A Tutorial on Ferroresonance

Elmo Price, ABB Inc.

Introduction

Ferroresonance occurs in power system circuits containing capacitance and a non-linear

transformer magnetizing inductance and is usually initiated with a transient disturbance such as
opening a switch. It usually results in overvoltages and/or high current spikes that may subject
system apparatus to dielectric and thermal stresses resulting in apparatus failure as well as
subjecting operating personnel to hazardous conditions. Also, protective relays that measure
these quantities are subject to incorrect operations causing unwanted outages. Today,
ferroresonance is a widely studied phenomenon in the power system, but due to its complexity
it is not well understood by those that do not study it in depth; hence the term "fuzzy­
resonance. " This paper will present a simple tutorial using phasors and a graphical approach to
explaining the ferroresonant operating states, circuit configurations and mitigation. It will
provide the basic concepts necessary to understanding more advanced investigations into
unique occurrences.
Transformer Characteristics

Transformer core steel magnetizing characteristics used in transformer design are usually
expressed as magnetic flux density vs. magnetic field strength (B vs. H). These parameters,
however, are not readily available for measurement with the manufactured transformer.
Therefore, the magnetizing curve is usually expressed as the applied voltage vs. magnetizing
current at the transformer's designed operating (fundamental) frequency. This is useful when
analyzing operation at the designed fundamental frequency, but falls short when evaluating the
transformer at other frequencies to which it may be exposed.
Design frequency V
or magnetizing flux </J

Saturation Vs.t, ¢s.t

'---- Anhysteretic
Curve

Magnetizing
Inductance

Magnetizing
Current

¢=kf v(t)·dt

Figure 1. Transformer characteristics

40th Annual Western Protective Relay Conference. Spokane, WA. October 1 5 - 17, 20 13.

978-1-4799-4739-3/14/$31.00 ©2014 IEEE 676 ProReJay 2014

Figure 1 shows the transformer characteristic expressed as magnetizing flux or fundamental
frequency voltage vs. magnetizing current. The fundamental voltage will produce the
corresponding flux per equation 1. Otherwise, the curve ¢vs. I is not tied to a specific
frequency and understanding the relationship defined in equation 1 between applied voltages of
any frequency and magnetizing flux will provide a better understanding to ferroresonant voltage
waveforms to be discussed later. In this equation k is a constant for the core and coil design,
v(t) is the applied voltage and rpo is the remnant flux.
J,t2
0= k tl vet) dt + 00 (1)
The non-linear anhysteric characteristic does not show the transformer's hysteresis and its
effect on magnetizing current and remnant flux, but is more suitable for the discussion and
analysis done here. There are two basic parts to the characteristic, the magnetizing inductance
and the air-core inductance. The magnetizing inductance is the inductance where the flux is
below rpSat, the saturation flux, and the transformer's core is not saturated. The air core
inductance is the inductance where the flux is above rpSat and the core is considered saturated.
This is the inductance of the winding being energized when not mounted to the transformer's
core. The saturation flux, rpSatt (or saturating voltage, VSat) is usually 140% or more of the
transformer's rated operating flux, rpR (or rated voltage, VR)
The transformer core design and winding connections also impacts the operating
characteristics, particularly in regard to ferroresonance. A few of the more common core
designs are shown in Figure 2.
�A �B �c
B-t-+-I_ C -t-+-I_

(a) Three single phase cores

B
�A ,------ �B �C
• • ....
A ... PB =- C P
I"""' F-
... F- .-
... � I"""'
... F'""
... ... F-
- 1"""'- - F-
'---

(b) Three phase five legged core form (c) Three phase three legged core form (d) Three phase shell form

Figure 2. Common transformer core designs

In the three-phase bank of single phase units of Figure 2(a) the phases are magnetically
decoupled while in the other three phase core designs the phases are magnetically coupled and
each phase is influenced by the other two. In a three-phase unit any flux unbalance in the
three phase legs will produce zero sequence flux and any interconnecting part of the core
(outer legs, top and bottom yokes, etc.) may become saturated. This is particularly true under
ferroresonant conditions. The phases may also be magnetically coupled through delta winding

677
connections. These inter-phase influences make the analysis of ferroresonant configurations
quite complex and stresses the importance of accurate modeling for the detailed study of these
configurations. Modeling is outside the scope of this paper, but more information may be found
in reference 3.
The Ferroresonant Circuit
Before addressing the ferroresonant circuit we will look at a simple RLC circuit of Figure 3(a)
where the inductance is linear (non-saturating). Therefore the voltage and current are linearly
related in a manner which is frequency dependent. If the switch is opened at time zero and
there is a charge on the capacitor, Vc(O), then there will be an oscillating voltage Vc at the
natural frequency fo (equation 3) due to the exchange of energy between L and C. If R = 0 then
the oscillation would continue indefinitely without decay of amplitude. In this circuit R is very
small and the circuit is under damped, therefore the oscillations will continue, but be dampened
and Vc will eventually go to zero with time.
R
a=- (2)
2L
1
fa = 2rr-JIZ (3)

Vc(O)
Vc = Vc(O) 'e-at 'cos (27rfo t)

c
Vc(O) t

(a) RLC circuit and transient response

�s >�
� �s 1 t

�s I

(b) RLC circuit with saturating inductance and transient response

Figure 3. RLC circuits with linear and non-linear inductance

Now consider Figure 3(b), which is the same circuit, but with non-linear inductance having the
characteristics of Figure 1. In this case there are two major natural frequencies to consider, Ji
for the magnetizing inductance in the non-saturated region andh for the air core inductance.
1
fl = 2rrN (4)

1
f2 = 2rrjL.;C (5)

678
When the switch is opened at time zero the voltage Vc begins to oscillate at the non-saturated
frequency fi. This frequency is very small and depending on the value of ( it is usually less than
a few Hz. Therefore, an oscillation at this frequency (almost dc) does not occur before
saturation flux of the Figure 1 characteristic occurs at r/Js assuming a design frequency of 50 or
60 Hz. This is graphically seen in Figure 3(b) transient response. When integrating equation 1
where v(t) = cos(21ifjt) the flux r/J (area under the voltage curve oscillating at fi) increases until
r/Js is reached and there is an oscillation at h during saturation, which is at a much higher
frequency that is perhaps 50 or more times fi. With this oscillation the inductor comes out of
saturation with a -Vc on the capacitor and the process is repeated. The magnitude of Vc will be
reduced by power losses contributed by R during the h oscillation. If there is no resistance (R
= 0) and saturation was truly symmetrical in the positive and negative direction the waveform

of figure 3(b) would be a uniform oscillating rectangular wave that would continue indefinitely.
The frequency of the oscillation would be primarily determined by the magnetizing
characteristics and to a much lesser extent 0. and 6.
t= 0
Vs(t)

