 5, the first four laminations(from top to bottom) form
824IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 18, NO. 3, JULY 2003
Fig. 5. Core with step-lap joint.
• Air gap (g). The air gap is the separation between lam-
ination and lamination in the direction of rolling. In the
practice, this value is less than 3 mm.
• Overlap (L). The overlap is the length between the half
points of the air gaps of two laminations contiguous in the
rolling direction. The typical range of this parameter is 1
to 2 cm.
• Lamination thickness (T). Grain oriented silicon steel is
graded according to the American iron and steel institute
are coated with C-2 coating or C-5 over C-2 coating. Typ-
ever C-5 is applied over C-2 coating, the thickness of
C-5 coating is approximately 0.0001 cm per surface for
wound-core distribution transformer . A safe value of
interlamination resistance must be maintained to prevent
stray losses in the core –.
In Fig. 5, the laminations are in succession in order to obtain
a higher mechanical stability. If the joints are rigid and strong, it
prevents them from coming apart under severe operation condi-
tions, and also diminishes the noise, from vibrations during the
operation of the transformer .
Most of the technical papers have analyzed the stacked trans-
former cores. Until now, little attention has been paid to wound
transformer cores. The objective of these measurements is to in-
vestigate experimentally the effects of core-parameter changes
on wound-core losses in a distribution transformer. This goal
was carried out by two experiments. In these experiments, the
number of laminations per step or book and the overlap length
were varied. The transformers were excited with a 60-Hzgener-
The following experiments were performed using a 15-kVA
transformer. The cores were assembled using six and 25 lami-
nations per step. The grain-oriented electrical steel laminations
(0.23 mm thick) produced a nominal core loss at 1.5 T and
60 Hz of 0.98 W/kg. The overlap and the air gap length were
mm andmm, respectively.
Table V shows that laminations per step do not have an im-
portant effect on the core losses. This experiment only tested
one core for each measurement.
In order to comply with manufacturing limitations, the
number of laminations per step is increased. Conversely,
NO-LOAD LOSSES FOR A 15-KVA TRANSFORMER WITH DIFFERENT
LAMINATIONS PER STEP
NO-LOAD LOSSES WITH DIFFERENT OVERLAP LENGTH
the overlap lengths are decreased and the net result of these
opposing factors results in no change.
These results agree with the conclusion of  where it was
stated that “when the number of laminations per one group is
increased core losses are slightly increased.”
A separate experiment examined a sample of nine 37.5-kVA
overlap length of 1 cm and three cores were manufactured with
an overlap length of 2 cm. The results are shown in Table VI. In
this experiment, only one core was tested in each measurement.
former with an overlap length of 2 cm. This is most likely due
to the increased area, where the flux is forced to pass perpen-
dicularly to the laminations, since the core steel is anisotropic
. The results of the no-load test vary with the temperature
of the transformer core . For this reason, the measurements
in Tables V and VI were carried out at the same temperature.
Sixty cores were manufactured to test the repeatability of the
This paper presented results of experimental investigations
regarding reduction of distribution-transformer losses. The
work contains experimental data that will be helpful for prac-
ticing engineers in the transformer industry. Experiments were
carried out in well-controlled conditions. First, a load loss test
was carried out under three different conditions: (a) tank walls
without shield, (b) tank walls with aluminum shield of 1.2 mm
of thickness, and (c) tank walls with aluminum shield of 10-mm
thickness. The electromagnetic shields of the transformer in
this experiment prevented the penetration of the magnetic stray
flux in the magnetic materials, where high losses would be
OLIVARES et al.: REDUCING LOSSES IN DISTRIBUTION TRANSFORMERS825
induced. In this case, an increase of the stray losses by 20.9%
was observed when the aluminum shield of 10 mm was not
used. On the other hand, there were not significant changes
in the losses when the 1.2-mm shield was used with respect
to unshielded case, since the depth of penetration was larger
than the shield thickness and the magnetic flux could reach the
carbon steel. The study also demonstrated that the dielectric
losses are important in no-load loss in the transformer when
the transformer insulations have a high water content. Until
now, little attention has been given to the design of wound
transformer cores. In the paper, results for the wound-core
distribution-transformer family were presented. It is known
that minimum losses occur when the rolling direction coincides
with flux magnetic lines, but this condition is not satisfied in
the core joints since the joints air gaps cause local disturbances
of magnetic flux. Two experiments were carried out, varying:
(a) the overlap length and (b) the number of laminations per
step. It is observed that the number of laminations per step does
not have much effect on the core losses. This is because as the
number of laminations was increased, the overlap length was
decreased in order to comply with manufacturing limitations.
