Ferroresonance in electromagnetic voltage

ers: A study based on nonlinear dynamics

Z.Emin
B.A.T.Al Zahawi
D.W.Auckland
Y. K. Tong

tively graded circuit breakers [3, 41, as was highlighted

Abstract: Ferroresonance in electromagnetic by recent site experience at a 275kV substation in Lon-
voltage transformers, fed through circuit breaker don.
grading capacitance, is studied using nonlinear This experience occurred under relatively innocuous
dynamics methods. The magnetising charact'eristic and commonplace operating conditions. Nevertheless,
of a typical l00VA voltage transformer is ferroresonance is by no means a widespread and domi-
represented by a single-valued two-term nant problem. This is because of the stochastic nature
polynomial of the order seven. The system of its initiation and subsequent build up, both of which
exhibits three types of ferroresonance: rely on the confluence of conditions as well as on fun-
fundamental frequency ferroresonance, damental circuit configuration. The realisation that fer-
subharmonic ferroresonance and chaotic roresonance could occur under very many operating
fcrroresonance, similar to high capacity power circumstances, given the right combination of condi-
transformers fed through capacitive coupling tions, has led to the call for a model which will accu-
from neighbouring lines or phases. Results also rately predict what these conditions might be, so that
show that while fundamental frequency and they can be suppressed or avoided. Conventional linear
subharmonic ferroresonancc can occur under mathematics is inappropriate in this situation and
commonplace operating conditions, chaotic states research has begun into the application of nonlinear
are unlikely in practice. dynamics, or chaos theory, to describe the phenome-
non.
Kieny [5] first suggested applying chaos to the study
of ferroresonance in electric power circuits. He studied
1 Introduction the possibility of ferroresonance in power systems, par-
ticularly in the presence of long capacitive lines as high-
Ferroresonancc is a complicated electrical phenomenon lighted by occurrences in France in 1982, and produced
which has long bcen a problem for power utilities. It a bifurcation diagram indicating stable and unstable
can occur in power transformers or reactors, and in areas of operation. Other studies, addressing the solu-
clectromagnetic voltage transformers fed through tion of the nonlinear equations for a typical ferroreso-
capacitive coupling from neighbouring sources. It can nant circuit containing power transformers, have also
be very dangerous, resulting in the generation of over- appeared in the literature. Araujo et al. [6] identified
voltages having magnitudes potentially several times three types of ferroresonant states: periodic, quasiperi-
the normal voltage ratings. In addition, the effect can odic and chaotic. Mozaffari et ul. [7] showed that the
give rise to dangerously high current transients, well in probability of chaos increases as losses decrease and
excess of the normal ratings of switching equipment. the nonlinearity of the transformer magnetic circuit
Ferroresonance in bulk power systems has been stud- increases.
ied in the litcrature since the latc 1930s [ I , 21. System Despite this interest in the application of nonlinear
capacitance together with the nonlinear inductance of dynamics to ferroresonance in power transformers, no
power transformers combine to induce parasitic reso- attempt has yet been made to employ chaos for the
nance which creates dangerously high transformer volt- study of voltage transformer ferroresonance. In this
ages. More recently, concern has been expressed about paper, voltage transformer ferroresonance is analysed
the possibility of ferroresonance in low capacity elec- from the standpoint of nonlinear dynamics. In particu-
tromagnetic voltage transformers (VTs) fed via capaci- lar, the effects of varying the value of system capaci-
0 I EE, 1997 tance to earth on the behaviour of the system are
IEE Pioccw/iizg.s onlinc no. 19971061
investigated.
Paper lirsl received 16th A p d 1996 and in revised form 15111 October
I996 2 Transformer and system configuration
%. Emin, B.A.T. AI Zahawi and D.W. Auckland arc with tbi: University
01 Manchester, Manchestcr School of Engineering. Simon Building, Fig. 1 shows the circuit diagram of system components
