Damping of resonances at energization of

transformers serving large pulse type loads

A. M. Miri

C. Sihler

Universitt Karlsruhe, Institut fr

Elektroenergiesysteme und Hochspannungstechnik (IEH)
Kaiserstr. 12, D-76128 Karlsruhe, Germany
miri@ieh.etec.uni-karlsruhe.de

Max-Planck-Institut fr Plasmaphysik (IPP)

EURATOM Association
Boltzmannstr. 2, D-85748 Garching, Germany
sihler@ipp.mpg.de

Abstract ASDEX Upgrade, an experimental tokamak for

nuclear fusion research, requires an electrical power up to a few
hundred MVA for a time period of 10 15 s. The power and
energy is provided by isolated networks based on flywheel
generators. About twenty power transformers must be energized
on a daily basis to get the pulsed power supply operational for
plasma experiments. Nuisance interactions of transformer inrush
currents with the protection system of the synchronous machines
and capacitor banks used for reactive power compensation must
be avoided. To investigate the effects of inrush transients in
present and future configurations, measurements were
performed and a simulation model based on an equivalent
magnetic circuit representation of the transformer was derived.
The paper describes the model, results of simulation and
measurement and shows which measures were taken to keep
transients in the isolated network at an acceptable level.
Transient analysis; transformer inrush currents; magnetic
equivalent circuit; resonances; reactive power compensation;
isolated network

I.

INTRODUCTION

Nuclear fusion research is at present focussed on

experimental devices of two types, the tokamak and the
stellarator. Both confine the plasma, a low-density gas which
has to be heated to an ignition temperature in the order of 100
million degrees, in a ring-shaped magnetic field cage [1]. The
ASDEX Upgrade tokamak has been operating in Germany
since 1990. It is investigating key questions of fusion research
under powerplant-like conditions. In this device the ringshaped hydrogen plasma carries an electric current up to 1.4
MA. The magnetically coupled poloidal field coils of the
experimental device, the largest one with a diameter of 7 m,
require an electric power up to several hundred MVA for 10 15 s [2].
The power supply of ASDEX Upgrade consists of the
flywheel generators EZ2 (1.5 GJ / 140 MVA), EZ3 (500 MJ /
144 MVA), EZ4 (650 MJ / 220 MVA), reactive power
compensation units and 11 thyristor converter units to provide
direct currents for the magnetic field coils. Before plasma
experiments can be performed, each flywheel generator is
accelerated within several minutes to maximum speed (up to
1650 r.p.m.) by means of a starting converter (generator EZ4)
and a separate driving machine (generators EZ2 and EZ3),
both being supplied from the utility supply (50 Hz) via a 16

MVA connection ("loading"). Then the generators are

electrically disconnected from the utility supply. During the
plasma experiment the flywheels are unloaded within seconds.
The busbar voltages are kept at a constant value (10.5 kV)
throughout the pulse. Thus, the energy needed for the
experiment is taken only from the flywheels and the pulsed
load is kept away from the utility supply. The large inductive
voltage required by the ASDEX Upgrade magnetic field coils
during plasma ignition and ramp-up of the plasma current
leads to a large reactive power component in the flywheel
generator networks during the phase when the physics
experiments are conducted.
Presently ASDEX Upgrade is strengthening its efforts
concerning investigations of advanced tokamak scenarios. To
fully exploit these operating modes, a reduction of apparent
power is required in the flywheel generator networks,
especially in the EZ3 network, in order to take full advantage
of the available flywheel energy. Satisfactory conditions for
long pulse operation can be achieved in compensating 120
MVAr by means of static VAr compensators. The
characteristics of the reactive power demand was the
determining factor for the decision to use switchable capacitor
banks for reactive power compensation, as shown in Fig. 1.
They are energized by line synchronised vacuum breakers.
This solution is far less expensive than thyristor controlled
compensation but requires an accurate tuning with regard to
the generator and load characteristics in order to keep transient
phenomena at an acceptable level [3].
Together with the high-voltage converters of the additional

