A transformer design optimisation tool for

oil immersed distribution transformers

R.P.W.S.Amarasinghe, W.G.K.P.Kumara, R.A.K.G.Rajapaksha, R.A.D.K.Rupasinghe, W.D.A.S.Wijayapala

Department of Electrical Engineering

University of Moratuwa
Katubedda, Sri Lanka
waruna_10@ elect.mrt.ac.lk, kelum_10@ elect.mrt.ac.lk, [email protected], dinusha_10@ elect.mrt.ac.lk and
[email protected]

Abstract—This paper presents a methodology to design transformers by use of fuzzy logic technique. H.Malik, et al
hermetically sealed shell type oil immersed distribution [14] present methodology to minimize the total mass of the
transformers with minimum evaluated cost according to the IEC core and wire material by satisfying constraints imposed by
standards. Distribution transformer model and design international standards and transformer user specification.
constraints are implemented as a user friendly software using
L.D.S.Coelho, et al [15] present methodology to solve
MATLAB. An objective function for total evaluated cost is
optimized subjected to ten constraints according to IEC 60076 in transformer design optimization problem using the differential
addition to geometrical constraints. A design example on a evolution (DE) algorithm.
630kVA transformer is presented for illustration. This paper proposes a design methodology which has
capability to minimize the total lifetime cost (objective
Index Terms— Cost function, Design optimization, Dielectric function) while meeting the conditions laid out by constraints
losses, Eddy currents, Hysteresis, Impedance, Inductance, for three-phase hermetically sealed oil immersed distribution
MATLAB, Transformers, Transformer cores. transformers with capacities from 50kVA to 2500kVA. The
objective of transformer design optimization (TDO) is to
design the transformer so as to minimize the transformer
I. INTRODUCTION manufacturing cost, cost of no-load loss and cost of full-load
Distribution transformers play a vital role in electricity loss subject to constraints according to the international
distribution networks. Transformer losses contribute to a standards and transformer user specification. It is also
significant portion of the total system loses and therefore necessary to consider the manufacturing limitations and to
keeping transformer losses under control is a target of network maintain the elegant appearance of the transformer. Following
operators. A transformer designed with better material, larger constraints are considered in the distribution transformer
cross sections in conductors and the core will have lower optimisation process; flux density constraint, induced voltage
losses but the cost of the transformer will be high. On the constraints, impulse voltage constraints, impedance voltage
other hand a cheap transformer can be manufactured with constraint, no-load loss constraint, full-load loss constraint and
average material and with minimum conductor and core cross total loss constraint. In this optimization methodology 8 major
sections but will have higher no-load losses and higher load design variables are considered for optimisation.
losses. A transformer having higher losses will have higher Most of the equations involved in the transformer
lifetime cost. Therefore a compromise has to be made and manufacturing process are non-linear multivariable functions
optimum transformer parameters have to be used for the and it is difficult to solve these equations using conventional
production of the transformer. By hand calculations this is an methods and have to follow the advance methods like
impossible task and since the invention of the digital Sequential Quadratic programming method (SQP). SQP has
computers, transformer design optimisation tools have been capability to solve non-linear multivariable functions under
developed by various manufacturers and such tools have been constraints on the variables.
the design monopoly of them. The first transformer design The proposed method uses SQP method and gives the
optimisation with the help of a digital computer was global minimum for objective function while satisfying the
implemented in 1955. Since 1955 there have been huge constraints. The presented system is executed and formed by a
revolutions in the transformer industry. In early days designers user interface with 3D visualization of the optimized
optimized its design based on trial and error. Based on those transformer developed in MATLAB.
results finally most of them derived their own methodologies
to optimize the transformer designs.
P.S.Georgilakis [12] presents an innovative method, II. METHODOLOGY
combining genetic algorithm and finite element method for the Table 1 defines the variables used in subsequent equations.
solution of transformer manufacturing cost minimization
problem. H.Malik, and R.K.Jarial, [13] present a new method A. Formulation of the objective function.
to estimate the weight and cost of main materials for

