Available online at www.sciencedirect.

com

Electric Power Systems Research 78 (2008) 441449

Evaluation of surge-transferred overvoltages in distribution transformers

M. Popov a, , L. van der Sluis a , R.P.P. Smeets b,c
a Delft University of Technology, Faculty of Electrical Engineering, Power Systems Laboratory, Mekelweg 4, 2628CD Delft, The Netherlands
b KEMA T&D Testing, Utrechtseweg 310, 6812 AR Arnhem, The Netherlands
c Department of Electrical Engineering, Eindhoven University of Technology, Eindhoven 5600 MB, The Netherlands

Received 3 July 2006; received in revised form 3 January 2007; accepted 27 March 2007
Available online 14 May 2007

Abstract
The paper presents an analysis of very fast-transient overvoltages that occur because of the capacitive surge transfer from the high-voltage (HV)
transformer winding to the low-voltage (LV) transformer winding.
The study is done on a 6.6 kV single-phase test transformer. By applying a pulse with a short rise time at the HV terminal, the voltage at the LV
side is measured and calculated. The voltage distribution along the LV winding is calculated by applying the transmission line theory, so that the
foils of the LV winding are represented as transmission lines. For the studied transformer with a nominal transformer ratio of 95.6, a maximum
voltage ratio of 3.3 was measured during the rate of rise of the applied impulse. The present paper also shows that in some specific cases, the
mutual inductance between the primary and the secondary side can be ignored during determination of the transients along the secondary winding.
The computation and measurement of the voltage at the LV side is validated by making use of the vector fitting method.
2007 Elsevier B.V. All rights reserved.

Keywords: Overvoltage; Fast surges; Transmission line modelling; Transformer

1. Introduction
Nowadays there are reports of transformer and motor fail-
ures during switching with vacuum circuit breakers (VCB) or
gas insulating switchgears (GIS). There have been a number
of studies done to determine the origin of the failure and in
the recent years such progress has been made by studying the
propagation of surges inside transformer and motor windings.
There are various reasons for the insulation failure. The insu-
lation can suffer because of the high amplitude of the voltage,
the high rate of rise or both. Very often the utilities ask not only
to provide a proper protection but also to find the reason for
the occurrence of these failures. For example in Ref. [1], it was
reported that the amplitude of the inter-turn voltage can reach
0.25 times the applied sinusoidal voltage at resonance frequency.
It is known that during switching off and switching on highly
inductive loads, because of the ability of the VCB to clear high-
frequency currents, multiple reignitions can take place. Multiple
reignitions contain oscillations with a broad frequency range.
Corresponding author. Tel.: +31 15 278 6219; fax: +31 15 278 1182.
E-mail address: M.Popov@tudelft.nl (M. Popov).
When they enter the transformer, a resonance might occur when
the frequency of the oscillation matches one of the resonance
frequencies of the transformer.
In Ref. [2], a transformer failure was reported during ener-
gizing a high power transformer. The reason was an external
resonance between the feeder cable and the transformer that
resulted in very high overvoltages on the LV transformer side.
So far there was significant research done on the investi-
gation of transients in transformers [38]. In Refs. [3,4], the
lumped parameter analysis was used for transformer modelling,
and [5,6] deal with the modelling of transformer based on the
well-known Rabins method. It was also shown that transformer
can be modelled by considering the windings or group of wind-
ings as transmission lines [7,8]. Furthermore, Wilcox et al. [9]
proposed a model that can be generally applied for multi-phase
multi-winding transformers. Recent research performed by Gus-
tavsen [10,11] resulted in a transformer model based on the
measured admittance matrix of the transformer. In this way, a
general representation of the transformer in a wide frequency
range can be done. This is called vector fitting and it can be used
to validate the measured results with full success.
Lightning surge studies are based on the observation of the
surge transfer from the HV to the LV transformer side. For exam-

0378-7796/$ see front matter 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.epsr.2007.03.015
442 M. Popov et al. / Electric Power Systems Research 78 (2008) 441449