Oxford Road, Manchesler M I3 9PL. UK at the aforementioned 275kV substation. VT is a low
Y.K. Tong IS with The National Grid Company plc, Technology and Sci- thermal capacity wound voltage transformer isolated
cnce Division, Kelvin Avcnuc, Lcatherhead. Surrey KT22 7S7, UK from sections of busbars via disconnector DS2. Ccu is
383
circuit breaker CB grading capacitance. C,, is the total determines the behaviour of the circuit parameters, i.e.
busbar capacitance to earth. Ferroresonance conditions voltages. currents, magnetic flux levels, as a function of
occurred upon closure of disconnector DSI with CB time. Because of the nonlinear nature of the trans-
and DS2 open, leading to a system fault caused by fail- former magnetising characteristics. tlie beha\-iour of the
ure of the voltage transformer primary winding. system is extreme11 sensitil-e to changes in system
parameters and initial conditions. A small change in
the \ d u e of system voltage, capacitance or losses may
lead to dramatic changes in the behaviour of the sys-
tem. Furthermore, nonlinear systems such as the one
[f%?i+i- circuit breaker described above inay give rise to chaotic behaviour pat-
terns in which the time response of the system appears
to be random.
In a real life situation, many of the parameters of the
voltage transformer
Cbb above circuit, such as the B-H characteristics of the
I 'Y transformer, transformer losses etc., are either
unknown or caiinot be measured with any degree of
accuracy. This would mean that a very large number of
simulations are required, to allow for all the inany pos-
sible variations, before a reasonable assessment of the
level of risk of ferroresonance could be made. This is
clearly unsatisfactory to the engineer or operator on
site who wants to know that he has an adequate safety
margin before he takes action.
A more suitable mathematical language for studying
ferroresonance and other nonlinear systems is provided
by nonlinear dynamic methods [9]. A brief background
of some of the mathematical tools used in the analysis
is given below.
Fig. 2 shows the basic ferroresonance equivalent cir-
cuit used in this analysis. E is the RMS voltage applied 3.2 Phase space
to the circuit, o is the supply frequency and h is the The phase space is an abstract mathematical space in
voltage transformer flux linkage. C,ri-ie is the circuit which the time behaviour of a system is represented by
breaker grading capacitance, C,rilu,lf is the total phase- the movement of a point representing tlie state varia-
to-earth capacitance including transformer winding bles of the system with time. For example, the voltage
capacitance. The resistor R represents transformer core transformer operating conditions at any given time can
losses shown in previous studies to be an important be specified by a point representing its terminal voltage
factor in determining the behaviour of the system [7. 81. V plotted against flux linkage A. As lime evolves, the
Transformer current is represented by the trans- initial state point in this phase space follows a trajec-
former nonlinear magnetisation characteristic, tory. If the trajectory closes on itself then we have a
expressed in the form of a power series [6]. By employ- periodic solution and this periodic trajectory is called a
ing per unit values (see Appendix), the transformer cycle. For a chaotic solution, the phase space will have
magnetisation curve was modelled by a single-value a trajectory that will never close onto itself.
seventh-order polynomial obtained by the least squares
curve fitting method from measured peak currents and 3.3 Poincare section
The Poincare section is a simple plot of the phase plane
voltages (apPiik= 42 VR,w.y/co)
trajectory recorded as a sequence of discrete points at
2 = ax bX7 + (1) constant time intervals. If the sampling frequency is
where a = 3.42, I? = 0.41, i is the transformer current in made to correspond to the frequency of the system's
p.u. values and h is the transformer flux linkage in p.u. forcing function (i.e. the frequency of the supply) then,
values. for a periodic waveform, the Poincare section will show
The time beliaviour of the basic ferroresoiiant circuit only one point.
is described by the differential equations The Poincare section can provide useful information
1 dV 1 about the ratio of the frequency of the forcing function
____
+

