You are on page 1of 7

Paper accepted for presentation at 2003 IEEE Bologna Power Tech Conference, June 23th-26th, Bologna, Italy

Numerical Calculations of Three-Phase


Transformer’s Transients
Amir Tokic, Non-Member, Vlado Madzarevic, Non-Member and Ivo Uglesic, Non-Member

on absolute values of the magnetic fluxes


Abstract—A three-phase, two-winding transformer’s model is Φ s ( j )1 ,Φ s ( j )2 ,...,Φ s ( j )N .
presented in the paper. The triplex core configuration is assumed
in the transformer’s model. Nonlinear magnetizing curves are
piecewise linearized. Input winding capacitances can be lumped
to the transformer’s terminals. A stiff differential system in state Φ s( j )
3
Lm( j )
space, describing transformer’s transient behavior, is solved by Φ s( j )
2
N

the L-stable backward differentiation formulas numerical rule. Φ s( j ) Lm( j )

magnetic flux Φ(j)


2
1
BDF compared to trapezoidal rule shows better stability Lm( j )
1
properties. A computer program is developed for generation of
variable state waveforms. The developed program is suitable for
simulations of low-frequency three-phase transformer transients Lm( j )
1
such as inrush currents and ferroresonance. Results of the −Φ s ( j )
Lm( j )
developed program are compared to MATLAB/Power System 2
1

Lm( j ) −Φ s ( j )
Blockset results. N
2
−Φ s ( j )
3

Index Terms—Inrush currents, ferroresonance, modeling, state


space methods, stiff differential equations, transformers, magnetizing current im(j)
transient analysis.
Fig. 1. Nonlinear magnetizing curve
I. INTRODUCTION

T HE main nonlinear element in the transformer’s model is


a ferromagnetic inductance L m . The inductance L m is
II. NUMERICAL CALCULATIONS OF THREE-PHASE
TRANSFORMER’S TRANSIENTS
defined by gradient dΦ / di m in any point of nonlinear
The simplified electrical circuit for analysis of low-
magnetizing curve: magnetizing current-magnetic flux. The
frequency three-phase transformer’s transients is shown in
curve is piecewise linearized, Fig.1, which gives input vectors
[ ]
T
Fig. 2.
of inductances Lm( j ) = Lm( j )1 , Lm( j )2 ,...,Lm( j )N and fluxes
Lumped elements Breaker Transformer
Φ ( ) = [Φ ( ) ,Φ ( ) ,...,Φ ( ) ]
Network
T
s j s j 1 s j 2 s j N
where the subscript j = 1,2 ,3
denotes phase number. Vectors Lm ( j ) and Φ s ( j ) are obtained
from the standard unloaded transformer’s tests, [1].
Magnetizing currents on the k-th linear region of magnetizing
curves, k = 1,2 ,..., N are calculated according to the equation Fig. 2. Simplified electrical circuit for transient analysis
[2]:
The power network is represented by voltage sources
Φ ( j) k −1  1 1 
i m ( j )k =
L m ( j )k
(
− sgn Φ ( j ) )∑ Φ s ( j )i 
 Lm( j )

L 
(1) [
E = [e1 ,e2 ,e3 ]T = Em sin ωt , Em sin ωt − 120o , Em sin ωt − 240o ( ) ( )]
T

i =1  m ( j )i +1 
i
and their corresponding network impedance z = R N + jωL N .
where: N represents a total number of the piecewise linear Lumped parameters pertain to parameters of overhead lines,
regions. During transformer’s transients inductances cables, capacitors of shunt filters, stray capacitances etc.
L m( j )1 , L m ( j )2 ,..., L m ( j )N are switched on and off depending The transformer’s parameters per phase are as follows: active
resistances Rtr ( j ) as well as leakage inductances Ltr ( j ) of
primary windings, iron core losses R m ( j ) and transformer’s
A. Tokic is with the Faculty of Electrical Engineering, Tuzla University, ferromagnetic inductances L m( j ) . Depending on number of
Tuzla, 75000, Bosnia and Herzegovina (e-mail: atokic2001@yahoo.com).
V. Madzarevic is with the Faculty of Electrical Engineering, Tuzla energy accumulator elements a state space vector could be
X = [x1 (t ), x 2 (t ),..., x n (t )]T , where
University, Tuzla, 75000, Bosnia and Herzegovina (e-mail:
vlado_madzarevic@yahoo.com). defined as x i (t ) ,
I. Uglesic is with the Faculty of Electrical Engineering and Computing, i = 1,2 ,..., n are required voltages and currents as well as the
Zagreb University, Zagreb, 10000 Croatia (e-mail: ivo.uglesic@fer.hr).