Reactive Power
Compensation Units
(4 x 30 MVAr)
Unit 1

Unit 2

Unit 3

Unit 4

Thyristor Converter Units

ASDEX Upgrade
Figure 1. Simplified representation of EZ3 network.

heating systems more than 20 converter transformers must be

energised on a daily basis to get the pulsed power supply
operational for plasma experiments. The converter
transformers are directly connected to the synchronous
machines. The resistance available to damp out the inrush
currents is small and the inrush transients from connecting the
unloaded transformers can be sustained for seconds. The
inrush current amplitudes can reach values in the range of 20
kA for large transformers of the experimental power supply.
In order to investigate the effects of inrush transients in
various configurations a detailed simulation model for one of
1 10

the power transformers was developed.

Inrush current measurements were performed at switchingon of the converter transformer at operating frequency (85
110 Hz in the flywheel generator networks). Because of the
lack of data with respect to the frequency dependence of the
B-H characteristic of the core material, the measurements
were performed at frequencies of 90 Hz and 100 Hz. The
measured inrush currents shown in Figures 2 and 3 are the

i1+5000
i2
i3-5000

1 10

i/A
i/A

i/A

i3+5000
i2
i1-5000

5000

5000

-5000

-1 10

INRUSH CURRENT MEASUREMENTS

II.

-5000

-1 10
0

0.1

0.2

0.3

0.4

0.5

0.6

i1+5000

1 10 4

0.7

1 10

0.1

0.2

0.3

0.4

0.5

0.6

i3+5000
i2
i1-5000

i2
i3-5000

5000

5000

i/A
i/A

i/A

i/A

0.7

-5000

-5000

-1 10 4

-1 10
0

0.05

0.1

0.15

t/s

Figure 2. Switching at 90 Hz network frequency

(at positive zero-crossing of phase 1)

0.2

0.05

0.1

0.15

t/s

Figure 3. Switching at 100 Hz network frequency

(at negative zero-crossing of phase 1)

0.2

resulting currents at switching-on of two identical starstar/star-delta transformers, each of them designed for 11
MVA in pulsed duty (nominal power in continuous duty about
2 MVA).

s1

Bsat=1.25Bn

A. Results from measurements at 90 Hz

The inrush currents in Figure 2 reach maximum values up to
5 kA. After the 6th current peak there is a difference in the
ideal inrush current time history which can be explained by
the generator voltage controller starting to influence the
transient response at about 70 ms.

n >> s

B. Results from measurements at 100 Hz

Whereas the inrush currents reached maximum values up to
5 kA at frequencies of 90 Hz, they were about 1.5 kA smaller
at a network frequency of 100 Hz (see Fig. 3).

Bsat=-1.25Bn

Inrush currents reach higher values at lower networkfrequencies (non-linear connection) since at lower frequencies
the transformers iron-core is saturated at lower magnetic field
strengths. The maximum amplitude of 5 kA is more than 20
times higher than the nominal current in continuous duty (i. e.
more than four times the nominal current in pulsed duty). In
previous measurements performed on a larger converter
transformer inrush currents with amplitudes up to 20 kA had
been observed in the system.
III.

MODELLING OF TRANSFORMER

An equivalent magnetic circuit representation of the

transformer was chosen. This way direct dependencies of the
transformer design and the inrush transients can be
investigated [4]. The electrical and magnetic transformer
circuits were modelled within a single, coupled circuit
simulation using the Simplorer code [5]. A major problem was