V to the actual frequency of the system. For example, if

io' dt + cshuibt)
~~(c,,,,,, the actual period of the ferroresonance voltage wave-
1
+ W(C,sm-ir, f
(ax + hX')
Cshunl)
form is 3 x 20nis, a condition referred to as subthird
ferroresonance, the PoincarC section would show three
points.
-
- &E cos fl A chaotic waveform will produce a Poincare section
+
C ~ C T C~5 h~u nS
t with a random set of points confined to a particular
dX dQ region of the plane.
- =V -
h t (2)
dt dt
4 Simulation results
3 Nonlinear dynamic systems
4.1 General
3. I GeneraI In the following analysis, instead of using actual values
The basic voltage transformer ferroresonance circuit of of circuit parameters E , w, C,serie,s,Cshuizietc., system ,

Fig. 1 was modelled by a differential equation that equations are made dimensionless by employing per
184 I E E Pi-o( -Goiw Ti-iiiisin Di\li;h., V d 144, N o 4, J u l j I997
unit values. Eqn. 2 may be written as 3 1.OT

"
= gcosH -10 -0.5 05 10
W
C -0
dX do L
.,-
L
0 -1.01
I
-
-=IT
(3)
dt dt = d v1

where g and q arc the driving force amplitude and Y

e flux linkageA ,p.u.
damping parameter, respectively, given by Fig.4 h'ornd .sine ~ t w mpon.ve.
e I'oincusb section
E = 1p.u. R = 225MR. c',,,,,,,= 0.SnF. C$,z,ii,i = 3nb. [ h ( r ) -- 0p.u., V(r) =
c . 5 e T2 r ,7 d2p.u. at f = 01
g=fiE (4)
CSCT1C6 + Csi,u n t
7
+ Csh?LTlt)
4 = Rw(Csertcs (3) a
W.
9
0 50 1
Eqns. 3 contain a nonlinear term and have no easily
available analytical solution. The cquations weire solved
numerically using an embedded Runge-Kut la---Fehl-
berg algorithm with adaptive stepsize control. A full
investigation of the behaviour of the system ai; a func-
tion of y and g would cause despair. Instead, values of
15 and (r) were fixed at 1 p.u., corresponding to AC sup-
ply voltage and frequency. is the CB grading
capacitance and its value o sly depends' on the
type of circuit breaker used. For this analysis,
was fixed at a typical value or 0.5nF. The value of
depends mainly on the length of busbar and the
C,5hu17t
number of items connected to it and may be estimated
at around l0pF per metre. In this analysis the value of
CtilunIwas varied over a range of realistic values which,
for the site under consideration, were assessed. to vary
between 0.1nF and 5nF. Transformer core loss was frequency, Hz
estimated between 50 W and 500 W. Calculations were
performed for a value of R = 225M9, corresponding
to a loss of 112w.
Results for four different parameter sets showing a
normal sinusoidal response plus the three possible types
of ferroresonance conditions are presented below. Solu-
tions wcrc obtained for initial values of V ( l ) =I 42 and
A([) = 0 at t = 0, representing circuit breaker operation
at maximum voltage. The waveforms and diagrams are
drawn for steady-state conditions after initial transients
L
a

g,
L

VI
c
0
7

-L
(7 -1
-2
-3
have disappeared.
f l u x Linkage A ,p.u.
4.2 Normal sinusoidal response Fig,7 Fundunientul frequency /hi.i.uiwonuncc, phuse plane d i n g m i
Figs. 3-6 show waveforms and diagrams for normal E = I p SI. R = 225MR. <'>,,,,( = 0.5iiF, C,,,,,,,, = 1.25nF. [7L(t) = 0p.u., Vil) =
~

d2p LI. at f = 01
sinusoidal conditions calculated for a value oT' C'~llu,zr
=
3nF. This corresponds to a lightly damped, lightly
driven system in which y = 247.4 and g = 0.20. The 3 3 T

phase plane diagram presented in Fig. 3 clearly shows

the closed trajectory characteristic of a periodic wave-
form. The corresponding Poincare section shows only
one point indicating that the resulting voltabre wave-
form is periodic with a frequency equal to the system
frequency, i.e. 50 cycles per second.
L
Y
f l u x l i n k a g e A , p.u
3
a
'T Fig.8 ~unu'umeniulfic~petzcyjei.i.oresonance, Poincurc; section
0 5-- E = l p U R = 225MR. C,,,,,., = 0.5llF, C,,,,,,,,= 1.25nF. [h(r)= Op.u.. V ( t )=
W
(51 d2p.u. at r = 01
-
e0 r + n __i

