2018 IEEE Electrical Power and Energy Conference (EPEC)

Modeling of Ferroresonance Phenomena

in MV Networks
Krzysztof Solak, Waldemar Rebizant
Faculty of Electrical Engineering
Wroclaw University of Science and Technology
Wroclaw, Poland
[email protected], [email protected]

Abstract— In this paper, investigation of ferroresonance different frequency components. Commonly, four types of
phenomena in medium voltage (MV) networks are presented and ferroresonance oscillations may be distinguished, [1]:
discussed. The main attention was paid to developing of adequate
digital simulation model for investigation of ferroresonance • fundamental mode: the signal waveforms are periodic
oscillations in MATLAB/Simulink program as well as to analysis and their spectra are discontinuous. The signal period is
of various conditions that may initiate ferroresonance oscillations. the same as power system period T1 which means that
In addition, two system grounding arrangements were taken into fundamental frequency component f1 dominates in signal
account with ungrounded and Petersen coil earthed neutral point. spectrum that additionally may contain large number of
In the paper simulative analyses of ferroresonance initiation harmonics (e.g. 2f1, 3f1, ...).
conditions as well as looking for signals reflecting this
phenomenon are described. The phase as well as open delta VT • subharmonic mode: for this condition the signal
voltages and their spectra were considered as signals for waveforms are also periodic, whereas the period of
ferroresonance detection. Additionally, the suppression possibility signal is a multiple of the system frequency period which
of ferroresonance oscillations was also studied. The results related brings period equal to nT1 (n is integer) and frequencies
to both ferroresonance initiation detection and its suppression are being equal f1/n. Therefore this mode is called
presented in the paper. subharmonic n or harmonic 1/n. The signal spectrum
comprises subharmonics f1/n (usually odd order, n = 3, 5,
Keywords— inductive voltage transformer; ferroresonance 7) and fundamental component f1 .
oscillations; medium voltage networks; simulation
• quasi-periodic mode: the signal is non-periodic for this
I. INTRODUCTION state and its spectrum is also discontinuous. The
frequency spectrum consists of a number of frequencies
The ferroresonance oscillations may occur in configurations (at least two components) which corresponds to a linear
where a nonlinear inductance (e.g. representing magnetizing combination of formula nfA + mfB (where coefficients n
branch of a voltage transformer, VT or CVT, power transformer, and m are integers and the ratio of fA/fB is a non-integer
etc.) is connected to the power system capacitance (e.g. value). Generally, for this mode the fundamental
capacitance of cable or transmission line, reactive power component and subharmonics may occur.
compensation capacitor bank, circuit breaker grading capacitors,
etc.), [1, 2]. The series (voltage) ferroresonance arises when • chaotic mode: for this state the signal waveform is non-
capacitance is series connected to the nonlinear inductance, periodic and its spectrum is continuous in broad band –
while the parallel (current) ferroresonance takes place for in the other words it looks like high level of noise
parallel configuration of capacitance and the nonlinear contained in frequency spectrum.
inductance. These nonlinear phenomena can occur in It can be concluded that non-linear ferroresonance
ungrounded or grounded neutral systems, both in medium oscillations can be characterized by overvoltage, overcurrent
voltage and high voltage power networks. Typical power system and waveform distortions by subharmonic and higher frequency
configurations, where ferroresonance oscillations are highly components. It may be dangerous for power system elements for
likely, are presented in details in [1, 3]. Generally, it increases the thermal (by overcurrent) and electrical (by
ferroresonance oscillations can be initiated by even small overvoltage) stresses that may destroy VTs or other equipment
change of system parameters or during transients, e.g. one- or as well as distort voltage and current measurement, which in turn
two-pole switching operations, fuse blowing, transient phase-to- may affect protection operation.
ground fault, lightning, transformer switching, loss of system
grounding [1, 2]. The ferroresonance phenomenon is complex In the literature one can find some practical examples of
due to its nonlinearity that brings several steady-state responses, destruction of power system equipment caused by
high level of signals (voltage and current) and content of many ferroresonance oscillations. The case described in [3] reported
destruction of 72 VTs in a 50 kV network in Norway. The
ferroresonance oscillations were initiated by clearing of a fault.