= BdA
A

= Rm =

B = H

= w i = Hds

Figure 4. Approximation used for non-linear magnetic resistances

how to model the non-linearity of the iron core, since not

much information was available on the core material. Finally,
the B-H characteristic of the core material was approximated
by straight lines as shown in Fig. 4. The reluctances of the
equivalent magnetic circuit can be derived by means of the
magnetic variables given in equation system (1).
The three-phase transformer shown in Fig. 5 [6, 9] was
modelled using the magnetic equivalent circuit representation
which is given in Fig. 6. The non-linear magnetic resistances
Rpab, Rpbc, Rsab and Rsbc represent the linking yoke sections of
length lJ. The non-linear magnetic resistances Rca, Rcb and Rcc
model the limb sections of length ls. In parallel to the limb
resistances the linear magnetic resistances Rama, Ramb and Ramc
have been installed which represent the stray fluxes between
the low-voltage windings and the transformer core. The flux
paths between the low-voltage and the high-voltage windings
are modelled by the linear magnetic resistances Rlka, Rlkb and
Rlkc. The stray fluxes that close in free space are represented
by the linear reluctances Rla, Rlb and Rlc. The coupling
elements in Fig. 6 are used for the mutual transformation of

la

lb

lc

(1)

= Rn

ama

pab

amb

sab

cb
lkb

pbc

amc

sbc

cc
lkc

fr - sat < < sat

= sat R n + ( sat ) R s

> sat

= sat R n ( + sat ) R s

< sat

Rs, Rn: Saturated, non-saturated reluctances of iron core

: Ampere-turns; : Flux; w: Turn number

ca
lka

Figure 5. Cross-section of a three-limb transformer

Figure 6. Magnetic equivalent circuit of the three-limb transformer shown in

Fig. 5
Figure 7. RPC Module (15 MVAr at 100 Hz)

electrical and magnetic energy.

The equation system completely describing the transformer
behaviour depends on the transformers vector group [7], [8].
In the general case the transformer is described by voltage
equation system (2):

[u ] = [R ] [i ] + [w] d
dt

(2)

[u], [i], and [d /dt]: Column vectors of the terminal voltages,

currents and flux changes in the windings. [R] and [w] are the
diagonal matrices representing the winding resistances and the
number of turns. The vector of magnetomotive forces
generated by the windings can be expressed by equation (3) or
(4), in which [Rm] is the matrix of equivalent magnetic
resistances and [] is the matrix of magnetic fluxes.

[ ] = [w] [i]

(3)

[ ] = [Rm ] [ ]

(4)

The remanence in the core was realised using magnetic

starting resistors. A qualitative comparison of inrush currents
simulated with this model assuming a remanence of 60 % and
results from literature showed good agreement [9].
IV.

On the one hand a minimisation of transient overvoltages

would require n 1. On the other hand n < 5 must be fulfilled
in order to prevent a resonant excitation of the RPC modules
by harmonics generated by the ASDEX Upgrade thyristor
converters. The filter reactor L was chosen so that the natural
frequency of each RPC module is c = 339 Hz, i.e. k = 3.8.
Fig. 8 shows the resonant frequencies of the RPC system if up
to 8 modules (4 units) are in operation.
During experiments which required high rotational speeds
of flywheel generator EZ3 (corresponding to network
frequencies > 105 Hz), resonances as shown in Fig. 9 were
observed while only one RPC module was in operation. Fig.
10 shows a detail of Fig. 9 which allows us to analyse the
frequencies of the resonant currents in the time domain. The
beat curve shown in Fig. 10 consists of two frequencies: 11 Hz
and 329 Hz which can be explained analytically from the
differential equations of an RLC equivalent circuit being
excited by an alternating voltage with frequency . If is
close to the resonant frequency c of the RLC circuit ( c),
the following formula for the current i can be derived [6]:
c
+ c
i = 2 reso sin
t sin
t
2

RESONANCE POINTS OF REACTIVE POWER

reso: Current amplitude in case of resonant excitation ( = c)

COMPENSATION MODULES

XC
n 2 = c =

3
X
''(m
+ k)
d
0
Xc : reactance of capacitor C; Xd: subtransient reactance of EZ3;
k = XL / 3 Xd; m: number of RPC modules