978-1-4799-1740-2/15/$31.00 ©2015 IEEE 

Authorized licensed use limited to: General Electric Global Research. Downloaded on March 01,2021 at 13:00:14 UTC from IEEE Xplore. Restrictions apply.
 The proposed transformer design methodology has ability to
design three phase hermetically sealed shell type oil immersed TEC=CTM+A (NLL) +C (LL) (1)
distribution transformers with following specifications. CTM=CMM +CRM + CLab (2)

kVA rating: 50kVA-2500kVA Low voltage winding's copper foil, high voltage winding's
Primary voltages: 33kV/11kV copper wire, core magnetic material, insulation paper,
Secondary voltage: 400V insulation duct strips, mineral oil and tank steel sheets are
considered as main materials which heavily affect the cost of
the transformer.
TABLE I CRM and CLab are consider as constants for design variables.
DEFINITIONS OF THE DESIGN VARIABLES
Therefore equations are derived to calculate main materials
Symbol Quantity Unit
TEC Total Evaluated Cost USD costs, NLL and LL using the following eight design variables.
CTM Cost of Transformer USD The variable are T(second step width), W(second step build),
Manufacturing H(first step width), n(LV turns), B(flux density),୪୴ (LV foil
A Cost of No load Loss W/USD thickness), J(current density) and †ୌ୚ (HV wire diameter).
NLL No Load Loss W
C Cost of load Loss W/USD Following assumptions are made for the derivations of the
LL Load Loss W equations. The magnetic flux density is constant throughout
CMM Cost of Main Material USD the core. Low voltage (LV) and high voltage (HV) windings
CRM Cost of Remaining Material USD are oval shaped, Material waste costs are ignored. High-
CLab Cost of Labour USD
f Frequency Hz voltage winding is manufactured with copper round
CSF Core Stacking Factor conductor, while low-voltage winding is manufactured with
VLV Voltage in secondary V copper foil.
nLV Turns in secondary
ȡCore Density of core material kg/mm3
J Current density in foil A/mm2 1) Mass of core
DHV Primary conductor diameter mm
HP Pressboard height mm Core cross-section area of the core for star connected
S Transformer capacity kVA secondary is calculated using basic equation [1]
LLV LV conductor length m
ȡcu Cu density kg/mm3
130034.VLV
LHV HV conductor length m Aeff = (3)
WLV LV copper mass kg f.nLV .B.CSF
WHV LV copper mass kg
WIP Insulation paper mass kg Mass of the core is calculated as follows [1],
WDS Duct strip mass kg
WCore Core steel mass kg
tLV,DDP LV diamond dot paper thickness mm Wcore =ሺ3.D+4.Hሻ.Aeff .ȡcore (4)
tHV,DDP HV diamond dot paper thickness mm
LLV,DDP LV diamond dot paper length mm The dimensions of the core are shown in figure 1.
LHV,DDP HV diamond dot paper length mm
ȡDDP Diamond dot paper density kg/mm3
TL Tank length mm DH
TH Tank height mm
TW Tank width mm
Tt Tank thickness mm DD
ȡt Tank steel density kg/mm3
ȇo Mineral oil density kg/mm3
LLLV Secondary load loss W
LLHV Primary load loss W
NLLG Guaranteed No-load loss W
LLG Guaranteed load loss W
TL Total loss W Fig. 1. Dimensions of the Transformer core
TLG Guaranteed total loss W
UG Guaranteed impedance voltage Shape of the winding is assumed as elliptical and
VPTLV LV voltage per turn V parameters of each ellipse are calculated using dimensions of
VPLHV HV voltage per layer V
InLV,max Maximum LV induced voltage V the core, thickness of the pressboard, thickness of a single LV
InHV,max Maximum HV induced voltage V foil, thickness of a single HV layer, Thickness of LV cooling
BILLV LV basic insulation level V duct, Thickness of HV cooling duct, thickness of LV paper
BILHV HV basic insulation level V insulation, thickness of HV paper insulation and thickness of
TLV Total LV turns
THV Total HV layers LV-HV barrier. Circumference of each ellipse is calculated
ImLV,max LV maximum impulse voltage V using each ellipse's parameters.
ImHv,max HV maximum impulse voltage V
USD United State Dollar By adding corresponding circumferences, equations are
derived for the length of LV copper foil, HV conductor length,
Transformers are designed at the minimum evaluated cost. LV insulation paper length, HV insulation paper length, LV
Formula for the objective function is given below. cooling duct, HV cooling duct, LV-HV barrier and press



Authorized licensed use limited to: General Electric Global Research. Downloaded on March 01,2021 at 13:00:14 UTC from IEEE Xplore. Restrictions apply.
board. Figure 2 shows the cross sectional view of the Mineral oil volume(VO) is derived by subtracting all parts
transformer core for an oval shaped winding. inside the tank from the volume of the tank. So mass of the
oil(WO) is derived by,

WO =VO ×ȡO (10)

