Transient Stability

Transient stability, as described in Chapter 2, is the ability of the power system

to maintain synchronism when subjected to a severe transient disturbance such as a
fault on transmission facilities, loss of generation, of loss of a large load. The system
response to such disturbances involves large excursions of generator rotor angles,
power flows, bus voltages, and other system variables. Stability is influenced by the
nonlinear characteristics of the power system. If the resulting angular separation
between the machines in the system remains within certain bounds, the system
maintains synchronism. Loss of synchronism because of transient instability, if it
occurs, will usually be evident within 2 to 3 seconds of the initial disturbance.
This chapter illustrates the nature of transient stability problems, identifies
factors influencing them, and describes modelling considerations and analytical
techniques applicable to transient stability analysis.

13.1 AN ELElVlENTARY VIEW OF TRANSIENT STABILITY [I -31

Consider the system shown in Figure 13.1, consisting of a generator delivering
power to a large system represented by an infinite bus through two transmission
circuits. An infinite bus, as described in Chapter 12 (Section 12.3), represents a
voltage source of constant voltage magnitude and constant frequency.
We will present fundamental concepts and principles of transient stability by
analyzing the system response to large disturbances, using very simple models. All
resistances are neglected. The generator is represented by the classical model (see
Chapter 5, Section 5.3.1) and the speed governor effects are neglected. The
corresponding system representation is shown in Figure 13.2(a). The voltage behind

Transient Stability

CCT 1
Et

@Jim
X2
Xtr
XI

Chap. 1

b,
/

5Infinite
5 bus
/

CCT 2

Figure 13.1 Single-machine infinite bus system

(a) Equivalent circuit

(b) Reduced equivalent circuit

Figure 13.2 System representation with generator represented by classical model

the transient reactance (Xi)is denoted by E'. The rotor angle 6 represents the angle
by which E leads EB. When the system is perturbed, the magnitude of E' remains
constant at its predisturbance value and F changes as the generator rotor speed
deviates from synchronous speed we.
The system model can be reduced to the form shown in Figure 13.2(b). It can
be analyzed by using simple analytical methods and is helpful in acquiring a basic
understanding of the transient stability phenomenon. This model is identical to that
shown in Figure 12.4 of Chapter 12. From Equation 12.71, the generator's electrical
power output is

Set. 13.1

A n Elementary View of Transient Stability

rnax

=-

E'EB
,X T

Since we have neglected the stator resistance, P, represents the air-gap power
as well as the terminal power. The power-angle relationship with both transmission
circuits in service (11s) is shown graphically in Figure 13.3 as curve 1. With a
mechanical power input of P,, the steady-state electrical power output P, is equal to
pm, and the operating condition is represented by point a on the curve. The
corresponding rotor angle is 6.,

I-

P, withL both circuits I/S

Figure 13.3 Power-angle relationship

If one of the circuits is out of service (OIS), the effective reactance XT is

higher. The power-angle relationship with circuit 2 out of service is shown in Figure
13.3 as curve 2. The maximum power is now lower. With a mechanical power input
of P,, the rotor angle is now 6, corresponding to the operating point b on curve 2;
with a higher reactance, the rotor angle is higher in order to transmit the same steadystate power.
During a disturbance, the oscillation of F is superimposed on the synchronous
speed oo, but the speed deviation (Ao,=dF/dt) is very much smaller than a,.
Therefore, the generator speed is practically equal to coo, and the per unit (pu) air-gap
torque may be considered to be equal to the pu air-gap power. We will therefore use
torque and power interchangeably when referring to the swing equation.

Transient Stability

830

Chap, 1

The equation of motion or the swing equation (see Chapter 3, Section 3.9)
"ay
be written as

where

P,
P,,
H
6
t

mechanical power input, in pu

= maximum electrical power output, in pu
= inertia constant, in MW-s/MVA
= rotor angle, in elec. rad
= time, in s
=

Response to a step change in P,

Let us now examine the transient behaviour of the system, with both circuits
in service, by considering a sudden increase.in the mechanical power input from an
initial value of P,, to P,, as shown in Figure 13,4(a). Because of the inertia of the

(a) Power-angle variations

(b) Rotor angle time response

Figure 13.4 Response to a step change in mechanical power input

Set, 13.1

A n Elementary Vie,w of Transient Stability

,,tor, the rotor angle cannot change instantly from the initial value of 60 to 6,
to the new equilibrium point b at which P, =P, I . The mechanical power
is now in excess of the electrical power. The resulting accelerating torque causes the
rotor to accelerate from the initial operating point a toward the new equilibrium point
b, tracing the P,-6 curve at a rate determined by the swing equation. The difference
between Pmland P, at any instant represents the accelerating power.
When point b is reached, the accelerating power is zero, but the rotor speed
is higher than the synchronous speed wo (which corresponds to the frequency of the
infinite bus voltage). Hence, the rotor angle continues to increase. For values of 6
higher than 61, Pe is greater than P,, and the rotor decelerates. At some peak value
,6, the rotor speed recovers to the synchronous value wo, but P, is higher than P,,,.
The rotor continues to decelerate with the speed dropping below oo; the operating
point retraces the P,-6 curve from c to b and then to a. The rotor angle oscillates
indefinitely about the new equilibrium angle 6 , with a constant amplitude as shown
by the time plot of 6 in Figure 13.4(b).
In our representation of the power system in the above analysis, we have
neglected all resistances and the classical model is used to represent the generator. In
effect, this neglects all sources of damping. Therefore, the rotor oscillations continue
unabated following the perturbation. In practice, as discussed in Chapter 12, there are
many sources of positive damping including field flux variations and rotor amortisseur
circuits. In a system which is small-signal stable, the oscillations damp out.

Equal-area criterion
For the system model considered above, it is not necessary to formally solve
the swing equation to determine whether the rotor angle increases indefinitely or
oscillates about an equilibrium position. Information regarding the maximum angle
excursion (6,) and the stability limit may be obtained graphically by using the powerangle diagram shown in Figure 13.4. Although this method is not applicable to
multimachine systems with detailed representation of synchronous machines, it helps
in understanding basic factors that influence the transient stability of any system.
From Equation 13.3, we have the following relationship between the rotor
angle and the accelerating power:

Now Pe is a nonlinear function of 6, and therefore the above equation cannot be

solved directly. If both sides are multiplied by 2d6/dt, then

Transient Stability

Chap. 13

Integrating gives

The speed deviation d6ldt is initially zero. It will change as a result of the disturbance.
For stable operation, the deviation of angle 6 must be bounded, reaching a maximum
value (as at point c in Figure 13.4) and then chaging direction. This requires the
speed deviation d6ldt to become zero at some time after the disturbance. Therefore,
from Equation 13.6, as a criterion for stability we may write

where 60 is the initial rotor angle and 6, is the maximum rotor angle, as illustrated
in Figure 13.4. Thus, the area under the function P, -P,plotted against 6 must be zero
if the system is to be stable. In Figure 13.4, this is satisfied when area A , is equal to
area A2. Kinetic energy is gained by the rotor during acceleration when 6 changes
from 60 to 6,. The energy gained is
61

El

/(pm-pe)d6 = area A,
60

The energy lost during deceleration when 6 changes from 61 to 6, is

As we have not considered any losses, the energy gained is equal to the energy lost;
therefore, area Al is equal to area A2. This forms the basis for the equal-area criterion.
It enables us to determine the maximum swing of 6 and hence the stability of the
system without computing the time response through formal solution of the swing
equation.

sec, 13.1

An Elementary View of Transient Stability

833

The criterion can be readily used to determine the maximum permissible

increase in Pmfor the system of Figure 13.1. The stability is maintained only if an
area A, at least equal to A I can be located above Pml. If A l is greater than A2, then
6rn>tiL, and stability will be lost. This is because, for 6 S L , Pmlis larger than Pe and
the net torque is accelerating rather than decelerating.
We will examine the mechanism of transient instability by considering next the
system response to a short-circuit fault on the transmission system, which is a more
common form of a disturbance considered in transient stability studies.
Response to a short-circuit fault

Let us consider the response of the system to a three-phase fault at location F

on transmission circuit 2, as shown in Figure 13.5(a). The corresponding equivalent
circuit, assuming a classical generator model, is shown in Figure 13.5(b). The fault
is cleared by opening circuit breakers at both ends of the faulted circuit, the faultclearing time depending on the relaying time and breaker time (see Section 13.5.2).
If the fault location F is at the sending end (HT bus) of the faulted circuit, no
power is transmitted to the infinite bus. The short-circuit current from the generator
flows through pure reactances to the fault. Hence, only reactive power flows and the
active power P, and the corresponding electrical torque Te at the air-gap are zero
during the fault. If we had included generator stator and transformer resistances in our
model, Pe would have a small value, representing the corresponding resistive losses.
If the fault location F is at some distance away from the sending end as shown
in Figures 13.5(a) and (b), some active power is transmitted to the infinite bus while
the fault is still on.
Figures 13.5(c) and (d) show Pe-6 plots for the three network conditions: (i)
prefault (both circuits in service), (ii) with a three-phase fault on circuit 2 at a location
some distance from the sending end, and (iii) postfault (circuit 2 out of service).
Figure 13.5(c) considers the system performance with a fault-clearing time of tcl and
represents a stable case. Figure 13.5(d) considers a longer fault-clearing time (tc2)such
that the system is unstable. In both cases P, is assumed to be constant.
Let us examine the stable case depicted by Figure 13.5(c). Initially, the system
is operating with both circuits in service such that Pe=P, and &=ti0. When the fault
occurs, the operating point suddenly changes from a to b. Owing to inertia, angle 6
cannot change instantly. Since P, is now greater than P,, the rotor accelerates until
the operating point reaches c, when the fault is cleared by isolating circuit 2 from the
system. The operating point now suddenly shifts to d. Now Pe is greater than P,,
causing deceleration of the rotor. Since the rotor speed is greater than the synchronous
speed a,, 6 continues to increase until the kinetic energy gained during the period of
acceleration (represented by area Al) is expended by transferring the energy to the
system. The operating point moves from d to e, such that area A2 is equal to area A l.
At point e, the speed is equal to ooand 6 has reached its maximum value 6,. Since
P, is still greater than P,, the rotor continues to retard, with the speed dropping below
uo.The rotor angle 6 decreases, and the operating point retraces the path from e to

834

Transient Stability

Chap. 13

d and follows the P,-6 curve for the postfault system farther down. The minimum
value of 6 is such that it satisfies the equal-area criterion for the postfault system. I,
the absence of any source of damping, the rotor continues to oscillate with constant
amplitude.

(a) Single-line diagram

(c) Response to a fault cleared

in tcl seconds - stable case

(b) Equivalent circuit

(d) Response to a fault cleared

in tc2seconds - unstable case

Figure 13.5 Illustration of transient stability phenomenon

S ~ C .13.1

An Elementary View of Transient Stability

835

With a delayed fault clearing, as shown in Figure 13.5(d), area A2 above Pm

is less than A l . When the operating point reaches e, the kinetic energy gained during
the accelerating period has not yet been completely expended; consequently, the speed
is still greater than coo and 6 continues to increase. Beyond point e, P, is less than Pm,
and the rotor begins to accelerate again. The rotor speed and angle continue to
increase, leading to loss of synchronism.

Factors influencing transient stability

From the above discussions, and referring to Figure 13.5, we can conclude that
transient stability of the generator is dependent on the following:
(a)

How heavily the generator-is loaded.

(b)

The generator output during the fault. This depends on the fault location and
type.

(c)

The fault-clearing time.

(d)

The postfault transmission system reactance.

(e)

The generator reactance. A lower reactance increases peak power and reduces
initial rotor angle.

(f)

The generator inertia. The higher the inertia, the slower the rate of change in
angle. This reduces the kinetic energy gained during fault; i.e., area At is
reduced.

(g)

The generator internal voltage magnitude (E'). This depends on the field
excitation.

(h)

The infinite bus voltage magnitude EB.

As a means of introducing basic concepts, we have considered a system having

a simple configuration and represented by a simple model. This has enabled the
analysis of stability by using a graphical approach. Although rotor angle plots as a
function of time are shown in Figures 13.4 and 13.5, we have not actually computed
them, and hence the time scales have not been defined for these plots. Practical power
systems have complex network structures. Accurate analysis of their transient stability
requires detailed models for generating units and other equipment. At present, the
most practical available method of transient stability analysis is time-domain
simulation in which the nonlinear differential equations are solved by using step-bystep numerical integration techniques. This leads us to the next section in which some
of the commonly used numerical integration methods are described.

836

Transient Stability

Chap. 13

13.2 NUMERICAL INTEGRATION METHODS [4-61

The differential equations to be solved in power system stability analysis are
nonlinear ordinary differential equations with known initial values:

where x is the state vector of n dependent variables and t is the independent variable
(time). Our objective is to solve x as a function of t, with the initial values of x and
t equal to xo and to, respectively.
In this section we provide a general description of numerical integration
methods applicable to the solution of equations of e above form. In describing these
methods, without loss of generality, we will treat quation 13.10A as if it were a firstorder differential equation. This simplifies presentation and makes it easier for a
novice reader to comprehend the special features of each method.
We will first describe the Euler method, which by virtue of its simplicity
serves as a good introduction to numerical integration, and then we discuss more
advanced methods.

13.2.1 Euler Method

Consider the first-order differential equation

with x=xo at t=to. Figure 13.6 illustrates the principle of applying the Euler method.

Tangent

Figure 13.6

Sec. 13.2

Numerical Integration Methods

837

At x=x,, t=to we can approximate the curve representing the true solution by
its tangent having a slope

Therefore,

The value of x at t=t,=to+At is given by

The Euler method is equivalent to using the first two terms of the Taylor series
expansion for x around the point (x,, to):

After using the Euler technique for determining x=xl corresponding to t=tl, we can
take another short time step At and determine x2 corresponding to t2=t,+At as follows:

By applying the technique successively, values of x can be determined corresponding

to different values of t.
The method considers only the first derivative of x and is, therefore, referred
to as a first-order method. To give sufficient accuracy for each step, At has to be
small. This will increase round-off errors, and the computational effort required will
be very high.
In the application of numerical integration methods, it is very important to
consider the propagation of error, which may cause slight errors made early in the
process to be magnified at later steps. Numerical stability depends on the propagation
of error. If early errors carry through but cause no significant further errors later, the
method is said to be numerically stable. If, on the other hand, early errors cause other
large errors later, the method is said to be numerically unstable.

Transient Stability

Chap. 13

13.2.2 Modified Euler Method

The standard Euler method results in inaccuracies because it uses the derivative
at the beginning of the interval as though it applied throughout the interval. ~h~
modified Euler method tries to overcome this problem by using the average of the
derivatives at the two ends.
The modified Euler method consists of the following steps:
(a)

Predictor step. By using the derivative at the beginning of the step, the value
at the end of the step is predicted

(b)

Corrector step. By using the predicted value of xf,the derivative at the end of
the step is computed and the average of this derivative and the derivative at
the beginning of the step is used to.find the corrected value

63'

If desired, a more accurate value of the derivative at the end of the step can be
calculated, again by using x=xf. This derivative can be used to calculate a more
accurate value of the average derivative which is in turn used to apply the corrector
step again. This process can be used repeatedly until successive steps converge with
the desired accuracy.
The modified Euler method is the simplest of predictor-corrector (P-C)
methods. Among the well known higher order P-C methods are the Adarns-Bashforth
method, Milne method, and Hamming method [4]. The applicability of these methods
to power system stability analysis has been investigated in reference 7 and has been
found to suffer from a number of limitations. They are not self-starting, need more
computer storage and require smaller time steps than the Runge-Kutta methods
described below.
13.2.3 Runge-Kutta (R-K) Methods [4,5]

The R-K methods approximate the Taylor series solution; however, unlike the
formal Taylor series solution, the R-K methods do not require explicit evaluation of
derivatives higher than the first. The effects of higher derivatives are included by
several evaluations of the first derivative. Depending on the number of terms
effectively retained in the Taylor series, we have R-K methods of different orders.

Set. 13.2

Numerical Integration Methods

839

second-order R-K method

Referring to the differential equation 13.1OB, the second-order R-K formula
for the value of x at t=t,+At is

where

This method is equivalent to considering first and second derivative terms in the
Taylor series; error is on the order of ~ t ~ .
A general formula giving the value of x for the (n+l)Ststep is

where

Fourth-order R-K method

The general formula giving the value of x for the (n+l)st step is

where
kl

= f(xn,tn)At

840

Transient Stability

Chap. 13

The physical interpretation of the above solution is as follows:

(slope at the beginning of time step)At
= (first approximation to slope at midstep)At
= (second approximation to slope at midstep)At
= (slope at the end of step)At
AX = 116 (k, +2k2+2k3+k4)
k,
k2
kj
k4

Thus Ax is the incremental value of x given by the weighted average of estimates

based on slopes at the beginning, midpoint, and end of the time step.
This method is equivalent to considering up to fourth derivative terms in the
Taylor series expansion; it has an error on the order of ~ t ~ .

Gill's version of fourth order R-K method (R-K-G)

With xo as the initial value of x at the begimng of a step and by using j=1,2,3
and 4 to denote four stages, each stage of the Gill method can be described as follows
141:

The values of the a, b, and c coefficients are as follows:

Solution at the end of a time step is given by x4. Initially qo=0, thereafter in advancing
the solution, qo for the next step is equal to q4 of the previous step.
The following are the advantages of Gill's version of the R-K method:
(a)

Roundoff errors are minimized (variable q is used for this purpose).

(b)

Storage requirements are less than for the original R-K method.

Care should be exercised in applying the R-K-G method when discontinuities

or sudden changes in the rate of change of variables occur. Within a time step, a
variable should not be limited, as a result of sharp nonlinearities; otherwise the
variable q takes on incorrect values.

Set, 13.2

Numerical Integration Methods

~ichardson's formula for accuracy check

,

The accuracy of results obtained with the above numerical integration methods
may be checked by using Richardson's formula which gives the accumulated
@*ncation error propagated in the coyse of integration. The difference between the
true value of a variable and the value obtained by a fourth-order Runge-Kutta method
using a step length of At is given by [4,7]

where
true)
x(At)
x(2At)

true value of x
= value computed with a step length of At
= value computed with a step length of 2At
=

13.2.4 Numerical Stability of Explicit lntegration Methods

Integration methods described in Sections 13.2.1 to 13-2.3 (i.e., the Euler,

predictor-corrector, and R-K methods) are known as explicit methods. In these
methods the value of the dependent variable x at any value of t is computed from
knowledge of the values of x from the previous time steps. In other words, %+, for
(n+l)ststep is calculated explicitly by evaluating f(x,t) with known x. These methods
are easy to implement for the solution of a complex set of system state equations.
A significant limitation of the explicit integration methods is that they are not
numerically A-stable [6]. Consequently, the length of the integration time step (At) is
restricted by the small time constants of the system.
Numerical stability is related to the stiffness of the set of differential equations
representing the system. The stiffness is associated with the range of time constants
in the system model. It is measured by the ratio of the largest to smallest time
constant, or more precisely by the ratio of the largest to smallest eigenvalues of the
linearized system.
Stiffness in a transient stability simulation increases with modelling detail. In
the overall system, not all the time constants may be readily apparent. Thus the
stiffness may be hidden and can only be established by computing the eigenvalues of
the linearized system.
Explicit integration methods have weak stability numerically; with stiff
systems, the solution "blows up" unless a small step size is used. Even after the fast
modes die out, small time steps continue to be required to maintain numerical
stability.

842

Transient Stability

Chap. 13

13.2.5 Implicit Integration Methods

Consider the differential equation

dx
dt

= f(x,t)

with x=xo at t = t o

The solution for x at t=t,=to+Atmay be expressed in integral form as

Implicit integration methods use interpolation functions for the expression

under the integral. Interpolation implies that the fknctions must pass through the yet
unknown points at time tl.
The simplest implicit integration method is the trapezoidal rule. It uses linear
interpolation. As shown in Figure 13.7, this implies that the area under the integral
of Equation 13.19 is approximated by trapezoids; hence the name - trapezoidal rule.