(a) RLCcircuit with saturating inductance (b) Equivalent of figure (a)

and driving voltage

t= 0 Vc C1I(C1+ C)
Vs (t) 2 ·v (t)
,l�
=

V"-F
""
'V' " '" -
"" "

$s '"j"
-$;-" '" $s $s
t

-$s -$s -$s

" ...... " ...... " ...... " "
"-"v "-"v "-"v '-' "

(c) Transient response and ferroresonant operating state

Figure 4. RLC circuits with non-linear inductance and driving source

The capacitor (1 is added to the circuit in Figure 4(a) to illustrate the effect of a driving voltage
that can sustain the oscillations. The equivalent circuit is shown in Figure 4(b) with the
equivalent voltage vs(t). The voltage VM, which is across R and LdL2' is Vc + vs(t). If the
equivalent voltage, vs(t}, is sufficient to provide the losses due to R, and depending on the initial
capacitor voltage and remnant flux at the inception of the transient (in this case opening the
switch) it is possible to set up a sustained condition - an operating state in the saturated region
of the transformer characteristics - known as ferroresonance.
Ferroresonant�odes

Transformer operation in the ferroresonant state is, as stated, dependent on a number of

factors such as the system voltage magnitude, the initial voltage on the capacitor, the initial

679
state of the magnetic characteristics of the transformer, the total loss in the ferroresonant
circuit and the point on wave of initial switching. It is normally initiated after some type of
switching event such as transformer energization, single-phase switching, fault clearing, blown
fuses. breakers opening, or loss of system grounding. Given the right conditions it can lock into
any of the following ferroresonant modes. The one common characteristic among these
ferroresonant modes is that they all contain the fundamental frequency driving voltage
component, which sustains it.

Fundamental Mode

The most frequent ferroresonant mode and simplest to analyze is the fundamental frequency
mode as shown in figure 5. The ferroresonant voltage waveform (VM) oscillates at the same
frequency as the driving voltage (Vs), which is also one of its components. The wave form's
positive and negative half cycles are generally symmetrical, but on a few occasions have been
observed non-symmetrical. The ferroresonant voltage is also of opposite polarity as the driving
(or normal state) voltage phase shifted only by the voltage drop across the circuit's resistance
R. The peak ferroresonant voltage can be much larger than the system voltage causing both
voltage and thermal stress issues. Sustained operation of surge protection (arresters) can lead
to failure. Ferroresonance also produces large current spikes that can be thermally damaging if
the condition is sustained.

Sub-harmonic Mode

The sub-harmonic mode includes those periodic oscillations that are greater than the
fundamental (Vs) period, T, and uniformly repeated as illustrated in Figure 6. The positive and
negative half cycles can be symmetrical or non-symmetrical. Sub-harmonic ferroresonant
voltages with a period of 3T (20 Hz.) are the most common sub-harmonic occurrences, but
those with a period of 2T (30 Hz.) have been frequently observed.

Quasi-periodic Mode

Some ferroresonance wave forms are as shown in figure 7 and have repeated patterns that
appear periodic, but have different peak magnitudes with each repetition and periods that are
not quite equal. They are referred to as quasi-periodic because of their similarity to a periodic
waveform.
Chaotic Mode

Hunting occurs in the period after the initial transient disturbance and before a ferroresonance
mode, which is discussed above, locks into a stable operating condition. This hunting for the
stable operating point appears chaotic and can take from less than a cycle to hundreds of cycles
to lock into the stable operation with repeated patterns. The hunting is characterized by non­
periodic waveforms and different voltage magnitudes bouncing around in chaos. On some
occasions this can go on indefinitely. The continuous hunting is usually the result of high
magnetic coupling between phases and initial conditions and is referred to as the chaotic mode
as illustrated in Figure 8. Usually a slight change in initial conditions of the ferroresonant circuit
will allow locking into one of the stable modes of repeated patterns.

680
 /"""'0. /"""'0. /"""'0. /"""'0. /"""'0. /"""'0. /"""'0. /"""'0. /"""'0.
Vs - - - - - - - - -

r-- .---- r-- .---- r-- .---- r-- .---- r--

Vc

--- ....... --- ....... --- ....... --- ....... ---

./ r-" r-' r-" r-' r-" r-' r-" r'-'" r-"

� ....... -.,. ....... -.,. � -.,. � -.,.

�
Figure 5. Fundamental mode ferroresonance

T
III( -I
VS �������r-

Vc �------+--�--�

Figure 6. Sub-harmonic mode ferroresonance - 20 Hz

681
T

Figure 7. Sub-harmonic mode ferroresonance - 30 Hz

Figure 8. Quasi-periodic mode ferroresonance

682
 Figure 9. Chaotic mode ferroresonance

Figure 10 is a copy of a vintage magnetic strip chart recording (circa 1976) from vt
ferroresonance testing in a 345 kV substation. During the testing several modes of
ferroresonance were observed, but at that time categorizing the ferroresonance modes was not
the objective. Several of the modes, however, are observed on this copy - fundamental, sub­
harmonic and chaotic (hunting) to be sure. Unfortunately an estimated 2 to 4 seconds of
recording were removed [from a long roll of recording paper] between events at t2 and t3. This
test was to determine if the calculated damping resistance of 1. 9 ohms would eliminate
ferroresonance of a specific vt. Refer to a later section and equation 25 for discussion of the
calculation method and also reference 2.