ple, surge-transferred overvoltages in single-phase installations this theory, transformer turns or a group of turns can be repre-
caused by lightning were investigated by Dugan and Smith [12]. sented by transmission lines. If N is the number of lines, and Z
-
In Ref. [13], it was shown that sometimes, simple models that and Y- are the impedance and the admittance matrix of the lines,
can be implemented into EMTP can be used to simulate the respectively, the relation between voltages and currents in the
surge-transferred overvoltages more accurately. lines is expressed through N voltage and N current equations:
The present paper describes a fundamental study of surge-
d2 V- d2 I-
transferred overvoltages in a distribution transformer. For this
2
= Z
- Y- V- , = Y- Z
- I- (1)
purpose a single-phase transformer under no-load with a trans- dx dx2
former ratio 6600/69 (V/V) is used [14]. Applying the modal theory [18], the relation between the
According to the IEC standard 60076-3, transferred overvolt- currents and the voltages can be represented by Eq. (2):
ages have capacitive and inductive character [15]. The capacitive     
transfer depends on the surge capacitance. The steeper the slope I- S A B V- S
= (2)
of the applied surge, the higher the effect of the surge capac- I- R B A V- R
itance. The inductive transfer of the surge voltage depends on
the flow of surge current in the HV winding, and it is less sensi- where
tive on the rate of rise of the applied voltage. However in Refs. A = Y- S- 1 coth( l)S- 1 , B = Y- S- 1 cosech( l)S- 1 (3)
[16,17], it is shown that depending on the transformer ratio, - - - -
the inductive-transferred voltages are not much smaller than the where I- S , I- R are current vectors at the sending and the receiving
capacitive-transferred voltages. end of the line, V- S , V- R the voltage vectors at the sending and
In the present case, the applied pulse at the HV side is with a the receiving end of the line, S- the matrix of eigenvectors of
the matrix Z 2
50 ns rise time and an amplitude of 50 V. The recorded current in - Y- , - the eigenvalues of the matrix Z - Y- and l is the
the HV terminal is with an amplitude of 0.25 A, and it is damped length of the line.
in less than 1 s after applying the pulse. A preliminary EMTP All parameters in (3) except l are frequency-dependent. In
study on the same transformer has shown that when ignoring this way, the general telegraphers equations, which are dif-
the surge capacitance, the voltage in the LV winding follows the ferential time-dependent equations, can be solved as ordinary
transformer ratio and it is approximately 100 times lower than frequency-dependent equations. The representation of the TLM
the applied voltage. for transformer modelling is represented in Fig. 1.
However, the recorded voltage at the LV terminal is much Applying the equality between the voltage and the current
higher than that determined by the transformer transfer ratio. at the receiving end of a specific line, and the voltage and the
Therefore, in this work, the full analysis is performed. The current at the sending end of the next line the following matrix
primary and the secondary transformer side are modelled by equation can be derived:
transmission lines. For the particular transformer, it is shown    
I- S1 V Si
that the inductive-transferred voltages are not high and can be = [F] - , i = 1, 2, . . . , n (4)
neglected. The voltages along the LV winding are calculated by 0 V- Rn
representing the turns as short transmission lines [14]. The com- where n is the total number of lines. In the present case
putations and measurements are verified by applying the vector n = N1 + N2 . Since, the transformer winding is grounded, V- Rn =
fitting method [10,11]. 0. To eliminate the divergence, the computations are done with
a small impedance Z = 109  connected to the ground. Elimi-
2. Transformer model based on transmission line nating the current I- S1 , the voltages in all lines can be calculated
analysis by
 
V- S1
For determination of the voltage along the winding, the trans- - Si ] = [ Hi1 H ] 0
[V (5)
mission line modelling (TLM) has been applied. According to

Fig. 1. Representing the transformer by transmission lines.

 M. Popov et al. / Electric Power Systems Research 78 (2008) 441449 443

where subscript i = 2, 3, . . ., n, and Hi1 is a vector of order

(n 1) 1. The other sub-matrix H does not have any physical
meaning because it is always multiplied by a zero vector. In (5),
the elements of Hi1 are

FFk+1,1
Hk = , k = 1, 2, . . . , n 1 (6)
FF1,1

where FF is the inverse matrix of the matrix F. Also, V- S1 is the

voltage at the sending end of the first transmission line that in
the present case is equal to the source voltage.
The time-domain solution that results from the inverse
Fourier transform is:

1 sin(/)
VSi (t) = VSi (b + j) e(b+j)t d (7)
2 /

for i = 2,3, . . ., n and the VSi (b + j) = Hi1 (b + j)VS1 (b + j).

In (7), the interval [, ], the smoothing constant b
Fig. 3. Description of the LV winding layers.
and the step frequency length d must be chosen prop-
erly in order to arrive at an accurate time-domain response.
Since the measurements are in a range of a few microsec- By substituting (9) in (8), V- S1 is
onds and the full voltage is achieved within 50 ns, the jCHL FF(1, 1)
time step is taken to be 10 ns. So the frequency interval is V- S1 = VS (10)
1 + jCHL FF(1, 1)
= 2 108 rad/s, and d = 2 104 rad/s. The constant b nor-
mally varies 103 < b < 105 , and for fast transients is taken to be 3. Transformer parameters
b = 105 . More explanation about the choice of these parameters
can be found in Ref. [18]. The studied transformer is a single-phase layer-type trans-
Let us consider that the mutual inductances between the pri- former. The secondary winding is a foil winding with 31 turns.
mary and the secondary windings are ignored, and only the surge The design of the transformer is such that the inner winding is
capacitance CHL between the primary and the secondary wind- the LV winding while the outer winding is the HV winding. The
ing is taken into account. In order to obtain the distribution of data and dimensions of the transformer are shown in Table 1.
voltages along the secondary winding, the transformer can be For the particular transformer, capacitances-to-ground of the
represented by Fig. 2. It can be seen that the primary winding transformer layers are small because only a small surface of
and mutual inductances between the HV- and the LV winding
are ignored. Table 1
Description of the LV winding is presented in Fig. 3. From Transformer geometry/data
(4) and (5) one notes that Transformer power 15 kVA
Transformer ratio 6600 V/69 V
V- S1 = FF(1, 1)IS1 (8) Short circuit voltage 310.3 V
Short circuit losses 332.5 W
The current in the capacitor CHL that is the current IS1 is No-load losses 57.2 W
No-load current 37.3 mA
Number of layers (HV side), N1 15
I- S1 = jCHL (VS VS1 ) (9) Number of turns in a layer 200
Inner radius of HV winding 73.3 mm
External radius of HV winding 97.4 mm
Inner radius of the LV winding 51 mm
External radius of LV winding 67.8 mm
Wire diameter 1.16 mm
Double wire insulation 0.09 mm
Distance between layers 0.182 mm
Coils height 250 mm
Top/bottom distance from the core 10 mm
Dielectric permittivity of oil 2.3
Dielectric permittivity of wire insulation 4
Number of foils (LV side), N2 31
Dielectric permittivity of foil insulation 3.0
Foil thickness 0.3 mm
Double foil insulation 0.24 mm
Fig. 2. Representing low-voltage winding by transmission lines.
444 M. Popov et al. / Electric Power Systems Research 78 (2008) 441449