0-7803-7967-5/03/$17.00 ©2003 IEEE


magnetic flux. Based on the defined vector X , the behavior a)
in electrical circuits could be described by system of linear Start
differential equations in state space on the k-th region:
dX/dt = Ak X + bk (2) Loading system
The next state variables are the magnetic fluxes: parameters
x i (t ) = Φ ( j ) (t ) , i = N − 2 , N − 1, N , j = 1,2 ,3 .
Steady state calculation
(initial conditions calculation)
Developed computer program resolves the equation system
(2) for separate linear regions. Input data are actual values of t = T0
the magnetic fluxes.

Considering that the system (2), for a real data model, defines t = t + dt
a system of “stiff” differential equations the rigidity of
differential equations makes explicit numerical tasks very State variable vector X(t) calculation
hard to solve the same equations successfully [3-4]. Explicit BDFp rule (4)
tasks applied to “stiff” equations are numerically unstable,
which implies an increase of truncation error and leads to
method divergence. Equations (2) traditionally are resolved by Moving on piece-wise regions of
magnetizing curves per phases 1,2,3
the absolutely stable trapezoidal rule [5]: (Routine)
−1
 h   h  h 
X n + 1 =  I − Ak    I + Ak  X n + [bk (tn ) + bk (tn +1 )] (3) No
 2   2  2  t ≤ T final

n = 0 ,1,2 ,... Yes


Depending on value of ratio eigenvalues of state matrixes
Ak : ξ = λ max / λ min , the rule (3) can be useed. However, Plotting of state
variable vector X(t)
for ξ → ∞ this rule breaks down and some other techniques
had to be used [3-5]. Trapezoidal numerical rule is not L-
stable and for matrixes Ak with large negative real part of End
b)
eigenvalues, this rule generates numerical oscillations. This
kind of problems can appear in numerical calculations of the Routine
transformer transients [6-7]. Avoiding numerical oscillations
is ensured by application of absolutely and L-stable backward j = 1: 3
differentiation formulas (BDF), numerical rule [6]. BDFp of
the p-th order applied to equation dX / dt = F ( X , t ) is
numerical rule based on the relation: No
Φ ( j ) ≥ Φ s ( j )k
p
1
∑m∇
m =1
m
X n +1 = hF ( X n +1 ,t ) (4) Yes

Depending on stability angle in the complex plane rule with No


Φ ( j ) ≥ Φ s ( j )k +1
order p ≤ 5 can be chosen. Order of truncation error of BDFp
method is h p +1 [8-9]. Yes
No
Note that BDFp method based on equations (4) completely Φ ( j ) ≤ Φ s ( j )k −1
avoids numerical oscillations for real model parameters for
acceptable integration steps, orders h = 10 −4 ÷ 10 −6 sec . Yes
Trapezoidal rule (3) reduces numerical oscillation with k = k +1 k = k −1
significantly decreased integration step in comparison to BDF
rule. For a complete suppression of the numerical oscillations
or for achievement of the same accuracy, the calculation time
with BDF rule is shorter then the time of calculation with the
trapezoidal rule. Return

A simplified flow chart named “Transf-Transient”, which Fig. 3. Simplified flow chart “Transf-Transient”
a) main program
satisfies earlier mentioned conditions, is shown on the Fig. 3. b) routine for movement on piece-wise region
III. TEST CASES r1 Rg Rg  r2 Rh Rh  1 / Lm(1) 0 0 
     k

This chapter presents low-frequency three-phase R1 = Rg r1 Rg  , R2 =  Rh r2 Rh  , LMk =  0 1 / Lm(2)k 0  ;
transformer’s inrush currents and ferroresonance transients   R R r   
analysis. Rg Rg r1   h h 2  0 0 1 / Lm(3)k 
Different approaches are used in simulation of low
frequency transformer’s transients, inrush currents and where: r1 = R N + R g , r2 = Rtr + R m + R h .
ferroresonace studies [10 -16]. k −1  
A. Three-phase transformer inrush currents
[
S k = s (1)k , s (2 )k , s (3 )k ]T
, s( j )
k
( )∑Φs( j)  L 1
= sgnΦ( j ) i