Figure 13.7

The trapezoidal rule for Equation 13.19 is given by

A genera1 formula giving the value of x at t=t,+, is

Set. 13.2

Numerical Integration Methods

We see that x, appears on both sides of Equation 13.2 1. This implies that the
variable x is computed as a function of its value at the previous time step as well as
the current value (which is unknown). Therefore, an implicit equation must be solved.
The trapezoidal rule is numerically A-stable [ 6 ] . The stiffness of the system
being analyzed affects accuracy but not numerical stability. With larger time steps,
high frequency modes and fast transients are filtered out, and the solutions for the
modes is accurate. For systems involving simulations in which time steps are
limited by numerical stability considerations rather than accuracy, implicit methods
are generally better suited than explicit methods.
The trapezoidal rule is a second-order method. Implicit integration methods of
higher order have been proposed in the literature on numerical methods. However,
they have not been widely used for power system applications since they are more
difficult to program and less numerically stable than the trapezoidal rule.
In the above description of different numerical integration methods, for
simplicity we have considered a first-order differential equation. When applied to the
analysis of power system stability, the system equations are organized as a set of firstorder differential equations. The rate of change of each state variable depends on other
state variables and is not an explicit function of time t. The following example serves
as a simple illustration of the application of numerical integration as well as the equalarea criterion for transient stability analysis.
+

Example 13.1
In this example, we examine the transient stability of a thermal generating station
consisting of four 555 MVA, 24 kV, 60 Hz units supplying power to an infinite bus
through two transmission circuits as shown in Figure E13.1. This system is the same
as the one considered in Example 12.2 of Chapter 12 in which we examined the
small-signal performance.

LT Trans.

4
7.

Y -

jw-

CCT 2

Infinite
5 bus
?5 EB

j0.5

,F

CCT 1

j0.93

QFigure E13.1

Transient Stability

Chap. 13

The network reactances shown in the figure are in per unit on 2220 MVA, 24 kV base
(referred to the LT side of the step-up transformer). Resistances are assumed to be
negligible.
The initial system-operating condition, with quantities expressed in per unit on 2220
MVA and 24 kV base, is as follows:
P=0.9

Q=0.436 (overexcited)

&=I .0~28.340 J ! ? ~ = o . ~ o o ~ ~ L o

The generators are modelled as a single equivalent generator represented by the

classical model with the following parameters expressed in per unit on 2220 MVA
24 kV base:

Circuit 2 experiences a solid three-phase fault at point F, and the fault is cleared by
isolating the faulted circuit.
(a)

Determine the critical fault-clearing time and the critical clearing angle by
computing the time response of the rotor angle, using numerical integration.

(b)

Check the above value of critical clearing angle, using the equal-area
criterion.

Solution
With the generator represented by the classical model, the system-equivalent circuit
is as shown in Figure E13.2.

Figure E13.2 Equivalent circuit

For the initial operating condition, the voltage behind Xi is

13.2

Numerical 1ntegration)Methods

845

Figure E13.3 shows the reduced equivalent circuits representing the three system
conditions: (i) prefault, (ii) during fault, and (iii) postfault. Also shown in the figure
are the corresponding expressions for the electrical power output as a function of 6.

(a) Prefault

(b) During fault

(c) Postfault
Figure E13.3 Reduced equivalent circuits and equations for power output

(a) Time response using numerical integration:

Equation 13.3 may be written in terms of the two first-order equations:

Transient Stability

346
where

lo

1.351

p~

1.1024

Chap. 13

before the fault

during the fault

after the fault

The initial values of 6 and Amr are 41.77" and 0 pu, respectively.
Any of the numerical integration methods described in Section 13.2 may be used to
solve Equations E13.1 and El 3.2. For illustration, let us consider the second-order RK method. The general formulas giving the values of Am, 6, and t for the (n+lyt step
of integration are as follows:

where

Figure E13.4 shows plots of 6 as a function of time, for the three values of faultclearing time (t,): 0.07 s, 0.086 s, and 0.087 s. The corresponding values of clearing
angle (6,) are 48.58", 52.04", and 52.30, respectively. These results were computed
using a time step (At) of 0.05 s throughout the solution. The time step was, however,
adjusted near the fault-clearing time so as to give the exact switching instant.
From the results, we see that the system is stable with tc=0.086 s (6,=52.04'), and is
unstable with tc=0.087 s (6,=52.30); the critical clearing time is, therefore,0.0865-tO.0005 s, and the critical clearing angle is 52.17"~0.13".

13.2

847

Numerical integratioh Methods

Figure E13.4 Rotor angle response for different values

of fault-clearing time

(71) Equal-area criterion:

The power-angle diagrams for the three network conditions are shown in Figure
E13.5. For the critically stable case, the maximum swing in 6 is given by
1.1024 sin6, =0.9

hence, 6, = 125.27"

fault

Figure E13.5 Power-angle diagram

848

Transient Stability

Chap. 1

From Figure E 13.5, for critical clearing,

A,

A,

or

Therefore, from the plots, we have

Thus the critical clearing angle is

This agrees with the value determined from time responses.

13.3 SIMULATION OF POWER SYSTEM DYNAMIC RESPONSE

13.3.1 Structure of the Power System Model

Analysis of transient stability of power systems involves the computation of

their nonlinear dynamic response to large disturbances, usually a transmission network
fault, followed by the isolation of the faulted element by protective relaying.
Figure 13.8 depicts the general structure of the power system model applicable
to transient stability analysis. This model structure is similar to that presented in
Chapter 12 (Figure 12.18) for small-signal stability analysis. However, for transient
stability analysis, nonlinear system equations are solved. In addition, large
discontinuities due to faults and network switching, and small discontinuities due to
limits on system variables, appear in the system model. Bus voltages, line flows, and
performance of protection systems are of interest, in addition to the basic information
related to the stability of the system.
As seen in Figure 13.8, the overall power system representation includes
models for the following individual components:
Synchronous generators, and the associated excitation systems and prime
movers
e

Interconnecting transmission network including static loads

set. 13.3

Simulation of Power System Dynamic Response

C--------------------------------------------------------------------

Stator equations and ER,EI

axes transformation I,, 1,

i
I

Transmission
network

[-.--.-------1. . . . . . . . . . . . . . . . . . . . . . . . .

**

Individual machine
reference fi-ame: d-q

) Other generators

I
I

equations

/ } Motors

i
:

1 1

'

including
static loads

governor

849

1I

Other dynamic
} devices, e.g.,
HVDC, svc

Common reference
fiame: R-I

* Algebraic equations
** Differential equations
Figure 13.8 Structure of the complete power system model
for transient stability analysis

Induction and synchronous motor loads

Other devices such as HVDC converters and SVCs

The model used for each component should be appropriate for transient stability
analysis, and the system equations must be organized in a form suitable for applying
numerical methods.
As we will see in what follows, the complete system model consists of a large
set of ordinary differential equations and large sparse algebraic equations. The
transient stability analysis is thus a differential algebraic initial-value problem.
13.3.2 Synchronous Machine Representation [8]

Synchronous machine modelling is presented in Chapters 3 to 5. The per unit

synchronous machine equations are summarized in Section 3.4.9. Modifications of
these equations necessary for representation in stability studies are discussed in
Section 5.1.
To illustrate the implementation of the generator model for transient stability
analysis, we assume that the generator is represented by a model with one d-axis and
two q-axis amortisseurs as shown in Figure 13.9. However, the equations presented
here can be readily modified to account for a model with an arbitrary number of
amortisseurs.

Transient Stability

Chap,

Figure 13.9 Synchronous machine equivalent circuits

The following is a summary of the synchronous machine equations as a set of
first order differential equations, with time t in seconds, rotor angle 6 in electrical
radians, and all other quantities in per unit.
Equations of motion

where

w,
Am,
p

27ch electrical radls

= pu rotor speed deviation
= derivative operator d d t
=

Rotor circuit equations

With rotor currents expressed in terms of rotor and mutual flux linkages (see
Equation 12.159 of Chapter 12), the rotor circuit dynamic equations are

set, 13.3

Simulation of Power System Dynamic Response

85 1

The d- and q-axis mutual flux linkages are given by

where

Lags

'19

'29

Here La& and Lagsare saturated values of the d- and q-axis mutual inductances given
by

and K, and Ksq are computed as a function of the air-gap flux linkage
described in Chapter 3 (Section 3.8.2).

vatas

Stator voltage equations

With stator transients @vd,pvq)
and speed variations (m/mO) neglected as
discussed in Chapter -5 (Section 5.1), the stator voltage may be written as follows:

852

Transient Stability

Chap.

with

Since we have neglected the effect of speed variations on the stator voltage,
6 =a/ao= 1.0 in the above equations. Consequently, 6L i =Xiand 615;' =X:. The above
equations are in the individual machine d-q referenc@frame which rotates with the
machine's rotor. For the solution of the interconnecting transmission network
equations, a synchronously rotating common R-I reference is used. The relationships
shown in Figure 13.10 are used to transform variables from one reference frame to
the other. The R-axis of the common reference frame also serves as the reference for
measuring the rotor angle 6 of each machine.

Figure 13.10 Reference frame transformation

and definition of rotor angle 6

For convenience in the organization of the complete set of algebraic equations,

the stator voltage equations are expressed in the common R-I reference fi-ame as
indicated in Figure 13.8. Use of Equations 13.31 to transform the stator voltage
equations 13.28 yields

Set. 13.3

Simulation of Power System Dynamic Response

The elements of the impedance matrix are given by

R,

(Xi -X:) sin6 cos6 +Ra

RII

(X: -Xi) sin6 cosS +Ra

X,

xicos26+x:sin26

X,

xisin26+x"cos26
4

earlier, 6 is assumed to be equal to 1.0 pu. The internal voltage components

are given by
AS noted

If subtransient saliency is negligible, L i =L:. Then

In this case, E i +jE; represents the voltage behind the subtransient impedance R, + j Y .
For network solution, the generator may be represented by either of the simple
equivalent circuits shown in Figure 13.1 1 .

(a) Thevenin' s equivalent

(b) Norton' s equivalent

Figure 13.11 Synchronous machine equivalent circuits

with negligible subtransient saliency

854

Transient Stability

Chap.

Variations in Lads and L, due to saturation may introduce a small amount of

subtransient saliency during a transient condition. This is usually insignificant and
may be ignored, if it is desired to take advantage of the computational simplicity
offered by the above equivalent circuits.
Active power and reactive power at the generator stator terminals are

The air-gap torque required for the solution of the swing equation (13.22) is

Since we have assumed 6 =a/ao= 1.O pu in the ststor voltage equations, in per unit the
air-gap torque is equal to the air-gap power (see Chapter 5, Section 5.1.2). Hence,

The field current in the reciprocal per unit system is given by

The per unit exciter output current Ifd (see Chapter 8, Section 8.6) is

Here we have considered a generator model with one d-axis and two q-axis
amortisseur circuits. For models with a different number of rotor circuits, changes to
the above formulation of machine equations are straightforward. However, we have
assumed equal mutual inductances between the armature and rotor circuits in each
axis. Reference 8 provides a description of the implementation of a model with
unequal mutual inductances for stability analysis, using the above general approach.

set 13.3

Simulation of Power System Dynamic Response

855

Initial values of generator variables:

As noted earlier, the transient stability analysis involves the solution of a large
of differential and algebraic equations with known initial values. Prefault powerflow analysis provides the initial values of network variables, including the active and
power outputs and voltages at the generator terminals.
The procedure for computing the initial values of the generator quantities for
known terminal conditions is presented in Section 3.6.5 of Chapter 3.

classical generator model and infinite bus

For a machine represented by the classical model, X,=X,=Xi,
and E" is
by E. The rotor angle 6 is the angle by which E leads the R-axis. The
magnitude of E' is constant throughout the solution. The R and I components of E'
change with 6, as determined by the solution of the swing equation.
For a node associated with an infinite bus, both the magnitude and angle of the
node voltage remain constant.

13.3.3Excitation System Representation

Models for different types of excitation systems in use are described in Chapter
8. We will illustrate the method of incorporating these models into a transient stability
program by considering the excitation system model shown in Figure 13.12. It
represents a bus-fed thyristor excitation system (classified as type STlA in Chapter
8) with an automatic voltage regulator (AVR) and a power system stabilizer (PSS).
A high exciter gain (KA)without transient gain reduction or derivative feedback is
used. The model of Figure 13.12 is identical to the excitation system model
considered in Chapter 12 (see Figure 12.14), except that we also account for the limits
on exciter and PSS outputs.
Terminal voltage
transducer
Et

1
* - .v1
l+sT,

Gain

Washout

Exciter

Phase
compensation

vs

VS max

Power system stabilizer

Figure 113.12 Thyristor excitation system with AVR and PSS

E~rnax

856

Transient Stability

Chap.

For a bus-fed (potential source) thyristor exciter, the voltages vary with the
generator terminal voltage (E,) and exciter output current (I/,):

From block 1 of Figure 13.12, we may write

From blocks 3 and 4,

with pAw, given by Equation 13.22. From block 5,

with pv, given by Equation 13.42. The stabilizer output v, is

vs = v3

with

From block 2, the exciter output voltage is

with

set. 13.3

Simulation of Power System Dynamic Response

857

Both limits considered here are windup limits. Modelling of non-windup limits
is discussed in Chapter 8.
The generator field voltage efd in the reciprocal per unit system is related to
exciter output voltage Efd (see Chapter 8, Section 8.6) as follows:

~n alternative method of treating blocks 4 and 5, in which the derivatives of the input
variables are not required, is described in Chapter 12, Section 12.3.4.
Initial values of excitation system variables:
For any given steady-state generator output, the field voltage efd is determined
by the generator equations (see Section 13.3.2). The excitation system quantities are
determined as follows:

The AVR reference is

Thus Vmf takes a value appropriate to the generator loading condition prior to the
disturbance.

Representation of field-sh orting circuit [8]

In the case of an ac or static exciter, the exciter current usually cannot be
negative. Therefore, a special field-shorting circuit in the form of a "crowbar" or a
varistor is provided to bypass the exciter and allow negative field current to flow (see
Chapter 8, Section 8.5.8).
The effects of field-shorting circuits are simulated by setting the field voltage
to zero and increasing the field resistance (Rfd) appropriately during the period when
the field current is negative. The resistance added is fixed for a crowbar. For a
varistor the resistance is a nonlinear function of the applied voltage. Once the field
current becomes positive, the field circuit resistance is restored to its normal value,
and the field voltage is established by the exciter output.

Transient Stability

Chap.

13.3.4 Transmission Network and Load Representation

The transients associated with the transmission network decay very rapidly.
fact, network transients will have died out by the time the solution of the swing
equation is advanced by one time step. Therefore, it is usually adequate to regard the
network, during the electromechanical transient conditions, as though it were passing
directly from one steady state to another. The fundamental frequency (50 or 60 H ~ )
variations may be regarded as microprocesses, and only the variations in the envelopes
(amplitude modulation) of current and voltage waveforms are considered for stability
analysis. For analysis of balanced conditions, a single-phase representation of the three
phases is used. Unbalanced faults are simulated by using symmetrical components as
described in Section 13.4.
Without the balanced steady-state representation of the transmission network,
stability analysis of large practical power systems would be impractical. For special
electromechanical problems requiring inclusion of transmission %&work and generator
stator transients, and three-phase representation, a program such as the electromagnetic
transients program (EMTP) may be used [9]. For conventional transient stability
analysis, the network representation is similar to that for power-flow analysis. As
described in Chapter 6 (Section 6.4.1.), the most convenient form of network
representation is in terms of the node admittance matrix. The structure and
formulation of this matrix are presented in Section 6.4.1.
The characteristics and modelling of loads are covered in Chapter 7. Dynamic
loads are represented as induction or synchronous motors, and their treatment is
similar to that of synchronous machines.
Static loads are represented as part of the network equations. Loads with
constant impedance characteristics are the simplest to handle and are included in the
node admittance matrix. Nonlinear loads are modelled as exponential or polynomial
functions of bus voltage magnitude and frequency. The net effect is that a nonlinear
static load model is treated as a current injection at the appropriate node in the
network equation. The value of the node current from ground into the network is

vi

where
is the conjugate of the load bus voltage, and PL and QL are portions of the
active and reactive components of the load which vary as nonlinear functions of VL
and frequency deviation. For an inductive load QL is positive.
The overall networkfload representation comprises a large sparse nodal
admittance matrix equation with a structure similar to that of the power-flow problem
discussed in Chapter 6, Section 6.4. The network equation is identical to Equation
6.87, which in matrix notation may be written as

<

set. 13.3

859

Simulation of Power System Dynamic Response

The node admittance matrix YN is symmetrical, except for dissymmetry introduced

by Phase-shifting transformers (see Section 6.2.3). Within the time frame of transient
stability simulations, transformer taps and phase-shift angles do not change. Therefore,
the elements of the matrix are constant except for changes introduced by networksdtching operations.
The effects of generators, nonlinear static loads, dynamic loads, and other
devices such as HVDC converters and regulated var compensators are reflected as
boundary conditions providing additional relationships between P and f at the
respective nodes.
In contrast to power-flow analysis (See Chapter 6), tie line power-flow control,
limits on generator reactive power output, and the slack bus make-up for the unknown
losses need not be considered in transient stability analysis.

Simulation of faults:

A fault at or near a bus is simulated by appropriately changing the selfadmittance of the bus.
Methods of simulating different types of fault will be discussed in Section
13.4.5. Depending on the type of fault, the equivalent negative- and zero-sequence
impedances seen at the fault point are computed, combined appropriately, and inserted
between the fault point and ground. This simply alters the self-admittance of the node
representing the fault bus.
For a three-phase fault, the fault impedance is zero andcihe faulted bus has the
same potential as the ground. This involves placing an infinite shunt admittance. In
practice, a sufficiently high shunt admittance (i.e., a very small shunt impedance) is
used so that the bus voltage is in effect zero. For example, a conductance G equal to
lo6 pu on 100 MVA base would effectively reduce the bus voltage to zero. The fault
is removed by restoring the shunt admittance to the appropriate value depending on
the postfault system configuration.
13.3.5 Overall System Equations

In Sections 13.3.2 and 13.3.3 we described the organization of synchronous

generator and excitation system equations. Modelling of turbine governing systems
is described in Chapter 9. Treatment of these models is similar to that of the
excitation system.
Other devices whose dynamics are represented in transient stability studies
include motors, HVDC converters and their controls, static var compensators, and
synchronous condensers. Models of these devices are described in Chapters 7, 10, and
11.
Equations for each of the generating units and other dynamic devices may be
expressed in the following form:

Transient Stability

Chap.

where
state vector of individual device
Id = R and I components of current injection from the device into the network
Vd = R and I components of bus voltage

xd

The overall system equations, including the differential equations (13.5 1) for
all the devices and the combined algebraic equations for the devices (13.52) and the
network (13S O ) are expressed in the following general form comprising a set of firstorder differential equations

and a set of algebraic equations

with a set of known initial conditions (xo,Vo), where

x
V
I

state vector of the system

= bus voltage vector
= current injection vector
=

Time t does not appear explicitly in the above equations. A wide range of approaches
has been reported in the literature for solving these equations, depending on the
numerical methods and modelling details used. Reference 10 provides a review of
these approaches. The many possible schemes for the solution of Equations 13.53 and
13.54 are characterized by the following factors:
(a)

The manner of interface between the differential equations 13.53 and the
algebraic equations 13.54. Either a partitioned approach or a simultaneous
approach may be used.

(b)

The integration method used, i.e., implicit method or explicit method.

(c)

The method used for solving the algebraic equations. As in the case of powerflow analysis described in Chapter 6, one of the following methods may be
used: (i) the Gauss-Seidal method based on admittance matrix formulation, (ii)

set. 13.3

Simulation of Power System Dynamic Response

861

a direct solution using sparsity-oriented triangular factorization, and (iii) an

iterative solution using the Newton-Raphson method.

~ 1 of1 the above approacheslmethods have been successfully used in production-grade

transient stability programs. In the next section, two basic solution schemes are
described. One scheme uses a partitioned solution and an explicit integration method,
and the other uses a simultaneous solution with an implicit integration method.