the winding is below the core. The greatest are the capaci- Table 3
tances between the foils and therefore they are responsible for Foil-to-foil capacitance at the LV-side (107 F)
the propagation of the surges along the winding. The capac- CLL1 0.1168738
itance matrices for the HV- and the LV side are formed as CLL2 0.1159321
follows: CLL3 0.1149904
CLL4 0.1140487
CLL5 0.1131070
Ci,i is the capacitance of layer i to ground and the sum of all CLL6 0.1121653
other capacitances connected to layer i, CLL7 0.1112235
Ci,j is the capacitance between layers i and j taken with a CLL8 0.110281
CLL9 0.109340
negative sign (i = j). CLL10 0.108398
CLL11 0.107456
The capacitance matrix has at the diagonal, upper diago- CLL12 0.106515
nal and lower diagonal elements non-zero values and all other CLL13 0.105573
CLL14 0.104631
elements are zeros. The full capacitance matrix of the all trans-
CLL15 0.1036898
former as shown in Fig. 1 can be written as CLL16 0.102748
  CLL17 0.101806
CHH CHL CLL18 0.100864
C= (11) CLL19 0.099922
CLH CLL
CLL20 0.098981
CLL21 0.098039
In (11), diagonal elements are capacitance matrices of the HV- CLL22 0.097097
and the LV side, respectively, whilst off-diagonal elements rep- CLL23 0.0961561
resent capacitance matrices between the HV windings and the CLL24 0.095214
LV windings. Since the present analysis is done on a layer- CLL25 0.094272
to-layer basis for the HV side, and turn-to-turn basis for the CLL26 0.093330
CLL27 0.092389
LV side, matrices CHH and CLL are square matrices of order CLL28 0.091447
N1 N1 and N2 N2 , respectively. Matrix CHL is a matrix of CLL29 0.090505
order N1 N2 . It can be seen that CLH = CTHL . Hence, the order CLL30 0.089564
of C is (N1 + N2 ) (N1 + N2 ). Also, all elements of the matrix CLL31 0.0887266
CHL are equal to zero except the element CHL (1,1) = CHL ; that
is the surge capacitance between the primary and the secondary
winding. Transformer capacitances are calculated by represent-
ing the adjacent layers (foils) as cylindrical capacitors ignoring If we neglect the core losses, then the inductance matrix can
the edge effects. The same results can be achieved by applying be calculated directly from the capacitance matrix as
formulas for plate capacitors. The calculated capacitances are
represented in Tables 2 and 3. C1 2
Correspondingly, the full inductance matrix of the trans- L= l (13)
v2s
former can be written as
 

LHH LHL where vs = c/ r , r is the relative dielectric permittivity and
L= (12)
LLH LLL l is the mean turn length. So, if the capacitance is expressed in
Farads, the inductance matrix results in Henry. Another way to
calculate the inductance matrix accurately, is to use the explicit
Table 2
Layer-to-layer capacitance at the HV-side (107 F)
Maxwell formulas [14,19]. Because in the present case, the LV
winding is made of foils, the inductance matrix LLL can easily be
CHH1 0.15249997 calculated by Eq. (13). The inductance matrices LHV and LHL
CHH2 0.15542747
CHH3 0.15835498
are calculated by making use of the Maxwell formulas. The
CHH4 0.16128248 primary winding has 15 layers with 200 turns/layer. Therefore,
CHH5 0.16420998 the inductance of a particular layer can be represented as a square
CHH6 0.16713748 matrix of the inductances of one layer. The total inductance of
CHH7 0.17006499 a layer will be equal to the sum of all elements of the matrix
CHH8 0.1729924
CHH9 0.1759199
[20]. If we set the LLH and LHL matrices to zero, then we get
CHH10 0.1788474 the case when the mutual inductances between the HV- and the
CHH11 0.1817749 LV winding are ignored. The calculated matrices LLL and LLH
CHH12 0.1847025 are shown in Appendix A. The matrix LHV can be found in Ref.
CHH13 0.1876300 [14].
CHH14 0.1905575
CHL 0.00362
The Z - and Y- matrices which are the self and mutual
impedances and the admittances between the lines are calculated
 M. Popov et al. / Electric Power Systems Research 78 (2008) 441449 445