T
7) No-Load loss

No-Load loss of a transformer include core loss, dielectric

loss, conductor loss in the winding due to excitation current
H and conductor loss due to circulating current in parallel
windings(IEEE 2002).
W
Transformer specific no-load loss (W/kg) for a given
magnetic flux density can be obtained from the no-load loss
curves given by magnetic material manufacturers.
Fig. 2. Cross sectional view of the transformer core Specific no-load loss (WNLL) curve for 0.23mm thick
Nippon silicon steel sheets at 50Hz is given by,
2) Conductor mass of windings
WNLL =0.8398.B3 -2.254.B2 +2.407.B-0.6884 (11)
Winding conductor masses are derived from [1],
NLL=WNLL .Wcore (12)
S.103 .LLV .ȡCU
WLV = (5) 8) Load loss
J.VLV .ξ3

ʌ.d2 L .ȡ
Load losses are incident to the carrying of a specific load.
WHV = HV HV CU (6) Load losses include I2R losses in current carrying parts
4
3) Insulation paper mass (winding, leads, busbars, and bushings), eddy current losses in
conductors and stray losses.
Insulation paper mass ( WIP ሻ is derived by adding LV
(WIP, LV ሻ and HV (WIP, HV ሻ diamond dot paper weights. RCU .LLV.. J.S.103
LLLV = (13)
VLV .ξ3

WIP =WIP,LV +WIP,HV (7)

4.RCU .LHV .S.103
=Hp .൫tLV,DDP .LLV,DDP +tHV,DDP .LHV,DDP ൯.ȡDDP LLHV = (14)
ʌ.d2HV .VLV .ξ3

4) Insulation duct strips mass LL=LLLV +LLHV (15)

Insulation duct strip mass(WDS) is derived by adding

cooling duct strips weights in LV(WC,LV) and HV(WC,HV) to B. Design Constraints formulation methodology
the LV-HV(WC,HV-LV) barrier duct strip weight.
1) No-load loss/Load loss/Total loss
WDS =WC,LV +WC,Hv +WC,HV-LV (8)
NLL<1.5×NLLG 
LL<1.5×LLG 
5) Tank steel sheet mass TL<1.1×TLG 
Dimensions of the steel tank are calculated using the Guaranteed loss values for a standard kVA rating are
dimensions of core, winding and clearances of HV to tank assigned in CENELEC 1992 and design tolerances according
from length side, HV to tank from width side, yoke to tank to the IEC 60076-1.
bottom, top to tap changer base, tap changer to tank cover and
height of tap changer. 2) Short-Circuit impedance
Mass of the tank(WT) is derived by [1],
Inductive part of short circuit impedance is calculated using
WT =2.ሺTL.TW +TL .TH +TH.TW ሻ.TT.ȡT (9) stored energy method [4]. Resistive part is calculated using
load loss [1] and total impedance (UK ሻ‫݄݄ܿ݅ݓ‬should satisfy
6) Mineral oil mass following condition.