Va

Vb

Vc

I ILl 60 Hz :
t1 - circuit breaker Event at t2 , Strip chart recording /"" Event at t3
52 opens 20 damping resistor paper removed 1.50 damping resistor manually
inserted with 52b inserted (0.50 removed)

Figure 10. Vt ferroresonance in a 345 kV substation

683
A Study of the Ferroresonant Circuit Using Phasor Analysis

Transformer normal and ferroresonant operating states at the fundamental frequency can be
graphically shown and analyzed using phasor analysis. Phasor analysis is the simplest means of
analysis and has proven very effective in analyzing circuits that can be reduced to a single
phase circuit that represents the ferroresonant system somewhat accurately. Due to the
variability of the non-linear circuit characteristics and the harmonics that are produced other
more complex forms of analysis are often preferred. [1]
Consider the circuit of figure l 1 (a), the basic ferroresonance circuit in its simplest form. A
simple set of simultaneous equations for the solution of VM as a function of 1M can be developed
where at fundamental frequency Xc is the capacitive reactance, which is shown negative, and
XM is the non-linear transformer magnetic characteristic. Plotting equations 6 and 7 in figure
l 1 (b) shows three intersections where VM and 1M are satisfied for both equations.
VM = XMIM (6)
VM = Vs + Xc 1M (7)

(a) Equivalent circuit

(b) Graphical solutions for VM and 1M

Figure 11. Graphical solution for transformer operating states at fundamental

frequency applied in a circuit with series capacitance

Point 1 is the normal operating state where the transformer is operating in the non-saturated
region of XM• Point 2 is the ferroresonant state where the transformer is operating in the
saturated region of XM• Operation at point 3 cannot occur. Basically, at points 1 or 2 for small
changes in current the voltage change acts in opposition to the current change and forces the
operation back to the operating point. At point 3 a small change in current changes the voltage,
which creates an additional current that is in the direction of the change; therefore, the current
avalanches away from the operating point. This is discussed in greater detail in Appendix A. rpM
is noted on the VM axis to remind the reader that the magnetizing current is the result of flux
being produced by the applied fundamental frequency voltage. When other sub-harmonic

684
voltage waveforms result during point 2 operations the value of flux, <rM, being produced is still
about the same. Therefore, study at fundamental frequency is appropriate.
Figure 1 2 shows the graphical solutions when considering the effect of resistance. Equation 6 is
still used, but the second equation, equation 1 1, is developed using Kirchhoff's Law and
summing the voltage drops around the series circuit as shown in equation 8. The equation is
complex as it has real and imaginary parts. From there it's a matter substitution, rearranging
terms and summing the squares to get to equation 1 1.
Vs - RIM - (-jXc)IM - jXMIM = 0 (8)
VM = XMIM
Vs = RIM + j(VM - XcIM) (9)
v/ = (RIM)2 + (VM - XcIM)2 ( 1 0)

VM = XcIM ± Vl- (RIM)2J ( 1 1)

Increasing R

(a) Equivalent circuit

.,­ ....
/'
-

.....
-

/ ,
I

, /
..... /'
....

__-----.,-
VS
� =+�V2-(Rl
x- )2 S M
-

Increasing R
(b) Graphical solutions forVM and 1M
/

Figure 12. Graphical solution for transformer operating states at fundamental

frequency applied in a circuit with series capacitance and resistance

685
Equation 1 1 is an ellipse that is symmetrical about the slope Xc. Plotting equations 6 and 1 1 in
figure 1 2(b) shows six intersections where VM and 1M are satisfied for both equations. Points 1,
2, and 3 are solutions for positive values of Vs and pOints 1', 2' and 3' are solutions for negative
values of Vs (-Vs). Similarly, pOints 1/1' are the normal operating state, pOints 2/2' are the
ferroresonant state and the operation at points 3/3' cannot occur. The ellipse contracts with
increasing R. This shows the possibility of eliminating ferroresonant point 2 operations with
sufficient loading. The ellipse expands with decreasing R. If R is equal to zero then equation 1 1
would be reduced to equation 1 2 (essentially the same as equation 7). This results in the two
straight lines on the figure 6(b) identified as the R = O.

( 1 2)

Single point operations

Figure 13 shows some possible single point operating state solutions depending on the value of
Xc relative to XM• In figure 13(a) where the magnitude of Xc is equal to or greater than the
non-saturated magnetizing reactance, XM1, there is only a point 2 solution and the transformer
will operate in an over excited state also referred to as ferroresonance. In figure 13(b) where
the magnitude of Xc is equal to or less than the saturated magnetizing (air core) reactance, XM2,
there is only a point 1 solution and there will be no occurrence of ferroresonance.

(a) Ferroresonant (over excited) state -IXcl2: XM1 (b) Normal state -IXcI:s XM2

Figure 13. Single point operating state solutions

Changing operating states under transient conditions

The transformer's normal operation is in the non-saturated region of the magnetizing

characteristics and ferroresonance is initiated after some type of transient event such as
transformer energization, single-phase switching, fault clearing, a blown fuse, breakers opening,
or loss of system grounding. Regardless of the event a transient change in voltage levels,
impedance parameters and/or circuit configuration forces the jump in operating state from

686
normal to ferroresonant. This is illustrated in figure 14 with a very simplified illustration of a
rather complex and dynamic process. The ferroresonant state, when it occurs, is a legitimate
operating state and will remain there without the impetus to stop it.
(2) (1) (3) (2) (3)

I I

State 2 is
forced (1) Initial normal state
(2) Transient period
(3) Final ferroresonant state

(a) Voltage transient - constant impedance (b) Impedance transient - constant voltage

Figure 14. Forcing the ferroresonant operating state

Damping the Ferroresonant Circuit

Figure 1 2(b) indicates that increasing circuit resistance (load) would eliminate operation in the
ferroresonant state. Load, however, must be added in parallel to the transformer magnetizing
reactance as it is impractical to add series load without affecting the system operation. Figure
lS(a) shows the ferroresonant circuit with parallel resistance. There are several ways to
approach the development of a circuit equation for VM as a function of 1M• The equation will be
complex and require a quadratic equation solution. Figure lS(b) shows the phasor diagram for
the circuit of figure lS(a) from which an equation for VM as a function of 1M can be derived.
vi = (Vecosa)2 +(VM - Vesina)2 (13)
Where cosa =
IR and sina = 1M
Ie Ie

vi = (V�:R) 2
+ ( VM _ V�;M)2 (14)

vi = Xll� + V� - 2VMXelM +xlth ( 1 5)

VM
Where IR-
-R

v:2 x�v� + V2M - 2VMXeIM +Xe

2 1M
2 ( 1 6)
S =
R2

V� [ (:e )2 + 1] - 2XelM VM +XllJ:r - vi = 0 (17)