Table 4 used. This is approximated by an augmented problem:

Measuring equipment
Pulse generator 500 V maximum voltage
N
c- n
  + D + sE

n=1 s a- n
Oscilloscope Le Croy 9304 A, 200 MHz, 100 MS/s
(s)Yt (s)
Voltage probes Philips 100:1; 20 M//24 pF N (17)
Impedance analyzer HP-4194 A (s)  c
-n + 1
s a- n
n=1
by [19]: From (17), the following equation is valid:
    N 
2 N
c- n  cn
-
- =
Z j +
0 d 2
L, Y- = (j + tan )C (14)
s a- n
+ D + sE =
s a- n
+ 1 Yt (s) (18)
i=1 i=1

where d is the distance between the layers, is the conductor
conductivity and tan is the loss tangent of the insulation.
Furthermore, the application of (14) means that the proximity
and dielectric losses are also taken into account.
The distance between the layers d will be different for the
diagonal and off-diagonal elements of the inductance matrix.
For matrix LLL , d = 0.272 mm, for LHH , d = 0.24 mm, and for
LHV , d = 5.6 mm that is the distance between the primary and
secondary winding.
4. Measurements
The listed equipment used for performing measurements of
the fast transients in the transformer and impedance charac-
teristics is given in Table 4. The time-domain measurements
are done with a digital oscilloscope. When dealing with fast-
transient measurements, one must be sure that the measurements
are correctly done and there is no EMC interference. On the other
hand, frequency-dependent impedance characteristic measure-
ments are straightforward and when these characteristics are
known, by applying the vector fitting method one can check if the
measurements are correctly done. The admittance characteristics
are measured by impedance analyzer HP 4194A.
5. Application of the vector tting method
When terminal voltages are of interest, the transformer can
be fully represented by its terminal admittance matrix Y- t which
defines the relations between the terminal voltages V- and the
terminal currents I- :
Eq. (18) is linear in its unknowns c- n , D, E, c-n . Writing (18)
for several frequency points gives the overdetermined lin-
ear problem Ax = B where the unknowns are in the solution
vector x.
The stage 2 provides the residue identification, by solving
the original Eq. (16) with the zeros of (s) as new poles a- n for
Yt (s). This again gives an overdetermined linear problem Ax = B
where the solution vector x contains the unknowns c- n , D and E.
The process continues until the convergence is achieved. More
about the solution of the problem and initial conditions can be
found in Ref. [10].
When the studied transformer is represented as a two-port
circuit like in Fig. 4, then the diagonal elements of the Y- t matrix
can be directly measured. The relation between the terminal
currents and voltages is
    
I1 Y11 Y12 V1
= (19)
I2 Y21 Y22 V2
In the studied case, the end of the primary and the secondary
winding of the transformer are connected [6]. This is also shown
in Fig. 4.
Since the matrix is symmetric, the off-diagonal elements can
be determined from the measured no-load admittance. When the
LV winding is open, the measured admittance at the HV terminal
is the no-load admittance Yin . Taking into account that the current
I2 = 0, from (19) it can be easily found that the off-diagonal
element is

Y12 = (Y11 Yin )Y22 (20)

I- (s) = Y- t V- (s) (15)

Y- t is a square matrix, the elements of which are frequency-

dependent and can be written as
Np
 c- n
Yt (s) = D + sE + (16)
s a- n
n=1

where Np is a positive integer, s = j and all other constants

are determined in two stages by applying the following fitting
procedure. In the stage 1, pole identification is done. First, start-
ing poles a- n are specified in (16), and Yt (s) is multiplied by an
unknown function (s). For (s), a rational approximation is Fig. 4. Measurement of the diagonal elements of the Y- t matrix.
446 M. Popov et al. / Electric Power Systems Research 78 (2008) 441449

Fig. 7. Synthesization of electrical network.

Fig. 5. Comparison between measured and fitted parameters of the Y- t matrix
(magnitude).
role. This is because the inductive transfer is determined by
the transformer ratio that for the studied transformer is 0.0104.
The capacitive transfer which can be approximately determined
as CHV /(CHV + C2 ) 0.52 is much more dominant in this case
[16,17].
Fig. 9 depicts the measured, computed and simulated results.
The TLM in this case is applied according to Fig. 2. The calcu-
lated voltage is in good agreement with the measured voltage.
The voltage at the LV terminal oscillates with a frequency of
about 1.3 MHz.
The study presented here shows that TLM is suitable for
analysis of surge-transferred overvoltages. Based on this
approach, the voltages in the other turns are calculated. The
ratio between the source voltage and the secondary voltage,
when the transformer is stressed by a steep impulse voltage
is UHV /ULV = 50/15 3.33. That means that for the studied
transformer, the voltage at the secondary side is about 28 times

Fig. 6. Comparison between measured and fitted parameters of the Y- t matrix

(phase).