13.3.6Solution of Overall System Equations

(a) Partitioned solution with explicit integration

In this approach, the algebraic and differential equations are solved separately.
Initially, at t=O-, the values of the state variables x and the network variables V and
I are known, and they form a consistent set; that is, the system is in steady state and
the time derivatives f(x,V) are equal to zero.
Following a disturbance, usually a network fault, the state variables x cannot
change instantly. The algebraic equations 13.54 are solved first to give V and I, and
the corresponding power flows and other non-state variables of interest at t=O'. Then
the time derivatives f(x,V) are computed by using the known values of x and V in
Equations 13.53. These then can be used to initiate the solution of the state variables
x by using any of the explicit integration methods described in Section 13.2.
We will illustrate this by considering Gill's version of the fourth order R-K
method. There are four stages (j=1 to 4) per time step At. Using the values of f(x,V)
computed at the beginning of a time step, k,, x,, and q, are computed according to
Equations 13.17 with j= 1. Then the algebraic equations 13.54 are solved with x =a,to
compute V1 and I,. These are in turn used in Equations 13.17 to compute k2, x2, and
q,. This process, involving alternating solutions of algebraic and differential equations,
is applied successively with x4 representing the solution at the end of each time step.
During a switching operation, the network variables change instantly, but not the state
variables.
Since the solution of differential equations requires values of network and state
variables only from the previous steplstage, the differential equations may be
partitioned in any way desired. For example, differential equations associated with
each device may be solved independently. This offers considerable programming
flexibility.
For solution of networks associated with large interconnected systems, the most
efficient method is to use sparsity-oriented triangular factorization.
The partitioned approach with explicit integration is the traditional approach
used widely in production-grade stability programs. Its advantages are programming
flexibility and simplicity, reliability, and robustness. Its principal disadvantage is
susceptibility to numerical instability. For a stiff system, a small time step is required
throughout the solution period, as dictated by the smallest time constant (or
eigenvalue).

862

Tran.sient Stability

Chap. 1

(b) Simultaneous solution with implicit integration 1141

In this approach, the state variables and the network variables are solved
simultaneously. We will illustrate this using the trapezoidal rule.
With X=H, and V=Vn at t=tn, the solution of x at t=t,+, =tn+Atis given by
applying the trapezoidal rule (Equation 13.21) to solve Equation 13.53:

From Equation 13.54, the solution of V at t=tn+,is

The vectors

and Vn+Iare unknown. Let

and

At solution,

Equations 13.59 and 13.60 are both nonlinear algebraic equations. Thus the
differential equations have been made algebraic by using an implicit formula. These
equations are very sparse; for computational efficiency it is necessary to take '
advantage of their special structure. Applying the Newton method (see Chapter 6) to
solve Equations 13.59 and 13.60, we may write for the (k+l)stiteration

The following equation is solved to obtain Ax:+, and AV~,+,:

Set. 13.3

Simulation of Power System Dynamic Response

The Jacobian in the above equation is computed at x = ~ k , and

+~
following structure:

v = v ~ ,It+has
~ .the

The matrices AD, BD, CD and YD are associated with the models for the dynamic
devices and nonlinear static loads. For a system with m such devices, they have the
following structures:

The solutions of Equations 13.59 and 13.60 are given in terms of the above matrices
by

In the above equations, k is the iteration counter and good starting values
v:+~)
are established by extrapolation. Also F:+~ and ~ k , , , are the residue vectors of the
states and current injections, respectively.
From Equation 13.64,
:
can be expressed as a function of AV~,+,

Transient Stability

Chap,

Substitution of Equation 13.66 into Equation 13.65 yields

Now AV:, and AX:,, can be calculated by solving Equations 13.66 and 13.67. Then
and v!:: are obtained from Equation 13.61.

$:

Treatment of discontinuities:
Equations 13.64 and 13.65 are valid only when the functions given by
Equations 13.57 and 13.58 are continuous and differentiable. At points of
discontinuity, such as network switching or limits on state variables, the exact
formulation by the Newton method or any method requiring derivatives would be
complicated [13]. This problem is dealt with in reference 14 as @llows:
For large network discontinuities, such as a network fault or switching
operations, the integration method is temporarily changed to the fourth order
Runge-Kutta method for one step at the point of discontinuity. This time step
has a zero step size and is used only for the calculation of the postfault
network conditions (the state vector is not updated). After this, the normal
trapezoidal integration is resumed.
For local non-differentiable functions, such as limits associated with
controllers, the device Jacobians are computed by neglecting their effect. This
is acceptable since they have only local impact and the overall convergence
will not be significantly affected.

Example 13.2
In this example, we analyze the transient stability of the system of Figure E13.1
(considered in Example 13.1) including the effects of rotor circuit dynamics and
excitation control. The system diagram is reproduced here as Figure E13.6 for,
reference.

Infinite
bus
EB

Network reactanc
Figure E13.6

2220 MVA base

13.3

Simulation of Power System Dynamic Response

865

Generator parameters:
The four generators of the plant are represented by an equivalent generator whose
parameters in per unit on 2220 MVA base are as follows:

The above parameters are unsaturated values. The effect of saturation is to be

represented by assuming that d- and q-axes have similar saturation characteristics with
A,,

0.031

B,,,

6.93

WTI

0.8

Vn=

Excitation system parameters:

The generators are equipped with thyristor exciters with AVR and PSS as shown in
Figure 13.12, with parameters as follows: -.

KA = 200
KmB = 9.5
Vs m m = 0.2

TR = 0.015 s
Tw = 1.41 s
Vsmin = -0.2

EFmm = 7.0
TI = 0 . 1 5 4 ~

EFmin= -6.4
T2 = 0.033 s

The exciter is assumed to be alternator supplied; therefore, EFmm and EFminare

independent of El.

Prefault system condition in pu on 2220 M A , 24 kV base:

0.9
= l.OL28.34O

El

0.436 (overexcited)
EB = 0.90081LO
=

Disturbance:
A three-phase fault on circuit 2 at point F cleared by isolating the faulted circuit
simultaneously at both ends.
Examine the stability of the system with the following alternative forms of excitation
control:
(0

Manual control, i.e., constant E/,

(ii)

AVR with no PSS

(iii)

AVR with PSS

Tran.sient Stability

Chap. 1

Consider the following alternative fault-clearing times:

(a)

0.07 s

(b)

0.10s

Solution

The generators of this example have the same characteristics as the generator
considered in examples of Chapters 3 and 4.
The per unit fundamental parameters as determined in Example 4.1 are as follows:

The initial per unit values of generator variables based on the results of Examples 3.2
and 3.3 are as follows:

The initial per unit values of exciter output current and voltages are

With EB assumed to be in phase with the R-axis, the initial value of the rotor angle

6,+28.34"

39.1+28.34"

67.44"
%

(a) Transient response with the fault-clearing time t, equal to 0.07 s:

The results of time responses computed with the three alternative excitation controls
are shown in Figures E13.7(a) to (d), which show time responses of 6, P,, E, and El,
respectively. These results were computed by using Gill's version of the R-K
integration method with a time step of 0.02 s.
From Figure E13.7(a), we see that with constant Efd, the system is transiently stable;
however, the level of damping of oscillations is low.
With a fast-acting AVR and a high exciter ceiling voltage, the first rotor angle swing
is significantly reduced. However, the subsequent swings are negatively damped; the
system loses synchronism in the third swing.

3.3

Simulation of Power System Dynamic Response

180

II

AVR with no PSS-/

I

V1

a>

120 -

.H

CQ

hl)

2
8

60 :
Prefault

L'

' I I'
\

I
I

Oo

2
3
Time t in seconds

Figure E13.7 (a) Rotor angle response with fault cleared in 0.07

Figure E13.7 (b) Active power response with fault cleared in 0.07

Transient Stability

Chap.

AVR and PSS

1.o

--..---------I

0.8 I

0.6 0.4 -

Constant Efd

\
1

t
I
I

,\

"/

AVR with no PSS

0.2 0

2
3
Time t in seconds

Figure E13.7 (c) Terminal voltage response with fault cleared in 0.07 s

Time t in seconds
Figure E13.7 (d) Exciter output voltage response with fault cleared in 0.07 s

Set. 13.3

869

Simulation of Power System Dynamic Response

With the PSS, the rotor oscillations are very well damped without compromising the
first-swing stability.
The results with regard to oscillatory stability (i.e., damping of rotor oscillations) are
consistent with those of Examples 12.4 and 12.5 of Chapter 12.
(b) Transient response with the fault-clearing time t, equal to 0.1 s:

The transient responses of rotor angle 6 with the three alternative forms of excitation
control are shown in Figure E13.8. With constant Ef, the generator is first-swing
unstable. With a fast-acting exciter and AVR, the generator maintains first-swing
stability but loses synchronism during the second swing. The addition of a PSS
contributes to the damping of second and subsequent swings.
It is evident from these results that the use of a fast exciter having a high-ceiling
voltage and equipped with a PSS contributes to the enhancement of the overall system
,
stability.

n s t i t E~

AVR with no PSS

I

\
\
\

I
I

;i /

'I\

i,/

'i

:i

1;

/;

I!

i
I
1

Prefault

I:;
1:
1:

/'

\',

ilk

'.

,--.

,'',

: ,',
:
;
%'

st

:
:

'\

:
:
!,

,AVR and PSS

,*--\

\
\

:;

',
, ,>'

i
,,

/'

%-.'

?.', :i
, ,

:I
I

'..'

:I

\i !I

$,, /;

'!

t,=O.l s

L'

i'

'4

Time t in seconds
Figure E13.8 Rotor angle response with fault cleared in 0.1 s

Example 13.3
For the system considered in Example 13.2, examine the accuracy and numerical
stability of Gill's version of the fourth order R-K method and the trapezoidal rule.
Consider the case with AVR and PSS, and the fault cleared in 0.07 s.

Transient Stability

Chap. 1

Solution

The system considered here includes the effects of rotor circuit dynamics and a fast
exciter with a PSS. This represents a very stiff system as is evident from the
eigenvalues computed in Example 12.5 of Chapter 12 (see Table E12.2).

R-K method:
Figure E13.9 shows plots of rotor angle transient responses computed by using four
different values of time step (At): 0.0 1 s, 0.03 s, 0.04 s, and 0.043 s.
With time steps of 0.01 s and 0.03 s, the results are practically identical. When the
time step is increased to 0.04 s, errors become noticeable. With a step size of 0.043 s,
the solution is numerically unstable.
In Table E13.1, the accuracy of results obtained with the R-K m@hod is checked by
applying Richardson's formula (see Equation 13.18) to the results obtained with step
sizes of 0.01 s and 0.02 s. The magnitude of the error is seen never to exceed 0.050,
The table indicates the manner in which the error propagates. It is evident that the
accumulated error oscillates, becoming alternately positive and negative.

01
0

2
3
Time t in seconds

I
5

Figure E13.9 Effect of integration step size on the results of R-K method

13.3

Simulation of Power System Dynamic Response

Table E13.1 Accuracy of results obtained with R-K method

Accumulated error

1
15

= -[6(At = 0.01 ) - 6 (At = 0.02)]

Trapezoidal rule:
The results obtained with step sizes of 0.01 s, 0.059 s, 0.09 s, and 0.093 s are shown
in Figure E13.10.
With step sizes of 0.01 s and 0.059 s, the results are practically identical. When the
step size is increased to 0.09 s, significant errors result; however, the general
characteristics of the overall response are still retained. With a step size of 0.093 s,
the errors become so large that the solution blows up after about 1.7 s.

Transient Stability

Chap. 1

Time t in seconds
Figure E13.10 Effect of integration step size on the results of trapezoidal rule
I

13.4 ANALYSIS OF UNBALANCED FAULTS

So far, we have considered only balanced operation of the power system; this
allowed for analysis of system performance by solving for the variables in only one
of the phases in a manner similar to single-phase circuits. The solution of unbalanced
three-phase circuits does not permit this simplification. A three-phase representation
can be used instead, but it complicates the problem enormously. The method of
symmetrical components is preferred because it offers considerable simplicity in the
analysis of power system stability problems involving unbalanced faults.
13.4. I Introduction to Symmetrical Components [ I 5-181

The general concept of symmetrical components was first developed by C.L.

Fortescue in 1918 [19]. He showed that an unbalanced system of n related phasors can
be resolved into n systems of balanced phasors called symmetrical components of the
original phasors.
In a three-phase system, a set of three unbalanced phasors can be resolved into
three balanced systems of phasors as follows:
(a)

Positive-sequence components consisting of a balanced system of three phasors

having the same phase sequence as the original phasors.

set. 13.4

873

Analysis of Unbalanced Faults

(b)

Negative-sequence components consisting of a balanced system of three

phasors having a phase sequence opposite to that of the original phasors.

(C)

Zero-sequence components consisting of three phasors equal in magnitude and

phase.

Figure 13.13 shows the three sets of balanced phasors that form the symmetrical
components of three unbalanced phasors.

(a) Positive-sequence
components

(b) Negative-sequence
components

(denoted by
subscript 1)

(denoted by
subscript 2)

(c) Zero-sequence
components

(denoted by
subscript 0)

Figure 13.13 Symmetrical components of three balanced phasors

Definition of operator a
The operator a denotes 120" phase shift. It is equal to the unit vector ej120 .
Recalling that eje=cosO+isin@,the unit vector representing a is given by

ej1200= cos120 +jsin120

=

-0.5 +j0.866

The operators a2 and a3 represent phase shifts of 240" and 360, respectively. Thus,
a 2 = e~2400=

and

-0.5-j0.866

874

Transient Stability

Chap.

Relationship between the original unbalanced phasors

and their symmetrical components
The total voltage of any phase is equal to the sum of the COITeSponding
components of the different sequences in that phase. Thus,

In matrix form,

Figure 13.14 depicts graphically the synthesis of a set of three unbalanced phasors in
accord with the above equations.

Figure 13.14 Unbalanced set of phasors reconstituted

from sequence components

Set, 13.4

Analysis of Unbalanced ,Faults

The inverse transformation is given by

similar transformations apply to currents. The sum of the three line currents is equal
to the neutral current. Thus,

Hence,

Egect of transformation on impedances

The impedances of system elements satisfy the equation

where Zabc is the impedance matrix giving self- and mutual impedances in and
between phases. ,
By using the transformation of Equation 13.70, Equation 13.74 can be
expressed in terms of the sequence components of voltages and currents:

For typical power systems Zabcis not diagonal but does possess certain symmetries.
These symmetries are such that the sequence impedance matrix ZOl2is diagonal,
either exactly or approximately. Sequence impedances of different types of system
elements will be discussed in Section 13.4.2.

Transient Stability

Chap.

Expression for power in terms of sequence components

Complex three-phase power is given by

In the above, the currents are line values and the voltages are line-to-neutral values.
It is interesting to note that there are no "cross" terms or components caused by
interaction between voltages of one sequence and currents of another sequence. The
factor 3 is reasonable when we think in terms of 9 components. This also means that
the transformation is not power invariant.

Why symmetrical components?

The resolution of three unbalanced phasors into symmetrical components
results in three balanced systems. In symmetrical power systems (systems in which
the three phases are identical), currents and voltages of different sequences do not
react to each other; that is, currents of one sequence induce voltage drops of like
sequence only.
Most of the apparatus used in practice is of the symmetrical type. This results
in considerable simplification of many types of unbalanced problems, such as those'
introduced by unbalanced faults or the opening of one phase.
Since in a symmetrical system there is no interaction between quantities of
different sequences, currents of any one sequence may be considered to flow in an
independent network associated with that sequence only. The single-phase equivalent
circuit composed of the impedances to currents of any one sequence is called the
sequence network for that particular sequence. To analyze the system performance
where there is an unbalanced fault, the sequence networks are interconnected at the
fault location to represent the interaction between quantities of different sequences due
to unbalance created by the fault.
The resolution of the problem into the sequence components has a further
advantage in that it isolates the quantities into components which represent a better

set, 13.4

877

Analysis of Unbalanced Faults

criterion of the controlling factors in certain phenomena. For example, in stability

investigations the synchronizing force between machines is affected principally by the
Positive-sequence quantities, the heating effects due to unbalanced currents depend on
negative-sequence currents, and the ground relays and grounding phenomena are very
closely associated with zero-sequence components. Thus, the analysis of many
problems is not only simpler, but the results are resolved into factors which may be
investigated separately.

Sequence impedances
The impedance of a circuit when positive-sequence currents alone are flowing
is called the impedance to positive-sequence currents or positive-sequence impedance.
similarly, impedance to negative-sequence currents is called negative-sequence
impedance, and the impedance to zero-sequence currents the zero-sequence
impedance.
A knowledge of the sequence impedances of the power system elements is
essential for the analysis of system performance during unbalanced faults.

13.4.2Sequence Impedances of Synchronous Machines [ I 51

Positive-sequence impedance
In Section 13.3.2, we discussed the detailed representation of synchronous
machines in the positive-sequence network appropriate for stability studies. Such a
model is accurate for the entire study period, including the subtransient, transient, and
steady-state conditions.
Figure 13.15 shows the simplified positive-sequence equivalent circuit of a
synchronous machine. As discussed in Chapter 5 (Section 5.3.3), the value of the
positive-sequence reactance Xl depends on the time frame of interest: subtransient,
transient, and steady state.

1
-

Reference bus

Figure 13.15 Positive-sequence equivalent circuit

of synchronous machine

Transient Stability

Chap. 1

For the time frame representing

the initial subtransient condition, XI =Xi;
the initial transient condition, Xl =XA;
the steady-state condition, X1=Xd.
The above assumes that saliency effects are negligible.
The positive-sequence resistance R1 is the ac resistance of the armature.
The equivalent circuit of Figure 13.15 represents one phase of the balanced
positive-sequence circuits. Its reference bus is the neutral of the genqtor.
Negative-sequence impedance

Negative-sequence reactance ( X j :
When fundamental frequency negative-sequence (reverse phase sequence)
currents are applied to the armature windings of a machine, with the rotor running at
synchronous speed and the field winding shorted through the exciter, the ratio of the
resulting negative-sequence fundamental frequency voltages to the currents gives the
negative-sequence impedance. The reactive part of this impedance is the negativesequence reactance X2.
In what follows, we will show that the negative-sequence reactance X, is equal
to the mean between the d- and q-axis subtransient reactances, Xz and X:. Saliency
effects result in somewhat different values of X, depending on whether sinusoidal
currents are circulated or sinusoidal voltages are applied.
With time t in seconds, the negative-sequence currents may be expressed as

Transforming into d, q, 0 components (see Chapter 3, Section 3.3), we have

set. 13.4

879

Analysis of Unbalanced Faults

The resulting field due to the negative-sequence stator currents rotates at synchronous
speed opposite to the direction of the rotor. Therefore, as viewed from the rotor, the
$ator currents appear to be double frequency currents id and i,. Currents of twice the
frequency are induced in the rotor circuits, keeping the flux linkages of the rotor
almost constant (since resistance is considerably smaller than inductance). The
flux due to the stator currents is forced into paths of low permeance which do not link
with any rotor circuit. These flux paths are the same as those encountered in
evaluating subtransient inductances. Therefore, the d- and q-axis components of stator
at 120 Hz encounter subtransient inductances L i and L: (see Chapter 4,
section 4.2). The corresponding per unit stator flux linkages are

Neglecting the stator resistance, with time t in seconds and a, equal to 1.O pu, the per
unit stator voltages are given by the following equations:

.
For fundamental frequency negative-sequence currents, w =ao.Hence,

e,

( 2 X l - X:) ,Z sin(20, t ++)

Similarly,

and

The corresponding phase voltages (negative sequence) may be determined by applying

the inverse Park's transformation (see Chapter 3 , \Section 3 -3):
,

Transient Stability

Chap.