1 
Lm( j )i +1 
i =1  m( j )i

During energization of unloaded transformers the inrush


Real data values for grounded three-phase power system, are:
current can reach the multiple value of the rated current and it
may last tens of seconds or even several minutes. - power supply parameters (Tuzla):
Fig. 4 presents typical electrical model for the transformer’s E m = 220 2 / 3 kV , L N = 0.281 H , R N = 8.82 Ω .
inrush current analysis.
- lumped element (overhead line Tuzla-Kladanj):
The behavior of electrical circuits in Fig. 4 is described by the
C = 4.128 pF .
equations in form (2), where the state variable vector is:
- transformer parameters (Kladanj 220 / 110 kV ):
X = [I N ,U C , I tr ,Φ ]T ,
• nominal power S tr = 200 MVA ,
while internal vectors are: • short circuit voltage u k % = 15 % ,
[ T
] [
I N = iN (1) , iN (2 ) , iN (3 ) , UC = uC (1) , uC (2 ) , uC (3 ) , ] T • resistance per winding phase R1 = 0.529 Ω ,

I tr = [itr (1) , itr (2 ) , itr (3 ) ] , Φ = [Φ(1) ,Φ(2 ) ,Φ (3 ) ] .


T T • leakage inductance L1 = 0.126 H ,
• iron core losses R m = 5.76 MΩ .
Ak is a 12-by-12 state matrix, b k is 12-by-1 free state Nonlinear magnetizing curve is represented with five
vector: piecewise linear regions, Table I:

 1 1  TABLE I
− L R1 − L U 0 0   1  MAGNETIZATION CURVE OF 200 MVA TRANSFORMER
 N N   L E 
 1 1   N 
 U 0 − U 0   o  i [p.u.] 0 0.005 0.015 0.03 0.075 1.0
Ak =  C C  , bk =  R .
 1 1 R   m
S  Φ [p.u.] 0 1.05 1.08 1.1 1.12 1.39
 0 U− R2 m LM k   Ltr k 
Ltr Ltr Ltr  
 
 0 0 RmU − Rm LM k  − R m S k  Computation with described model without transformer
remanent magnetic flux and with the integration step
dt = 0.5 ⋅ 10 −4 sec gives eigenvalues ratio:
U is an 3-by-3 unit matrix, 0 is an 3-by-3 zero matrix, o is an
3-by-1 zero vector, while matrix R1 , R 2 and L M k are: ξ = λ max / λ min = 9.0623 ⋅ 10 8 .

RN LN iN(1) itr(1) Rtr Ltr

im(1)
e1(t) Rm Lm

t=T0 im(2) im(3)


Lm Lm
e3(t) e2(t)
Rg iN(2) itr(2) Rtr Ltr
RN LN
Rm Rm
Rh
RN LN iN(3) itr(3) Rtr Ltr

C uC(1) C uC(2) C uC(3)

Fig. 4. Three-phase transformer’s energization-equialent model


Numerical oscillations cannot be avoided when traditional Transformer’s inrush currents are in a good agreement, Fig. 6.
trapezoidal rule is applied (3), Fig. 5.a. BDFp numerical rule However, for the transformer primary voltages
(4) gives a stable numerical solution for the same integration MATLAB/PSB shows numerical oscillations, while “Transf-
step, Fig. 5.b. Transient” program results is avoiding them, Fig.7.

a) a)
'Trapezoidal rule' 'Transf-Transient'
1000 1500
ph.1 ph.1
ph.2 ph.2
750 1000 ph.3
ph.3

inrush currents [A]


inrush currents [A]

500
500

250
0
0

-500
-250

-500 -1000
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
time [sec] time [sec]
b) b)
'BDF rule' 'MATLAB/PSB'
1000 1500
ph.1 ph.1
ph.2 ph.2
750
inrush currents [A]

ph.3 1000 ph.3


inrush currents [A]