Figs. 5 and 6 present the elements of the Y- t terminal admit-

tance matrix fitted by the vector fitting method. These elements
are fitted by Np = 80 fractions as shown in the admittance rational
function (16). Based on these results, each admittance rational
function is synthesised by RLCG parallel elements and imple-
mented in the Alternative Transient Program (ATP) [11]. The
equivalent RLCG network is shown in Fig. 7. The elements of
the parameters are given in Appendix B.

6. Results and discussion

Fig. 8 shows the difference between the computed results

(according to Fig. 1) when a surge with a rise time of 50 ns is
applied at the transformer terminal. It can be seen that for the
studied case, the mutual inductance between the primary and Fig. 8. Comparison between the computed voltages at the LV terminal with and
the secondary side of the transformer does not play a significant without mutual inductances according to Fig. 1.
 M. Popov et al. / Electric Power Systems Research 78 (2008) 441449
Fig. 9. LV-terminal voltage comparison; measured, simulated by vector fitting
and simulated by TLM (Fig. 2).
higher in amplitude than the rated secondary voltage. This is
dangerous not only for the transformer itself, but also for the
equipment connected to the transformer low-voltage terminals.
The studied case presents the worst case scenario when a steep
surge is applied directly to the transformer terminal and the
transformer is unloaded. This is for instance possible in arc
furnace systems where the transformer is installed close to the
switchgear. When a cable is connected to the primary side of
the transformer, the voltages will probably be less steep than in
the case presented here.
Fig. 10. Computed inter-turn voltages.
447
When fast surges reach the transformer terminal, inter-turn
and inter-layer insulation might suffer severe stress because of
the amplitude and the steepness of the voltage transients. Apply-
ing the TLM, inter-turn voltages can be computed as shown in
Fig. 10.
It can be seen that the maximum inter-turn voltage is 0.8%
from the applied voltage. This means that in case of an applied
surge impulse with a voltage equal to the nominal voltage of
6.6 kV, a maximum inter-turn voltage of about 50 V can be
expected. Since the nominal voltage on the LV winding is 69 V,
it is clear that a value of approximately 50 V/turn is dangerous
for the LV winding.
7. Conclusion
An analysis of surge-transferred overvoltages in transform-
ers by applying the transmission line theory is presented. The
inter-turn voltages of the LV transformer side are calculated
and they are much higher than the voltages that normally occur.
It is shown that the surge-transferred overvoltages maybe one
of the reasons for transformer failure as the induced voltage
on the LV terminal during fast transients is much higher than
the nominal voltage. In this work, the HV- and the LV side
are analyzed by representing the windings with transmission
lines. Since the inductive-transferred ratio is equal to the trans-
former ratio (69/6600), the influence of this coupling can be
neglected and the voltages in the LV winding can be studied with-
out taking into account the mutual inductance. The experience
[16,17] shows that when the nominal ratio of the transformer
(UnHV /UnLV ) is lower, then inductive-transferred overvoltages
must not be neglected. Therefore, for a particular transformer
it is important to compare the coefficient of the inductive-
transferred voltage to the coefficient of the capacitive-transferred
voltage. The computed and measured terminal voltage of
the LV side is verified by making use of the vector fitting
method.
Acknowledgements
The authors would like to acknowledge the Dutch Scientific
Foundation NWO for financial research support of this project
under grant VENI, DET.6526. The authors also thank Kema
High Voltage Laboratories and the High Current Laboratory at
the Eindhoven University of Technology for providing the nec-
essary equipment and for using their facilities. The suggestions
and comments from Professor A. Ametani, Doshisha University,
Japan are highly appreciated.
448 M. Popov et al. / Electric Power Systems Research 78 (2008) 441449

Appendix A

A.1. L22 matrix ( 108 H)