Denoting the negative-sequence stator voltage by ea2 and noting that

synchronous speed, we have

,-

Fundamental frequency

Third harmonic

Similarly,

We see that the voltages have a third harmonic negative-sequence component which
is a function of subtransient saliency (X:-X;).This component is small; unless the
system resonates near this harmonic, it may be neglected. As far as the fundamental
frequency negative-sequence currents are concerned, the reactance is equal to the
average of the subtransient reactances:

Since the rnmf wave due to negative-sequence stator currents moves at twice the
synchronous speed with respect to the rotor, it alternately meets permeances of the
two rotor axes, corresponding to Xi and X;.Hence the value of X2 lies between Xj
and X;. With sinusoidal negative-sequence currents applied, the negative-sequence
reactance is the arithmetic mean of Xi and X;.This would apply when the external
circuit has a large reactance in series.
If sinusoidal negative-sequence voltages are applied to the stator, the resulting
currents in the stator windings can be shown to have a fundamental frequency
component proportional to

a n d a third harmonic component

set. 13.4

88 1

Analysis of Unbalanced Faults

<").

--Proportional to 2(x,'.

If only the fundamental is considered, the effective

oegative-sequence reactance is

When X$=X;,
there is no third harmonic component. In this case, the voltages
and currents would have only fundamental frequency components, and X,=X$ =X;.

Negative-sequence resistance (Rd:

The power associated with the negative-sequence current may be expressed as
a resistance times the square of the current. This resistance is designated as the
negative-sequence resistance R,.
As far as the negative-sequence stator currents are concerned, the rotor circuits
appear as short-circuited windings. The situation is similar to that of an induction
machine. Therefore, the physical nature of the negative-sequence resistance is best
visualized by analyzing the performance of an induction motor. Figure 13.16 shows
the equivalent circuit of an induction machine with symmetry in both axes. For the
case with asymmetry in the two'axes, the circuit represents the average of the two
axes. With negative-sequence stator currents, slip s =2. Therefore, the equivalent
resistance representing power transfer from rotor to shaft is Rr(l -s)/s = -R,/2.
The total rotor copper loss is represented by Rr. Half of the rotor loss is
supplied from the stator and the other half by the shafr torque (represented by -RJ2)
through the rotor by conversion of mechanical power.

X,
Xr
R,

magnetizing reactance
= rotor leakage reactance
= stator resistance

Xz
Rr
s

stator leakage reactance

= rotor resistance
= slip
=

Figure 13.16 Induction motor equivalent circuit

882

Transient Stability

Chap,

The real part of the equivalent impedance seen from the stator terminals is the
negative-sequence resistance R2. Since X, is large, it may be neglected. Hence,

Figure 13.17 shows the equivalent circuit of a synchronous machine appropriate

for use in negative-sequence networks. Since the generated voltages in the machine are
of positive sequence only, the negative-sequence equivalent circuit contains no emf
but is represented by the impedance (R2+jX2)offered to the flow of neBive-sequent;
currents. The reference bus of the equivalent circuit is the neutral of the generator.

Reference bus
Figure 13.17 Negative-sequence equivalent circuit
of a synchronous machine

Zero-sequence impedance
The machine must be star-connected; otherwise, the term zero-sequence
impedance has no significance as no zero-sequence current can flow.
Zero-sequence reactance (Xo):
When zero-sequence currents are applied to the armature, the instantaneous
values of currents in the three phases are equal. With the armature windings infinitely
distributed so that each phase produces a sinusoidal spatial distribution of the rnmf,
the net field produced by the three phases with equal instantaneous currents is zero.
No flux is produced across the air-gap. The only reactance associated with any phase
winding is due to slot leakage flux and end-winding leakage flux. However, in an
actual machine, the winding distribution is not perfectly sinusoidal, and the flux due
to the three phases will have a small value. Therefore, the actual X,is slightly higher
than for the ideal case and depends on pitch and breadth factors of the winding.
Zero-sequence resistance (Rd .
The zero-sequence resistance of a synchronous machine is somewhat larger
than the positive-sequence resistance for reasons discussed below.

set, 13.4

Analysis of Unbalanced Faults

883

The spatial distribution of stator windings is usually not purely sinusoidal, but
of higher harmonics. The spatial third harmonic contribution to the mmf due
to stator currents i,, ib ic is

MMF

K[iacos30+i,cos3(0-1200)+i,cos3(0+1200)]

m e n the stator currents form a balanced set, such as the positive- or negativesequence, the net mmf due to spatial winding distribution harmonics is zero. However,
for zero-sequence currents, with ia=ib=ic=Imcosotand the rotor travelling at
synchronous speed (@=at),the mrnf due to third harmonic spatial distribution is

MMF

3
2

= -I,

[cos(38 +at)+ cos(38 - ot)]

Thus the mmf due to third harmonic spatial winding distribution, when zero-sequence
currents flow in the stator windings, consists of second and fourth harmonic
components. These induce currents i~ the rotor and cause rotor heating. Therefore, the
zero-sequence resistance (R,) is slightly higher than the positive-sequence resistance
(RI). The difference is usually not significant.
Figure 13.18(a) shows the flow of zero-sequence currents in the three phases
and the neutral of the synchronous machine. The corresponding (single-phase)
equivalent circuit is shown in Figure 13,18(b).

40

(a) Current paths

oc

Reference bus0

(b) Equivalent circuit

Figure 13.18 Zero-sequence current paths and equivalent

circuit of a synchronous machine

884

Transient Stability

Chap. 1

The neutral impedance 2, carries a current equal to the sum of the zerosequence currents in the three phases. This is reflected as 32, in the equivalent circuit
of Figure 13.18(b); which represents only one phase. The reference bus for the zerosequence equivalent circuit is the ground at the generator. If the neutral is isolated
from the ground, 2, =aand I,, =O.
13.4.3 Sequence lmpedances of Transmission Lines

Positive- and negative-sequence impedances

In a symmetrical three-phase static circuit, the positive- and negative-sequence
impedances are identical because the impedance of such a circuit is independent of
the phase order of the applied voltages. The equivalent circuits of transmission lines
developed in Chapter 6 are applicable to the positive-sequence as well as the negative.
sequence networks.
Zero-sequence impedance
When only zero-sequence currents flow in a transmission line, the currents in
the three phases are identical. The sum of these currents returns through the ground
and/or overhead ground wires. Since the three-phase currents are in phase, the
magnetic field produced is very different from that produced by positive- or negativesequence currents. It requires consideration of the current distribution in the earth. The
net effect is that the zero-sequence reactance of transmission lines tends to be 2 to 3
times the positive-sequence reactance [15- 171.
Zero-sequence currents flowing through ground result in voltage drops due to
the impedance of the ground. Evidently, the ground is not necessarily at the same
potential at all points. The reference bus of the zero-sequence network does not
represent the ground potential at any particular point but is the reference ground for
zero-sequence voltages throughout the system. heref fore, the impedances of the
ground and ground wires are included in zero-sequence impedances of the
transmission lines. This ensures that the zero-sequence voltages at the terminals are
correctly given when referred to the reference bus.
13.4.4 Sequence lmpedances of Transformers

Positive- and negative-sequence impedances

Transformers, like transmission lines, are non-rotative or static. With the
assumption that symmetry exists between different phases, the impedances to balanced
three-phase currents do not depend on the phase sequence. Therefore, the equivalent
circuits and the associated impedances derived in Chapter 6 for the two-winding and
three-winding transformers are equally valid for the positive-sequence and the
negative-sequence networks.

See. 13.4

Analysis of Unbalanced Faults

zero-sequence impedances

The impedance offered by a transformer to the flow of zero-sequence currents

depends on the winding connections. Figure 13.19 gives the zero-sequence equivalent
circuits for different winding connections applicable to two-winding and three-winding
transformers.
For zero-sequence currents to flow through windings on one side of the
transformer and into the connected lines, a return path must exist through which a
completed circuit is provided. Additionally, there must be a path for the corresponding
current in the coupled windings on the other side.
For illustration, let us consider the winding connections for case 1 in Figure
13.19. The Y-connected primary (P) windings with the neutral grounded provide a
*ath to the ground. The A-connected secondary (S) windings provide a path for the
zero-sequence currents, which circulate within the A. This serves to balance the
currents in the primary windings; however, the zero-sequence currents circulating
within the secondary windings cannot flow in the lines connected to them. The
impedance 2, represents the total per-phase leakage reactances and resistances of the
primary and secondary windings and is equal to the positive-sequence impedance. The
neutral of the Y-connected primary is grounded through an impedance 2,; therefore,
the effective zero-sequence impedance as seen from the primary side is Zo+3Zn.Since
a A-connected circuit provides no path for zero-sequence currents flowing in the line,
the impedance as seen from the secondary side is infinite.
If the windings on one side are Y-connected with neutral ungrounded (as in
cases 2 and 5), zero-sequence currents cannot flow in the windings on either side.
Viewed from either the primary side or the secondary side, the zero-sequence
impedance is infinite.
Cases 6 to 8 in Figure 13.19 consider three-winding transformers where the
effects of the leakage reactances and resistances of the three windings are represented
by a star-connected equivalent (see Chapter 6). Connection of this equivalent to the
primary (P), secondary (S), and tertiary (T) terminals, and the reference bus depends
on the winding connections as shown in the figure.

13.4.5 Simulation of Different Types of Faults

In this section, we consider the simulation of a power system that is essentially
symmetrical but is rendered unbalanced by a fault at a particular location on the
system. Such faults can be analyzed by appropriately interconnecting the positive-,
negative-, and zero-sequence networks at the fault point as described below.
Figure 13.20 identifies the fault currents and voltages at fault location F. The
currents flowing into the fault from the three phases a, b, and c are represented by I,,
Ib, and I,, respectively; V,, Vb and V, represent voltages to ground at F.

Transient Stability

Zero-seq uence
equivalent circuit

Winding connection

Case

Chap. 1

1.

0 Ref.

0----------

41

Ref.

lo
o

Figure 13.19 Zero-sequence equivalent of three-phase twowinding and three-winding transformer banks

Ref.

0 Ref.

Sec. 13.4

Analysis of Unbalanced Faults

Figure 13.20 Designation of fault currents and voltages

(a) Single line-to-ground fault

Let us assume that phase a is shorted to ground at the fault point F, as shown
in Figure 13.21(a). From the figure, we see that

va = 0

I,

Ic

In terms of the symmetrical components' (see Equations 13.69 and 13.71), we have

va2j I

Neg. seq. net.

-R/

Zero seq. net.

(a) Fault schematic

(b) Sequence network connection

Figure 13.21 Simulation of a single line-to-ground fault

888

Transient Stability

Chap, 1

The connection of the sequence networks to satisfy the above relationships is

in Figure 13.21(b). The three sequence networks are connected at the fault point (F)
and the reference bus (R) of each sequence network so as to form a series connectioa
(b) Line-to-line fault

From Figure 13.22(a),

In terms of symmetrical components,

The sequence network connection reflecting the above relationships is shown in Figure
13.22(b). Since the fault does not involve ground, the zero-sequence network is absent,

i'Ial

'''a2

xF

xF

Pos. seq. net.

R

(a) Fault schematic

Neg. seq. net.

R

(b) Sequence network conne~tion

Figure 13.22 Simulation of a line-to-line fault

(c) Double line-to-ground fault

From Figure 13.23(a),

In terms of symmetrical components,

set. 1 3.4

889

Analysis of Unbalanced Faults

(a) Fault schematic

(b) Sequence network connection

Figure 13.23 Simulation of a double line-to-ground fault

The corresponding sequence network connection is shown in Figure 13.23(b).

(d) Three-phase fault

To complete the picture with regard to simulation of different types of faults
in terms of sequence networks, let us examine a symmetrical three-phase fault as
shown in Figure 13.24(a).
From Figure 13.24,

vI Lq.
F

pos.

(a) Fault schematic

net.

(b) Sequence network connection

Figure 13.24 Simulation of a three-phase fault

Transient Stability

Chap. 13

In terms of symmetrical components,

There are no zero-sequence or negative-sequence emfs, and the voltages V, and Vo2
at the fault point F are each equal to zero. Therefore, zero- and negative-sequence
currents do not flow anywhere in the system. As expected, only the positive-sequence
network is involved. The fault is simulated by shorting the fault point F and the
reference bus R of the positive-sequence network, as shown in Figure 13.24(b).

Formation of sequence networks

We see from the above analyses that a fault on a power system can be
analyzed by appropriately interconnecting the sequence networks of the system. The
solution of the resulting network gives the symmetrical components of voltages and
currents throughout the system.
The positive-, negative-, and zero-sequence systems represent. symmetrical
systems. Therefore, each of these systems may be represented by an equivalent singlephase network. Each element of the power system is represented by a model with an
appropriate degree of precision.
In each sequence network, the sequence voltages are referred to the reference
(or zero potential) bus for the network. In the positive- or negative-sequence systems,
all neutral points are at the same potential. Therefore, the neutrals of each of these
systems can be connected to a common bus which is the reference or zero-potential
bus for its sequence network.
As stated earlier, the ground is not necessarily at the same potential at all
points in the system. Therefore, the reference bus for the zero-sequence network is not
representative of the ground at any particular point; instead, the reference bus serves
as a reference for zero-sequence voltages at all points of the system. The individual
equipment models are so constructed that the zero-sequence voltages to ground are
correctly represented when referred to the reference bus of the network.
r

Representation in stability studies

The positive-sequence network is represented in detail with equipment models
as described in Chapters 3 to 11.
The negative- and zero-sequence voltages and currents throughout the system
are usually not of interest in stability studies. Therefore, it is unnecessary to simulate
the complete negative- and zero-sequence networks in system stability simulations.
Their effects may be represented by equivalent impedances (2,and 2,) as viewed at
the fault point F. These impedances are combined appropriately depending on the type

set. 13.4

891

Analysis of Unbalanced Faults

of fault and inserted as the effective fault impedance Zef in the positive-sequence

as shown in Figure 13.25.

Fault type

ValI I

Pos. seq. network

R
Double line-to-ground

z2z0
2 2 +2 0

Three-phase

Figure 13.25 Representation of faults in system stability studies

This representation gives the correct voltages and currents in the positivesequence network.
There are no negative- or zero-sequence voltages generated in the system. The
negative- and zero-sequence currents that flow during unbalanced faults are driven by
the positive-sequence voltage sources. The positive-sequence power outputs computed
represent the actual power supplied electrically by the machines. This includes all
negative- and zero-sequence resistance losses supplied electrically.
Negative-sequen ce braking torque

As discussed in Section 13.4.2, only half of the negative-sequence rotor losses

in synchronous machines are supplied electrically. The other half are supplied
mechanically by the shaft power or torque. This represents the component of the total
mechanical torque that is converted to supply rotor losses. It acts to decelerate the
rotor and is called the negative-sequence braking torque. This effect could be
significant for machines close to a fault, particularly those with solid rotors or salient
pole rotors with high-resistance damper windings. The method of accounting for this
braking torque is described below.
Referring to Figure 13.16, with the magnetizing reactance X, neglected, the
negative-sequence braking torque is

892

Transient Stability

Chap. 1

where R2 is the negative-sequence resistance of the machine and Ra is its stato,

resistance.
The machine negative-sequence current Ia2can be determined from the negative.
sequence network. Depending on the network configuration, Ia2can be estimated as
fraction of the machine positive-sequence current [20].
The total accelerating torque acting on the rotor is

The swing equation (Equation 13.22) is appropriately modified to include T, .

Example 13.4
Figure E13.11 shows the system representation applicable to a 1,000 M V A , 20 kV,
60 Hz generating unit.

HI

~ ' 6

XI=Xo=O.
15

Lbl

X, =0.6,Xo=1.8
F

5I ~ I ~ I I I I L ~ .

2 bus
X,=Xo=O.l
2~ ~
/

CCT 2

5- X , =0.6,Xo=1.8
Figure E13.11

~ 1 . 0

The transmission data shown on the figure are in per unit on 1,000 MVA, 20 kV base.
Network resistances are assumed to be negligible.
The generator data in per unit on the rating of the unit are as follows:

The neutrals of the generator stator and the step-up transformer HT windings are
solidly grounded. The neutral of the star-connected winding of the transformer near
the infinite bus is ungrounded.
-(a)

A double line-to-ground (LLG) fault occurs on circuit 2 at point F as shown

in Figure E13.11.

13.4

893

Analysis of Unbalanced Faults

(i) Neglecting all resistances, find the value of the effective fault impedance
Z,/which, when inserted in the positive-sequence network, represents the
unbalanced fault.
(ii) If the initial generator output conditions are

compute the magnitudes of positive-, negative-, and zero-sequence

currents throughout the network immediately after the fault occurs,
neglecting the effect of generator resistance.
(iii) Compute the negative-sequence braking torque corresponding to the
above condition.
(b)

How would the above results change if a line-to-ground fault or a line-to-line

fault were to occur at point F?

Solution
For conditions immediately after the fault occurs, the effective positive-sequence
machine reactance is X" = X j=Xg.
The three-sequence networks of the system are shown in Figure E13.12. Since
X j =X," =X,, the negative- and positive-sequence reactances are equal for each of the
elements in the system.
(a) Double line-to-ground fault

(i) Each sequence network represents a balanced system. The unbalanced effect
introduced by the fault at location F is represented by interconnecting the sequence
networks as shown in Figure E 13.13.

XI1=0.25
A- mrm
Gen.

I$&,,
0.6

HI

0.15

r
QTJT-n

Trans.

H2

CCT 1

F
V

0.6

B-

.".. CCT 2

0.1

Trans.

0 2 B
I
--

Zero potential bus

(a) Positive-sequence network

Figure E13.12 (Continued on the next page)

Transient Stability

Infinite
bus

Zero potential bus

(b) Negative-sequence network

I
-

Zero potential bus

(c) Zero-sequence network

Figure E13.12 (Continued)

$+

5fl

r,',
Fx
Negative
sequence

Fx

Positive
sequence
Zero potential
bus

Zero potential
$
-lr
bus

rr.
Figure E13.13

--.

!i
Fx
Zero
sequence
Zero potential

9'bus

1 3.4

Analysis of Unbalanced Faults

895

The effective negative-sequence and zero-sequence impedances as measured at point

F are shown in Figure El 3.14.

(a) Effective negative-sequence impedance

(b) Effective zero-sequence impedance

Figure E13.14
The effective fault impedance representing the L-L-G fault is

(ii) Since prior to the fault P=Q=O,

The positive-sequence network with the effective fault impedance inserted at F

reduces to the form shown in Figure E13.15(a).

Chap. l 3

Transient Stability
Generator
~

j0.4

j0.4

,-.

Tv
V

=. O m
l OO

j0.0857

j0.2

Infinite bus

+;;;

=>

0 1 . 0 ~ 0 ~

--

Figure E13.15 (a)

Because of network symmetry the solution is considerably simplified. The positive,
sequence component of the fault current is

The total positive-sequence fault current splits into negative- and zero-sequence
components as shown in Figure E13.15(b).

Figure E13.15 (b)

Because of the network symmetry about the point F, one-half of the fault current of
each sequence component flows from the generator, and the other half from the
infinite bus. Owing to the generator step-up transformer winding connections, the
zero-sequence current is restricted to the transformer; no zero-sequence current flows
through the generator. Figure E13.16 shows current flows in different parts of the
sequence networks.

?-,en.

1.~~7:
Po;itive
sequence

Negative
sequence

sequence

3.5
-

Figure E13.16

13.4

Analysis of Unbalanced Faults

(iii) The negative-sequence braking torque' is given by

Tb2 = ( 4 - ~ ~ ) 1 : 2

(0.063- 0.003)0.75~= 0.034 pu

(b) Line-to-ground fault

(i) The effective fault impedance is

2,

Z2+Zo = j(0.2+0.15)

= j0.35

pu

(ii) Figure E l 3.17 shows current flows in different parts of the sequence networks.

Positive
sequence

~~$-=
1.82
x

F
Negative
sequence

0.9 111

0.91

F
I

1.82

Zero
sequence

Figure E13.17

(iii) The negative-sequence braking torque is

Transient Stability

Chap. 13

Line-to-line fault
(i) The effective fault impedance is
2,

2,

j0.2

(ii) Figure El 3.18 shows current flows in the sequence network.

Figure E13.18

(iii) The negative-sequence braking torque is

TM

(4- R,) I:

= (0.063 - 0.003) 1.252

13.4.6 Representation of Open-Conductor Conditions

open-conductor conditions may be caussd by broken conductors or a deliberate

single-phase switching operation. Such faults may involve the opening of one phase
or two phases of a three-phase circuit.