Authorized licensed use limited to: General Electric Global Research. Downloaded on March 01,2021 at 13:00:14 UTC from IEEE Xplore. Restrictions apply.
 0.9×UG <UK <1.1×UG  Conventional transformer design optimization method is
known as heuristic technique that assigns many alternative
values to the design variables so as to generate a large number
3) Induced voltages in LV and HV of alternative designs and finally to select the design that
satisfies all the problem constraints with minimum evaluated
The thickness of insulation papers between the winding cost.
layers must withstand the induced voltage test[1]. This is not very accurate, efficient and quick for
4.VPTLV <InLV,max multivariable non-linear objective function with multivariable
4.VPLHV <InHV,max non-linear constraints. So in this study, mathematical
optimization method, Sequential Quadratic Programming is
4) Impulse voltages in LV and HV suggested as it is a powerful technique for solving non-linear
constrained optimization problem which is implemented using
The thickness of insulation papers between the winding MATLAB.
layers must withstand the impulse voltage test[1].
2.BILLV The principal idea of sequential quadratic programming
<ImLV,max (SQP) is the formulation of a quadratic programming sub
TLV
problem for the non-linear objective function based on a
2.BILHV quadratic approximation and by linearizing the nonlinear
<ImHV,max constraints through its partial derivatives. This sub problem
THV
find better approximation to the design variables (xk+1) than
current approximation (xk) and converges it to an optimal
solution through iterative sequence.
5) No-load current
MATLAB Optimization tool provides built-in function
No-load current is equal to the vectorial sum of the
called ‘Fmincon’ where SQP algorithm can be easily
magnetising current (Im) and hysteresis and eddy current (Ih+e).
implemented. In this design optimization method, aim is to
Transformer specific magnetising power (VA/kg) for a given
minimize the objective function that gives total evaluated cost.
induction can be obtained from the curves given by magnetic
Figure 3 shows the flow diagram of the design optimization
material manufacturers. Hysteresis and eddy current
criteria. In this optimization criterion, after selecting the
component is obtained through the no-load loss.
transformer capacity, frequency, LV and HV voltages and
In the design no-load current is constrained to 2% of full load
insulation levels, objective function runs with the “Fmincon”.
current at 112.5% rated voltage.
Then it gives the optimum major design variable values which
gives lowest Total Owning Cost (TOC). Since these variable
6) Efficiency and voltage regulation
values do not exactly match with the standard materials which
are available in the market, it is necessary to match the
Ș (Efficiency) is computed for full-load at unity power
optimum variable values with the available values to find a
factor by[1]:
large number of combinations for the major variables and then
S×100% those combinations are checked with the design constraints.
Ș= (16) Combinations which satisfy all of the constraints are used to
S+LL+NLL
calculate the TOC and finally the solution with the lowest
The Voltage regulation (¨V) is computed for full-load at TOC is selected for the design of the transformer.
unity power factor by[1]:
III. DESIGN EXAMPLE: 630KVA TRANSFORMER
e2x
¨V= ቀer + ቁ ×100% (17) Table 2 shows the optimum values of the major design
2
Where, variables for a 630kVA, 33kV/400V transformer. Table 3
shows the cost of the transformer main materials, Cost of the
LL
er = and ex =ටU2K -e2r no-load loss and Cost of the full-load loss in optimum
S
transformer design. Total owning cost of this transformer is
USD 21105.80. For this design following constraints are
Efficiency of the designs is higher than 98% and voltage
imposed, Guaranteed Load Loss(LLG)<8100W, Guaranteed
regulation is lower than 4%.
NO Load Loss (NLLG)<1290W, Guaranteed Total
Loss(TLG)<6886W,4%<Uk <5%, InLV,max = 10kV
7) Maximum flux density ,InHV,max =30kV , ImLV,max =23.5,ImHV,max =50,B<1.6T.U
nder the above constraints design software gives the optimum
The maximum flux density of the core is required to be transformer design output data which results in the lowest
smaller than the saturated flux density. TOC. Table 4 shows the other parameters that designer should
be concerned in any design. Due to the 3-D visualization
C. Transformer design optimization(TDO) methodology capability that software it gives the optimum transformer
design in 3-D. Designer can a clear idea about what does it



Authorized licensed use limited to: General Electric Global Research. Downloaded on March 01,2021 at 13:00:14 UTC from IEEE Xplore. Restrictions apply.
look like and whether any changes in the dessign are required. The study was based on hermetically sealed type oil immersed
Figure 4 and Figure 5 show the 3-D view of the optimum three phase distribution transform mer design optimization
transformer for 630kVA. method. The proposed method haas many advantages since
transformer manufacturers can dessign the transformer with
TABLE II lowest evaluated cost while enhan ncing the life time of the
MAJOR DESIGN VARIABLE VALUES IN OPTIIMUM DESIGN
transformer by adhering to relevantt transformer standards. In
Variable Values
this paper a methodology is described to optimize the total
T(mm) 117.00
evaluated cost function restricted by a set of constraints. The
W(mm) 40.23 main advantage of the software pacckage are, i)Even relatively
H(mm) 191.00 less experienced engineers can eassily handle, ii)The process
nLV 19.00 uses Matlab , ensuring convergeence to global optimum
B(T) 1.45 iii)Program contains tools for thee 3D visualization of the
୪୴(mm) 0.90 optimum design iv)Program facilitates updating of
J(A) 2.10
transformer materials v)User friendlly input/output interface.
†ୌ୚ (mm) 2.00

TABLE III
630KVA, 33KV/415V TRANSFORMER DESIGGN AT 50 HZ
Item unit cost Amount Cost(USD)
(kg/USD)/
(W/USD)
Core material 2.50 1136.2kg 2,840.0
LV copper 8.50 188.17 kg 1,599.5
HV copper 8.50 262.2 kg 2,228.7
Insulation 4.50 43.32 kg 194.9
Mineral oil 1.72 537.4 kg 924.3
No load loss 7.58 712.03 W 5,397.2
Full load loss(@ 1.52 5211.32 W 7,921.2
75°C)
Total owning cost(USD) 21105.8 Fig. 4. 3-D view of the optimum transformeer for 630kVA, 33kV/415V

TABLE IV
IMPORTANT DESIGN PARAMETE
ERS
Efficiency(%)-full load 98.80
Impedance Voltage (%) 4.66
Voltage Regulation (%) 3.40
No load Current (%) 0.87