1.72
1.73 1.79
1.75 1.80 1.86
1.76 1.82 1.87 1.93
1.78 1.83 1.89 1.95 2.00
1.79 1.85 1.90 1.96 2.02 2.08
1.81 1.86 1.92 1.98 2.04 2.09 2.15
1.82 1.88 1.94 1.99 2.05 2.11 2.17 2.23
1.84 1.90 1.95 2.01 2.07 2.13 2.19 2.25 2.31
1.85 1.91 1.97 2.03 2.09 2.15 2.21 2.27 2.33 2.39
1.87 1.93 1.99 2.05 2.11 2.17 2.23 2.29 2.35 2.41 2.48
1.89 1.94 2.00 2.06 2.12 2.19 2.25 2.31 2.37 2.44 2.50 2.56
1.90 1.96 2.02 2.08 2.14 2.20 2.27 2.33 2.39 2.46 2.52 2.59 2.65
1.92 1.98 2.04 2.10 2.16 2.22 2.29 2.35 2.41 2.48 2.54 2.61 2.68 2.74
1.94 2.00 2.06 2.12 2.18 2.24 2.31 2.37 2.43 2.50 2.57 2.63 2.70 2.77 2.84
1.95 2.01 2.07 2.14 2.20 2.26 2.33 2.39 2.46 2.52 2.59 2.65 2.72 2.79 2.86 2.93
1.97 2.03 2.09 2.16 2.22 2.28 2.35 2.41 2.48 2.54 2.61 2.68 2.75 2.82 2.89 2.96 3.03
1.99 2.05 2.11 2.17 2.24 2.30 2.37 2.43 2.50 2.57 2.63 2.70 2.77 2.84 2.91 2.98 3.05 3.13
2.00 2.07 2.13 2.19 2.26 2.32 2.39 2.45 2.52 2.59 2.66 2.73 2.80 2.87 2.94 3.01 3.08 3.15 3.23
2.02 2.09 2.15 2.21 2.28 2.34 2.41 2.48 2.54 2.61 2.68 2.75 2.82 2.89 2.96 3.03 3.11 3.18 3.26 3.33
2.04 2.10 2.17 2.23 2.30 2.36 2.43 2.50 2.57 2.64 2.70 2.77 2.85 2.92 2.99 3.06 3.14 3.21 3.29 3.36 3.44
2.06 2.12 2.19 2.25 2.32 2.39 2.45 2.52 2.59 2.66 2.73 2.80 2.87 2.94 3.02 3.09 3.16 3.24 3.32 3.39 3.47 3.55
2.08 2.14 2.21 2.27 2.34 2.41 2.48 2.54 2.61 2.68 2.75 2.82 2.90 2.97 3.04 3.12 3.19 3.27 3.35 3.42 3.50 3.58 3.66
2.10 2.16 2.23 2.29 2.36 2.43 2.50 2.57 2.64 2.71 2.78 2.85 2.92 3.00 3.07 3.15 3.22 3.30 3.38 3.45 3.53 3.61 3.69 3.78
2.12 2.18 2.25 2.32 2.38 2.45 2.52 2.59 2.66 2.73 2.80 2.88 2.95 3.02 3.10 3.17 3.25 3.33 3.41 3.49 3.56 3.65 3.73 3.81 3.89
2.13 2.20 2.27 2.34 2.40 2.47 2.54 2.61 2.69 2.76 2.83 2.90 2.98 3.05 3.13 3.20 3.28 3.36 3.44 3.52 3.60 3.68 3.76 3.84 3.93 4.01
2.15 2.22 2.29 2.36 2.43 2.50 2.57 2.64 2.71 2.78 2.86 2.93 3.00 3.08 3.16 3.23 3.31 3.39 3.47 3.55 3.63 3.71 3.80 3.88 3.97 4.05 4.14
2.17 2.24 2.31 2.38 2.45 2.52 2.59 2.66 2.74 2.81 2.88 2.96 3.03 3.11 3.19 3.26 3.34 3.42 3.50 3.58 3.66 3.75 3.83 3.92 4.00 4.09 4.18 4.27
2.19 2.26 2.33 2.40 2.47 2.54 2.61 2.69 2.76 2.83 2.91 2.98 3.06 3.14 3.21 3.29 3.37 3.45 3.53 3.62 3.70 3.78 3.87 3.95 4.04 4.13 4.22 4.31 4.40
2.22 2.28 2.35 2.42 2.50 2.57 2.64 2.71 2.79 2.86 2.94 3.01 3.09 3.17 3.24 3.32 3.40 3.48 3.57 3.65 3.73 3.82 3.90 3.99 4.08 4.17 4.25 4.35 4.44 4.53
2.24 2.31 2.38 2.45 2.52 2.59 2.66 2.74 2.81 2.89 2.96 3.04 3.12 3.20 3.28 3.36 3.44 3.52 3.60 3.68 3.77 3.85 3.94 4.03 4.12 4.20 4.29 4.39 4.48 4.57 4.67

A.2. L21 matrix ( 105 H)