Equation 17 is also an ellipse and is symmetrical about the slope m with VM axis intercepts ±VQ•
Plotting equations 6 and 17 in figure lS(c) shows the normal and ferroresonant operating state
intersections. If R is increased to infinity (opened) then equation 17 would be reduced to

687
equation 1 2. On the other hand the ellipse shrinks and its slope decreases with decreasing R.
This shows the possibility of eliminating ferroresonant point 2 operations with sufficient parallel
loading by reducing R.
Points 2' and 3 can be analytically derived using the straight line approximation of the nonlinear
magnetizing reactance curve's saturated region (air core reactance) as shown in figure 1 6.
(18)
Rearranging equation 1 8 in terms of 1M and then substituting into equation 17 and
understanding that the slope Xc is greater than the slope XM2 for the pOints 2'and 3 intersections
to occur we can develop equation 1 9.
V� [( ) (
xc
R
2
+ XC-XM2 ) ] VM [-2Vsat (
2
+ Xc
XM2
XC-XM2
XM2
)] (
+ XcVsat
XM2
) -vl=O
2
(19)
XM2

The solutions for VM of equation 19 can be derived using the quadratic formula.

-b+vb2 -4a c
Vm = --=-=-a-
2 - - (20)

Where a =
(:c)2 (XC;:2M2)2
+
(2 1)

M2 )
-2 Vsat� (X C X
b =
XM2 XM2 (22)
-

(XcVsat)2 -vi.
C =
XM2 (23)

The maximum value of resistance, RMax, can be determined by graphically shrinking the ellipse
until there is only a single solution with equation 18 representing the saturated region. The
single solution is also the unstable point 3 solution for which operation will not occur forcing the
point 1 solution for normal operation. Analytically, a single point solution will only occur when
the term under the radical in equation 20 is equal to zero.
b2 - 4ac = 0 (24)
If the term is negative then the solution is complex and indicates no intersections. Therefore,
any value of R less than RMax will have no intersections in the saturated region. An equation for
RMax can be derive by substituting equations 2 1, 22 and 23 into equation 24 and solving for R.
After a lot of careful and tedious term manipulation and reduction equation 25 is derived.

(25)

In cases where Xc is much greater than XM2 equation 25 can be greatly simplified to equation
26.

R Max - XCVSat
-
Vs (26)

688
 (a) Equivalent circuit

VM - Vcsin(a)
Vcsin(a) \ � I IR
I
I
I
I
I
I
I
I
I
I

Vc
Vccos(a)
-------- 'I
(b) Phasor diagram

I
/
/
Y-VA�[( � ) I]
c
2
+ _ 2XcIMVM + X�l�[ - VS2 = 0

(c) Graphical solutions for VM and 1M

Figure 15. Graphical solution for transformer operating states at fundamental

frequency applied in a circuit with series capacitance and parallel resistance

VM =XMiM +Vsal
VSat -

VQ � �
V� [( ; ) ]
X
' +1 -2XCVMIM +xU! -vi �O

Figure 16. Graphic solution for eliminating ferroresonance

689
Other analysis methods

Phasor analysis as done above is the simplest method of ferroresonance analysis to use for
developing a basic understanding of the phenomena. It has been used effectively where simple
and accurate ferroresonant circuits could be defined. Due to the variability of the non-linear
circuit characteristics and the harmonics that are produced other more complex forms of
analysis are often preferred. Also, more complex configurations will require more accurate
modeling and the use of time domain EMTP analysis. Other methods of analysis are discussed
in detail in reference 1.
System Ferroresonant Configurations

The basic ferroresonant circuit involves a transformer's non-linear magnetizing reactance, very
low core and load losses (burden), a system voltage source and a series capacitance, which is
usually coupled with a shunt capacitance, between the source and non-linear reactance. Figures
4(a) and 4(b) represent the basic circuits. There are many possibilities. Following is a brief
discussion of some rather common cases.

Unbalanced Switching

This condition occurs more frequently on distribution systems where open phases result from
blown fuses and/or single phase switching of transformers. It may also happen in three phase
switching applications where unintended unbalanced switch operation occurs, e.g. a stuck pole
on a breaker opening operation. This condition is illustrated in figures i7(a) and i7(b) where
the ferroresonance circuit can be easily visualized. The source voltage is supplied by the
remaining connected phase(s). The series capacitance is associated with the conductor
characteristics ... cable, transmission line, bus, etc., and may be either phase-to-phase, phase­
to-ground, or both.

-------
- --- ----------
--

(a) Coupling through phase-to-phase capacitance with grounded transformer

------------

(b) Coupling through phase-to-ground capacitance with ungrounded transformer

Figure 17. Unbalanced switching conditions

690
Voltage Transformer Applications on Ungrounded Systems

Although not always readily apparent, phase-to-ground connected voltage transformers (vts)
applied to ungrounded system configurations are prone to ferroresonance. Such configurations
should be avoided and if temporary as a result of system operation appropriate precautions
taken. Two applications as shown in figure 1 8 are discussed here, a single phase-to-ground vt
and a bank of three phase-to-ground vts with unbalanced phase-to-ground line capacitance.
Sequence networks are used to derive the ferroresonant circuit.

- -

(a) Single phase-to-ground vt (b) Three phase-to-ground vt with unbalanced capacitance

Figure 18. Vts applied on ungrounded systems

Zs

-3XUb

Zso Zso

(a) Single phase vt sequence networks (c) Three phase vt sequence networks

(b) Equivalent circuit of (a) with Zs = 0 (d) Equivalent circuit of (c) with Zs = 0
Figure 19. Sequence network interconnects and equivalent
ferroresonant circuits for figure 18 systems

691
A single phase vt applied on an ungrounded system as shown in figure 1 8(a) is subject to
ferroresonance due to the phase-to-ground capacitance and the unbalanced effect of the single
phase transformer. The positive, negative and zero sequence circuits are connected with 3XM
as shown in figure 19(a). Since the source impedance magnitude, Zs, is very small compared to
Xc and XM it can be neglected [and assumed a short]. The resulting equivalent ferroresonant
circuit is shown in figure 1 9(b).
A three phase bank of voltage transformers applied on an ungrounded system as shown in
figure 18(b) is subject to ferroresonance due to the unbalance of the phase-to-ground
capacitance that usually exists. To simplify the circuit for analysis the phase-to-ground
unbalance is shown as -XUb lumped on one phase while the balanced capacitive reactance is
shown as -Xc. The positive, negative and zero sequence circuits are connected with -3XUb as
shown in figure 1 9(c). Since the source impedance magnitude, Zs, is very small compared to Xc
and XM it can be neglected. The resulting equivalent ferroresonant circuit is shown in figure
19(d).