500
500
250

0
0

-250 -500

-500
0 0.01 0.02 0.03 0.04 0.05 -1000
time [sec] 0 0.02 0.03 0.01
0.04 0.05
time [sec]
Fig. 5. Inrush currents: a) trapezoidal and b) BDF rule
Fig. 6. Three-phase transformer inrush currents:a) ‘’Transf-Transient’’ and
b) ‘’MATLAB/PSB’’ results
Situation is additionally deteriorated for equivalent circuits
modeling of isolated system with relatively large impedance
R g and R h when trapezoidal rule is used. Due to increasing a)
5 'MATLAB/PSB'
values of eigenvalues ratio ξ = λ max / λ min , numerical 2.5
x 10

oscillations will be amplified. BDF rule solves the same 2


transformer voltage [V]

problem without numerical oscillations and with acceptable 1.5


steps of integration.
1
0.5
Further, for verification of the developed algorithm “Transf-
Transient” program calculation results are compared to results 0
from MATLAB/Simulink/Power System Blockset [17] and -0.5
presemted in Fig. 6. From MATLAB/PSB solver library -1
ode15s (stiff/NDF) numerical method is chosen as the most
-1.5
suitable. In this case, the transformer is energized at the
-2
moment T0 = 1.5 m sec , and the remanent magnetic fluxes are
-2.5
assumed to be per phases Φ r (1) = 0.68Φ nom , 0 0.0025 0.005 0.0075 0.01 0.0125 0.015
time [sec]
Φ r (2 ) = −0.68Φ nom , Φ r (3 ) = 0 . Integration step
dt = 0.5 ⋅ 10 −4 sec is chosen.
b)  1 1 
− L R1 − L U o 0 0
5 'BDF rule'
x 10   1 
2.5  N N  L E 
2  CC 0 o C C1 0   N 
   o 
1.5  
1 , bk =  0  .
transformer voltage [V]

Ak =  oT oT 0 − u oT 
1 C
   Rm 
0.5  1 1 T 1 Rm   Sk 
 0 U 1 u − R2 LMk   Ltr 
0 Ltr Ltr Ltr Ltr
  − R S 
-0.5  0 0 o RmU − Rm LMk   m k
 
-1
-1 5
-1.5 U is an 3-by-3 unit matrix, 0 is an 3-by-3 zero matrix, o is an
-2 3-by-1 zero vector, while matrix R1 , R 2 , L M k , C C , C C1 ,
-2.5 U 1 and u are:
0
0.0025 0.005 0.0075 0.01 0.0125 0.015
time [sec]
r1 Rg Rg  r2 Rh Rh  1 / Lm(1) 0 0 
Fig. 7. Transformer voltage in phase 1: a) ‘’MATLAB/PSB’’ and b) ‘’Transf-
     k

Transient’’ results R1 = Rg r1 Rg  , R2 =  Rh r2 Rh  , LMk =  0 1 / Lm(2)k 0  ,
  R R r   
B. Three-phase transformer ferroresonance Rg Rg r1   h h 2  0 0 1 / Lm(3)k 
1 / CN 0 0   0 0 0 
Ferroresonance is a nonlinear phenomenon that can lead to   1  
very large system overvoltages. The feroresonance not only CC =  0 1 / (CN + C) 0  , CC =  0 − 1 / (CN + C) 0 ,
characterises the jump to a higher current fundamental  0 0 1 / (CN + C)  0 0 −1 / (CN + C)
frequency state but also bifurcations to subharmonic, quasi-
0 0 0 
periodic and even chaotic oscillations in any circuit containing
a nonlinear inductor. Here is considered a ferroresonance of U 1 = 0 1 0  , u = [1 0 0 ] ;
three-phase transformer that appears during switching-off one 0 0 1
of the phases, Fig. 8.
where:
Similar to earlier considerations (Chapter A) the behavior of
electrical circuits in Fig. 8 is described by the equations in the r1 = R N + R g , r2 = Rtr + R m + R h .
state space form (2), where the state variable vector is: k −1  
[
X = I N ,U C , utr (1) , I tr ,Φ ]T . [
S k = s(1)k , s(2 )k , s(3 )k ]T
( )∑Φs( j)  L 1
, s( j )k = sgnΦ( j ) i

1 
Lm( j )i +1 
i =1  m( j )i
Now, internal vectors are:
[ ] [ ]
I N = iN (1) , iN (2 ) , iN (3 ) T , UC = uC (1) , uC (2 ) , uC (3 ) T , For grounded three-phase supply network and isolated
I tr = [itr (1) , itr (2 ) , itr (3 ) ] , Φ = [Φ(1) ,Φ(2 ) ,Φ (3 ) ] .
T T transformer neutral point, real data values are [18]:
E m = 110 2 / 3 kV , L N = 0.0193 H , R N = 0.605 Ω .
Ak is an 13-by-13 state matrix, b k is 13-by-1 state vector: - lumped elements:
C = 6 µF , C = 0.78 µF .