1.082 1.063 1.045 1.027 1.009 0.991 0.973 0.956 0.939 0.922 0.905 0.888 0.872 0.855 0.839
0.823 0.807 0.792 0.776 0.761 0.746 0.731 0.716 0.702 0.687 0.673 0.659 0.645 0.631 0.617
0.605 1.079 1.060 1.042 1.024 1.006 0.988 0.971 0.954 0.936 0.919 0.903 0.886 0.869 0.853
0.837 0.821 0.805 0.790 0.774 0.759 0.744 0.729 0.714 0.700 0.685 0.671 0.657 0.643 0.629
0.616 0.604 1.072 1.054 1.035 1.017 1.000 0.982 0.965 0.948 0.931 0.914 0.897 0.880 0.864
0.848 0.832 0.816 0.801 0.785 0.770 0.755 0.740 0.725 0.710 0.696 0.681 0.667 0.653 0.639
0.626 0.612 0.600 1.065 1.047 1.029 1.011 0.994 0.976 0.959 0.942 0.925 0.908 0.892 0.875
0.859 0.843 0.827 0.811 0.796 0.780 0.765 0.750 0.735 0.720 0.706 0.691 0.677 0.663 0.649
0.635 0.622 0.608 0.597 1.058 1.040 1.023 1.005 0.987 0.970 0.953 0.936 0.919 0.903 0.886
0.870 0.854 0.838 0.822 0.806 0.791 0.776 0.761 0.746 0.731 0.716 0.702 0.687 0.673 0.659
0.645 0.632 0.618 0.605 0.593 1.052 1.034 1.016 0.999 0.981 0.964 0.947 0.930 0.914 0.897
0.881 0.865 0.849 0.833 0.817 0.802 0.786 0.771 0.756 0.741 0.726 0.712 0.698 0.683 0.669
0.655 0.642 0.628 0.615 0.601 0.590 1.045 1.028 1.010 0.993 0.975 0.958 0.941 0.925 0.908
0.892 0.875 0.859 0.843 0.828 0.812 0.797 0.782 0.766 0.752 0.737 0.722 0.708 0.693 0.679
0.665 0.652 0.638 0.624 0.611 0.598 0.586 1.039 1.021 1.004 0.987 0.969 0.952 0.936 0.919
0.903 0.886 0.870 0.854 0.838 0.823 0.807 0.792 0.777 0.762 0.747 0.732 0.718 0.704 0.689
0.675 0.661 0.648 0.634 0.621 0.607 0.594 0.583 1.033 1.015 0.998 0.981 0.964 0.947 0.930
0.914 0.897 0.881 0.865 0.849 0.834 0.818 0.803 0.787 0.772 0.757 0.743 0.728 0.714 0.699
0.685 0.671 0.658 0.644 0.631 0.617 0.604 0.591 0.580 1.026 1.009 0.992 0.975 0.958 0.941
0.925 0.908 0.892 0.876 0.860 0.844 0.829 0.813 0.798 0.783 0.768 0.753 0.738 0.724 0.710
0.695 0.681 0.668 0.654 0.640 0.627 0.614 0.601 0.588 0.576 1.020 1.003 0.986 0.969 0.952
0.936 0.919 0.903 0.887 0.871 0.855 0.839 0.824 0.808 0.793 0.778 0.763 0.749 0.734 0.720
0.706 0.691 0.677 0.664 0.650 0.637 0.623 0.610 0.597 0.584 0.573 1.014 0.997 0.980 0.963
0.947 0.930 0.914 0.897 0.881 0.866 0.850 0.834 0.819 0.804 0.789 0.774 0.759 0.744 0.730
0.716 0.701 0.687 0.674 0.660 0.646 0.633 0.620 0.607 0.594 0.581 0.570 1.008 0.991 0.974
0.957 0.941 0.925 0.908 0.892 0.876 0.861 0.845 0.829 0.814 0.799 0.784 0.769 0.755 0.740
0.726 0.711 0.697 0.684 0.670 0.656 0.643 0.629 0.616 0.603 0.590 0.578 0.566 1.002 0.985
0.968 0.952 0.935 0.919 0.903 0.887 0.871 0.856 0.840 0.825 0.809 0.794 0.780 0.765 0.750
0.736 0.722 0.707 0.693 0.680 0.666 0.652 0.639 0.626 0.613 0.600 0.587 0.574 0.563 0.996
0.979 0.963 0.946 0.930 0.914 0.898 0.882 0.866 0.851 0.835 0.820 0.805 0.790 0.775 0.760
0.746 0.732 0.717 0.703 0.689 0.676 0.662 0.649 0.635 0.622 0.609 0.596 0.584 0.571 0.560
 M. Popov et al. / Electric Power Systems Research 78 (2008) 441449 449

Appendix B [10] B. Gustavsen, A. Semlyen, Rational approximation of frequency domain

responses by vector fitting, IEEE Trans. Power Deliv. 14 (3) (1999)
10521061.
The branch can be represented by an electrical network as
[11] B. Gustavsen, Computer code for rational approximation of frequency
shown in Fig. 7. R0 and C0 are calculated as dependent admittance matrices, IEEE Trans. Power Deliv. 17 (4) (2002)
1 10931098.
C0 = E, R0 = (B.1) [12] R. Dugan, S.D. Smith, Low-voltage-side current-surge phenomena in
D single-phase distribution transformer systems, IEEE Trans. Power Deliv. 3
Each real pole gives an RL-branch: (2) (1988) 637647.
[13] R.J. Bachega, M.L.B. Martinez, Transformer modeling for transferred volt-
a 1 ages, in: IEEE/PES Transmission Distribution Conference & Exposition:
R1 = , L1 = (B.2) Latin America, 2004, pp. 355359.
c c
[14] M. Popov, L. van Der Sluis, R.P.P. Smeets, J. Lopez-Roldan, Analysis of
Each complex conjugate pair: very fast transients in layer-type transformer windings, IEEE Trans. Power
Deliv. 22 (1) (2007) 238247.
c + jc a ja
+ (B.3) [15] IEC60076-3: Power Transformers. Part 3. Insulation Levels, Dielectric
s (a + ja ) s (a ja ) Tests and External Clearances in Air, http://www.iec.ch, IEC 2000.
[16] A. Ametani, et al., Field test and EMTP simulation of transient voltage
Gives an RLC-branch: when clearing a transformer secondary fault, IEEJ Trans. B-118 (4) (1998)
1 381388.
L=  (B.4) [17] A. Ametani, et al., Theoretical analysis of transient recovery voltage when
2c clearing a transformer secondary fault, Electr. Eng. Jpn. 134 (3) (2001)
R1 = (2a + 2(c a + c a )L)L (B.5) 4452.
[18] J.P. Bickford, N. Mullineux, J.R. Reed, Computation of Power System
1 Transients, IEE, Peter Peregrinus Ltd., 1976, ISBN 0901223859.
= (a + a + 2(c a + c a )R)L
2 2
(B.6) [19] Y. Shibuya, S. Fujita, High frequency model of transformer winding, Electr.
C
Eng. Jpn. 146 (3) (2004) 815.
G = 2(c a + c a )CL (B.7) [20] F. de Leon, A. Semlyen, Reduced order model for transformer transients,
IEEE Trans. Power Deliv. 7 (1) (1992) 361369.