Vt ferroresonance with circuit breaker grading capacitors

High voltage circuit breakers generally require multiple breaker contacts in series to divide the
fault interrupting transient voltage stress across each breaker contact evenly. Voltage grading
capacitors are connected across each contact as shown in figure 20(a) to assure an even
voltage distribution during the interruption process. The net capacitance across the breaker
depends on the number of breaking contacts and the value of the grading capacitor. A typical
value might be 1300 pf resulting in 650 pf across the breaker with two series breaker contacts.
The grading capacitors also serve to improve the breaker's current interruption capability
(rating) and may be sized appropriately. TRV (transient recovery voltage) capacitors may be
applied phase-to-ground and/or across the breaker contact. Therefore, it is possible that single
contact breakers, which do not need grading capacitors, may have a TRV capacitor across the
contact.

r--------------I
I Circuit Breaker I
I I
Isolated Bus

���
I Grading Capacitors
L... ______________
I
I
Transformer
(a) Circuit breaker grading capacitors (b) Ferroresonant bus configuration
VEq - equivalent system source voltage, Cs - total series o1gen breaker capacitance,
CB - bus capacitance, Vs - coupled bus voltage in normal operating state

Figure 20. Ferroresonant bus configuration with circuit breaker grading capacitors

A number of breakers may be connected to the bus during the bus isolation, e.g. breaker failure
lockout, and the net series capacitance, (s, can couple a large voltage, Vs, to the bus with the
bus capacitance, (B' Depending on the aforementioned conditions this configuration may result
in transformer ferroresonance. Generally, induction voltage transformers, vts, are most likely

692
affected; however, power transformers have been affected where a large number of breakers
are connected to the isolated bus.

Ccvt ferroresonance

Coupling capacitor voltage transformers are by the nature of their design very prone to
ferroresonance and ccvt manufacturers provide various designs of ferroresonance suppression
circuits (FSCs) applied on a secondary winding. The ccvt equivalent circuit and typical FSCs are
shown in figure 2 1.
Primary

Tuning Reactor

/ Step-down

L / Transformer
MOV MOV

T Secondary
Filter circuit Saturating reactor

....----'
2
C .--- J [$] (active) damping circuit
(passive)

(b) Typical Ferroresonance suppression circuits

with surge arrestor
(a) CVT equivalent circuit

Figure 21. Ccvt equivalent circuit with FSCs

The filter FSC is a fundamental frequency (50 or 60 Hz) blocking filter that allows damping of
harmonics and sub-harmonic modes of ferroresonance. This leaves primarily the fundamental
frequency voltage operating in the normal state. Selection of the blocking filter with a narrow
bandwidth at fundamental frequency is desirable. The saturating reactor circuit uses a
saturating reactor to switch in damping resistance when secondary voltages ( and flux density)
become excessive. It responds to all harmonic and sub-harmonic voltages. The addition of
MOVs provides additional secondary surge protection and reduces the ferroresonant elimination
time. While the FSCs are effective at eliminating ferroresonance they do temporarily affect ccvt
voltage accuracy during system faults and protective relay operation. This must be evaluated.

Auxiliary Transformers

Auxiliary transformers connected to the secondary of a vt or ccvt may be subject to

ferroresonance. The auxiliary vt is a non-linear burden that will affect the output of the vt or
ccvt. This is particularly true in the case of ccvts where capacitance is a major part of the ccvt
transformer's primary circuit. The auxiliary vt's saturation characteristics should be selected
such that saturation does not occur before the vt or ccvt to which it is applied. Sometimes a
higher voltage rating may be preferred - e.g. 230 : 230 V auxiliary vt instead of 1 1 5 : 1 1 5. V.

Transformer connected to an isolated line

A transformer that is connected to an isolated line (de-energized and not grounded) that is
physically parallel and adjacent to an energized line for part of its length is subject to
ferroresonance if sufficient capacitive coupling exists between the two circuits. Refer to figure

693
22. The capacitive (electrostatic) coupling would depend primarily on the geometry between
the two circuits and the length of the lines.

-f-of---L---I-
----, .---
---�
---o-f-
capacitive coupling

OPEN

I
(a) Vt or tapped transformer

(b) Transformer terminated line

Figure 22. Isolated transmission line ferroresonant configurations

Mitigation of ferroresonance
The mitigation of transformer ferroresonance usually involves one or more of the following :
• Correcting a voltage unbalance condition
• Changing the transformer magnetic design
• Inserting damping resistance
• Detuning the circuit

Correcting voltage unbalance

Voltage unbalance can be corrected by changing operating procedures or by applying protective

relays to sense and remove the unbalanced condition that drives ferroresonance. For example,
single phase switching of three phase distribution transformers may be changed to three phase
operation; or a voltage relay that is used to sense excessive zero sequence voltage and trip a
circuit breaker. Opportunities to correct voltage unbalance should be investigated. There are,
however, many applications where steady state unbalanced voltages cannot be avoided and
alternative solutions are required.

Changing the transformer magnetic design

Improvements in transformer core steel design have allowed transformers to be designed at

higher rated flux levels while producing acceptable excitation current and core losses. This
allows rated voltage, VR, operation closer to the saturation level VSat such that the ratio VSav'VR is
relatively small. Refer back to figure 1. This is perhaps the reason for the apparent increase in
the occurrence of and the subsequent study of ferroresonance.

694
 vR - rated (applied system) voltage
Vs - equivalent voltage
VSaIA - saturation voltage for
characteristic A
VSatB - saturation voltage for
characteristic B

(a) Magnetizing characteristic A (b) Magnetizing characteristic B

Figure 23. Ferroresonance operation comparison

Figure 23 shows two different magnetizing characteristics. For characteristic A the VSatA/VR
ratio is 1.38 and for characteristic B the VSatB/VR ratio is 2. 17. Comparing the two characteristics
it is readily observed that when using characteristic B a much higher voltage transient is
required to force a jump to the ferroresonant operating state (refer to simple analysis of figure
14(a)). Also referring back to equation 25 it is observed that characteristic B has less damping
(R M ax is larger) to eliminate ferroresonance.
Modification of the transformer's magnetizing characteristic may not always be an economically
practical solution, particularly three phase power transformers. It should be looked at in critical
cases. In the case of vts and distribution transformers selecting a voltage rating greater than
the applied voltage and maintaining the correct turns ratio may be practical. Also, investigating
the differences in magnetizing characteristics between vendors might be worthwhile. Reference
4 provides a discussion that compares three vt characteristics from different vendors with
simulation studies.