RN LN itr(1) Rtr Ltr

iN(1) im(1)
e1(t) Rm Lm

t=T0 im(2) im(3)


Lm Lm
e3(t) e2(t)
Rg itr(2) Rtr Ltr
RN LN
Rm Rm
iN(3) iN(2) Rh
RN LN itr(3) Rtr Ltr

CN uC(1) CN uC(2) CN uC(3) C utr(1) C uC(2) C uC(3)

Fig. 8. Three-phase transformer ferroresonance-equialent model


- transformer parameters: same way it is possible to realize a program for analyzing any
• nominal power S tr = 60 MVA , other low-frequency transient processes for a non-linear
• short circuit voltage u k % = 12 % , transformer. The stiff differential equation system is solved by
the BDFp rule. It is noticed that the program avoids numerical
• resistance per winding phase R1 = 0.02 Ω , oscillations during solution of real problems in cases when the
• leakage inductance L1 = 0.0771 H , use of traditional trapezoidal rule is inappropriate. Developed
• iron core losses R m = 100835 Ω . program is compared to MATLAB/Simulink/Power System
Blockset results. Comparison shows that developed program
• inceptive saturation point Φ s = 1.2 p.u .
has similar computing time and in some cases better numerical
stability properties then MATLAB/PSB.
The simulation of feroresonance is conducted for the model
depicted on the Fig. 8. with the iintegration step
dt = 10 −5 sec . The phase 1 is switched-off at the moment
T0 = 35 m sec and the results are shown in the Fig. 9. V. REFERENCES
The last figure shows good correspondence of transformer
voltages achived with “Transf-Transient” and [1] W. L. A. Neves and H. W. Dommel, “Saturation curves of delta-
MATLAB/PSB programs. connected transformers from measurements”, IEEE Trans. Power
Systems, vol. 8, pp. 417-425, May 1993.
5
x 10 'Transf-Transient' [2] A. Tokic, V. Madzarevic and I. Uglesic, “Computer modelling of non-
4 linear transformer iron core in calculations of transiets”, in Proc. 2002
ph.1
ph.2
ICEM Inernational Conf. on Electrical Machines, pp. 440-445. Brugge,
3 Belgium, Aug. 2002.
ph.3
[3] H.W. Dommel, Electromagnetic Transients Program, Reference
2
manual, (EMTP Theory Book), Bonneville Power Administration,
transformer voltages [V]

1
Portland, Oregon, 1986.
[4] R. C. Aiken, Stiff Computation, Oxford University Press, New York,
0 1985.
[5] J. R. Cach, Stable Recursion with Applications to the Numerical
-1 Solution of Stiff Systems, Academic Press, New York, 1979.
[6] A. Tokic, I. Uglesic and F. Jakl, “An algorithm for calculations of low-
-2 frequency transformer transients”, to be published in Proc. 2003 IPST
International Conf. on Power Systems Transients, Hong Kong, June
-3 2003.
[7] J. M. Vakilian, R. C. Degeneff, M. Kupferschmid, “Computing the
-4 internal. transient voltage response of a transformer with nonlinear core
0 0.02 0.04 0.06 0.08 0.1
time [sec] using Gear's method”, IEEE Trans. Power Delivery, vol. 10, pp. 1836-
5 'MATLAB/PSB' 1842, Oct. 1995.
x 10 [8] E. Hairer, G. Wanner, Solving Ordinary Differential Equations II: Stiff
4
ph.1 and Differential-Algebraic Problems, Springer-Verlag, Berlin, 1991.
3 ph.2 [9] R. Fazio, “Stiffness in numerical initial-value problems: A and L-
ph.3
transformer voltages[V]