(a) One open conductor

Figure 13.26(a) shows a section of a three-phase system with phase a open
between points X and Y. Let va, vb, and v, be series voltage drops in phases between
X and Y in phases a, b, c, respectively; I , Iband I, are the line currents. From Figure
13.26(a),

Set. 13.4

Analysis of Unbalanced Faults

'a I

"a1

'a 1

X.
*Y
Pos. seq. net.
R
52

"a2

'a2

X*
.Y
Neg. seq. net.
R

"a0

(a) Schematic diagram

Zero seq. net.
R
(b) Connection of sequence networks

Figure 13.26 Simulation of one open-conductor condition

In terms of symmetrical components,

The connections of sequence networks so as to satisfy the above equations are shown
in Figure 13.26(b).
(b) Two open conductors

We now consider the condition with phases b and c open between points X
and Y as shown in Figure 13.27(a). The applicable equations are

In terms of symmetrical components,

Transient Stability

Chap,

(a) Schematic diagram

'a2

X.
.Y
Neg. seq. net.
R

xi bY
Zero seq. net.
R

Val

"a0

__C

Ial.

'a2

X.
.Y
Pos. seq. net.
R

(b) Connection of sequence networks

Figure 13.27 Simulation of two-open-conductors condition

The sequence network connections that satisfy the above equations are shown in
Figure 13.27(b).
From Figure 13.26 we see that, for the one-phase open-circuited condition, the
three-sequence networks are connected in parallel at points X and Y; therefore, even
if the zero-sequence circuit as seen from X and Y has infinite impedance, some power
is transferred through the line between X and Y.
For the case with two phases open-circuited as in Figure 13.27, the threesequence networks are connected in series; unless the zero-sequence circuit provides
a path having finite impedance, no power can be transferred through the line.
Example 13.5
The system shown in Figure El 3.1 1 (Example 13.4) is subjected to a single line-toground fault at point F on circuit 2. The fault is cleared by single-phase switching
which disconnects the faulted conductor at both ends simultaneously. Show how the
network may be represented in stability studies during the period when one conductor
of circuit 2 is switched out.

13.4

Analysis sf Unbalanced Faults

Solution

This situation differs slightly from that

depicted in Figure 13.26. As shown in
Figure E13.19, the impedances of
conductors of phases b and c need to be
accounted for. This is achieved by including
the impedances associated with circuit 2 in
the sequence networks.

X
I

i-i

I,

0.15

H*

"a

Figure E13.20 shows the sequence networks

of the system of Figure E l 3.1 1 -,and their
interconnection to represent the condition
with one phase of circuit 2 open at both
ends.

Y
I

Figure E13.19

0.1

CCT 2

0.6

w
77

"t

Positive-sequence network
'a2

0.25
Gen.

CCT 1

0.6
0

0.15
Trans.

0.1
Trans.

CCT 2

0.6
CCT I

Negative-sequence network

1.8
0.15

r'Er'-%z

CCT 2

1.8
CCT 1

Zero-sequence network

Figure E13.20

-3-

0.1
Trans.

/
/

902

Transient Stability

Chap. 13

(a) Reduction of negative-sequence network

(b) Reduction of zero-sequence network

(c) Total effective series impedance

Figure E13.21 Reduction of negative- and zero-sequence networks

As shown in Figure E13.21, the effective negative- and zero-sequence impedances as
viewed from points X and Y are

and the net effective impedance that has to be inserted between points X and Y of the
positive sequence network is

Thus, from the viewpoint of power transfer between the generator and the infinite bus,
the effect of opening one phase of circuit 2 is the same as increasing the reactance of
I
the circuit by 0.7024 pu.

Sec. 13.5

Performance of Protective Relaying

13.5 PERFORMANCE OF PROTECTIVE RELAYING

Protective relays detect the existence of abnormal system conditions by
monitoring appropriate system quantities, determine which circuit breakers should be
opened, and energize trip circuits of those breakers. In order to perform their functions
satisfactorily, relays should satisfy three basic requirements: selectivity, speed, and
reliability.
Since transient stability is concerned with the ability of the power system to
maintain synchronism when subjected to a severe disturbance, satisfactory
performance of certain protection systems is of paramount importance in ensuring
system stability. Protective relays must be able to distinguish among fault conditions,
stable power swings and out-of-step conditions. While the relays should initiate
circuit-breaker operations to clear faulted elements, it is important to ensure that there
are no further relaying operations that cause unnecessary opening of unfaulted
elements during stable power swings. Tripping of unfaulted elements would weaken
the system further and could lead to system instability.
On the other hand, if the system is unstable we need to know how it separates
and whether the point of separation is acceptable. Special relaying may be required
to cause separation as desired.
One of the important aspects of transient stability analysis, then, is the
evaluation of the performance of protective systems during the transient period,
particularly the performance of relaying used for protection of transmission lines and
generators.
13.5.1 Transmission Line Protection [21-231

There are a variety of line protection schemes and practices used by utilities
to meet particular system requirements. The following factors influence the choice of
the protection scheme:
Type of circuit: single line, parallel line, multiterminal, magnitude of fault
current infeeds, etc.
Function of line, its effect on service continuity, speed with which fault has to
be cleared
Coordination and matching requirements
Three basic types of relaying schemes are used for line protection: (a)
overcurrent relaying, (b) distance relaying, and (c) pilot relaying [22].

904

Transient Stability

Chap. 1

(a) Overcurrent relaying

Overcurrent relaying is the simplest and cheapest form of line protection. T ~ ,
basic forms of relays are used: instantaneous overcurrent relay and time overcurrent
relay. They are difficult to apply where coordination, selectivity, and speed are
important. Changes to their settings are usually required as the system configuration
changes [22].
Overcurrent relaying cannot discriminate between load and fault currents
therefore, when used for phase-fault protection, it is applicable only when th;
minimum fault current exceeds the full load current.
Overcurrent relaying is used principally on subtransmission systems and radial
distribution systems. Since faults on these systems usually do not affect system
stability, high-speed protection is not required.
Overcurrent relaying schemes, which utilize communication channels to
interchange intelligence between terminal stations, have been developed to overcome
the above limitations. The two most common schemes are pilot wire relaying and
phase-comparison relaying schemes.

(b) Distance relaying

A distance relay responds to a ratio of measured voltage to measured current.
The relay operates if the ratio, which represents the effective impedance of the
network, is less than the relay setting. Impedance is a measure of distance along the
transmission line and, therefore, the relay is known as a distance relay.
The impedance approach provides an excellent way of obtaining discrimination
and selectivity, by limiting relay operation to a certain range of the impedance. The
following are the different forms of distance relays:
L

Impedance relay
Reactance relay
Mho relay
Modified mho and impedance relays, and combinations thereof

The triggering characteristics of distance relays may be shown conveniently

on a complex impedance (R-X) plane [24] as illustrated in Figure 13.28.
The impedance characteristic is a circle on the R-X plane; the reactance
characteristic is a horizontal line on the R-Xplane; the rnho characteristic is a straight
line on the G-B (admittance) plane, and a circle through origin on the R-X
(impedance) plane. The relay operates when the impedance measured by the relay
falls within the characteristic.

Sec. 13.5

Performance of Protective Relaying

(a) Impedance

(b) Reactance

(c) Mho

(e) Offset mho

(f) Lens

Directional

(d) Impedance with

directional element

Figure 13.28 Distance relay characteristics displayed on a coordinate system

with resistance ( R ) as the abscissa and reactance (X) as the ordinate

The mho characteristic is inherently directional; i.e., it detects faults only in

one direction. The reactance and impedance characteristics detects faults in all four
quadrants. Therefore, some type of directional supervision is generally used with these
relays. Combinations of similar elements with different settings andor elements with
different characteristics are used to provide selkctivity and coordination.
Distance relaying is the most widely used form of protection for transmission
lines. It is easy to apply and coordinate, provides fast protection, and is less affected
by changes in system conditions.
Figure 13.29 illustrates the application of distance relaying. At each end of the
line, three separate sets of relays are arranged to provide three protective zones as
shown. The first and second protective zones provide primary protection for the
protected line, and the third zone provides remote backup for the adjacent line.
For example, the first-zone relays at A are set to protect about 80% of line AB
without any time delay. This ensures high-speed protection of the line section and
provides about a 20% margin of safety to avoid incorrect relay operation caused by
equipment errors for faults beyond the end of the line section.
The second-zone relays at A are set to reach beyond the end of the line, a
typical setting being about 120% of the impedance of the protected line AB. This
ensures protection of the remaining 20% of the line not protected by the first-zone
relays. Since the second-zone relays will overreach line AB and operate for faults on

906

Transient Stability

2ndzone
i
I

1St zone

---------------I

Chap. 1

Relay characteristic
at breaker A
Relay characteristic
bat breaker B

,---------d

Distance

Figure 13.29 Distance relay characteristic

a section of the adjacent line CD, they are set to trip after a time delay (typically 0.3
to 0.5 seconds) thereby providing time coordination with the first-zone relays at c
protecting the adjacent line CD.
The third-zone relays at A provide backup protection for the adjacent line CD.
They are set to reach beyond line CD, so that relay operation is assured for faults on
the line. Time coordination with primary protection of line CD is achieved by having
a time delay of about 2 seconds for the third-zone protection.
A protection scheme similar to that at A is provided at B, but set to reach in
the opposite direction. Thus a fault on line AB opens breakers at both ends of the
line.
The number of relays at each end of a line is determined by the requirement
that three-phase, phase-to-phase, phase-to-ground, and double phase-to-ground faults
be covered. Usually, two sets of relays are provided: one set for phase faults and the
other for ground faults.
The reactance type relays are used for short lines, and the mho type relays for
long lines. Lens type relays are used when needed to restrict the tripping area of mho
units for protection of long lines.
Straight-distance relaying provides adequate protection for many situations.
However, it is not satisfactory when simultaneous high-speed tripping of both ends
of the line is critical to maintain system stability. Pilot-relaying schemes, described
below, are better suited for such applications.
(c) Pilot-relaying schemes

Pilot-relaying schemes utilize communication channels (pilots) between the

terminals of the line that they protect. Relays located at each of the terminals
determine whether the fault is internal or external to the protected line, and this
information is transmitted between the line terminals through the communication
channel. For an internal fault, circuit breakers at all terminals of the protected line are
tripped; for an external fault the tripping is blocked.

Set. 13.5

Performance o f Protective Relaying

907

The communication medium for transmitting signals between terminals may be

Pilot wire (metallic wires), power-line carrier, microwave, or fibre optics.
The pilot relay systems can be categorized by the manner in which the
transmitted signal is used and by the fault detector principle:
The transmitted signal may be used for blocking, unblocking, or transfer trip
operations. In a blocking system, the transmitted signal blocks the tripping
function of local relays that are normally set to trip a breaker. In a transjkr trip
system, a signal is sent by relays at one end of the line to either cause or
permit the breakers at the other end to trip. An unblocking system is a
combination of the above two systems. In this system, the tripping fhnction of
the local relays is normally blocked; the signal transmitted from the remote
terminal unblocks them when tripping is required. In the event that the
communication channel is lost for any reason, local relays are permitted to trip,
thus protecting against a fault that may occur during this period.
The detection of the fault location (i.e., internal or external to the protected
line) may be based on either the directional comparison or the phase
comparison principle, depending on the type of sensing relay used.
In a scheme using the directional comparison principle, distance relays and
directional relays are used to detect internal and external faults.
In a scheme using the phase comparison principle, the fault location (i.e.,
whether the fault is internal or external) is detected by comparison of the
relative phase displacement of currents entering and leaving the protected line.
Several different pilot-relaying schemes using a combination of the above
categories are possible. We will describe here two such commonly used schemes. For
simplicity we will consider only two terminal lines.

Permissive overreaching scheme:

Figure 13.30 illustrates the principle of operation of this scheme.

Station A

Z,
station
L
B
~

Figure 13.30 Permissive overreaching relay

Transient Stability

Chap.

Each terminal station of the line has the following relays:

Underreaching zone 1 phase and ground directional distance relays coverin

2!
about 80% of the impedance of the protected line. Operation of these relays
results in tripping of local breakers. These relays operate instantaneously with
no communication required from the remote end.
Overreaching zone 2 phase and ground directional distance relays covering
about 120% of the impedance of the protected line.
Operation of these relays results in a permissive signal being sent to the
remote end. Provided a permissive trip signal is received from the remote end,
a zone 2 operation also results in the local breakers being tripped and a
transfer trip signal being sent to the remote end.
If the apparent impedance remains inside the zone 2 relay characteristic for a
fixed time (typically 0.4 seconds), local breakers will be tripped without
receiving a permissive trip signal. In this case no transfer trip signal is sent.
Typical zone 1 and zone 2 relay characteristics at each station are shown in Figure
13.31. If the apparent impedance measured by the relays at a station is within zone
1 characteristics, then breakers at that station are tripped instantaneously.
Station B

Figure 13.31 Relay characteristic at station A

If the apparent impedances measured by the relaysat both ends are within zone
2 characteristics, then breakers at both ends open at high speed.
If the apparent impedance is within zone 2 characteristics of the relays at only
one station, the breakers at that station are tripped with a time delay of 0.4 seconds,
provided the impedance remains within zone 2 for 0.4 seconds.
To illustrate how this scheme operates, consider the faults F,, F,, and Fj as
shown in Figure 13.32.

Set. 1 3.5

Performance of Protective Relaying

Figure 13.32 Fault locations F1, F2 and F3

For the fault at F, within the zone 1 characteristics of the relays at both ends,
breakers at the two ends are tripped without intentional time delay. For the fault at F2
within zone 1 characteristic of the relay at B and zone 2 characteristic (but outside
zone 1) of the relay at A, breakers at B trip instantaneously; the zone 2 relays at both
are picked up and they send a.permissive signal to each other. The zone 2 relay
at A, upon receiving permissive signal fiom B (channel time not more than 20 ms),
trips the breakers at A.
For the fault at F3 (outside of the line section), the zone 2 relay at A is picked
up. As the zone 2 relay at B is not picked up, no tripping occurs at B and no
permissive signal is sent to A. The zone 2 relay at A trips the breakers only if the
fault is not cleared by the protective relays of line section BC within 0.4 seconds.
Directional comparison scheme:
The principle of operation of a directional comparison scheme with a blockingtype pilot is illustrated in Figure 13.33. The pilot channel is used to block tripping for
faults external to the protected line.
Each terminal station of the line has three sets of relays:
Underreaching zone 1 phase and ground directional distance relays covering
about 80% of the impedance of the line. These relays operate instantaneously
to trip the local breakers.
Overreaching zone 2 phase and ground directional distance relays set to reach
beyond the remote terminal so as to cover about 120% of the line impedance.
This ensures that all internal faults are detected. These relays trip local
breakers after a delay of about 25 ms (pickup time of the coordinator timer),
if no blocking signal is received from the remote terminal. Zone 2 relays trip
local breakers if the fault is not cleared in about 0.4 s, irrespective of any
blocking signal received from the remote end.
Zone 3 reverse blocking directional distance relays set in a direction opposite
to the protected line. They detect external faults and send a blocking signal to
the remote end to prevent tripping.

Transient Stability

Station A

Station B

Chap,

station c

(a) Protective zones of relays at stations A and B, protecting

line AB, and at station C protecting line BC

(b) Relay characteristics at station B,

protecting line from B to A

Figure 13.33 Directional comparison relaying scheme

Let us examine the operation of this scheme for four different fault locations
F,, F2, FS and F, identified in Figure 13.33(a). For a fault at location F1 identified in
Figure 13.33, zone 1 relays at both stations A and B operate and trip breakers at the
two ends with no delay.
For a fault at F2, the zone 1 relays (2;')at station B trip the local breakers
instantaneously; the zone 2 relays at station A, having received no blocking signal
from the zone 3 relays (2;')at station B, trip the breakers at A after a short time delay
of about 25 ms associated with the coordinating timer. This time delay is the waiting
period in case a blocking signal is sent from the other end. The zone 2 relays (2;)
at station C also detect the fault at F2, but are prevented from tripping at high speed
by the blocking signal received from the reverse sensing zone 3 relays (2;")
at station

Set 13.5

Performance o f Protective Relaying

91 1

0. However, a local timed trip results, if the fault is not cleared and the zone 2 relay
at station C continues to detect the fault after about 0.4 s.
For a fault at Fj, the zone 2 relays (2;) at station A detect the fault but are
Pgvented from tripping by the blocking signal received from the zone 3 relays (2;')
at
B. Similarly, a fault at F4 is detected by the zone 2 relays (2;) at station B,
but is blocked by signals from the zone 3 relays (2;) at A.

13.5.2 Fault-Clearing Times

The removal of a faulted element requires a protective relay system to detect
that a fault has occurred and to initiate the opening of circuit breakers which will
isolate the faulted element from the system. The total fault-clearing time is, therefore,
made up of the relay time and breaker-interrupting time. The relay time is the time
from the initiation of the short-circuit current to the initiation of the trip signal to the
circuit breaker. The interrupting time is the time from the initiation of the trip signal
to the interruption of the current through the breaker. On high voltage (HV) and extrahigh voltage (EHV) transmission systems, the normal relay times range from 15 to
30 ms (1 to 2 cycles) and circuit-breaker interrupting times range from 30 to 70 ms
(2 to 4 cycles).
Local breaker failure backup protection is usually provided on each breaker on
critical high-voltage transmission systems. If a breaker fails to operate at a local
station, infeed to a fault through the stuck (failed) breaker will be interrupted by
sending trip signals to adjacent zone breakers and transferring trip signals to remote
end breakers.
An example of a normally cleared fault is given in Figure 13.34. Bus A has
2-cycle air-blast breakers, and bus B has 3-cycle oil breakers. The system nominal
frequency is 60 Hz and the communication medium is microwave. A fault within zone
1 characteristic of the relay at bus A and within zone 2 (but outside zone 1)
characteristic of the relay at bus B #isconsidered. With the typical relaying times
shown in Figure 13.34, the fault is cleared from bus A in 64 ms and from bus B in
104 ms.
An example of a stuck breaker fault cleared by backup protection is given in
Figure 13.35. A fault on line BC within zone 1 characteristic of the relay at bus B and
within zone 2 (but outside zone 1) characteristic of the relay at bus C is considered.
Breaker 4 at bus B fails to operate, and the fault is cleared from bus B by sending trip
signals to breaker 3 at bus B (local backup) and to breakers 1 and 2 at bus A (remote
backup). With the typical breaker-interrupting and relaying times shown in Figure
13.35, the fault is cleared from bus C in 104 ms, from bus B in 157 ms, and from bus
A in 180 ms. We see from this example that the effect of a breaker-failure is not only
to delay fault-clearing, but to isolate from the system additional elements adjacent to
the faulted element.

Transient Stability

Bus A

Bus B

Local (bus A)
breakers 1 and 2
Primary relay time
(fault detection)
Auxiliary relay(s) time
Communication time
Breaker trip module
Breaker-clearing time

Chap, 1

Remote (bus B)
breakers 3 and 4

25 ms
3 ms
-

3 ms
33 ms (2 cycles)

Total time

9 ms
17 ms (microwave)
3 ms
50 ms (3 cycles)

64 ms

Fault cleared from bus A in 64 ms

Fault cleared from bus B in 104 ms

Notes: (i)

For purposes of illustration, 2-cycle breakers have been assumed at A

and 3-cycle breakers at B.

(ii)

Communication time depends on channel medium used. With power line

carrier the time may be longer.

Figure 13.34 Typical fault-clearing times for a normally cleared fault

Performance of Protective Relaying

Sec. 13.5

Bus A

Bus B
Bus C

Stuck breaker

0
Breaker 4 assumed to be stuck
Breakers 1, 2, 3, 4, and 5 assumed to be 2-cycle air-blast breakers (33 ms)
Breakers 6 and 7 assumed to be 3-cycle oil breakers (50 ms)
Local breaker Remote breakers Local backup Remote backup
6 and 7
breaker 3 breakers 1 and 2
5
Primary relay time
(at bus B)

25 ms

25 ms

25 ms

25 ms

Auxiliary relay(s)
time

3 ms

9 rns

6 ms

12 ms

17 ms

17 ms

3 ms

3 ms

3 ms

3 ms

33 ms (2 cyc.)

