IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO.

2, APRIL 2004 637

A Simple Method for Calculating Core Temperature

Rise in Power Transformers
Simon A. Ryder, Member, IEEE, and Ian J. Vaughan

Abstract—This paper presents a new method for calculating Thermal conductivity of cooling liquid (W/mk).
core temperature rise in large power transformers. A general for- Thermal conductivity of cooling liquid (W/m/K).
mulation of the problem is made using the electrical analogy. This Thermal conductivity of the core material in the
formulation extends to the L- and T-joints. An expression for the
thermal resistance of the core material is found using analytical plane of the laminations (W/m/K).
methods. A comparison of the results of the new method with mea- Thermal conductivity of the core material perpen-
sured values on two power transformers is presented. dicular to the plane of the laminations (W/m/K).
Index Terms—Magnetic cores, power transformer, power trans- Specific loss (of core material) [see below for
former thermal factors, temperature. primes] (W/kg).
Thermal resistance (see below for subscripts)
[km /W].
NOMENCLATURE Ratio of active to nominal core cross-section.
The following symbols are used in the equations which Thickness (of block of generic material, see text for
follow. a more detailed explanation) [in meters].
Cubic expansivity (of cooling liquid) [per Kelvin]. When applied to thermal resistances (as a first subscript) the
Core interior temperature gradient (see below for subscripts COR and OIL refer to thermal resistances of the core
primes) [in Kelvin]. material and the boundary layer in the cooling liquid, respec-
Temperature drop in the cooling liquid boundary tively. When applied to thermal resistances (as a second sub-
layer (in Kelvin). script) or to ratio of total to exposed surface area, the subscripts
Average temperature gradient across the cross-sec- and refer to the directions in the plane of the laminations
tion of the core (see text for a more detailed expla- and perpendicular to the plane of the laminations, respectively.
nation) [K}. When applied to core temperature gradients or specific
Plate edge temperature gradient (see below for losses, a single prime denotes the L-joint and a double prime
primes) [in Kelvin]. the T-joint. When applied to a thermal resistance for the cooling
Weighted average of plate edge and plate surface liquid, the primes denote that this was calculated using the
temperature gradients (see text for a more detailed appropriate temperature drop in the cooling liquid boundary
explanation) [K]. layer.
Plate surface temperature gradient (see below for The average plate width between ducts is given by
primes) [in Kelvin].
Dynamic viscosity (of cooling liquid) [kg/m/s].
Density of core material (kg/m ).
Density of cooling liquid (kg/m ). A typical three-phase transformer core is shown in Fig. 1.
Stacked depth between cooling ducts (in meters). There are a total of six joints between the vertical limbs and
Cross-section between cooling ducts (m ). horizontal yokes. The four joints between the end limbs and the
Average plate width between cooling ducts (m) [see yokes are termed L-joints. The two joints between the center
below for calculation method]. limb and the yokes are termed T-joints.
Specific heat capacity at constant pressure (of Losses are lowest in the limbs and yokes (linear parts of the
cooling liquid) [J/kg/K]. core), higher in the L-joints, and highest in the T-joints [7].
Ratio of exposed to total surface area (see below for In this paper, terms of the kind “core temperature gradient”
subscripts). are understood to mean the difference between the core tem-
Acceleration due to gravity (m/s ). perature and the local cooling liquid temperature. The term
“core interior temperature gradient” is understood to mean
the temperature gradient of the hottest part of the core. The
Manuscript received September 30, 2002. term “plate edge temperature gradient” is understood to mean
S. Ryder was with Alstom Transformateurs de Puissance, Saint-Ouen, France
(now Areva Transformateurs de Puissance) when the work described in this the temperature gradient of the edges of the core laminations.
paper was done. He is now with Doble PowerTest, Guildford GU3 ILZ, U.K. The term “plate surface temperature gradient” is understood
(e-mail: sryder@doble.com). to mean the temperature gradient of the flat surfaces of the
I . Vaughan is with Areva T&D Transformers, Stafford ST17 4LN, U.K.
(e-mail: ian.vaughan@areva-td.com). core laminations. The meaning of these last three terms is
Digital Object Identifier 10.1109/TPWRD.2003.820222 illustrated in Fig. 2.
0885-8977/04$20.00 © 2004 IEEE
638 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO. 2, APRIL 2004