Inserting damping resistance

A common way to eliminate ferroresonance is to insert damping resistance across the

secondary wye connected phase-to-ground windings or in the corner of a broken delta winding
as shown in figure 24. This applies to both power and instrument transformers. The damping
resistor is often part of ferroresonance suppression devices (FSD) involving a damping resistor
with other devices controlling it.

695
 A Burden A

B B Burden

c c

-
(a) per phase damping with a (b) zero sequence damping
wye grounded secondary with a broken delta secondary

Figure 24. Ferroresonance mitigation with damping resistance

Zero Sequence Damping

Figures 17 and 18 illustrate typical unbalance conditions. Ferroresonance for these conditions is
usually the result of excessive zero sequence flux that occurs in the core with the voltage
unbalance. This is particularly true in the case of three phase cores. In single phase cores a
large component of the phase-to-ground voltage is zero sequence. Therefore, eliminating the
zero sequence voltage may be sufficient to eliminate the ferroresonant condition. The
implementation of operating procedures to prevent unbalance voltages or restore balanced
voltages may be sufficient, but in cases where a voltage unbalance cannot be avoided zero
sequence damping as shown in Figure 24(b) is appropriate.
Unbalances also occur on systems due to the phase-to-ground capacitive unbalance as
illustrated for the ungrounded system of figure 18(b). The equivalent circuit is derived from
figures 19(c) and 19(d) and is predominately zero sequence as the positive and negative
sequence XC and XM are shorted by the soutce impedance. Again, damping as shown in Figure
24(b) would be appropriate. Figures 19(c) and 19(d) are shown without a secondary winding
connection. Modifying these figures with the secondary broken delta winding and resistor is as
shown in figure 25.
Figure 26 shows the resulting phase and zero sequence (3Vo) coupled voltages after clearing a
three phase bus and inserting grading capacitance similar to the single phase representation of
figure 20(b). In this case vt ferroresonance did not occur, but in other bus clearing trips it did.
The coupled phase voltage is approximately 25% of the pre-trip (system) voltage and the zero
sequence voltage (3Vo) is about 8%. 3Vo is an indication of the bus capacitance unbalance. In
this case where 3Vo is relatively small zero sequence damping per figure 18(b) will probably not
be sufficient and phase damping per figure 24(a) will be required. Ferroresonance has often
been mitigated in such applications with zero sequence damping, but in many cases phase
damping was required. About the only generalization that can be made is that the greater the
unbalance (the higher 3Vo) the more likely zero sequence damping will be successful.

696
 -3XUb

RD
3

(a) Sequence network connection (b) Equivalent circuit

Figure 25. Sequence network interconnects and equivalent ferroresonant

circuit for ungrounded system of figure 18(b) with damping resistor, RD, of figure 24(b)

Figure 26. Coupled bus vt voltage with all breakers open

For zero sequence damping, sizing the damping resistor ohmic value, RD, with the phasor or
other method of analysis is done and then it is evaluated for meeting thermal requirements.
The thermal requirements depend on the maximum normal state (non-ferroresonant) 3Vo that
will appear across RD during the unbalanced condition and the time that it is there. This can be
the maximum time it takes protective relays to clear the unbalanced condition to an indefinite
time where manual correction is required. In the latter case a damping resistor that can sustain
a continuous current of 3Vo/RD may be permanently connected. If not then an alternate means
of ferroresonance damping is required. These are explored in a later discussion.
The effect of RD being permanently connected must also be investigated on other devices
measuring and using the broken delta 3Vo quantity - e.g. vt polarizing voltage for directional

697
relays, generator resistance grounding and ground fault protection, etc. During balanced
conditions 3Vo will be zero so there will be no effect on phase measurements.

Phase damping

Phase damping is the application of damping resistance placed across the secondary winding as
shown in figure 24(a). This connection cannot generally be a permanent connection for either
vts or power transformers. The damping resistor must usually be switched in and out in some
manner to address the ferroresonance or some other means of ferroresonance suppression.

Ferroresonance Suppression Circuits

Figure 27 shows a number of devices that are used in ferroresonant suppression circuits. Each
device is not necessarily a complete solution, but may complement other approaches. The
availability of properly rated devices and their impact on the operation and performance of
connected apparatus and system must be investigated.

Ls

MOV Switched Filter circuit Saturating reactor PTe thermistor

resistor damping circuit with triac switch

Figure 27. Ferroresonance damping devices

MOV - metal oxide varistor

The MOV provides some level of damping when it is conducting. It is able to limit, or assist in
limiting; the fundamental frequency ferroresonant modes where voltages are above the MOV
rating, but cannot prevent the transition to or the elimination of sub-harmonic modes where
maximum sub-harmonic voltage levels are below the required MOV rating. Therefore, they are
not suitable to apply alone to dampen ferroresonance. They may, however, be added in parallel
to the filter circuit or saturating reactor circuit to enhance performance. Also, MOVs cannot
continuously absorb energy so their application should be evaluated.

Mechanically switched resistor

A mechanically switched resistor is effective at inserting the damping resistance and eliminating
ferroresonance. The problem is that opening the switch may produce a transient that puts the
transformer back into the ferroresonant state. Therefore, the mechanical switch must remain
closed until the ferroresonant circuit is corrected (eliminated). For example, consider the case
of vt ferroresonance due to circuit breaker grading capacitors shown in figure 20. If all the
breakers connected to the bus are opened then a capacitance is inserted in series with the vt
and the ferroresonant circuit is created. The resistor is switched in with a series arrangement
of the breakers' 52b auxiliary contacts. If one breaker is closed and system voltage is applied to
the bus then the ferroresonant circuit is eliminated. This configuration is okay as long as the vt
and resistor are appropriately rated. (This may potentially be a good solution with lEe 6 1 850.)