5 0 m s (3 cyc.)

33 ms

33 ms

Communication
channel time
Breaker failure
timer setting
Breaker-tripping
module time
Breaker time

--

Total time

64 ms

104 ms

--

157 ms

180 ms

Fault cleared from bus C in 104 ms

Fault cleared from bus B in 157 ms
Fault cleared from bus A in 180 ms
Notes: (i)

(ii)

Breaker failure timer setting has been assumed to be 90 ms for the 2-cycle
breaker 4. This could vary from one application to another. For a 3-cycle oil
breaker, a typical value is 150 ms.
Communication channel time depends on channel medium used.

Figure 13.35 Tv~icalfault-clearing: times for a stuck breaker fault

Transient Stability

Chap.

13.5.3 Relaying Quantities during Swings

The performance of protective relaying schemes during swings (electromechanical oscillations) and out-of-step conditions may be illustrated by considering
the simple two-machine system of Figure 13.36. The machines are assumed to be
represented by voltages of constant magnitudes behind their transient impedances (2
A
and 2'). The effect of swings on relaying quantities is analyzed by considering the
voltage, current, and apparent impedance measured by a relay at bus C (terminal of
machine A).

C
~ e n - ~ o +1-

Line

D
I

0Gen-B

(a) Schematic diagram

(b) Equivalent circuit

Figure 13.36 Two-machine system

EA and EB are voltages behind the transient impedances of the two machines.
EB is assumed to be the reference phasor, and 6 represents the angle by which EA
leads EB.
The current I is given by

The voltage at bus C is

The apparent impedance measured by an impedance relay at C protecting line CD is

given by

Sec. 13.5

Performance of Protective Relaying

During a swing, the angle 6 changes. Figure 13.37 shows the locus of 2, as a function
of 6 on an R-X diagram, when EA=EB.

Electrical centre

'

j C Z c when 6=90

%?-&Lc~
_---

*--

Note:

-------

Zc when 8=60

Origin is assumed to be at C, where the relay is located.

Figure 13.37 Locus of Zc as a function of 6, with EA=EB

91 6

Transient Stability

Chap. 1

When the voltage magnitudes EA and EB are equal, the locus of Zc is seen to
be a straight line which is the perpendicular bisector of the total system impedance
between A and B, i.e., of the impedance ZP The angle formed by lines from A and
B to any point on the locus is equal to the corresponding angle 6.
When 6=0, the current I is zero and Zc is infinite. When 6=180, the voltage
at the electrical centre (middle of total system impedance) is zero. Therefore, the
relay at C in effect will see a three-phase fault at the electrical centre. The electrical
centre and impedance centre coincide in this case.
If EA is not equal to EB, the apparent impedance loci are circles with their
centres on extensions of the impedance line AB. When EA>EB,the electrical centre
will be above the impedance centre; when EA<EB,the electrical centre will be below
the impedance centre. Figure 13.38 illustrates the shape of the apparent impedance
loci for three different values of the ratio EA/EB.

Electrical centre (6 = 180")

Figure 13.38 Loci of Zc with different values of EA/EB

For generators connected to the main system through a weak transmission

system (high external impedance), the electrical centre may appear on the transmission
line. On the other hand, for the case where a generator is connected to the main
system through a strong transmission system, the electrical centre will be in the stepup transformer or possibly within the generator itself.
We should recognize that the electrical centres are not fixed points since the
effective machine reactance (assumed to be fixed at Xi in our analysis) and the
magnitudes of internal voltages vary during dynamic conditions.
The voltage at the electrical centre drops to zero as 6 increases to 180' and
then increases in magnitude as 6 increases further until it reaches 360'. When 6
reaches 180, the generator will have slipped a pole; when 6 reaches the initial value
where the swing started, one slip cycle will have been completed.

See. 13.5

Performance of Protective Relaying

91 7

After a system disturbance, the rates of change of the angle, voltage, and
current are quite variable. At first the rotor angle changes quickly, slowing down as
the peak angle is reached for a stable case; then the angle decreases and oscillates
with smaller changes until equilibrium is reached. If the system is unstable, the angle
increases gradually until it reaches 180, when a pole is slipped. Unless the system
is separated by protective systems, this is followed by repeated pole slips in rapid
The voltages and apparent impedances at points near the electrical centre
oscillate rapidly during pole slipping.
Actual loci of apparent impedances measured by distance relays are more
than for the idealized case we have considered; they depend on variations
in internal machine voltages, voltage regulator and speed governor actions, and
interactions between all the machines in the system as influenced by the
interconnecting network. Such loci can be readily determined for any given situation
by using a transient stability program. However, analysis of idealized cases involving
simple-system configurations is helpful in understanding the results obtained with
complex simulations.
Generalized impedance diagram
The concept of using the impedance diagrams for the analysis of distance relay
performance was initially developed by C.R. Mason and J.H. Neher [25,26]. They
considered the "swing line" on the R-X diagram for the case EA=EBand analyzed the
performance of different relays by plotting their characteristics on the same R-X
diagram. This was indeed a pioneering step in understanding the performance of
protective systems during power swings.
This concept was subsequently extended by other workers such as Edith Clarke
[27] by developing impedance loci for various values of the ratio EA/EB.They showed
that the impedance locus was a circle for each value of the ratio EA/EB,centred on the
system impedance line with radii and offset determined by the values of the voltage
ratio as shown in Figure 13.39. The specific case of EA/EB=I is the limiting case with
infinite radius and offset.
In reference 27, Edith Clarke also developed another family of curves
representing apparent impedance loci for constant angular separation (6). If the angle
between EA and EB is held constant while the voltage-magnitude ratio is varied, the
apparent impedance will trace a portion of a circle which passes through A and B, and
whose centre lies on the perpendicular bisector of the system-impedance line. These
characteristics are shown in Figure 13.40. Line AB is seen to be a portion of the circle
(with infinite radius) which represents O0 and 180" angular separation. It also
separates the right and left portions of each constant-6 circle into two characteristics
such that the separation angle for the parts differs by 180". For example, the 90"
circle on the right becomes the 270" circle on the left.
The two families of characteristics shown in Figures 13.39 and 13.40 can be
combined to form the generalized per-unit impedance diagram [24,27]. These
characteristics are helpful in visualizing and analyzing how the apparent impedance
varies in the general case when both the angular separation and the ratio of source
voltages are varied.

Transient Stability

Figure 13.39 Impedance loci for constant voltage ratio

1"

Figure 13.40 Loci of apparent impedance for constant angular separation

Sec. 13.5

919

Performance of Protective Relaying

13.5.4 Evaluation of Distance Relay Performance during Swings

It is clear from the loci of apparent impedance derived in Section 13.5.3 that
the impedance measured by a distance relay during electromechanical oscillations
be within the relay characteristics. The relay should, however, trip only for
unstable conditions.
The performance of the relays during power swing conditions can be
determined by calculating the impedances measured by the relays during the step-bystep stability simulations and comparing this with the operating characteristics of the
relays.
For line pq of Figure 13.41, the apparent impedance measured by a relay at p
is given by

For evaluating the performance of distance relays, the locus of apparent

impedance so computed is plotted on the R-X diagram along with the relay
characteristics as shown in Figure 13.42.

Figure 13.41

I
Figure 13.42

The performance of protective relaying, using a timed second zone and not
using communication between line terminals, does not depend on relaying at the
opposite end of the circuit. Tripping occurs if the impedance locus at either end
remains in the second-zone characteristic sufficiently long to allow timed operation
or if the first-zone characteristic is entered.
Performance of schemes using pilot relaying will depend on impedances
measured at both ends.

Transient Stability

Chap,

13.5.5 Prevention of Tripping during Transient Conditions

Requirements for prevention of tripping during swing

conditions generally fall
.into two categories:
(a)

Prevention of tripping during stable swings while allowing tripping fa

unstable transients.

(b)

Prevention of tripping during unstable transients and forcing of separation at

another point.

Prevention of tripping during stable transients

The "rnho" relays have smaller tripping characteristics than impedance and
reactance relays set to reach the same distance. They may still cover too wide a range
to avoid tripping of long lines during swings. If swings up to 120" are to be
permitted, the mho relay characteristic is satisfactory for lines whose impedance is
less than about half the total impedance of the two generators.
The angular range covered by the mho relay may be reduced by the use of two
"ohm" units having linear characteristics as shown in Figure 13.43. The ohm units are
similar to reactance units except that they respond to impedances at angles of +I600
and -30" instead of +90. These angles are chosen to make the characteristics of the
ohm units parallel to the boundaries of fault area.
The mho unit controls the reach, and the ohm units control the angular range
of tripping on swings. The ohm units are referred to as "blinders." If the line flow is
only in one direction, it is sufficient to use only one blinder (02)Instead of using
blinders, relays with shaped characteristics such as the "lens'? characteristic shown in
Figure 13.28 or the "peanut" characteristic shown in Figure 13.44 may be used for
line protection. Such characteristics can be readily realized by using solid-state relays.

' ~ a u l tarea

Tripping can occur

only for impedance
between Ol and 0 2 ,
and within M
Trip coil

Figure 13.43 Reduction of mho relay angular range

Sec. 13.5

Performance of Protective Relaying

Figure 13.44

Out-of-step blocking and tripping relays

In some cases, it may be desirable to prevent tripping of lines at the natural

separation point and choose the separation point so that (a) load and generation are
better balanced on both sides, or (b) a critical load is protected, or (c) the separation
is at a corporate boundary.
In certain instances, it may be desirable to trip faster in order to prevent the
voltage from declining too far. Out-of-step relaying, either blocking or tripping, is
used for this purpose.
'.

Principle of out-of-step relaying:

The movement of the apparent impedance under out-of-step conditions is slow
compared to its movement when a line fault occurs. Therefore, a transient swing
condition can be detected by using two relays having vertical or circular
characteristics on an R-X plane as shown in Figure 13.45. If the time required to cross
the two characteristics (00S2, 00S1) exceeds a specified value, the out-of-step
function is initiated.
In an out-of-step tripping scheme, local breakers would be tripped. Such a
scheme could be used to
e

Speed up tripping to limit voltage decline

Ensure that a selected line is tripped instead of other circuits that may be more
critical

The out-of-step tripping relays should not operate for stable swings. They must
detect all unstable swings and must be set so that normal load conditions are not
picked up.

922

Transient Stability

(a) Vertical characteristics

Chap. I

(b) Circular characteristics

Figure 13.45 Out-of-step relaying schemes

In an out-of-step blocking scheme, zone 2 and perhaps zone 1 relays are

prevented from initiating tripping of the line monitored, and transfer trip signals are
sent to open circuits at a remote location in order to cause system separation at a more
preferable location.
The out-of-step blocking relays must detect the condition before the line
protection operates. To ensure that line relaying is not blocked for fault conditions,
the setting of the relays must be such that normal load conditions are not in the
blocking area.
13.5.6 Automatic Line Reclosing

The majority (60 to 80%) of transmission line faults are of a transitory nature.
An example of such a fault is an insulator flashover due to high transient voltages
induced by lightning. After the line is de-energized long enough for the fault source
to pass and the fault arc to de-ionize, the line may be reconnected. Therefore,
common practice is to reclose the circuit breakers automatically to improve service
continuity.
Reclosing may be either single-shot (one attempt) or multishot (several
attempts) with time delay between each attempt. If the fault persists after two or three
attempts, the operator may attempt to reclose manually after some delay. If this fails,
it is indicative of a permanent fault and the line is taken out of service for repair. The
first attempt to reclose may be either high-speed or with time delay.
High-speed reclosure refers to the closing of circuit breakers after a time just
long enough to permit fault-arc de-ionization. The reclosure can be completed in less
than 1 second. The following are the benefits of high-speed reclosure:
e

Reduction of the likelihood of system difficulties resulting from several

outages in rapid succession, usually caused by severe lightning, wind, or icing,
by speeding up restoration of the system to the prior level of security.

Sec. 13.5

Performance of Protective Relaying

923

Minimization of the effect of line outage on critical load areas and restoration
of normal voltages.
However, high-speed reclosure may not be acceptable in all cases. Reclosure into a
permanent fault may cause system instability. Many utilities operate on the premise
that all circuit faults are permanent and therefore the system must successfully
withstand reclosing into a fault. In such cases, high-speed reclosure does not
contribute to increased flexibility of system operation.
In addition, high-speed reclosure of lines close to steam turbine generating
units may increase shaft fatigue. This is discussed further in Chapter 15.
In view of the above considerations, timed and supervised reclosure is often
used on the bulk transmission system. This ensures that specific conditions are
satisfied and a preferred sequence of reclosing of circuit breakers is followed. The
general procedure is as follows:
Potential is applied by closing one pre-selected breaker at one terminal after
a time delay of 5 to 10 seconds.
At terminals where out-of-synchronism conditions are a significant probability
(e.g., at points in the system with few lines), the other circuit breakers
associated with the line are automatically closed after the line has successfully
remained on potential with acceptable angles across open breakers (measured
by synchrocheck relays) for a period of at least 1 second.
Where the probability of an out-of-synchronism condition existing between
terminals is minimal, the circuit is automatically placed back on load by
closing all remaining breakers associated with the circuit after it has
successfully remained on potential for about 0.5 seconds.
13.5.7 Generator Out-of-Step Protection

For situations where the electrical centre is out in the transmission system, the
detection of an out-of-step condition and the isolation of unstable generator(s) are
accomplished by line protection.
However, for situations where the electrical centre is within the generator or
step-up transformer, a special relay must be provided at the generator. Such a situation
occurs when a generator pulls out of synchronism in a system with strong
transmission. A low excitation level on the generator (EA<EB)also tends to contribute
to such a condition.

Effect of generators operating in out-of-step condition

During an out-of-step condition, there are large cyclic variations in currents
and voltages of the affected machine with the frequency being a function of the rate

924

Transient Stability

Chap, 13

of slip of its poles. The high amplitude currents and off-nominal frequency operation
could result in winding stresses and pulsating torques that can excite potentially
damaging mechanical vibrations. There is also a risk of losing the auxiliaries of the
.
affected unit as well as the auxiliaries of nearby stable units.
In order to avoid these adverse effects on the unit and the rest of the system
it is desirable to have an out-of-step relay that will trip the unit; the tripping should
be restricted to disconnection of the machine from the system by tripping the
generator breaker, rather than initiating a complete shutdown of the unit. This allows
resynchronization of the generator as soon as conditions stabilize.
.

Relays for out-of-step tripping of generators

The basic schemes available are similar to those used to detect out-of-step
conditions on the transmission system. There are, however, no industry standards or
commonly used practices with regard to generator out-of-step protection. Reference
28 provides the results of a survey of utility practices and a description of different
possible schemes. We will discuss two of the schemes here.

(a) Mho element scheme:

The principle on which this scheme is based is illustrated in Figure 13.46. A
mho relay monitors the apparent impedance (looking into the network) at the HT
terminal (H) of the unit transformer and is set to reach into the local generator.
The relay will immediately trip the generator when the apparent impedance
measured at the HT bus enters the offset mho characteristic. The setting objective
(achieved by choosing the size of the circle) is to allow tripping only for unstable
swings. Typically, the angle Zit at the point where the swing ,impedance enters the
relay characteristic is set to about 120, the maximum angular separation of the
machine from the system which may occur without loss of synchronism.
If the relay circle is too large, the protection may trip the generator for stable
swings. If the relay circle is too small (6c large), the scheme may not trip the
generator for unstable swings. In addition, if 6, is too large the tripping can occur
when the angular separation approaches 180"; this should be avoided since it subjects
the circuit breaker to the maximum recovery voltage during interruption.
To minimize the probability of false tripping for a loss-of-potential condition,
an overcurrent fault detector is sometimes used to supervise the mho characteristic.

(b) Blinder scheme:

This scheme consists of two impedance elements, referred to as blinders, and
a supervisory relay with an offset rnho characteristic. The characteristics are shown
in Figure 13.47.
There are many possible ways of applying blinders. In one scheme, the
generator is tripped when the apparent impedance locus is within the mho

Sec. 13.5

Performance of Protective Relaying

Infinite

B/
/

t$ 5 bus

2s

--C

zaPP
-

(a) System schematic

(b) System equivalent circuit

locus

.(

YA

(c) Relay characteristic and swing locus as measured at the HV bus

Figure 13.46 Generator out-of-step protection
using an mho element scheme
characteristic and crosses both blinders, the time for crossing being in excess of a set
value (typically 0.15 s).
The scheme offers more-selectivity than the simple mho element scheme. It is
also easy to set the scheme so that circuit breaker operation is effected at a favourable
swing angle. In view of the time delay involved in detecting the out-of-step condition,
it is easy to coordinate with the transmission line protection; this permits the reach to
extend into the system beyond the HT bus (H) of the step-up transformer, as depicted
in Figure 13.47.
In Figure 13.47, the relay is assumed to be located at the HT bus of the stepup transformer. It may just as conveniently be located at the generator terminals, in
which case the origin of the R-X diagram shifts to point L.

Transient Stability

Chap. 13

Blinder element

Figure 13.47 Generator out-of-step protection using a blinder scheme

Example 13.6
For the following unstable case simulated in Example 13.2, plot the apparent
impedance measured by a distance relay located at the HT bus:
-

Exciter on manual control (constant Eg); fault duratidn equal to 0.1 s.

Identify the location of the electrical centre and compute the effective machine
reactance.
Solution

Figure E13.22 shows the locus of apparent impedance computed during the timedomain simulation.
From the apparent impedance locus plot, we see that the electrical centre (C) lies
between the HT bus (H) and the infinite bus (B). The reactance between points H and
C, as measured on the plot, is equal to 0.055 pu. Therefore,

Sec. 13.5

Performance o f Protective Relaying

Solving, we get

This is equal to the average of X:=0.23 and XJ=0.25.

0.3 0.2 -

./'(prefault)
r t=1.0 s

Electrical
centre

0-1 - t=1.71 s

Equivalent circuit

t=1.1 s (fault cleared)

I .b R in Pu

0.0
X,=O. 15
-0.2 -0.3 -0.4 -

AXJ

Figure E 13.22 Apparent impedance locus

13.5.8 Loss-of-Excitation Protection

Loss of excitation of a synchronous machine can be caused by a wide variety

of abnormal field circuit conditions. This may range from a complete short-circuit of
the field winding to an effectively open-circuited field winding.
When a generator loses excitation, it operates as an induction generator,
running above synchronous speed. The excitation is now supplied from the power
system; hence, the machine draws reactive power from the system. This may cause
severe system voltage reductions.
Continued operation of the machine as an induction generator may cause
damage due to stator and rotor heating. The stator current may be as high as 2 to 3
times rated, depending on slip. The induced slip frequency rotor currents may cause

928

Transient Stability

Chap, 1

rotor overheating in steam-turbine generators. The time to reach dangerous

overheating is on the order of several minutes, depending on the slip.
Loss of generator excitation may cause widespread voltage reduction if the
power system to which the generator is connected is weak. This may, in turn, cause
loss of load and loss of synchronism of other generators in the system. To Prevent
this, quick-acting loss-of-excitation relaying should be provided to trip the main and
field breakers of the faulty generator.
The older loss-of-excitation protection consisted of undercurrent relays
connected in the field circuit. Such relaying could not distinguish between purposely
reduced excitation during light load conditions and accidental loss of excitation,
The relay equipment widely used now for loss-of-excitation protection is a
directional-distance relay connected to the generator terminals. It is based on the
principles developed by C.R. Mason in reference 29. In order to understand the basis
for this relay, we need to first understand the loss-of-excitation characteristics of the
generator.

Loss-of-excitation (LOE) characteristics [23,24,29]

The response of the machine following the loss of excitation depends on the
initial loading condition.
Figure 13.48 shows the general shapes of the paths traced by the apparent
impedance as measured at the generator terminal when the field circuit is shortcircuited for three different initial power output conditions. The apparent impedance
loci have been plotted on an R-X diagram on a per unit machine impedance base.
Each characteristic is terminated just before the rotor angle 6 is 180". Beyond this
point, the impedance changes become erratic due to pole slipping.