978-1-5386-5419-4/18/$31.00 ©2018 IEEE

 2018 IEEE Electrical Power and Energy Conference (EPEC)

In [4] the example where riser pole arrester was destroyed in 12

kV distribution network in Thailand is noticed. The operation of
disconnector breaker and the connection of capacitance of cable
with station service transformer caused non-linear oscillations.
The explosion of the potential transformer in 230 kV HVDC
converter station in Manitoba is presented in [5] – the
ferroresonance was due to switching operation. The destruction
of a voltage transformer in 275kV power system in England is
described in [6] – here the non-linear phenomenon was initiated
by switching operations and configuration with circuit breaker
grading capacitor. In [7] the ferroresonance oscillations with
overvoltages (2 p.u.) and loud humming noise (heard by the
local operator) in a 400kV substation in Ireland were reported.
The ferroresonance case induced by switching operation in 230
Fig. 1. Basic MV power system for ferroresonance studies.
kV cogeneration substation which leads to catastrophic
substation damage (destruction of VT and other equipment) is
noticed in [8].
From above-cited examples it can be concluded that when
ferroresonance oscillations occur VTs are usually destroyed.
Furthermore, these non-linear phenomena are generally caused
by operations of circuit breakers, they can happen at any
moment and for random conditions. The problem with
ferroresonance oscillations in power systems is still not solved
because of its complex behavior due to its non-linear and
random nature. Therefore, there is still a need for analysis of
ferroresonance phenomena and study of suppression methods of
these oscillations.
The following sections of this paper demonstrate results of
ferroresonance oscillations in MV network and typical methods
of suppression of this phenomenon. In Section 2, the developed
MATLAB/Simulink digital model of the power system for
ferroresonance studies together with analysis of obtained non- Fig. 2. Saturation characteristic (flux – magnetizing current) of a VT.
linear oscillations are described. Next (Section 3), selected
methods of ferroresonance oscillations suppression are The magnetizing resistance Rm of VTs is 40 MΩ, which means
presented that are based on introduction of a damping resistance that active power losses in iron core are smaller than 5W. It
in open delta VT circuit. Lastly, Section 4 presents the follows that the magnetizing resistance may be disregarded
conclusions of this work. under ferroresonance study; it may cause very little damping of
the oscillations only. The source impedance ZS parameters were
calculated for assumed short-circuit capacity of 150 MVA. It
II. FERRORESONANCE STUDIES FOR MEDIUM VOLTAGE was assumed that damping resistance Rd =105 Ω and grounding
POWER SYSTEM MODEL resistance Ro=10-5 Ω.
For the analysis of ferroresonance phenomenon a simplified The ferroresonance oscillations may be initiated when the
model of MV power system shown in Fig. 1 was considered and VT iron core inductance becomes saturated while being
implemented in MATLAB/Simulink [9]. The model from Fig. 1 connected to the circuit capacitance and the VT terminal voltage
represents a distribution network with isolated or Petersen coil suddenly increases, e.g. by clearing of a short-circuit. The
grounded neutral where VTs are in parallel with the zero ferroresonance cases under simulation were initiated by
sequence network capacitance. In such a network configuration transient (lasting for 0.1s) L1-phase-to-ground fault via small
ferroresonance oscillations have quite large probability of fault resistance 0.1Ω. In addition, the phase-to-ground
occurrence. The VTs were modeled as single-phase saturable capacitive reactance of the network was assumed to be changed
transformers and their parameters were: transformation ratio in the range 0.005<Xc/Xm<15.0.
20/0.1/3 kV and nominal power 100 VA. The primary windings
of VTs were connected in grounded Wye, while the VTs The following network grounding arrangements were taken
secondary windings were connected in open delta. The VT into account:
model includes a nonlinear magnetizing characteristic, which is
presented in Fig. 2. From the considered VT saturation • ungrounded neutral system (isolated network),
characteristic results that magneti-zing reactance for the non- • Petersen coil grounded neutral system, with
saturated region (first segment) equals to Xm = 26.67 MΩ, while compensation factor 90% (overcompensated) up to
for the saturated part (second segment) the magnetizing 110% (undercompensated).
reactance is is Xms=26.67 kΩ.
 2018 IEEE Electrical Power and Energy Conference (EPEC)