698
Filter circuit

Solutions that have been applied to ccvts are appropriate for conventional vt or other
transformer applications, but may require different component ratings. The filter circuit is a
fundamental frequency blocking filter that allows damping of harmonic and sub-harmonic
modes of ferroresonance. This leaves primarily the coupled fundamental frequency voltage
operating in the normal state. Selection of the blocking filter with a narrow bandwidth at
fundamental frequency provides better suppression. The addition of a MOV provides additional
secondary surge protection and reduces the ferroresonant damping time. The filter circuit
affects the transient response voltage accuracy during system faults and thus protective relay
operational security. This must be evaluated.

Saturating reactor

A saturating reactor is used to switch in damping resistance when the secondary voltage (flux
density) becomes excessive. It responds to all harmonic and sub-harmonic voltages as well as
the fundamental. This solution has been broadly applied to ccvts and is also appropriate for
conventional vt or other transformer applications, but may require different component ratings.
The operation is relatively simple. Refer to figure 28. The saturating reactor is selected to have
a saturation voltage, VSatR, above, but very close, to the vt rated secondary voltage, YR. During
normal operation at rated voltage the non-saturated magnetizing reactance of the saturating
reactor is so large compared to the damping resistor, RD, that almost the entire vt secondary
voltage is across the reactor. The reactor has little to no effect on the vt burden. When the
ferroresonant circuit of figure 28(a) is formed inserting the series capacitance and vt
ferroresonance is attempted the reactor saturates before the vt and inserts the damping
resistance, RD, into the circuit.

Vt

VSat
_1-_- Reactor

(a) Ferroresonant circuit

(b) Saturating reactor

Figure 28. Saturating reactor application

699
The reactor being saturated is in ferroresonance, however the damping resistor, if calculated
correctly, will force normal operation in a matter of cycles. Be sure to include the reactor's
internal resistance when calculating RD. It is often not negligible. In some cases the resistance
in the reactor is not sufficiently low such that a parallel resistor may be useful to provide more
damping. The parallel resistance must be included in the vt's maximum burden or other
limitations dictated for its application.

PTC Thermistor

Refer to figures 24(b), 25 and 27. This approach is generally limited to applications where
damping resistance is applied in the corner of a broken delta secondary vt winding to eliminate
ferroresonance. This is particularly applicable to ungrounded systems where ferroresonance
occurs and the damping resistor cannot meet the thermal requirements imposed by the normal
sustained zero sequence voltage (3Vo) after the ferroresonance is eliminated. RTh is the
thermistor resistance and RD is additional damping resistance. V represent a voltage threshold
circuit such that if VT is exceeded the triac is switched on closing the damping circuit. VT is set
to allow expected 3Vo unbalance when operating in the non-ferroresonant state. RTh, the PTC
(positive temperature coefficient) thermistor, initially has a very small resistance that will begin
to increase after one to two seconds depending on the current through it. Depending on the
thermistor characteristics RTH can increase to several hundred ohms or more.
Consider an ungrounded system where an operating condition allows the capacitive unbalance
to create ferroresonance. The zero sequence ferroresonant voltage appears across the broken
delta winding and closes the damping circuit eliminating ferroresonance forcing a normal state
operation in a matter of cycles. Then V decreases below VT and the triac is turned off. When
an event occurs, say a ground fault on an ungrounded system a very large 3Vo that is above the
VT setting will exist in the non ferroresonant state until the ground fault is removed. The large
3Vo would create a large damping current that would thermally damage the vt or RD and other
parallel connected loads. For these cases the triac will close the damping circuit and eliminate
the ferroresonance and after a second or more RTh will begin to increase substantially reducing
the current "soft" way (gradual without transients) and the potential of thermal damage and
preventing the reoccurrence of ferroresonance. The thermistor remains closed until the ground
fault is removed.
This approach involves the application of electronics and its application is currently limited to
medium voltage. At higher voltage installations there is still a reluctance to use electronics that
could possibly provide better mitigation solutions. It is expected that more of these solutions
will become available and will be combined with substation automation functions and IEC6 1850.

Detuning the Ferroresonant circuit

It is possible to have applications where the resistance in the circuit may be too large for
effective damping or the slope of the circuit capacitive reactance Xc forces a single point
ferroresonant solution. Consider the circuit of figure 24(a). The equivalent capacitive reactance,
Xc (Cs \\ CB and CSec) is represented in figure 24(c) by the lines with slopes A, B and C for
different values of CB and/or CSec• Curve A represents the single point solution where
ferroresonance cannot be damped. The circuit must be altered or detuned to avoid these
conditions. Slope A represents the XC value that results in a single point ferroresonant
operating state is at point 2, the intersection of slope A with the saturated region of the
transformer's magnetizing characteristic. Increasing the phase-to-ground capacitance, CB, or
adding a secondary capacitance, CSec, across the secondary winding will decrease the value

700
(slope of line) of Xc. As capacitive reactance is decreased toward XM2 to a point where the
ferroresonant operating point would be in severe saturation (slope B), the occurrence of
ferroresonance is reduced or may even not be obtainable. Note from equation 25 that as Xc
approaches the value of XM2 that RMax increases indicating that less damping is required. This is
shown in figure 29(b). If Xc continues to decrease to a value less than XM2 then no damping is
required as there is only the normal operating point. As with any potential solution the
availability of properly rated detuning capacitors and their impact on the operation and
performance of connected apparatus and system.

Vs

Cs

(a) Ferroresonant circuit

c

90
n
80

70

50

40

30
RMax
20

10

Increasing C B o r C , (decreasing Xc) _

(b) M aximum damping resistance (c) Increasing Cs or Cs

Figure 29. Detuning the ferroresonant circuit

Conclusions
Ferroresonance is a widely studied phenomenon but it is still not well understood because of its
complex behavior. It is "fuzzy-resonance. " A simple graphical approach using fundamental
frequency phasors has been presented to elevate the readers understanding. Its occurrence
and how it appears is extremely sensitive to the transformer characteristics, system parameters,
transient voltages and initial conditions. More efficient transformer core material has lead to its
increased occurrence and it has considerable effects on system apparatus and protection.
Power system engineers should strive to recognize potential ferroresonant configurations and
design solutions to prevent its occurrence.