References
[1] S. Fujita, N. Hosokawa, Y. Shibuya, Experimental investigation of high
frequency oscillation in transformer windings, IEEE Trans. Power Deliv.
13 (4) (1998) 12011207.
[2] G.C. Paap, A. Alkema, L. van der Sluis, Overvoltages in power transform-
ers caused by no-load switching, IEEE Trans. Power Deliv. 10 (1) (1995)
301307.
[3] R.C. Degeneff, W.J. McNutt, W. Neugebauer, J. Panek, M.E. McCallum,
C.C. Honey, Transformer response to system switching voltage, IEEE
Trans. Power Apparatus Syst. PAS-101 (6) (1982) 14571470.
[4] R.C. Dugan, R. Gabrick, J.C. Wright, K.V. Pattern, Validated techniques
for modeling shell-form EHV transformers, IEEE Trans. Power Deliv. 4
(2) (1989) 10701078.
[5] T. Van Craenenbroeck, J. De Ceuster, J.P. Marly, H. De Herdt, B. Brouwers,
D. Van Dommelen, Experimental and numerical analysis of fast transient
phenomena in distribution transformers, in: Proceedings of the IEEE PES
Winter Meeting, Singapore, 2000, p. 6, CD-ROM.
[6] J. Lopez-Roldan, H. de Herdt, T. Sels, D. van Dommelen, M. Popov, L. van
der Sluis, J. Declercq, Analysis, simulation and testing of transformer insu-
lation failures related to switching transients overvoltages, in: Presented at
CIGRE, Paper 12-116, Paris, 2002.
[7] M. Popov, L. van der Sluis, G.C. Paap, H. de Herdt, Computation of very
fast transient overvoltages in transformer windings, IEEE Trans. Power
Deliv. 18 (4) (2003) 12681274.
[8] K.J. Cornick, B. Fillat, C. Kieny, Muller W, Distribution of very fast tran-
sient overvoltages in transformer windings, in: Presented at CIGRE, Paper
12-204, Paris, 1992.
[9] D.J. Wilcox, W.G. Hurley, T.P. McHale, M. Conion, Application of modi-
fied modal theory in the modeling of practical transformers, in: Proceedings
of IEE, vol. 139, pt. C (6), 1992, pp. 513520.
Marjan Popov graduated with a Dipl.-Ing. and M.S. in Electrical Engineering
from the Sts. Cyril and Methodius University in 1993 and 1998, respectively,
and gained a Ph.D. at Delft University of Technology in 2002. From 1993 until
1998 he was a teaching and research assistant at the Faculty of Electrical Engi-
neering at the University of Skopje, spending 1997 as a visiting researcher at
the University of Liverpool. He is currently assistant professor in the Electri-
cal Power Systems Group at the Power Systems Laboratory at Delft University
of Technology. His research interests are in arc modelling, transients in power
systems, parameter estimation and relay protection. He is a senior member of
IEEE.
Lou van der Sluis was born in Geervliet (the Netherlands) on 10 July 1950. He
obtained his M.Sc. in electrical engineering from Delft University of Technology
in 1974. He joined the KEMA High Power Laboratory in 1977 as a test engineer
and was involved in the development of a data acquisition system for the High
Power Laboratory, computer calculations of test circuits and the analysis of test
data by digital computer. He became a part-time professor in 1990 and, 2 years
later, was appointed full-time professor in the Power Systems Department at
Delft University of Technology. Prof. van der Sluis is a senior member of IEEE
and former chairman of CC-03 of Cigre and Cired to study the transient recovery
voltages in medium and high-voltage networks. He is currently a member of
Cigre WG A3-20 for modelling power systems components.
Rene Peter Paul Smeets graduated with an M.Sc. in Physics from Eindhoven
University of Technology (the Netherlands) in 1981. He obtained a Ph.D. at
the same institution in 1987 for research on vacuum arcs. He remained at Eind-
hoven as an assistant professor until 1995. During 1991, he was with the Toshiba
Corporations Heavy Apparatus Engineering Laboratory in Kawasaki (Japan),
and, in 1995, joined KEMA in Arnhem (the Netherlands). At present, he man-
ages the R&D activities of KEMAs High Power Laboratory. In 2001, he was
appointed part-time professor at Eindhoven University of Technology. He is a
senior member of IEEE.