IV. CONCLUSION
Most of the TDO methodologies concern on minimization
of transformer main material cost under connstraints. It’s the
conventional method where customer has too evaluate cost of
losses. Summation of it with initial buying cost is the cost,
customer has to pay throughout the life tim me which is not
optimized. In this study we have developed an algorithm and
software to design oil immersed distribution ttransformers with
minimize total evaluated cost. So using this tool transformer Fig. 5. 3-D view of the optimum transformer for 630kVA, 33kV/415V
manufactures able to design distribution transsformers with the
cost satisfying manufacture as well as custom
mer.
Due to the transformer manufactures cconcern about to
reduce the total manufacturing cost of transfoormer. So most of
the designs come up with higher flux denssities and current
densities to reduce material cost. Considering life time of the
transformer it is not an economical to custoomer to buy low
cost transformer with high loss. So minim mization of total
evaluated cost is the most suitable critteria for design
transformers.



Authorized licensed use limited to: General Electric Global Research. Downloaded on March 01,2021 at 13:00:14 UTC from IEEE Xplore. Restrictions apply.
 Start
[3] E.I Amoiralis, P.S.Georgilakis , M. A.Tsili, A.G. Kladas, A.T.
Souflaris, “A complete software package for transformer design
optimization and economic evaluation analysis.”, in Materials
Science Forum., 2010,pp. 535-546
Select transformer [4] N.Jahromi ,J.Faiz, H.Mohseni.( 2003 ,Apr), A fast method for
calculation of transformers leakage reactance using energy
method. International Journal of Engineering. [Online]. 16(1), pp
.41-48 .Available: http://www.ije.ir/abstract/%7BVolume:16-
Transactions:B-Number:1%7D/=40
Load two step objective [5] IEC Power transformers: general, IEC 60076-1, 2001
function and constraints [6] IEC Power transformers: temperature rise, IEC 60076-2, 2001
[7] IEC Power transformers: insulation levels and dielectric tests, IEC
60076-3, 2001
[8] IEC Loading guide for oil-immersed power transformers, IEC
Run with “Fmincon” 60354, 2001
[9] Transformers.Siemens.Berlin,Germani[Online]. Available:
https://www.swe.siemens.com/belux/portal/nl/aanbod/tip/producte
n/Documents/chap_5_1438063.pdf
Optimized [10] Martin J. Heathcote., “The J & P Transformer Book,” A
variables PRACTICAL TECHNOLOGY OF THE POWER
TRANSFORMER, 12th ed, London:Reed Educational and
Professional Publishing Ltd, 1998.
Optimized variables are [11] I.Dasgupta ,” Design of Transformers,”, New delhi:Tata
McGraw-Hill Education, 2002, pp 66-277
[12] P.S.Georgilakis, " Recursive genetic algorithm-finite element
method technique for the solution of transformer manufacturing
Values put into an array cost minimisation problem,” in Electric Power Applications,
IET , 2009,pp. 514 – 519
[13] H.Malik, R.K.Jarial," Fuzzy-Logic Applications in Cost Analysis
Check next of Transformer's Main Material Weight,” in CICN., Gwalior .,
combinatio 2011,pp. 386 - 389
Select combination [14] H.Malik, B.A.Kr, A.K.Yadav, R.K.Jarial," Application Research
Based on Fuzzy Logic to Predict Minimum Loss for Transformer
Design Optimization,” in CICN., Gwalior ., 2011,pp. 207– 211
[15] L.D.S.Coelho, V.C.Mariani, M.V. Ferreira da Luz, J.V. Leite,"
Yes Novel Gamma Differential Evolution Approach for Multi
Reject Check violation objective Transformer Design Optimization,” in Magnetics, IEEE
values of constraints Transactions, 2013,pp. 2121 - 2124

No Check next
combination
Run objective function

F(n)=objective function

No
F(n)<F(n+1)

Yes

End
Minimum=F(n)

Fig. 3. Simplified flow diagram for transformer design optimization method

REFERENCES
[1] P.S. Georgilakis, “Spotlight on Modern Transformer Design,”
,London: Springer-Verlag London Limited, 2009.
[2] J.C.O.Galvan,P.S.Georgilakis ,R.E.Perez
,E.C.Littlewood.(2011,Dec), Optimal design of single-phase
shell-type distribution
transformers based on a multiple design method validated
by measurements. Electrical Engineering . [Online]. 93( 4) , pp.
237-246. Available:
http://link.springer.com/article/10.1007%2Fs00202-011-0211-
9#page-1



Authorized licensed use limited to: General Electric Global Research. Downloaded on March 01,2021 at 13:00:14 UTC from IEEE Xplore. Restrictions apply.