P -1 0 -05 05 1 0
L
a,
0 5-- 4.3 Fundamental frequency ferroresonance
k Figs. 7-10 show wavelorins and diagrams for a funda-
101
c
mental frequency ferroresonance condition calculatcd
for a value of C,ltunt= 1.25nF. Compared with the pre-
vious example, damping and driving force amplitude
are increased by a factor of two so that y = 123.7 and g
= 0.4. The phase plane diagram and the corresponding
3x5
Poincare section presented in Figs. 7 and 8, respec-
tively, again show the characteristics of a periodic
waveforin with a frequency equal to the system fre-
quency, i.e. 50 cycles per second. Operation in the fer-
roresonance region is demonstrated by the high
amplitude of the transformer voltage waveform and by
the presence of high frequency harmonic components
?I-/*.
cn

0 100 200
:

300
in the power density spectrum of the voltage waveform. frequency, Hz
Fig. 14 Sd3hili./?loizk
E = 1 p . t . R = 225MR.
4 2 p . u at / = 01

4.4 Subharmonic ferroresonance

Figs. 11-14 show waveforms and diagrams for a value
of CIiiiiilr
= 0.19nF ( q = 48.8 and g = 1.02). The result-
ing waveform is still periodic, but with a period of
40ins or twice the period of the supply cycle.
For this type of ferroresonance, the phase plane dia-
gram shows two separate trajectories closing on thern-
selves. Two points occur on the corresponding

pi/:-^
Poincare section. The power density spectrum of the
voltage waveform contains high frequency Components
with a sharp peak at 50Hz, the system drive frequency.
, 4.5 Chaotic ferroresonance
An example of chaotic ferroresonance conditions is
presented in Figs. 15-18 showing waveforms and dia-
0 100 200 300 grams calculated for values of E = Ip.u., Csiiunt=
frequency, Hz 0.1 nF, C,,,,.,,,,= 3nF and R = 1900MQ corresponding
Fig. 10 Fiindumentul frequency ,[email protected],
lioii'er .cpectr~iin
E = 1 p , ~R = 225MQ, = I,ZjnF, [)L(r) = 0p.u.. v(/)= to a loss Of O1lly l 3 and no
d2p.u at / = 01 1850.4, g = 1.37). Such high values of C,
unlikely in practice, but the waveforms demonstrate the
7
possibility, in theory at least, of chaotic ferroresonance.
a
J
R

c
ffl -41
e
L

C
i?
f l u x linkage h , p . u i
flux linkage h , p U
Fig. 15 phase pkmr diagvum
C%ciotic~ri.or.esorztmcr,
E = Ip.ti R = 1900MQ. C,,,,,, = 3nF. C,,,,,,,, = O.lnF, [A(/)= Op.u.. V(i) =
d 2 p 11 a1 f = 01

3 3
Q. a 8T
W a
W
ol 2 --
0
5
0
5 0

-ffl
ffl
c

e f l u x linkage h , p U
e
Y
Y
f l u x l i n k a g e A , p.u.
Fig. 12 Suhhar.monicj~~rrove,~onaizce,
PoincmP sectioiz Fig. 16 Ciinoric ferroiwononcc: foincur6 .SeCtlOi7
E = l p u R = 225MQ, C,,,,,, = 0.5nF, C,,,,,,,,= 0.19nF. [h(r)= Op u.. V ( / )= E = 1p.u. R = 1900MQ. C,,.,,,., = 3nF, C,,!,,,,, = 0 InF, [h(t)= Op.u., V ( t ) =
d2p.u. at / = 01 d2p.u. at / = 01

Fig. 13 Subl7aimonic fiwovesonmce, time series Fig. 17 Chilotic ferroi.esonance, tinzr srries
E = 1p.u. R = 225MQ, C,,,,,, = 0.5nF, C,,,,,,,, = 0.19nF. [A(r) = 0p.u.. i'(f) = E = 1p.u. R = l900MQ. C,,,,,, = 3nF. C,/,,,,,l= 0.lnF. [h(l)= Op.u.. V ( i ) =
~