L-Joint
flow equation in the core cross section. As can be imagined,
T-Joint
the algebra involved in this method is extremely complicated,
as is the resulting expression for the temperature distribution in
the core. Furthermore, this method can only be applied to find
temperature rise in linear parts of the core and not in the joints.
An alternative advanced method, proposed by Rele and
Palmer [13], uses a conducting paper analog to represent the
core. Again, this type of modeling can only be used to find
temperature rise in linear parts of the core and not in the joints.
A further possibility would be to use finite element mod-
eling. This would be applied to its best advantage when coupled
to the output of an advanced core loss distribution calculation.
Very detailed, and potentially very accurate, results could be ob-
tained in this way. A model of this type has been proposed by
Limb teNyenhis et al. [5].
Yoke
While finite element modeling has undoubted merits, the de-
tail required to construct a successful model and the time taken
Fig. 1. Sketch of three limb, three-phase transformer core showing L- and
T-joints. to do so make it difficult to apply early in the design process. It is
more suitable for final design verification and or developing new
designs. However, an extension of the simple method first pro-
posed by Küchler using the electrical analogy would allow the
temperature rise in the core joints to be estimated without these
problems. This method could be used throughout the design and
optimization process, and not merely for design verification.
Thus, a reasonable approach would be to use the simple method
Core interior for most design process and one of the advanced methods for
final verification.
In this paper, the author describes a simple method offering
Plate edges
three major improvements over the methods found in the lit-
erature: 1) the new method uses an improved and clarified gen-
Plate surfaces
eral formulation; 2) the treatment of the cooling liquid boundary
Fig. 2. Cross section of a linear part (limb or yoke) of the core.
layer is more rigorous; and 3) the method can be used to find the
temperature rise in the core joints and not just in the linear parts
of the core.
I. INTRODUCTION In the explanation of the new method, which follows, the
author begins with a general formulation of the heat transfer

T HE problem of calculating accurately the core temperature

rise in large power transformers is not a particularly new
one. Its importance may not be obvious since the most widely
equations for a transformer core. The effects of both the
core material and the cooling liquid boundary layer are then
examined individually. A comparison with test results from two
used specifications for large power transformers do not put spe- transformer cores is made. Finally, conclusions are drawn and
cific limits of the core temperature rise [1], [3]. However it is recommendations for future work are made. Recommended
obviously necessary to limit the core temperature to values that values for the thermodynamic properties of materials and an
cause no damage to the core itself, adjacent materials, or the example calculation are presented as appendices.
oil. It has been shown that oil may be degraded by core temper-
atures as low as 110 C–120 C [4], [5]. This has led experts in
the field to suggest that 130 C would be a reasonable limit for II. GENERAL FORMULATION
the core temperature [5].
The simplest method for calculating core temperature found Using the electrical analogy, the losses distributed through
in the literature is that proposed by Küchler [6] (and subse- the core may be replaced by a point source lying at the center
quently republished in its entirety by Karsai [7]). Here the tem- of the part of the core considered. The core material and the
perature distribution in the core cross section is found using a cooling liquid boundary layer can then be represented as sep-
simple superposition of two one-dimensional (1-D) solutions to arate thermal resistances. This process is shown diagrammati-
the heat transfer equation. This method can only be applied to cally in Fig. 3. Expressions for the core interior, plate edge, and
find the temperature rise in linear parts of the core, and not in plate surface temperature rises may then be derived using Ohm’s
the joints. and Kirchoff’s laws.
A more advanced method initially proposed by Cockcroft [8] This gives rise to the following core interior temperature gra-
and subsequently improved by Bucholz [9], Roth [10], [11], and dient in the linear parts of the core: [See (1) shown at the bottom
Higgins [12] uses a solution to the two-dimensional (2-D) heat of the next page.]
RYDER: A SIMPLE METHOD FOR CALCULATING CORE TEMPERATURE RISE IN POWER TRANSFORMERS 639

the average temperature gradient across the cross-section of the

core is given by

(5)

Beginning again with the L-joints, the core interior tempera-

ture gradient is given by (6) shown at the bottom of the page.
Similarly, for the plate edge and plate surface temperature
Fig. 3(a). Cross section of a linear part (limb or yoke) of the core.
gradients
ROILI RCORI RCORI ROILI
(7)

(8)
ROILT RCORT RCORT ROILT

(ab)(ρqs) Beginning again with the T-joints, the core interior tempera-
ture is given by (9) shown at the bottom of the page.
Similarly, for the plate edge and plate surface temperature
gradients

(10)
Fig. 3(b). Equivalent circuit using the electrical analogy.