(5)

Susceptance / S

Each reactive power compensation (RPC) unit in Fig. 1

consists of two 15 MVAr modules. The design of the modules
is shown in Fig. 7. The arrangement in Fig. 7 can be
transformed into a single-phase equivalent network with the
following per-module natural frequency n:

(6)

Power
Frequency

m= 87654 3 2

1
Thyristor
Converter
Harmonics

-2

-4

m= 8765 4 3 2
0

100

200
Frequency / Hz

300

1
400

500

Figure 8. Susceptance functions of m = 1...8 RPC modules in the EZ3

network ( = c suceptance )

Frequencies above 300 Hz can be explained by a third

harmonic in the flywheel generator network which can be
caused by an asymmetrical load on generator EZ3 due to
inaccuracies in the gate control of one of the high current
converters. The curves displayed in Figs. 9-10 were measured
at a system frequency of 106 Hz, i. e. the third harmonic of
that system frequency is 3 = 318 Hz. During ASDEX
Upgrade shot # 13373 only one RPC unit was connected, thus
a resonant frequency c = 340 Hz can be calculated by means
of equation (5). I. e. the following parameters of the sine
functions in equation (6) can be derived
3 c
2

11 Hz

3 + c
2

329

which leads to the beat curve shown in Fig. 9.

Because of the variable frequency and harmonics, a
resonant excitation of RPC modules can occur within the
whole operational frequency range (85-110 Hz), depending on
the number of modules connected during the load pulse (see
Fig. 8). Under these conditions the most promising design
improvement was to install additional resistors. They were
connected in parallel to the RPC filter reactors in late 2000.
They have a resistance value of 70 , and provide damping
tuned to the 3rd harmonic [3]. They also provide sufficient
damping in cases where transformer inrush currents cause a
resonance in RPC modules which has been investigated
numerically with the model presented in section III.
V.

SIMULATION RESULTS

100 ms

Figure 10. Resonant excitation of RPC module (AUG # 13373):

Detail of Fig. 9.

interactions of inrush currents with the RPC modules in the

IPP variable frequency network. The lowest natural frequency
of the RPC system is in the order of the 2nd harmonic. With 8
RPC modules in operation, resonances can be caused by
transformer inrush currents which is demonstrated in Fig. 11.
Fig. 12 shows the inrush currents causing these resonances
in the RPC modules. The simulations considered the
additional damping resistances installed. Nevertheless it can
be seen in Fig. 12 that no exponential decay of the inrush
currents occurs. In one phase the inrush current amplitude
even increases during the first few periods after connecting the
transformer.
Fig. 13 shows the first second after connecting the power
transformer. In this simulation the generator voltage control
has not been considered (worst case analysis). Although the
resonant excitation of RPC modules leads to a temporary
increase of the transformer inrush current in one phase, it is
clear from these figures that the additional damping
resistances installed in the RPC modules provide sufficient
damping for this scenario

The simulation model has been used to investigate

IC11

IC12

VI.

CONCLUSION

Inrush transients of power transformers in a flywheel

generator network were investigated by simulation and
measurement. An equivalent magnetic circuit model of one of
the power transformers was derived for analysing the effects
of inrush currents in future configurations. This model was
chosen so that direct dependencies of the transformer design
and the inrush transients can be investigated. Simulations were
performed to analyse possible interactions of transformer
55.000k
kA

170.0m 200.0m

100 ms 250.0m

300.0m

350.0m

400.0

170.0m 200.0m

250.0m

300.0m

350.0m

400.0

2.500k
2.5
kA

= 2,4 kA
= 630 A

IC13

IC13

-2.500k
-2.5 kA

-5.000k

-5 kA

Figure 9. Resonant excitation of reactive power compensation module:

Measured capacitor currents (in the three phases of one module)