The voltages in three phases and in open delta VT

connection were analyzed. For the purpose of signals spectrum
calculation 240-point DFT (Discrete Fourier Transform) was
applied. It was assumed that sampling frequency was 1 kHz,
which means that 240 points (samples) in terms of time
correspond to 0.24 s. In such a case the spectrum consists of
spectral lines regularly spaced at the frequencies distant by
1000/240=4.17 Hz. The figures below show the spectra of the
voltages calculated for the time just after fault clearing.
Figs. 3-6 present the voltage spectra for transient phase-to-
ground fault clearing induced ferroresonance in ungrounded
neutral system for assumed range of Xc/Xm ratio. It can be
observed that frequencies between 0 Hz and 150 Hz dominate
after ferroresonance oscillations inception. Additionally, one
can see that the components of frequency less than or equal to
fundamental component 50 Hz (subharmonics) dominate in Fig. 5. Spectrum of voltage in phase L3; for ungrounded neutral system.
voltage signals for small values of Xc/Xm ratio, while the
frequencies greater than 50 Hz (especially third harmonic)
predominate for Xc/Xm ratio greater than 7. It is observed that the
highest amplitude is associated with fundamental component
and for extreme cases it is greater than 2.5 p.u., but only in
phases L2 and L3 (see Figs. 4 and 5). The amplitude of
fundamental component in phase L1 (Fig. 3) is close to 0.7 p.u.

Fig. 6. Spectrum of voltage in open delta VTs connection; ungrounded neutral

system.

that is significantly lower than in other phases. Such unbalance

is caused by two factors. First, the energy stored in capacitances
of phases L2 and L3 is much higher than in phase L1 since
during the L1-phase-to-ground fault the voltage in phase L2 and
Fig. 3. Spectrum of voltage in phase L1; ungrounded neutral system. L3 is square root of 3 times higher than nominal voltage, while
the voltage in phase L1 is close to 0. Second, when fault
occurred the iron core inductance in phases L2 and L3 becomes
saturated due to suddenly increased voltage, which promotes
initiation of ferroresonance oscillations. Moreover, it can be
seen that the amplitude of fundamental frequency component in
open delta VTs connection (Fig. 6) is close to zero for Xc/Xm ratio
less than 7, which means that voltage phasors in three phases for
fundamental component are nearly symmetrical and thus sum of
voltage phasors may be quite small. For Xc/Xm ratio greater than
7 the voltage phasors in three phases for fundamental frequency
component are non-symmetrical and the voltage amplitude in
open delta VTs connection may be as high as 5.5 p.u., which is
an extremely dangerous level.
In addition, for low values of Xc/Xm ratio the subharmonic
mode (with subharmonics of 3rd, 5th, 7th order) is observed, while
for higher values of Xc/Xm ratio the fundamental mode (with 3rd
harmonic), quasi-periodic mode and chaotic mode (especially
Fig. 4. Spectrum of voltage in phase L2; ungrounded neutral system. for Xc/Xm ratio close to 15) are detected. In Fig. 7 the
 2018 IEEE Electrical Power and Energy Conference (EPEC)

subharmonic mode for Xc/Xm=0.3 is presented with 3rd Fig. 9 presents spectrum of voltage for Xc/Xm=10.5 and it is
subharmonic dominating in voltage signal. The fundamental clearly seen that this is a quasi-periodic mode. The chaotic mode
mode (Fig. 8) was obtained for Xc/Xm=14 and here the was achieved for Xc/Xm=15 (Fig. 10) with the voltage spectrum
fundamental component and 3rd harmonic predominate in distributed continuously from 0 to 300 Hz.
voltage signal after ferroresonance inception.

Fig. 10. Spectrum of voltage in phase L2 (after fault clearing) for Xc/Xm = 15.
Fig. 7. Spectrum of voltage in phase L2 (after fault clearing) for Xc/Xm = 0.3.
The voltage spectra for transient phase-to-ground fault for
the assumed range of Xc/Xm ratio in the network with Petersen
coil grounded neutral (overcompensated – compensation factor
of 90 %) are depicted in Figs. 11-12, for selected phase L1 and
VT open triangle connection only. It is clearly seen that
ferroresonance oscillations did not occur – only fundamental
frequency component predominates in voltage spectrum after
switching off the fault. Analysis of graphs presented in Fig. 11-
12 revealed that observed slight disturbances are caused by
capacitors’ discharge after fault clearing. The capacitors become
discharged through Petersen coil and not through the iron cores
of VTs. One can conclude that ferroresonance oscillations are
difficult to be initiated in power system with Petersen coil
grounded neutral. Additional tests were performed for
compensation factor of 98%, 102% and 110% and for these
parameters ferroresonance also did not occur.

Fig. 8. Spectrum of voltage in phase L2 (after fault clearing) for Xc/Xm = 14.

Fig. 11. Spectrum of voltage in phase L1 (after fault clearing) for Petersen coil
grounded neutral system.
Fig. 9. Spectrum of voltage in phase L2 (after fault clearing) for Xc/Xm= 10.5.
 2018 IEEE Electrical Power and Energy Conference (EPEC)

Fig. 13. Methods of ferroresonance oscillations suppression.