Similarly for the plate edge and plate surface temperature (11)
gradients

(2) III. CORE MATERIAL

An expression for the thermal resistance of the core material
(3) may be derived by considering a solid slab of material in which
heat is uniformly generated. Let all but two opposite edges of
The weighted average temperature gradient at the plate edges the block be perfectly insulated and the remaining two edges
and plate surfaces is given by (4), at the bottom of the page. held at the same temperature.
Making use of the assumption that the temperature distribu- It is a simple matter to apply Fourier’s law to find the tem-
tion in the core is approximately parabolic, it can be seen that perature distribution within this block. From this, the following

(1)

(4)

(6)

(9)
640 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO. 2, APRIL 2004

expression for the effective thermal resistance of the block can TABLE I
be derived: COMPARISON OF CALCULATED AND MEASURED CORE
TEMPERATURE GRADIENTS

(12)

Note that the above thermal resistance is based on the max-

imum temperature of the block and not the average. [If the av-
erage temperature of the block were used, then the denominator
of the constant on the right-hand side of (12) would be six].
The above can now be used to derive expression for the
thermal resistance of the core. This results in

(13)
(14)

IV. BOUNDARY LAYER TABLE II

THERMAL CONDUCTIVITY OF CORE MATERIAL
The author has already investigated the thermal resistances of
cooling liquid boundary layers in the course of work on winding
temperature gradients [14]. For the case in question (natural
flow), he found the following expression:

the quantity concerned were made). There is generally good

(15) agreement between the calculations and the measurements,
especially for the core interior temperature gradients. The
The above equation can now be used to derive expression for errors are largest for the plate surface temperature gradients.
the thermal resistance of the two cooling liquid boundary layers. This suggests that the translaminar conductivity of the core
This results in may be underestimated.

(16) VI. CONCLUSION

The method described in this paper allows the core temper-
(17) ature gradient in power transformers to be calculated with rea-
sonable accuracy and little effort. It is difficult to say whether
the methods could successfully be applied by other transformer
Naturally, the above can also be used to calculate the thermal manufacturers. The formulation and the method for calculating
resistance of the boundary layers in the cooling liquid at the L- the thermal resistance of the core material should be of universal
and T-joints. It is necessary to substitute the appropriate values validity. The method for calculating the thermal resistance of the
for the plate edge or plate surface temperature gradient. boundary layer may not be applicable to all types of construc-
The thermal resistance of the boundary layer can be seen to tion and all sizes of transformer. Other manufacturers might be
depend on the temperature drop across it. Thus, one cannot be able to use the same approach but with a different version of
calculated without the other. For computer applications, it is (15) and different values for the material properties.
easy to set up an iteration to perform the necessary calculations.
For “pencil and paper” calculations, the iteration is laborious
and it may be preferable to guess a value for the temperature VII. RECOMMENDATIONS
drop across the cooling liquid boundary layer and afterwards The author has obtained good results using simple methods.
verify that it was not too much in error. The results could be improved if a more accurate value for the
translaminar thermal conductivity were available.
V. COMPARISON
APPENDIX A
To assist in the development of the method described in this
THERMODYNAMIC PROPERTIES OF DIELECTRIC MATERIALS
paper, the author’s company made detailed core temperature rise
measurements on two medium-power transformers—one rated Recommended values for the thermal conductivity of the core
at 30 MVA and the other at 60 MVA. A sample calculation based material may be found in Table II. These were originally taken
on the 60-MVA transformer is presented in Appendix B. from [4].
A comparison of the test results with the results of calculations Recommended values for the thermodynamic properties of
using the method proposed in this paper may be found in Table I. oil may be found in Table III. These were originally taken from
(A blank column indicates that no reliable measurements of a table in [4], and are for a reference temperature of 75 C.
RYDER: A SIMPLE METHOD FOR CALCULATING CORE TEMPERATURE RISE IN POWER TRANSFORMERS 641

TABLE III These, in turn, give the following value for the core interior
THERMODYNAMIC PROPERTIES OF OIL temperature gradient:

Assuming that the specific losses in the T-joints are 2.5 times
the specific losses in the linear parts of the core [7], the temper-
ature gradients of the plate edges and surfaces as obtained by
iteration are
APPENDIX B
WORKED EXAMPLE
The example chosen is the core of the 60-MVA transformer. These give the following thermal resistances:
The core cross-section is 0.153 435 m . It is divided by a single
duct. The stacked depth on either side of the duct is 295 mm. OIL
The equivalent plate width is
OIL
These, in turn, give the following value for the core interior
The specific losses in the core are 0.740 W/kg (supplier’s temperature gradient:
data). The density of the core material is 7650 kg/m and the
ratio of nominal to effective cross-sectional area is 0.97. The
ratio of exposed to total surface area at the plate edges and sur-
faces is assumed to be 0.6. ACKNOWLEDGMENT
For the core material, this gives The authors gratefully acknowledges the contributions of H.
Karadogan of Areva Elektrik (Gebze, Turkey) and J. Dykes, for-
COR merly of Areva T&D Transformers (Stafford, U.K.).
COR The authors would like to thank Areva Transformers de Puis-
The temperature drops in the two boundary layers are equal sance, by whose permission this work has been put into the
to the temperature gradients of the plate edges and surfaces. By public domain.
iteration, these are
REFERENCES
[1] K. Karsai, D. Kerényi, and L. Kiss, Large Power Transformers, Ams-
terdam: Elsevier, 1987, pp. 29–80.
[2] IEEE Standard General Requirements for Liquid-Immersed Distribu-
These give the following thermal resistances: tion, Power, and Regulating Transformers.
[3] Power Transformers–General, 2nd ed: IEC Publication 76-1, 1993.
OIL [4] T. V. Oommen, R. A. Ronnou, and R. S. Grigis, “New mechanism of
moderate hydrogen generation in oil filled transformers,” in CIGRE,
OIL paper 12-206, 1998.
[5] , “Hydrogen generation of some oil-immersed cores of large power
These, in turn, give the following value for the core interior transformers,” presented at the Int. Conf. Doble Clients, Boston, MA,
temperature gradient: 1998.
[6] E. G. teNyenhuis, R. S. Girgis, G. F. Mechler, and G. Zhou, “Calculation
of core hot-spot temperature in power and distribution transformers,”
IEEE Trans. Power Delivery, vol. 17, pp. 991–995, Oct. 2002.
Finally, the weighted average temperature gradient at the [7] R. Küchler, Die Transformatoren, Berlin: Springer, 1956, pp. 252–260.
[8] J. D. Cockcroft, “The temperature distribution in a transformer or lam-
edges and surfaces of the laminations and the average tempera- inated core of rectangular cross-section in which heat is generated at
ture gradient across the cross-section of the core are as follows: a uniform rate,” in Proc. Cambridge Philosophical Soc., vol. 22, July
1925, pp. 759–772.
[9] H. Von Buchholz, “Die zweidemensionale Wärmestronüng des Be-
harrungszustande in rechteckigem querschnitt Querschnitt geblätterter
Eisenkörper bei flächenhaft, unsteig oder steig verteilten Wärme-
quellen,” Zeitschrift für Angewandte Mathematik und Mechanik, vol.
Assuming that the specific losses in the L-joints are 1.5 times 14, no. 5, pp. 285–294, 1934.
the specific losses in the linear parts of the core [7], the temper- [10] E. Roth, “Introduction à l’étude analytique de l’échauffement des ma-
ature gradients of the plate edges and surfaces as obtained by chines électriques,” Bull. Soc. Franc. Elect., vol. 7, pp. 840–954, 1927.
[11] P. Hammond, “Roth’s method for the solution of boundary-value prob-
iteration are lems in electrical engineering,” Proc. Inst. Elect. Eng., vol. 114, no. 12,
pp. 1969–1976, Dec. 1967.
[12] T. J. Higgins, “Formulas for calculating temperature distribution
in transformer cores and other electrical apparatus of rectangular
cross-section,” AIEE Trans., vol. 65, pp. 190–194, 1945.
These give the following thermal resistances: [13] A. Rele and S. Palmer, “Cooling of large transformer cores,” IEEE
Trans. Power App. Syst., vol. PAS-91, pp. 1527–1535, 1972.
[14] S. A. Ryder, “A simple method for calculating winding temperature gra-
OIL dient in power transformers,” IEEE Trans. Power Delivery, vol. 17, pp.
OIL 977–982, Oct. 2002.
642 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO. 2, APRIL 2004

Simon A. Ryder (M’03) was born in the U.K. in 1973. He received the M.Eng. Ian J. Vaughan (A’01) was born in 1957 and continued his education at the
degree in engineering science from St. John’s College, Oxford, U.K., in 1996. University of Aston, Birmingham, U.K.
Currently, he is with Doble PowerTest, Guildford, U.K. He was with Areva Currently, he is a Project Manager and Developer of transformer design soft-
(formerly Alstom) from 1997 to 2003. He was also a Development Engineer ware with Areva T&D, Stafford, U.K., where he is a designated expert. He was
at Areva Transformateurs de Puissance (formerly Alstom Transformateurs de involved in transformer design and development of calculation and software for
Puissance), Saint-Ouen, France, when the work described in this paper was the past 19 years.
done. His main research interests are transformer condition and life manage-
ment, including the use of frequency response analysis.
Mr. Ryder is a member of the IEE (U.K.).