701
Appendix A. Stable operating point analysis

Referring to the circuit of figure l (a) two equations for VM can be derived as
VM = Vs + Xcl M (1)
VM = XM I M (2)
Where Vs is the source voltage, Xc is the capacitive (negative) reactance and XM is the non­
linear magnetizing reactance. Please note that the minus sign is shown on -jXc to illustrate that
it is an impedance vector in the opposite direction of XM and that the voltage Vc always opposes
(in opposite direction) VM. The VM vs. 1M characteristics of both equations are plotted on figure
l (c) to derive possible solutions for VM and 1M at their intersections. As can be observed there
are three possible solutions at pOints 1, 2, and 3. However, further analysis is required to
determine each point's stability as an operating solution.

(a) Equivalent circuit

Vc

Region A Region B
XM > Xc Xc > XM
(b ) Phasor diagram for loop voltage summation of (a) (c) Graphical solutions for VM and 1M

Figure 1. Graphical solution for series LC circuit with non-linear inductance

In Region A of figure l (c) the value (slope) of XM is greater than that of Xc (shown as the
dashed line). This will cause the current 1M to be lagging the source voltage Vs. The phasor
diagram for this region is shown in figure l (b). This shows that for the loop voltage summation
Vs + Vc - VM = 0 (3)
in the circuit of figure l (a), VM acts in opposition to source voltage Vs and Vc acts in the
direction of Vs. In Region B the magnitude of Xc is greater than that of XM. This will cause the
current 1M to be leading the source voltage V S . 1M is now negative of the previous lagging value.
This shows that for equation 3 to be satisfied that VM acts in the direction of the source voltage
Vs and that Vc acts in opposition to Vs.
At point 1, which is in Region A, a small increase in current causes VM to increase more rapidly
than Vc (XM has larger slope). The net difference (VM-VC) is inductive and acts in the opposition

702
to Vs. The circuit impedance is increasing thus forcing a reduction in current back to the
intersection. For a small decrease in current, VM decreases more rapidly than Vc and the net
difference (VM-Vc) is less inductive and still acts in opposition to Vs, but less. The circuit
impedance decreases forcing an increase in current back to the intersection. Without a
significant change (a transient) in current or voltage operation cannot be moved from point 1;
therefore, it is a stable operating point.
At point 2, which is in Region B, a small increase in current causes Vc to increase more rapidly
than VM (Xc has larger slope). The net difference (VM-VC) is capacitive and acts in the opposition
to Vs. The circuit impedance (capacitive reactance) is increasing thus forcing a reduction in
current back to the intersection. For a small decrease in current, Vc decreases more rapidly than
VM and the net difference (VM-VC) is less capacitive and still acts in opposition to Vs, but less.
The circuit impedance (capacitive reactance) decreases forcing an increase in current back to
the intersection. Without a significant current or voltage change (damping) operation cannot be
moved from point 2; therefore, it is a stable operating point.
At point 3, which is in Region A, a small increase in current causes Vc to increases more rapidly
than VM (Xc has larger slope). The net difference (VM-VC) is capacitive and acts in the direction
of Vs. The circuit impedance decreases thus resulting in an increase of current and an increase
in the voltages away from the intersection. As the voltages increase the current continues to
increasing and so on. For a small decrease in current, Vc decreases more rapidly than VM and
the net difference (VM-VC) is more inductive and acts in opposition to Vs. The circuit impedance
increases forcing a decrease in current away from the intersection. Since any small change in
current will avalanche away from point 3 it is an unstable operating point and cannot exist.

703
References

1. Rezaei-Zare, A. and Iravani, R. (20 13). "Ferroresonance in Power and Instrument

Transformers. In: Su, C. Q. Electromagnetic Transients in Transformer and Rotating
Machine Windings. Hershey, PA: IGI Global. Chapter 5.
2. Price, E. "Voltage transformer ferroresonance in transmission substations. " 25 th Annual
Conference for Protective Relay Engineers. Texas A & M University, March 1977.
3. R. Rudenberg. "Transient Performance of Eledric Power Systems. " McGraw Hill Company.
New York, 1950.
4. Piasecki, W., Stosur, M., Florkowski, M., Fulczyk, M., Lewandowski, B., "Mitigating
ferroresonance in HV inductive transformers. " International Conference on Power Systems
Transients. Kyoto, Japan. June 3 - 6, 2009.
5. Mork, B. "Understanding and Dealing with Ferroresonance. " Minnesota Power Systems
Conference, November 7-9, 2006.
6. Horak, J. "A review of ferroresonance. " 30 th Annual Western Protective Relay Conference.
Spokane, Washington. October 2 1 -23, 2003.
7. Hayes, S. and Vaziri, M. "A case study of ferroresonance in a CCVT secondary circuit and its
impact on protective relaying. " 33rd Annual Western Protective Relay Conference. Spokane,
Washington. October 17- 19, 2006.
8. Durbak, D. "Ferroresonance damping devices solves Amtrak's problem. " Power technology.
October 2004.
9. Jacobson, D. and Menzies, R. "Investigation of Station Service Transformer Ferroresonance
in Manitoba Hydro's 230 kV Dorsey Converter Station. " International Conference on Power
System Transients. Rio de Janeiro. June 24 - 28, 200 1.
1 0. Martinez, E., Antonova, G, and Olguin, M. "Ferroresonance phenomenon in CFE, its origin
and effects. " 39rd Annual Western Protective Relay Conference. Spokane, Washington.
October 1 6- 1 8, 20 1 2.
1 1. Piasecki, W. et al. "Ironing out resonance - Ferroresonance prevention in medium voltage
transformers. " ABB Review. April 2005.
1 2. ABB Inc. "Type FSR-983 outdoor current transformer. " ABB Instrument transformers and
sensors.
Biography

Elmo Price received his BSEE degree in 1 970 from Lamar University in Beaumont, Texas and
his MSEE degree in Power Systems Engineering in 1 978 from the University of Pittsburgh. He
began his career with Westinghouse in 1970, which was consolidated into ABB in 1 988. He has
worked in many engineering and management aSSignments with the primary focus on
protective relays. He is currently a Senior Consultant supporting product engineering and
application. Elmo is a registered professional engineer and a Life Senior member of the IEEE
and the IEEE Power System Relay Committee and serves on Line Protection Subcommittee. He
has several patents and has written numerous papers primarily focusing on power system
protection applications.

704