Figure 11. Simulated capacitor currents at resonant excitation with 8 RPC

modules connected (c = 217 Hz) due to transformer inrush current

15 kA

1515
kAkA

0.170

0.400

0.600

400
ms
0.800

1.000 1.070

0.00k
5

0
100

-15kkA
-15
A

5.00k

170.0m

200.0m

ms

250.0m

300.0m

350.0m

-5.00k

0.00k

400.0m

5.00k

-15 kA

inrush currents and a 120 MVAr reactive power compensation

(RPC) system upgraded by additional damping resistances .
The simulations show that the RPC system can be excited to
resonances by inrush currents of large transformers. However,
the damping provided by the 24 resistances installed in the
RPC system is sufficient. Even when large transformers are
connected to the isolated network the transients can be kept at
an acceptable level.

0.170

0.400

0.600

0.800

1.000 1.070

Figure 13. Extended time scale

of simulation result shown in Fig. 12
(similar time scale as in Figures 2 and 3, above)

Figure 12. Transformer inrush currents causing a resonant excitation of

RPC modules

VII. APPENDIX

XL
Xd
[]
[d /dt]
[i]
[R]
[Rm]
[u]
[w]

reactance of RPC filter reactor

generator subtransient reactance
matrix of magnetic fluxes
column vector of magn. flux changes
column vector of terminal currents
diagonal matrix representing the winding
resistances
matrix of equivalent reluctances
column vector of terminal voltages
diagonal matrix representing the number of turns

TABLE OF SYMBOLS

0
3
c
A
B
C
H
reso
l
k
m
n
QN
Rm
w
Xc

magnetic flux
magnetomotive force
network frequency
natural frequency of RPC module
network angular frequency
fundamental angular network frequency
third harmonic of angular network frequency
natural angular frequency of RPC module
cross sectional area
magnetic induction
capacitor
magnetic field strength
current amplitude at resonant excitation of an
equivalent RLC circuit
yoke section length
ratio of reactances
number of RPC modules connected
per-unit natural frequency of SVC
nominal reactive power (at 100 Hz)
reluctance (magnetic resistance)
turn number
reactance of capacitor

REFERENCES
[1]

[2]

[3]

[4]

[5]
[6]
[7]
[8]
[9]

T. Hamacher, A.M. Bradshaw, "Fusion as a future Power Source:

Recent Achievments and Prospects", Proc. 18th World Energy
Congress, Buenos Aires, World Energy Council, London, 2001
M. Blaumoser, M. Kottmair, A. Wieczorek,
th O. Gruber, ASDEX
Upgrade power supply system, Proc. 14 Symposium on Fusion
Technology, Avignon, France, 1986, pp. 915-920
C. Sihler, M. Huart, B. Streibl, D. Hrabal, H. Schmitt, Transient
Performance of Vacuum-Switched Static VAR Compensators Optimised
for Large Inductive Loads, Proc. Int. Conf. On Power System
Transients, Brazil, 2001, pp. 481-486
R. Yacamini, H. S. Bronzeado, Transformer Inrush Calculations Using
a Coupled Electromagnetic Model, IEE Proc., Science, Measurements
and Technologies, Vol. 141, No. 6, Nov. 1994, pp. 491-498,
SIMEC GmbH, Simplorer Code User's Manual, Version 4.2, Chemnitz,
Germany, 2000
A. M. Miri, Ausgleichsvorgnge in Elektroenergiesystemen, Springer
Verlag, Berlin/Heidelberg, ISBN 3-540-67735-6, 2000
R. Kchler, Die Transformatoren, Springer-Verlag, Berlin/Heidelberg,
1966
W. Schmidt, ber den Einschaltstrom von Transformatoren, ETZ-A,
Band 82, 1961
A. M. Miri, C. Mller, C. Sihler, Modelling of Inrush Currents in
Power Transformers by a Detailed Magnetic Equiv. Circuit, Proc. Int.
Conf. On Power System Transients, Brazil, 2001, pp. 215-220