-2-o

Figure 13.48 Typical LOE characteristics

100% initial output

50% initial output
30% initial output

Sec. 13.5

929

Performance of Protective Relaying

When excitation is lost, the field flux linkage and the effective internal voltage
gadually decrease depending on the field circuit time constant. This results in a
gadual reduction in active power output, accompanied by an increase in rotor angle.
In addition, the machine absorbs increasing amounts of reactive power as the field
flux decays. Referring to the two-machine equivalent circuit of Figure 13.36, what we
have is a situation where the internal voltage EA is decreasing and the rotor angle 6
is increasing. The loss-of-excitation characteristic is, therefore, some combination of
the two families of curves shown in Figures 13.39 and 13.40. The actual path traced
depends on the relative rates of variation in EA and 6, which in turn depend on the
initial generator output, transmission strength, and field circuit time constant. The
overall effect on the apparent impedance at the machine terminals is as depicted in
Figure 13.48.
With a high initial output, the rotor angle advances more rapidly and the
endpoint shown in the figure is reached more quickly, typically within about 10 s. The
slip is also relatively high, and therefore the impedance locus terminates at a value
corresponding to the average of subtransient d- and q-axis reactances.
On the other hand, with a low initial output, the rotor angle advances slowly
and the slip at the endpoint is small. The impedance locus terminates at a value
corresponding to the average of d- and q-axis synchronous reactances.
The locus of the endpoints of the loss-of-excitation characteristic is shown in
Figure 13.49. It ranges between the subtransient reactances and the synchronous
reactances. Although these extreme values are not actually reached, they are helpful
in identifying the LOE relay characteristic. By encompassing the locus of the
endpoints as shown in Figure 13.49, the relay will operate before the generator first
starts to slip poles.

-R
Nearly equal to X:
Locus of endpoints
of LOE characteristics

ly equal to Xd
Relay characteristic
Figure 13.49 LOE relay and system characteristics

930

Transient Stability

Chap. 13

Loss-of-excitation relays
The characteristics of an LOE relay proposed by Mason in reference 29 are
shown in Figure 13.49. It is an offset mho relay. The offset is equal to one-half ofthe
transient reactance of the generator. The total reach of the relay is equal to the
synchronous reactance Xd.
The offsetting of the relay characteristic from the origin provides selectivity
against improper operation during power swings or out-of-step conditions. In
extreme case of zero external impedance, the electrical centre (intersection of the
swing locus with the system impedance locus) is at the centre of the generator
impedance Xj.Therefore, the LOE relay characteristic is offset by XJ12.
The offset mho relay with the above characteristic has been widely used and
has generally performed satisfactorily.
To provide greater selectivity against false operation during underexcited
operation, stable transient swings, or system disturbances causing underfrequency
operation, the use of two offset rnho relays with characteristics as shown in Figure
13.50 is proposed by J. Berdy in reference 30.
Reference 3 1 describes a more discriminating relay developed at Ontario
Hydro in 1978. This relay can prevent unit trip for recoverable excitation failures and
trip the unit more rapidly for unrecoverable failures. The relay uses generator field
and terminal voltages as inputs. It operates on the general principle that when the
terminal voltage is high the field voltage should be low and, conversely, when the
terminal voltage is low the field voltage should be high. For a detected excitation
problem, the relay will initially try to transfer voltage control to an alternate
AVRIexciter. If this does not rectify the problem, the relay trips the unit. Figure 13.51
shows a block diagram of the relay. The relay also includes selective overexcitation
protection as shown by the dashed blocks in the figure.

,,

Diameter 1.0 pu;

instantaneous
Diameter Xd;
time delay of 0.5 to 3.0 s

-xl
Figure 13.50 LOE protection with two mho relays

Sec. 13.5

Performance of Protective Relaying

I Overvoltage I---;
i I Timer /-----lANd1-,
I
I------

!--1

detector

Undervoltage - Timer
detector

voltage

,----__-------__-

Relay
driver

II

(Excitation
regulator
transfer)
1

---II Overvoltage !--------------I detector j

C----------------

I;

AC
generator
terminal
voltage

DC field

.--I

1
r
'

Undervoltage
detector

Relay
Timer driver

Dashed blocks show overexcitation protection.

-G
Timed trip
(Generator
trip)

Figure 13.51 A discriminating loss-of-excitation relay [3 11

Example 13.7
In this example, we examine the response of the system of Example 13.2 to loss of
generator excitation.
(a)

If the field of the equivalent generator representing the four units of the plant
is suddenly shorted (directly, with zero external resistance), determine the
locus of apparent impedance measured by an LOE relay located at the
generator terminals. Assume that the LOE relay has a characteristic similar to
that shown in Figure 13.49. Show the corresponding time responses of
generator terminal voltage, active and reactive power outputs, and rotor angle.

(b)

Determine the locus of apparent impedance if the field of only one of the four
generators in the plant is shorted (directly). The other generators operate
normally with excitation control.

Assume that in each of the above cases the initial system operating condition is the
same as in Example 13-2.

Solution
(a) The apparent impedance locus computed by using a transient stability program
is shown in Figure E13.23. The generator is represented in detail including
amortisseur circuits, and shorting of the field is simulated by setting the field voltage
E/, to zero at t=1.0 s.

Transient Stability

Chap. 13

generator rated values

Figure E13.23 Apparent per unit Z measured by an LOE relay

when field windings of all four units are shorted

The corresponding time response plots of Et, P, Q and 6 are shown in Figure E13.24.
These plots are helpful in explaining the shape of the apparent impedance locus, if we
note that
L

where

From Figure E13.23, we see that the apparent impedance locus enters the LOE
characteristic at t=3.4 s, i.e., 2.4 s after the field is shorted. As we see in Figure
E13.24(b), this corresponds to the time when the generator 'has slipped one complete
cycle. As the generator continues to slip poles, P, Q, and E, oscillate in rapid
succession.
(b) Figure E13.25 shows the apparent impedance locus when the field of one of the
generators is shorted. The other three generators continue to operate normally with
AVR and PSS controlling their excitation.

13.5

Performance o f Protective Relaying

Time in seconds
(a) Time response' of terminal voltage E,

Time in seconds
(c) Time response of active power P

Time in seconds
(b) Time response of rotor angle 6

Time in seconds
(d) Time response of reactive power Q

Figure E13.24 Time responses of generator variables

following shorting of field windings

Transient Stability

Chap. 13

I Base MVA and kV e q u a l 2

Igenerator rated values

Figure E13.25 Apparent per unit Z measured by an LOE relay when

the field winding of one of the generators is shorted

13.6 CASE STUDY OF TRANSIENT STABILITY

OF A LARGE SYSTEM
This case study is intended to demonstrate the analysis of a large practical
system for which a disturbance results in transient instability. The results show how
the unstable portion of the system is isolated from the rest of the system through the
actions of protective relaying. In addition, two possible methods of maintaining
stability are illustrated.

The system
The area of interest for the system under study is shown in Figure 13.52. It
consists of an 8 unit 7,000 MW nuclear plant connected to the rest of the system
through two 500 kV double circuit lines and three 230 kV double circuit lines. The
system model, which represents a portion of a large intercohnected power system,
,- consists of
o

e
o

2279 buses, 467 machines, and 6581 branches

All machines and associated controls modeled in detail
Loads modeled as voltage dependent (P=50%1+50%Z7 Q=100%2)

Sec. 13.6

Case Study of Transient Stability of a Large System

Transient Stability

936

Chap, 1

The contingency

A double line-to-ground (LLG) fault occurs on the 500 kV double-circuit line

comprising circuits 1 and 2 in Figure 13.52. The location of the fault is shown in the
figure as junction X, which represents the location nearest the plant where the two
circuits share the same tower. The switching sequence is as follows:
Time (ms)
0

Event
No disturbance

100

Apply LLG fault at junction X on circuits 1 and 2

164

Local end clearing:

Open breakers at bus 1 for circuit 1
Open breakers at bus 2 for circuit 2
This occurs 64 ms after the fault is applied, and this time
is computed as the sum of fault detection time (25 ms),
auxiliary relay time (6 ms), and breaker-clearing time
(33 ms = 2 cycles). At this time, the fault remains
connected on the ends of circuits 1 and 2 at junction X.
Remote end clearing:
Open breakers at bus 4 for circuit 2
Open breakers at bus 3 for circuit 1
Clear fault (the line is isolate,d)
This occurs 87 ms after the fault is applied, and the time
is calculated as the sum of fault detection time (25 ms),
auxiliary relay time (12 ms), communication time (17 ms;
microwave), and breaker-clearing time (33 ms = 2 cycles)

5000

Terminate simulation

Simulation

The 5-second simulation was conducted using explicit numerical integration

(fourth order R-K) and an integration time step of 0.01 second.
Figure 13.53 shows the absolute rotor angle plots for unit G3 and for a unit
located in the external system. Absolute angle is measured relative to the
synchronously rotating reference frame, and it is clear from plots that the external
system unit remains synchronized while the unit G3 loses synchronism and becomes
monotonically unstable. The responses of the other seven units at the nuclear plant is
similar to that of G3. As units G1 to G8 become unstable, the rest of the system

Sec. 13.6

Case Study of Transient Stability of a Large System

501

-50.
0.0

0.5

1.O

/Remote

1.5

generator

2.0

2.5

937

3.0

Time in seconds
Figure 13.53 Rotor angle time response

becomes generation deficient, and hence the absolute angles of all machines in the
system drift slightly. In this regard, using relative angles rather than absolute angles
is ofien a better choice since it permits one to easily observe the relative motion of
rotors between machines. The LLG fault results in the loss of one 500 kV doublecircuit line, leaving the entire 7,000 MW plant connected to the system through a
significantly weaker (i.e., higher impedance) path consisting of one double-circuit 500
kV line and three 230 kV double-circuit lines. The plant is unable to transfer full
output through this path, and the resulting accelerating power leads to instability.

How does the system come apart as a result of instability?

Figure 13.54 shows the characteristic for the out-of-step protection on unit G3.
The locus of the effective network impedance measured at the unit terminals is
superimposed on the circle. We see that the impedance locus does not enter the circle
except during the brief period when the fault is on; therefore, the protection will not
trip the units. The point of system separation is clearly beyond the reach of the out-ofstep protection.

Transient Stability

-0.1
-0.15

-0.1

-0.05

0.0

0.05

0.1

Chap, 1

0.15

Resistance R in pu
Figure 13.54 Unit G3 out-of-step protection

Figure 13.55 and Figure 13.56 show the line protection performance for relays
on circuit 3 at buses 1 and 7, respectively. The apparent impedances measured at each
end looking toward the opposite ends are shown on the plots. The relays are mho
distance relays and have zone 1 coverage of about 75% of line length and zone 2
overreach of about 125% of line length. These plots show that the apparent impedance
enters the zone 2 relays at bus 1 and enters the zone 1 and zone 2 relays at bus 7. The
zone 1 relay at bus 7 would trip circuit 3 at bus 7 and send a transfer trip signal to
the breakers at bus 1 which would then trip circuit 3 at bus 1. This logic also holds
true for the companion 500 kV circuit (#4) which would be tripped in an identical
manner. Following the loss of the 500 kV circuits (at approximately 0.8 s), the
remaining 230 kV circuits would become extremely overloaded and would be lost
through protection actions, thereby completely isolating the unstable plant from the
system.
From Figure 13.55, we can determine that the impedance swing crosses the
circuit at a point about 84% of the line length from bus 1. This point represents the
electrical centre following the disturbance and is theoretically where separation occurs
(although the entire circuit is tripped). Figure 13.57 shows, plots of bus voltage
measured at buses 1 and 7, and at the electrical centre. The voltage at the electrical
centre leads the other voltages in collapsing as instability occurs.

Case Study of Transient Stability of a Large System

13.6

Resistance R in pu
Figure 13.55 Line protection (circuit 3) at bus 1

Relay characteristic

Resistance R in pu
Figure 13.56 Line protection (circuit 3) at bus 7

Transient Stability

Chap. 13

----------

1.0 0.8

0.6

0.4

\
'

+
'

-.--_
--___

++
+$,

Bus 7
,/'

2%.
Electrical centre
\'.

-.
---_

Y,

\'\

\',

Y-,

+.

\'

0.2
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

\\
\ \'
\\

0.8

Time in seconds
Figure 13.57

Methods of maintaining stability

As discussed in Chapter 17, there exist many possible ways of improving
transient stability (including adding transmission facilities, fast-,valving, faster fault
clearing, etc.). Here, we consider the following two methods:
(a)

Reduction of the pre-contingency output of the plant. If one unit is taken off
line (with output redispatched to other units in the external system), the total
plant output will be reduced sufficiently to ensure stability following the LLG
fault. Curve 2 in Figure 13.58 shows the absolute rotor angle for unit G3 for
this case. It shows that the plant remains in synchronism. This solution to
instability is, however, generally costly because available energy is kept idle
or bottled in the plant.

(b)

Rejection of generation following the disturbance. If the output of the plant

can be rapidly reduced following the contingency, the accelerating power can
be reduced sufficiently to maintain synchronism. Curve 3 in Figure 13.58
shows the unit G3 rotor angle response with two units (G7 and G8) tripped at
173 ms (73 ms following fault application).

Sec. 13.7

Direct Method of Transient Stability Analysis

250

/t 8 units, no GR

V3

a,

2
gJ 200 -

i
:

(curve 1)

+-,

41 100-

\
\

Q)

\'

'

/'

'

0**

+-,

--__0'

50-

GR

-_

7-----

8 units, 2 with GR
(curve 3)

generation rejection

Time in seconds
Figure 13.58 Unit G3 rotor angle responses with
and without generation rejection

13.7 DIRECT METHOD OF TRANSIENT STABILITY ANALYSIS

The direct methods determine stability without explicitly solving the system
differential equations. This approach is academically appealing and has received
considerable attention since the early work of Magnusson [32] and Aylett [33] who
used transient energy for assessment of transient stability. The energy-based methods
are a special case of the more general Lyapunov's second method or the direct
method (see Chapter 12, Section 12.1.4), the energy function being one possible
Lyapunov function [34-371.
This section describes the basic concepts on which the direct methods are
based. A detailed treatment of the subject is beyond the scope of this book. Interested
readers may consult references 38 and 39, books dedicated entirely to the analysis of
power system transient stability by the direct methods.
13.7.1 Description of the Transient Energy Function Approach
Rolling ball analogy [40,4q

The transient energy approach can be described by considering a ball rolling

on the inner surface of a bowl as depicted in Figure 13.59. The area inside the bowl
represents the region of stability, and the area outside is the region of instability. The

942

Trans.ient Stability

Chap. 13

rim of the bowl is irregular in shape so that different points on the rim have different
heights.

SEP
w

Figure 13.59 A ball rolling on the inner surface of a bowl

Initially the ball is resting at the bottom of the bowl, and this state is referred
to as the stable equilibrium point (SEP). When some kinetic energy is injected into
the ball, causing it to move in a particular direction, the ball will roll up the inside
surface of the bowl along a path determined by the direction of initial motion. The
point where the ball will stop is governed by the amount of kinetic energy initially
injected. If the ball converts all its kinetic energy into potential energy before reaching
the rim, then it will roll back and eventually settle down at the stable equilibrium
point again. However, if the kinetic energy injected is high enough to cause the ball
to go over the rim, then the ball will enter the region of instability and will not return
to the stable equilibrium point. The surface inside the bowl represents the potential
energy surface, and the rim of the bowl represents the potential energy boundary
surface (PEBS).
Two quantities are required to determine if the ,ball will enter the instability
region: (a) the initial kinetic energy injected and (b) the height of the rim at the
crossing point. The location of the crossing point depends on the direction of the
initial motion.

Application to power systems

The basis for the application of the TEF method to the analysis of power
system stability is conceptually similar to that for a ball rolling in a bowl. Initially the
system is operating at a stable equilibrium point. If a fault occurs, the equilibrium is
disturbed and the synchronous machines accelerate. The power system gains kinetic
and potential energy during the fault-on period and the system moves away from the
SEP. After fault clearing, the kinetic energy is converted into potential energy in the

Sec. 13.7

Direct Method of Transient Stability Analysis

943

same manner as the ball rolling up the potential energy surface. To avoid instability,
the system must be capable of absorbing the kinetic energy at a time when the forces
on the generators tend to bring them toward new equilibrium positions. This depends
on the potential energy-absorbing capability of the postdisturbance system. For a
given postdisturbance network configuration, there is a maximum or critical amount
of transient energy that the system can absorb. Consequently, assessment of transient
stability requires
(a)

Functions that adequately describe the transient energy responsible for

separation of one or more synchronous machines from the rest of the system

(b)

An estimate of the critical energy required for the machines to lose

synchronism

For a two-machine system the critical energy is uniquely defined, and the TEF
analysis is equivalent to the equal-area criterion described in Section 13.1. This is
illustrated in Figure 13.60, which shows two plots, both having rotor angle (6) as the
ordinate [40]. The upper plot illustrates the equal-area criterion in which the critical
clearing angle (63 is established by equality of areas A l and A2. The lower plot
illustrates the transient energy method which can be used to specify the critical
clearing angle in terms of potential and kinetic energy. The kinetic energy gained
during the fault-on period is added to the potential energy at the corresponding rotor
angle, and the sum is compared to the critical potential energy to determine stability.
Given a disturbance, there is a stable equilibrium point for the postfault
system. A region of attraction can be defined for this postfault SEP as shown in
Figure 13.61. Any postfault system trajectory with the state of the system at fault
clearing (xcl) inside this region of attraction will eventually converge to the SEP, and
the system is said to be stable. On the other hand, if xcl lies outside the region of
attraction, the postfault system will not converge to the stable equilibrium point, and
the system is said to be unstable.
The state of the system at fault clearing (x,J can be described by the value of
the energy function evaluated at xcl, i.e., V(xcl). Hence the direct method solves the
stability problem by comparing V(xcl) to the critical energy V., The system is stable
if V(xcl) is less than V, and the quantity V,-V(xcl) is a good measure of system
relative stability. This quantity is defined as the transient energy margin.
The quantity V(xcl) measures the amount of transient energy injected into the
system by the fault while the critical energy measures the strength of the postfault
system. More precisely, the critical energy measures the energy-absorbing capability
of the postfault system.
Referring to Figure 13.61, if the rotor oscillates within the range 6,, to 6,,, the
system will remain transiently stable. If the rotor swings beyond this range, the system
will become unstable. Hence the two points 6,1 and 6,2 on the potential energy curve
form a boundary to all stable rotor angle trajectories. This boundary is called the
potential energy boundary surface (PEBS) and the points on this boundary are local
potential energy peaks.