d2p.u. at t = 01 d 2 p . u at i = 01

386 I E E Pi-oi.-Gener. Tiim.vn. Diairih , Vi)/ 144, No 4, July 1907

 resonance. The work described in this paper, together
with recent studies [ 5 4 ] of ferroresonance as a nonlin-
ear phenomenon, does not yet provide this comprehen-
sive understanding. Many questions are still to be
answered. Nevertheless, a useful contribution has been
made that will provide a foundation for future develop-
0 100 200 300 ment.
frequency, Hz
Fig. 18 Clzaotic~rroresc~nanc~, power Jpectrum 6 Acknowledgments
E = 1 p u R = 1900MB, C,,,,,, = 3nF, CAhunr
= 0 InF. [A(t) = 01’U , V(t) =
d2p U at t = 01 The authors wish to thank The National Grid Com-
pany plc for its support. In particular, the authors are
Repeated simulation failed to produce any chaotic grateful to Mr D. Boyle for his valuable contribution.
states for realistic values of model parameters, totally 7 References
in agreement with practical site experience.
Figs. 15-18 show all the characteristics of chaos. An 1 ODESSEY, P.H., and WEBER, E.: ‘Critical conditions in fcr-
irregular and unpredictable time behaviour, a phase roresonance’, A I E E Trans., 1938, 57, pp. 444452
2 CLARK, E., PETERSON, H.A., and LIGHT, P.H.: ‘Abnormal
plane trajectory that never closes on itself, a random voltage conditions in three-phase systems produced by single-
set of points confined to a particular region of the phase switching’, A I E E Trans., 1941, 60, pp. 329-339
Poincare section and a broadband power spectrum 3 AGGARWAL, R.P., SAXENA, N.S., SHARMA, R.S.,
KUMAR, S., and KRISHAN, S.: ‘Failure of electromagnetic
with a sharp component at 50Hz. voltage transformer due to sustained over-voltage on switching --
an in-depth field investigation and analytical study’, IEEE Trans.,
1981, PAS-100, pp. 4448-4455
5 Conclusions 4 CRANE, D.R., and WALSH, G.W.: ‘Large mill power outages
caused by potential transformer ferroresonance’, JEEE Trans.,
Low capacity electromagnetic voltage transformers fed 1988, IA-24, pp, 635-640
5 KIENY. C.: ‘Aoolication of the bifurcation theorv in studving
through circuit breaker grading capacitance have been and understand& the global behaviour of a ferrokonant ;le:
shown to exhibit fundamental frequency, subharmonic tric power circuit’, JEEE Trans., 1991, PD-6, pp. 866-872
and chaotic ferroresonance conditions similar to high 6 ARAUJO. A.E.A., SOUDACK, A.C., and MARTI, J.R.: ‘Fer-
roresonance in power systems: chaotic behaviour’, ZEE Proc. C,
capacity power transformers fed via capacitive coupling 1993, 140, ( 3 ) , pp. 237-240
from neighbouring sources. Repeated solution of the 7 MOZAFFARI, S., HENSCHEL, S., and SOUDACK, A.C.:
system’s nonlinear differential equation has shown that ‘Chaotic fcrroresonance in power transformers’, ZEE Proc. C,
1995, 142, (3). pp. 247-250
a change in the value of the equivalent circuit capaci- 8 MARTI, J.K., and SOUDACK, A.C.: ‘Ferroresonance in power
tance to earth, possibly as a result of a change in sys- systems: fundamental solutions’, IEE Proc. C, 1991, 138, (4), pp.
321-329
tem configuration, can give rise to different lypes of 9 BAKER, G.L., and GOLLUB, J.P.: ‘Chaotic dynamics: and
ferroresonance voltages. It has also been shown that introduction’ (Cambridge University Press, 1994)
while fundamental frequency and subharmonic
ferroresonance conditions may occur under common- 8 Appendix
place operating conditions, chaotic ferroresonance The base values used in the above analysis are
states are not likely to occur under practical site condi- base voltage = RMS phase voltage = (2751d3)kV
tions.
A comprehensive understanding of the possibilities base volt-amperes = transformer VA rating = lOOVA
that exist for ferroresonance is very desirable for engi- base impedance = (base voltage)*/base volt-amperes =
neers so that they can operate their systems outside 252MQ
dangerous regions and can plan the expansion of sys- base angular frequency = AC supply angular frequency
tems without enhancing the possibility of ferro- = 2$50)rad/s

IEE Proc.-Gmer. Tsunsm. Dictrih., Vol. 144, No. 4,July I997 387