The test showed that transformers had higher losses when an
overlap length of 2 cm is used.
The authors want to thank Mr. A. L. Von Holle and F.
Gaudino (Ak Steel, USA), Mr. A. Trujillo (EMSA, Mexico),
Mr. P. Subbaraman (GE, India), Mr. J. Avila and Mr. D. Posadas
(Prolec-GE, Mexico), Mr. M. Alley and Mr. N. Clark, (Virginia
Tech), and Mr. E. Melgoza (Ph.D. Student, University of Bath)
for their helpful discussion.
 T. H. Harrison and B. Richardson, “Transformer Loss Reduction,” Ses-
sion, Cigre 12-04, 1988.
 R. Beaumont, “Losses in Transformer and Reactors,” Session, Cigre
 P. R. Barnes, “The Feasibility of Replacing or Upgrading Utility
Distribution Transformers During Routine Maintenance,” Rep.,
ORLN-6804/R1, Apr. 1995.
 D. J. Allan, “Power transformers- the second century,” Power Eng. J.,
pp. 5–14, Jan. 1991.
 B. Szabados, “A new approach to determine eddy current losses in the
tank walls of a power transformer,” IEEE Trans. Power Delivery, vol.
PWRD-2, pp. 810–816, July 1987.
 R. S. Girgis, K. Gramm, E. G. teNijenhuis, and J. E. Wrethag, “Exper-
imental investigations on effect of core production attributes on trans-
former core loss performance,” IEEE Trans. Power Delivery, vol. 13,
pp. 526–531, Apr. 1998.
 G. F. Mechler and R. S. Girgis, “Magnetics flux distributions in trans-
former core joints,” IEEE Trans. Power Delivery, vol. 15, pp. 198–203,
 E. G. teNyenhuis, G. F. Mechler, and R. S. Girgis, “Flux distribution
 M. A. Jones, A. L. Moses, and J. E. Thompson, “Flux distribution and
power loss in the mitered overlap joint in power transformer cores,”
IEEE Trans. Magn., vol. MAG-9, pp. 114–122, June 1973.
 F. Loffler, H. Pfützner, T. Booth, C. Bengtsson, and K. Gramm, “Influ-
ence of air gaps in stacked transformer cores consisting of several pack-
ages,” IEEE Trans. Magn., vol. 30, pp. 913–915, Mar. 1994.
 Z. Valkovic and A. Rezic, “Improvement of transformer core magnetic
properties using the step-lap design,” J. Magn. Magn. Mater., vol. 112,
pp. 413–415, 1992.
 J. C.Olivares,
Guadalajara, Jalisco, Mexico: EMSA, 1998.
 F. J. Vogel and E. J. Adolphson, “A stray loss problem in transformer
tanks,” in Amer. Inst. Elect. Eng. Trans., Aug. 1954, pp. 760–764.
 P. Moreno, “Extra Low Frequency Magnetic Fields Shielding With Fi-
nite Width Planar Shields,” Ph.D., Univ. Washington, 1997.
 K. V. Namjoshi and P. P. Biringer, “Power efficiency of eddy current
shielding of structural steel in transverse magnetic field,” IEEE Trans.
Magn., vol. 25, pp. 2829–2831, Sept. 1992.
former—a review,” in IEEE Power Eng. Soc. Winter Meeting, Paper no.
0.7803-5938-0/00, Singapore, Jan. 2000, pp. 2269–2274.
 Y. Inui, S. Saito, K. Okuyama, and K. Hiraishi, “Effects of tank shields
on magnetic fields and stray losses in transformer windings,” in IEEE
Power Eng. Soc. Summer Meeting Conf., Vancouver, Canada, 1973.
 K. Kazmierski, M. Kozlowski, J. Lasocinski, I. Pinkiewicz, and J.
Turowski, “Hot Spot Identification and Overheating Hazard Preventing
When Designing a Large Transformer,”, Cigre 12-12, 1984.
 C. Yongbin, Y. Junyou, Y. Hainian, and T. Renyuan, “Study on eddy
Trans. Magn., vol. 30, pp. 3068–3071, Sept. 1994.
 D. A. Koppiker, S. V. Kulkarni, P. N. Srinivas, S. A. Khaparde, and
R. Jain, “Evaluation of flitch plate losses in power transformer,” IEEE
Trans. Power Delivery, vol. 14, pp. 996–1001, July 1999.
 A. A. Berezovskii, “Calculation of stray losses in transformer tanks,”
Elektrichestvo, no. 9, pp. 421–448, 1966.
 P. D. Agarwal, “Eddy-current losses in solid and laminated iron,” in
Amer. Inst. Elect. Eng. Trans., May 1959, pp. 169–181.