100Lms
Rd1 A = (1)
N2
or it can be calculated from:
Fig. 12. Spectrum of voltage in open delta VT connection (after clearing fault)
for Petersen coil grounded neutral system. 3 3U S2
Rd 1B = (2)
Pe
III. INVESTIGATION OF FERRORESONANCE OSCILLATIONS
SUPPRESSION where: N – turn ratio of VTs, Lms – saturated inductance of VT
iron core in millihenries (here: 84893.25 mH), US – secondary
The possibility of ferroresonance oscillations suppression voltage of VT (here: 100/3 V), Pe – rated thermal limit burden,
was also investigated and the results are presented below. In i.e. the full power the voltage transformer can supply without
practice, the ferroresonance oscillations in ungrounded neutral overmuch high temperature rise. The rated thermal limit burden
system can be suppressed in the following manner (Fig. 13): can be determined by multiplying the total rated burden of VT
A. The open triangle circuit is closed by a resistance Rd1 being (here: 100 VA) by square of a voltage factor equal to 1.9 (for
sufficient to prevent ferroresonance; the triangle is opened ungrounded neutral system) [12].
when a ground fault occurs and then correct measurement of The damping resistance calculated with (1) is Rd1A=23.58Ω,
the zero sequence voltage can be performed; while the suppressing resistance according to (2) is Rd1B=16Ω.
B. The open triangle circuit is permanently closed by a These values of damping resistance were taken into account
resistance Rd1 being sufficient to prevent the ferroresonance; during testing of suppression method C. In addition, closing
correct measurement of zero sequence voltage is based on times (time when damping resistance is connected to VT open
processing of phase voltages (not from the voltage on triangle) tclose = 0.02, 0.4 and 2 s were considered.
resistance Rd). The ferroresonance suppression effectiveness is presented in
C. The triangle circuit is normally open and voltage becomes Fig. 14 for closing time of suppressing resistance tclose=0.02s
measured to detect ferroresonance oscillations; then the (one fundamental period). The presented graph shows that
suppressing resistance Rd1 is connected for a certain time. ferroresonance oscillations were not suppressed effectively for
the whole assumed range of Xc/Xm ratio, especially for Xc/Xm less
In the considered methods (points A÷C) an additional than 4, where subharmonics dominate. It can be noticed that
resistance is used which is connected in parallel to the damping better results were obtained for lower suppressing resistance
resistance Rd. Generally, the suppressing resistance Rd1 should Rd1B=16 Ω (red curve).
be very low (compared to damping resistance Rd) in order to
effectively suppress ferroresonance oscillations. However, low
value of suppressing resistance Rd1 can cause quite a large
current flow (which increases thermal stresses to the VTs or this
would probably blow the fuse) and further disturb the
measurement of zero-sequence voltage (as a result the phase-to-
ground fault cannot be correctly identified). Therefore,
suppression of ferroresonance in this manner should be very
carefully studied. Of these three methods the method C is most
preferred since the suppressing resistance is connected to open
triangle only when the ferroresonance oscillations take place.
This method is advantageous since the VTs are thermally
stressed only for a short period of time, otherwise than the other
methods (A and B). However, this manner requires an algorithm
for detection of ferroresonance oscillations.
In the references [10], [11] it is proposed to calculate the Fig. 14. Ferroresonance oscillations initiated by transient phase-to-ground fault
– effectiveness of suppression vs. Xc/Xm ratio for tclose= 0.02 s.
value of damping resistance Rd1 according to:
 2018 IEEE Electrical Power and Energy Conference (EPEC)

fail because high content of fundamental frequency component

was also observed after ferroresonance inception. The
subharmonic mode, fundamental mode, quasi-periodic mode
and chaotic mode of ferroresonance may be distinguished
through careful analysis of resulting voltage spectra. From the
test results it could be noticed that initiation of ferroresonance
oscillations in Petersen coil grounded networks is very difficult
(here it has not been obtained at all).
Additionally, the ferroresonance oscillations suppression
possibility was studied. The testing results proved that it is
possible to effectively suppress ferroresonance oscillations by
temporary switching on the suppressing resistance in the open
triangle VT circuit. The success in ferroresonance suppressing
depends on the value and closing time of suppressing resistance.
It could be noticed that the most difficult to suppress are
Fig. 15. Ferroresonance oscillations initiated by transient phase-to-ground fault ferroresonance cases for low values of Xc/Xm ratio, or - in the
– effectiveness of suppression vs. Xc/Xm ratio for tclose= 0.4 s.
other words - for subharmonic mode when odd subharmonics
dominate in voltage signals. The best results (suppression of all
ferroresonance cases) were obtained for suppressing resistance
16 Ω and closing time tclose=2s.