Transient Stability

Infinite
bus

t
(a) System configuration

Power 1

(b) Power-angle relationships

Energy

Critical energy

PE(6,)

(c) Energy-angle relationships

Figure 13.60 Illustration of equivalence of transie~t

energy method and equal-area criterion

Chap. 13

Sec. 13.7

Direct Method o f Transient Stability Analysis

Potential
energy v
,

Figure 13.61 Region of stability and its local approximation

The boundary of the stability region is usually approximated locally by a

constant energy surface {x 1 V(x)= K ) as shown in Figure 13.61, where K represents the
critical energy V, of the postfault system.
13.7.2 Analysis of Practical Power Systems [40-491

The application of the direct method to practical power systems is difficult. A

number of simplifying assumptions are necessary. To date, the analysis has been
mostly limited to power system representation with the generators represented by the
classical model and loads modelled as constant impedances [42-451. Recently, there

946

Transient Stability

Chap. 13

have been several attempts to extend the method to include more detailed models
[46,47]. This, however, is still at a developmental stage and f~rtherresearch efforts
are required to improve the accuracy and reliability of the method with detailed
models. Here, we will limit our discussion to the simplified system models.
With the generator transient reactances and load admittances included in the
node admittance matrix, we may write

where YR is the reduced admittance matrix with all nodes other than the generator
internal nodes eliminated, EG is the generator internal source voltage vector, and rG
is the generator current vector.
Let the internal voltage of the ith generator in phasor notation be given by

and the ith row, jthcolumn element of YR by

For a system with n machines, the active output of the ith generator is given by

For the application of the TEF method, it is convenient to describe the transient
behaviour of the system with the generator rotor angles expressed with respect to the
inertial centre of all the generators. The position of the centre of inertia (COI) is
defined as

where HT is the sum of the inertia constants of all n generators in the system.
The motion of the COI is determined by

Sec. 13.7

Direct Method of Transient Stability Analysis

and

with

where

P,,

mechanical input power of the i" machine

D:.
= E.E.G~
I/
1 I
oo = synchronous speed in elec. radls

Amco,

per unit speed deviation of COI from synchronous speed

The motion of the generators with respect to the COI can be expressed by defining
=

6,-GcOI

rad

(13.93)

and

The equations of motion of the ithmachine in the COI reference frame are

where oi is the pu speed of the ith machine with respect to the COI.
The energy function V describing the total system transient energy for the
postdisturbance system is defined as

Transient Stability

Chap. 13

where
8:

= angle of bus i at the postdisturbance SEP

= 2Himo = per unit moment of inertia of the

ith generator

The transient energy fuction is composed of the following four terms:

(a) -1C J , ~ : : change in rotor kinetic energy of all generators in the COI reference
2
frame
(b)

ZPm~(0, - 0:) :

(c)

xx cv(cosOv- cos0;)

I .

change in rotor potential energy of all generators relative to COI

: change in stored magnetic energy of all branches

(d) x x ~ ~ c o s 0 ~+ej)
d (:0 change
~
in dissipated energy of all branches
The first term is called the kinetic energy (Vke)and is a function of only generator
speeds. The sum of terms 2, 3, and 4 is called the potential energy (Vpe) and is a
function of only generator angles.
The transient stability assessment involves the following steps:
1.

Calculation of the critical energy V,

2.

Calculation of the total system energy at the instant of fault-clearing Vcl

3.

Calculation of stability index: V,-V,.

index is positive.

The system is stable if the stability

Time-domain simulation is run up to the instant of fault clearing to obtain the angles
and speeds of all the generators. These are used to calculate the total system energy
(VcJ at fault clearing.
The calculation of V, i.e., the boundary of the region of stability, is the most
difficult step in applying the TEF method. Three different approaches are briefly
described below.

Sec. 13.7

Direct Method of Transient Stability Analysis

949

(a) The closest unstable equilibrium point (UEP) approach

Early papers on the application of the TEF method for transient stability
analysis used the following approach to determine the smallest V:,
Step 1: Determine all'-the UEPs. This is achieved by solving the postdisturbance
system steady-state equations with different initial values of bus angles.
Step 2: Calculate system potential energy at each of the UEPs obtained in step 1.
The critical energy V, is given by system energy at the UEP which results
in the minimum potential energy.
Referring to the rolling ball analogy, the closest UEP is analogous to the lowest point
on the rim of the bowl. The critical energy at the closest UEP is equal to the potential
energy of the ball when it is at the lowest point on the rim. This assumes that the
direction of initial motion is always toward the lowest point on the rim. Similarly, for
a power system this approach computes the critical energy by implicitly assuming the
worst fault location; hence, the results are very conservative.
(b) The controlling UEP approach [40, 481

The degree of conservatism introduced by the closest UEP approach is such

that the results are usually of little practical value. The controlling UEP approach
removes much of this conservativeness by computing the critical energy dependent
on the fault location. This approach is based on the observation that the system
trajectories for all critically stable cases get near to those UEPs that are closely related
to the boundary of system separation. The UEPs are called the controlling or relevant
UEPs.
The computation process in this approach consists of two steps: (1)
identification of the mode of disturbance and (2) calculation of the controlling UEP.
Mode of disturbance (MOD):
The mode of disturbance identifies the machines that are severely disturbed by
a given disturbance. These machines are most likely to lose synchronism with the rest
of the system if the disturbance is severe enough to cause system instability. One way
to determine the MOD is to perform time-domain simulations and identify which
machines go unstable. An automatic procedure for determining the MOD is described
in reference 49 based on the normalized V., This procedure requires a considerable
amount of CPU time when detailed generator models are used.
Controlling e7EP calculation:
The SEP or UEP is a solution of the power-flow equations (with i=l to n):

Transient Stability

Chap, 1

There are many solutions to the above equations, one of which is the SEP. All other
solutions are UEPs. The controlling UEP is calculated by solving the following
minimization problem:

min

F=C&

i : l }

An approximate UEP based on the MOD is used as a starting point for the solution,
Numerical problems are sometimes encountered when one is solving for the
controlling UEP based on the above approach. A robust procedure based on the
potential energy boundary surface crossing point suggested in reference 44 is
described below.
(c) The boundary of stability-region-based controlling UEP (BCU) method
Earlier CUEP methods face serious convergence problems when solving for
the controlling UEP, especially when the system is highly stressed or highly
unstressed, or the mode of system instability is complex. These problems usually arise
if the starting point for the CUEP solution is not sufficiently close to the exact CUEP.
Some of the convergence problems can be overcome by the BCU method which has
the capability of producing a much better starting point for the CUEP solution.
In reference 44, the boundary of the stability region is defined as the union of
the stable manifolds of the UEPs on the boundary. Any trajectory starting from a
point on the boundary will converge to one of the UEPs as time increases. By making
use of this property, the BCU method computes the controlling UEP of a power
system through computing the CUEP of its associated reduced system.
The dynamic behaviour of a multimachine power system represented by
classical generator models and constant impedance loads is described by Equations
13.95 and 13.96. The associated reduced system model for this multimachine power
system is defined as

The equilibrium points of the reduced system are the same as those of the original
system. The BCU method computes the CUEP as follows:

Sec. 13.7

Direct Method of Transient Stability Analysis

951

(i)

For a given system and a contingency, a sustained fault trajectory in time

domain is simulated until it reaches the boundary of stability region of the
reduced system (the PEBS). The system state at this time instant is referred to
as the exit point.

(ii)

Starting with the exit point, the reduced system is simulated in time domain
until the quantity C lgi 1 reaches its first local minimum. The rotor angles at
this point constitute the starting point for the CUEP solution.

(iii)

The controlling UEP is computed using any robust numerical technique. The
UEP so obtained is the controlling UEP relative to the fault-on trajectory.

The BCU method appears to be promising and can be easily extended to

accommodate more detailed models. More testing is required to verify its accuracy
and convergence characteristics. More importantly, the boundary of stability region
for power systems modelled in detail needs to be investigated.

(4 Sustained-fault approach

[42]

The calculation of the controlling UEP for large systems is very timeconsuming. Under severe system conditions, the solution process may either fail to
converge or converge to the wrong UEP. The sustained-fault approach was developed
to avoid the calculation of the controlling UEP.
The critical energy is determined as follows:

1.

A time domain simulation with a sustained fault is run until the system crosses
the potential energy boundary surface (PEBS). The crossing of the PEBS is
indicated by the potential energy reaching its maximum.

2.

The potential energy at the crossing of the PEBS is taken as the critical energy
for that particular fault location.

Example 13.8
In this example, we illustrate the use of the TEF method for transient stability analysis

by considering the simple system of Example 13.1.

For the single-machine infinite bus system of Figure E13.1,
(a)

Write the dynamic equations for the postdisturbance system corresponding to

Equations 13.95 and 13.96.
%\

(b)

Write the expression for the system energy function.

(c)

Calculate the postdisturbance system SEP, UEP, and the critical energy

V'.

Transient Stability
(d)

Chap. 13

Calculate the energy at fault clearing with tc=0.07 s, 0.086 s, and 0.087 s.
Determine the system stability for each of the three f3ult durations.

Solution
(a) The postdisturbance reduced equivalent circuit of the system is shown in Figure
E13.26.

Figure E13.26 Postdisturbance equivalent circuit

Because bus 2 represents an infinite bus, M, =

and w l = 8 1 / w o = A o , .

and A oCol
= 0 . Consequently, el = 6

Since GI2=0,D12=O and P A =P,, =0.9. The postdisturbance dynamic equations of the
system, corresponding to Equations 13.95 and 13.96, are

(b) From Equation 13.97, the system transient energy function is

(c) At the postdisturbance SEP, we have

Hence,

0;

0.9552 rad

54.73'

For this simple system there is only one postdisturbance UEP, which is given by

13.7

Direct Method of Transient Stability Analysis

0;

n-8;

2.1864 rad

953

125.27"

The critical energy, which equals the system potential energy at the postdisturbance
UEP, is

v,

-p;,(e; - e;)- c,,( C O S ~-; C O S ~ ; )

-0.9(2.1864 -0.9552) - 1.1024(-0.5775 -0.5775)

0.1651

(d) To compute the energy at fault clearing, we will use the generator angles
computed in Example 3.1.
(i) With tc=0.07 s, 0;=48.58"=0.8479 rad, and

The total energy at fault clearing is

Since Vcl<Vcr, the system

0.165 1 -0.1107 =0.0544.

is

stable; the

stability

margin

is

Vcr-Vc,=

(ii) With tc=0.86 s, 0;=52.04"=0.9083 rad, and

The total energy at fault clearing is

The system is stable; however, the stability margin is now reduced to

0.165 1 -0.1620 =0.0031 .
(iii) With tc=0.087 s, 8~=52.3"=0.9128rad, and wc=O.O1l186 pu. The total energy at
fault clearing is

Transient Stability

Chap. 13

Since VcpVcr,the system is unstable.

The above results are consistent with the time-domain simulation results of Example
13.1.
I
13.7.3 Limitations of the Direct Methods

In spite of the many significant accomplishments in recent years in the

application of the direct methods, modelling limitations and unreliabilit~ of
computation techniques continue to be major impediments to their widespread
practical use. The direct methods are vulnerable to numerical problems when one is
solving stressed systems. The use of sophisticated and robust CUEP solution
techniques usually imposes a heavy computational burden, making the method slower
than the time-domain simulation.
The best course appears to be the use of a hybrid approach in which the
transient energy calculation is incorporated into the conventional time-domain
simulation [50-541. This enhances the capability of time-domain simulations by
computing the stability margin and minimizing the effort required in determining the
stability limits.

REFERENCES
[I]

W.D. Stevenson, Elements of Power System Analysis, 3rd Edition, McGrawHill, 1975.

[2]

R.T. Byerly and E.W. Kimbark, Stability of Large Electric Power Systems,
IEEE Press, 1974.

[3]

E.R. Laithwaite and L.L. Freris, Electric Energy: Its Generation, Transmission
and Use, McGraw-Hill (UK), 1980.

[4]

A. Ralston and H.S. Wilf, Mathematical Methods for Digital Computers, John
Wiley & Sons, 1962.

[5]

B. Carnahan, H.A. Luther, and J.O. Wilkes, Applied Numerical Methods, John
Wiley & Sons, 1969.

[6]

C.W. Gear, Numerical Initial Value Problems in Ordinary ~ g e r e n t i a l

Equations, Prentice-Hall, 1971.

References

955

[7]

Kundur Prabhashankar, Digital Simulation and Analysis of Power System

Dynamic Performance, Ph.D. thesis, University of Toronto, 1967.

[a]

P. Kundur and P.L. Dandeno, "Implementation of Advanced Generator Models

into Power System Stability Programs," IEEE Trans., Vol. PAS- 102, pp. 20472052, July 1983.

[9]

H.W. Dommel, "Digital Computer Solution of Electromagnetic Transients in

Single and Multiphase Networks," IEEE Trans., Vol. PAS-88, pp. 388-399,
April 1969.

[lo]

B. Stott, "Power System Dynamic Response Calculations," Proc. IEEE, Vol.

67, pp. 219-241, February 1979.

11

P.L. Dandeno and P. Kundur, "A Non-iterative Transient Stability Program

Including the Effects of Variable Load-Voltage Characteristics," IEEE Trans.,
Vol. PAS-92, pp. 1478-1484, September 1973.

[12]

H.W. Dornmel and N. Sato, "Fast Transient Stability Solutions," IEEE Trans.,
Vol. PAS-91, pp. 1643-1650, JulyIAugust 1972.

[13]

M. Stubbe, A. Bihain, J. Deuse, and J.C. Baader, "STAG - A New Unified

Software Program for the Study of the Dynamic Behaviour of Electrical Power
Systems," IEEE Trans., Vol. PWRS-4, No. 1, pp. 129-138, 1989.

[14]

EPRI Report, Programmers Manual for ETMSP Version 3, Prepared by

Ontario Hydro, 1992.

[15]

C.F. Wagner and R.D. Evans, Symmetrical Components, McGraw-Hill, 1933.

[16]

Edith Clarke, Circuit Analysis of AC Power Systems, Vol. I, John Wiley &
Sons, 1943.

1171

P.M. Anderson, Analysis of Faulted Power Systems, Iowa State Press, Ames,
Iowa, 1973.

[18]

Westinghouse Electric Corporation, Electric Transmission and Distribution

Reference Book, East Pittsburgh, Pa., 1964.

[19]

C.L. Fortescue, "Method of Symmetrical Coordinates Applied to the Solution

of Polyphase Networks," AIEE Trans., Vol. 37, Pt. 11, pp. 1027-1140, 1918.

[20]

T.M. 0'Flaherty and A. S. Aldred, "Synchronous Machine Stability under

Unsymmetrical Faults," Proceedings IEE, Vol. 109A, p. 43 1, 1962.

Transient Stability

Chap. 13

Westinghouse Electric Corporation, Applied Protective Relaying, a new "Silent

Sentinels" publication, Newark, 1976.
H.M. Rustebakke (editor), Electric Utility Systems and Practices, John Wiley
& Sons, 1983.

C.R. Mason, The Art and Science of Protective Relaying, John Wiley & Sons,
1956.
General Electric Company, Use of the R-X Diagram in Relay Work,
Philadelphia, Pa., 1966.
C.R. Mason, "Relay Operation during System Oscillations," AIEE Trans., Val.
56, pp. 823-832, 1937. Discussion by J.H. Neher, pp. 1513-1514. Closure by
C.R. Mason, Vol. 57, pp. 111-114, 1938.
J.H. Neher, "A Comprehensive Method of Determining the Performance of
Distance Relays," AIEE Trans., Vol. 56, pp. 833-844, 1937.
Edith Clarke, "Impedances Seen by Relays during Power Swings with and
without Faults," AIEE Trans., Vol. 64, pp. 372-384, 1945. Discussion by A.J.
Mccomell, p. 472, June Supplement 1945.
IEEE Working Group Report, "Out of Step Relaying for Generators," IEEE
Trans., Vol. PAS-96, pp. 1556-1564, SeptemberIOctober 1977.
C.R. Mason, "A New Loss-of-Excitation Relay for Synchronous Generators,"
AIEE Trans., Vol. 68, pp. 1240-1245, 1949.
J. Berdy, "Loss of Excitation Protection for Modern Synchronous Generators,"
IEEE Trans., Vol. PAS-94, pp. 1457-1463, SepternberlOctober 1975.
D.C. Lee, P. Kundur, and R.D, Brown, "A High Speed, Discriminating
Generator Loss of Excitation Protection," IEEE Trans., Vol. PAS-98, pp.
1895- 1898, November/December 1979.
P.C. Magnusson, "Transient Energy Method of Calculating Stability," AIEE
Trans., Vol. 66, pp. 747-755, 1947.
P.D. Aylett, "The Energy Integral-Criterion of Transient Stability Limits of
Power Systems," Proc. IEE, Vol. 105c, No. 8, pp. 527-536, September 1958.
G.E. Gless, "Direct Method of Lyapunov Applied to Transient Power System
Stability," IEEE Trans., Vol. PAS-85, No. 2, pp. 159-168, February 1966.

References

957

A.H. El-Abiad and K. Nagappan, "Transient Stability Regions of Multimachine

Power Systems," IEEE Trans., Vol. PAS-85, No. 2, pp. 169-178, February
1966.
M. Ribbens-Pavella, "Critical Survey of Transient Stability Studies of MultiMachine Power sysiems by Lyapunov's Direct Method," Proc. 9th Annual
Allerton Conference on Circuits and System Theory, pp. 75 1-767, October
1971.
A.A. Fouad, "Stability Theory-Criteria for Transient Stability," Proc.
Engineering Foundation Conference on System Engineering for Power, pp.
42 1-450, Henniker, N.H., 1975.
M.A. Pai, Energy Function AnaZysis for Power System Stability, Kluwer
Academic Press, 1989.
A.A. Fouad and Vijay Vittal, Power System Transient Stability Analysis Using
the Transient Energy Function Method, Prentice-Hall, 1992.
T. Athay, V.R. Sherkat, R. Podmore, S. Virmani, and C. Puech, "Transient
Energy Analysis," Final Report of U.S. Department of Energy Contract No.
EX-76-C-01-2076, Prepared by Systems Control, Inc., June 1979.
,

Chi-Keung Tang, "Evaluation of the Direct Method for Power System

Transient Stability Analysis," M.Eng. degree thesis, University of Toronto,
August 1984.
N. Kakimoto, Y. Ohsawa, and M. Hayashi, "Transient Stability Analysis of
Electric Power System Via Lure' Type Lyapunov Functions, Parts I and 11,"
Trans. IEE of Japan, Vol. 98, No. 516, MayIJune 1978.
A.A. Fouad and S.E. Stanton, "Transient Stability of Multimachine Power
System, Parts I and 11," IEEE Trans., Vol. PAS-100, pp. 3408-3424, July
1981.
H.D. Chiang, F.F. Wu, and P.P. Varaiya, "Foundations of the Potential Energy
Boundary Surface Method for Power System Transient Stability Analysis,"
IEEE Trans., Vol. CAS-35, pp. 712-728, June 1988.
M.A. El-Kady, M.M. Abu-Elnaga, R.D. Findlay, and C.G. Bailey, "Sparse
Formulation of the Direct Method of Transient Stability for Applications on
Micro-Computers," Proc. 29th Midwest Symposium on Circuits and Systems,
pp. 519-522, Lincoln, Neb., 1986.

Transient Stability

Chap. 1

A.A. Fouad, V. Vittal, Y-X. Ni, H.M. Zein-Eldin, E. Vaahedi, H.R. Pota, K.
Nodehi, and J. Kim, "Direct Transient Stability Assessment with Excitation
Control," IEEE Trans., Vol. PWRS-4, No. 1, pp. 75-82, February 1989.
V. Vittal, N. Bhatia, A.A. Fouad, G.A. Maria, and H.M. Z e i n - ~ l d i ~ ,
"Incorporation of Nonlinear Load Models in the Transient Energy Function
Method," IEEE Trans., Vol. PWRS-4, No. 3, pp. 1031-1036, August 1989.
T. Athay, R. Podmore, and S. Virmani, "A Practical Method for the Direct
Analysis of Transient Stability," IEEE Trans., Val. PAS-98, No. 2, pp. 573584, MarchIApril 1979.
A.A. Fouad, V. Vittal, and Taekyoo Oh, "Critical Energy for Transient
Stability Assessment of a Multimachine Power System," IEEE Trans., Val.
PAS- 103, pp. 2 199-2206, 1984.

P. Kundur, "A Practical View of the Applicability of the Direct Methods,"

Minipaper presented at the IEEE Panel Session on Application of Direct
Methods to Transient Stability Analysis of Power Systems, IEEE Trans., Val,
PAS-193, pp. 1634-1635, July 1984.
P. Kundur, "Evaluation of Methods for Studying Power System Stability,"
Proc. International Symposium on Power System Stability, Ames, Iowa, 1985.

B. Gao, "Assessment of Power System Transient Stability Using Energy

Functions," M.A. Sc. thesis, University of Toronto, 1986.
G. Maria, C. Tang, and J. Kim, "Hybrid Transient Stability Analysis," IEEE
Trans., Vol. PWRS-5, No. 2, pp. 384-393, May 1990.

C.K. Tang, C.E. Graham, M. El-Kady, and R.T.H. Alden, "Transient Stability
Index from Conventional Time Domain Simulation," Paper 93SM487-9PWRS,
presented at the 1993 IEEE PES Summer Meeting, Vancouver, July 1993.