 R. Feinber, Ed., Modern Power Transformer Practice.
Macmillan, 1979, pp. 128–129.
 P. Georgilakis, N. Hatziargyriou, and D. Paparigas, “Al helps reduce
transformer iron losses,” IEEE Comput. Applicat. Power, vol. 12, pp.
41–46, Oct. 1999.
 E. Wasilenko and M. Olesz, “On-site loss tangent measurements of high
voltage insulation,” in Proc. Dielectric Materials, Measurements and
Applicat., 1992, Sixth Int. Conf., 1992, pp. 170–173.
 H. Kurita, T. Hasegawa, and K. Kimura, “Dielectric loss of high
voltage/high frequency transformers used in switching power supply
for space,” in Proc. 19th Annu. IEEE Power Electron. Specialists Conf.
Rec., vol. 2, 1988, pp. 1120–1126.
 D. J. Ward, “An analysis of loss measurements on older distribution
transformers,” IEEE Trans. Power Apparat. Syst., vol. PAS-103, pp.
2254–2261, Aug. 1984.
 F. Gaudino, Letter (Private Communication), July 6, 1999.
 J. P. Barton, “Interlamination resistance,” in Amer. Inst. Elect. Eng.
Trans., Sept. 1944, pp. 670–672.
 A. C. Beiler and P. L. Schmidt, “Interlaminar eddy current loss in lami-
nated cores,” Amer. Inst. Elect. Eng. Trans., vol. 66, pp. 872–878, 1947.
 N. S. Sheth, “Development in core plate varnishes for the electrical in-
dustry,” Elect. India, pp. 3–6, Nov. 1991.
 A.J. Pansini,
ment.Englewood Cliffs, NJ: Prentice-Hall, p. 25.
 T. Nakata, N. Takahashi, and Y. Kawase, “Magnetic performance of
step-lap in distribution transformer cores,” IEEE Trans. Magn., vol.
MAG-18, pp. 1055–1057, Nov. 1982.
 C. Lee and H. K. Jung, “Two-Dimensional analysis of three-phase
transformer with load variation considering anisotropy and overlapped
stacking,” IEEE Trans. Magn., vol. 36, pp. 693–696, July 2000.
 R. M. Bozorth, Ferromagnetism: IEEE Press, IEEE Magne. Soc., 1978,
Juan Carlos Olivares (M’02) was born in Zamora, Michoacan, Mexico, on
March 26, 1969. He received the Bachelor’s and Master’s degrees from the
Instituto Tecnologico de Morelia, Mexico, in 1995 and 1997, respectively. He
is currently pursuing the Ph.D. degree in electrical engineering at Centro de
Investigación y de Estudios Avanzados (CINVESTAV), Guadalajara Campus,
Currently, he is a Visiting Scholar at Virginia Tech., Blacksburg, VA. He
worked in Electromanufacturas, Mexico, from 1997 to 1999. His main inter-
ests are related to transformers.
826IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 18, NO. 3, JULY 2003 Download full-text
Yilu Liu (SM’99) received the Ph.D. degree from The Ohio State University,
Columbus, in 1989.
Currently, she is a Professor of Electrical Engineering at Virginia Tech.,
Blacksburg, VA. She also has 17 years of experience in transformer modeling
José M. Cañedo (M’91) received the Ph.D. in 1985 from the Moscow Power
Currently, he is a Professor at Centro de Investigación y de Estudios Avan-
zados (CINVESTAV), where he has been since 1997. His main areas of interest
are control of power systems and electrical machines.
Rafael Escarela-Pérez (M’97) was born in Mexico City, Mexico, in 1969. He
received the B.Sc. degree in electrical engineering from Universidad Autonoma
Metropolitana, Mexico City, in 1992, and the Ph.D. degree from Imperial Col-
lege, London, U.K., in 1996.
His research interest includes the modeling of electrical machines.
from K.U. Leuven in 2000 on the finite element solution of coupled thermal-
electromagnetic problems and related applications in electrical machines and
drives, microsystems, and power quality issues.
Currently, he is a postdoctoral research fellow of the Belgian “Fonds voor
Dr. Driesen received the 1996 R&D-award of the Belgian Royal Society of
Electrotechnical Engineers (KBVE) for his Master Thesis on power quality
Pablo Moreno (M’00) received the Ph.D. degree in electrical engineering from
Washington State University, Pullman.
Currently, he is Research Professor in Centro de Investigación y de Estudios
Avanzados (CINVESTAV), Guadalajara Campus, Mexico. His main interests
are related to electromagnetic wave propagation and electromagnetic compati-