REFERENCES
[1] P. Ferracci, “Ferroresonance, Cahier technique no 190”, Groupe
Schneider, March 1998.
[2] M. R. Iravani, et al, Slow Transients Task Force of the IEEE Working
Group on Modelling and Analysis of System Transients Using Digital
Programs, “Modelling and Analysis Guidelines for Slow Transients – Part
III: The Study of Ferroresonance”, IEEE Trans. on Power Delivery, Vol
15, No 1., January 2000, pp. 255-265.
[3] J. Horak, "A review of ferroresonance", 57th Annual Conference for
Protective Relay Engineers, 30 March - 1 April 2004, pp. 1-29.
[4] C. B. K.Pattanapakdee, "Failure of Riser Pole Arrester due to Station
Service Transformer Ferroresonance," Proceedings of the Internatioinal
Fig. 16. Ferroresonance oscillations initiated by transient phase to ground fault Conference on Power Systems Transients (IPST'07) Lyon, France, 4-7
– effectiveness of suppression vs. Xc/Xm ratio for tclose= 2 s. June 2007.
[5] D. A. Jacobson, D. R. Swatek, and R. Mazur, "Mitigating potential
Fig. 15 presents the effectiveness of suppression for the transformer ferroresonance in a 230 kV converter station," Proceedings
closing time tclose=0.4s. It can be observed that ferroresonance of the IEEE T&D Conference, 1996, pp. 269-275.
oscillations were suppressed effectively for nearly whole range [6] Z. Emin, B.A.T Al Zahawi, Y. K. Tong, “Voltage Transformer
of Xc/Xm ratio. In addition, the effectiveness of suppression for Ferroresonance in 275kV Substation”, Proceedings of the Eleventh
International Symposium on High Voltage Engineering, Volume: 1,
damping resistance Rd1A is worse than for Rd1B. The results of London (UK), August 1999, pp. 283 - 286.
ferroresonance suppression for closing time tclose=2s is depicted [7] M. Val Escudero, I. Dudurych, M. Redfern, “Characterization of
in Fig. 16. It can be noticed that here the ferroresonance Ferroresonant Modes in HV Substation with CB Grading Capacitors,”
oscillations were suppressed for almost all cases and for both Electric Power Systems Research, Vol. 77, No 11, Sept. 2007, pp. 1506-
values of damping resistance similar results were achieved. 1513.
[8] W.S. Viltech, M.V. Haddad, “Voltage Transformer Ferroresonance in
Cogeneration Substation”, Conference Record of 1992 Annual Pulp and
IV. CONCLUSION Paper Industry Technical Conference, Portland, OR, 8-12 June 1992,
In the paper the ferroresonance phenomena in MV networks pp.148-158.
are discussed. For the sake of simulation analyses an appropriate [9] SimPowerSystems. User Guide, The MathWorks, Inc. Natick, MA.
digital model of MV networks with isolated neutral and with [10] R. F. Karlicek, E. R. Taylor JR., "Ferroresonance of Grounded Potential
Transformers on Ungrounded Power Systems", Power apparatus and
neutral grounded via Petersen coil was developed and systems, part iii. Transactions of the American Institute of Electrical
thoroughly tested. It was shown that the ferroresonance Engineers, Volume: 78, Issue: 3, August 1959, pp. 607-614.
oscillations can occur in MV network with isolated neutral for a [11] M. Sanaye-Pasand, A. Rezaei-Zare, H. Mohseni, Sh. Farhangi,
wide range of capacities and it can be initiated by transient "Comparison of Performance of Various Ferroresonance Suppressing
phase-to-earth fault. In addition, from the analyzed voltages Methods in Inductive and Capacitive Voltage Transformers", Proceedings
spectra presented it turned out that after non-linear oscillation of the IEEE Power India Conference, New Delhi, 2006.
inception the dangerous overvoltage (even 2.5 p.u.) can take [12] IEC 60044-2, ‘‘Instrument transformers - Part 2: Inductive voltage
transformers’’, 2003.
place which may eventually lead to damage of the VTs. If the
network protection is based on zero-sequence quantities it may