stability of numerical methods”, Internat. Journal of Mathematical


2 Education in Science and Technology, vol. 32, pp. 752-760, Nov. 2000.
[10] J. J. Rico, E. Acha and M. Madrigal, “The study of inrush current
1 phenomenon using operational matrices”, IEEE Trans. Power Delivery,
vol. 16, pp. 231-237, Apr. 2001.
0
[11] M. Rioual, C. Sicre A., “Energization of a no-load transformer for power
-1 restoration purposes: Impact of the sensitivity to parameters”, in Proc.
2001 IPST Inernational Conf on Power Systems Transients, pp. 221-
-2 227. Rio de Janeiro, Brasil, June 2001.
[12] C. G. A. Koreman, “Determination of the magnetizing characteristic of
-3 three-phase transformers in field tests”, IEEE Trans. Power Delivery,
vol. 4, pp. 1779-1785, July 1989.
-4 [13] K. S. Smith, L. Ran, B. Leyman, “Analysis of transformer inrush
0 0.02 0.04 0.06 0.08 0.1
time [sec] transients in offshore electrical systems”, IEE Proc. Gener. Transm.
Distrib., vol. 146, pp. 89-95, Jan. 1999.
Fig.9. Transformer voltages, ferroresonance event in phase 1: "Transf-
[14] B. A. Mork, “Application of nonlinear dynamics and chaos to
Transient" and "MATLAB/PSB"
ferroresonance in distribution systems”, IEEE Trans. Power Delivery,
vol. 9, pp. 1009-1017, Apr.1994.
Note that equialent model in Fig. 8. can decribe all types of [15] N. Janssens, Th. Van Craenenbroeck, D. Van Dommelen, F. Van De
three-phase transformer ferroresonance represented in [19- Meulebroeke “Direct Calculation of the stability domains of three-phase
ferroresonance in isolated neutral networks with grounded-neutral
20]. voltage transformers”, IEEE Trans. Power Delivery, vol. 11, pp. 1546-
1553, July1996.
IV. CONCLUSIONS [16] D. A. Woodford, “Solving the ferroresonance problem when
compensating a dc converter station with a series capacitor”, IEEE
The developed program presented in this paper is suitable Trans. Power Systems, vol. 11, pp. 1325-1331, Aug. 1996.
[17] Power System Blockset User’s Guide, Natick: TEQSIM International,
for simulations of low-frequency three-phase transformer Hydro-Quebec & The MathWorks, Aug. 2001.
transients such as inrush currents and ferroresonance. In the
[18] A. Schei., A. Ekstrom, “Stresses on metal oxide surge arresters in
HVAC systems by temporary and transient overvoltages and related
tests”, in Proc. 1986 CIGRÉ International Conference on Large HV
Electrical Systems, Paris, 1986, pp. 2-20.
[19] P. Ferracci, (1998, March), “Ferroresonance”, Groupe Schneider: Cahier
technique No 190, pp. 1-28. [Online]. Available:
http://www.schneiderelectric.com/en/pdf/ect190.pdf,
[20] Working Group, “Modeling and analysis guidelines for slow transients –
part III: The study of ferroresonance”, IEEE Trans. Power Delivery, vol.
15, pp. 255-265, Jan. 2000.

Amir Tokic was born in Tuzla, Bosnia and


Herzegovina, in 1970. He received B.Sc. and
M.Sc. degrees from the Faculty of Electrical
Engineering in Tuzla, 1996 and Faculty of
Electrical Engineering and Computing in
Zagreb, 2001, respectively. He is currently
working towards a Ph.D. degree at the
University of Zagreb, Croatia. His areas of
interest include power system transients,
nonlinear elements modeling and power quality.
Vlado Madzarevic was born in Tuzla, Bosnia
and Herzegovina, in 1953. He received M.Sc.
and Ph.D. degrees from the Faculty of Electrical
Engineering in Sarajevo, 1987 and Faculty of
Electrical Engineering in Tuzla, 1998,
respectively. He is currently working as a
Professor at the Tuzla University, Bosnia and
Herzegovina. His areas of interest include
numerical calculations of forces and torques in
electrical machines and electromagnetic fields.

Ivo Uglesic was born in Zagreb, Croatia, in


1952. He received Ph.D. degrees from the
Faculty of Electrical Engineering and
Computing in Zagreb, 1988. Presently, he is a
Professor of the Department of High Voltage
and Power Systems of the University of Zagreb,
Croatia. His areas of interest include high
voltage engineering and power transmission.

You might also like