INDIAN INSTITUTE OF TECHNOLOGY. MADRAS 600036 DEC. 27-30
Transient Stability of Power Systems using
Lyapunov Functions
Rajesh Joseph Abraham and V.G. Rau
Abstract. —‘Translent stability studies of power systems are
very important to ensure that the system can withstand the
transient conditions following major disturbances. This paper
throws light upon the different methods for constructing
Lyapunoy functions for transient stability of power systems,
Index Terms- Lyapunov Direct Method, Lyapunov Funetion,
‘Transient Stability
1. INTRODUCTION
JOWER system stability is that attribute of the system or
part of the system which enables it to develop restoring
forces between the elements thereof equal to or greater than
the disturbing forces so as to restore a state of equilibrium,
Depending upon the type of disturbance, there are three types
of stability : (1) Steady state/Dynamic stability - the ability of
the system to restore its initial conditions after a small
disturbance (2) Transient stability - the ability of the system to
restore its initial conditions after a large disturbance or to go
to.a condition close to the initial one (3) Resultant stability ~ if
after a large disturbance, the synchronous operation of the
system is violated and restored after an allowable interval of
time, the system is said to have a resultant stability.
‘Transient stability studies are helpful in determining such
things as the nature of the relaying system needed, critical
clearing time of circuit breakers, voltage level of and transfer
capability between systems. Traditionally, transient stability
analysis has been conducted off-line using step-by-step time-
domain simulations of a wide range of possible system
conditions. This is one reason why direct numerical
integration of the system equations of motion becomes
increasingly unattractive for large systems as the required
computer time and memory size For such an analysis becomes
prohibitively expensive. Each one of these may give
completely different transient stability results that transient
stability should be investigated for all a priori selected
disturbances and all adapted steady state conditions.
Manuscript recived May 28, 2008
Rajesh loseph Abraham is Research Scholar with the Department of
leteal aginorig, Indian laste of Technology, Kharagpur ~721 302
India ema aye itgpemet i,
'V-G. Rau isa Profesor with the Deparment of Eletical Enginevng
Indian Institute of Technology, ‘Kharagpur — 721 302, India (email
‘gree itp.eme in)
‘Transient stability studies are very important and essential
to ensure that the system can withstand the transient condition
following major disturbances like different faults or changes
in system loading, system generation, network connections
tc, Mathematical modeling of a power system is essential for
transient stability analysis. Mathematical modeling involves
the formulation of performance equations in the form of
differential and algebraic equations for synchronous
‘machines, excitation systems, speed governing systems, loads
and transmission networks. Accurate modeling of the
disturbance is of extreme importance for correct transient
stability assessment in power systems. State space models
derived from the above mathematical equations are used in
transient stability analysis, Equal Area Criterion [1], Transient
Energy Method [2], Energy Integral Criterion [3], Phase Plane
‘Techniques [4] and the Lyapunov Direet Methods (LDM) are
direct methods commonly used for transient stability analysis
of power systems, Recent developments in the LDM have led
to its being increasingly accepted as a convenient tool for
fransient stability analysis. In this paper, different techniques
for generating Lyapunov functions for the transient stability
studies of power systems is looked into.
LDM requires the power system equations in state space
form, The LDM, as itis known in contol literature, applied to
the power system transient stability problem involves the
following essential steps: (1) Development of a mathematical
‘model of the power system characterizing its €ynamis and its
state space representation about the post fault equilibrium
point (2) Construction of a Lyapunov function for the model
developed in step 1 (3) Estimation of the boundary of the
stability region (4) Integration ofthe equations of the faulted
system until the system trajectory reaches the boundary, thus
obiaining the critical clearing time
I, CONSTRUCTION OF LYAPUNOV FUNCTION
A. Lyapunov Direct Method
‘The central idea of LDM is to detect stability or instability
for a given system by using the properties of an auxiliary
function of the system state variables, V(X), also known as the
Lyapunov function. A dynamical system represented in the
state space form as = F(X), F(0)=0 is said 10 be
asymptotically stable in the large if the
properties are satisfied for a scalar function
follow
ray[NATIONAL POWER SYSTEMS CONFERENCE, NPSC 2004
(A) VOX) > O forall X40; VO) =0
(2) 7 (X) <0 and F(X) does not vanish identically for ¢ > 0
along any other solution of the system excepting X
(3) VOX) > was |x|] +o.
It may be noted that the stability conditions of Lyapunov are
sufficient and may not be necessary. Consequently, the failure
to find the VOO, F(X) pair which satisfies the Lyapunov
conditions does not necessarily mean that the system is
unstable.
The construction of a suitable Lyapunov funetion is an
Important factor in the application of the direct method to the
westigation of transient stability of power systems. A large
number of methods are available for the construction of
Lyapunov function in the control literature. Various methods
employed for constructing Lyapunov functions for power
system transient stability study can be categorized into two
‘groups. (1) methods applicable only for single-machine
infinite bus systems, and (2) methods applicable for both
single-machine and multi-machine power systems. Among the
methods in the first category, only a few can handle the higher
‘order models, taking into account the effects of all control
‘equipments. Also, these methods become tedious when more
‘accurate representation of control equipment is warranted.
B. Stability Regions
Identification of stability regions occupies a position of
central importance in the practical application of the
Lyapunov method to power system transient stability analysis.
‘The stability region around the origin of the post-fault
‘equilibrium state X = 0 for any Lyapunov function V(X) is,
defined by an inequality V(X) < Vg where Va, is a constant
to be computed. The constant Va; is obtained by evaluating
the function V(X) = Vg corresponding to the unstable
‘equilibrium point, Le. the lowest saddle point. For a system
described by the state equation X= FOX), the determination
‘of unstable equilibrium points requires the solution of non-
linear algebraic equations F(X) = 0.
IIL, METHODS OF CONSTRUCTING LYAPUNOV.
FUNCTIONS
A. Trial and Error Method
In the Trial and Enror Method [5, 6, 7, 8], a quadratic
function with unknown coefficients is assumed as. the
Lyapunov function. These coefficients are determined by trial
‘and error. Different V functions have been proposed for
mple models of single-machine infinite bus systems and an
‘energy-type function is obtained for multi-machine power
systems. This method is time-consuming and requires
‘engineering experience, Hence itis very difficult to use it for a
higher-order multi-machine power system model.
B. Carnuright’s Method
Cartwright’s method involves the generation of a Lyapunov
function for the linearised power system model, and its
generalisation for the nonlinear model. Routh - Hurwitz
stability conditions are to be satisfied for the linear case.
Dharma Rao (9] applied this method to generate Lyapunov
functions for the basic Second order model of the single-
machine infinite-bus system and Lyapunov functions of the
following type were generated,
() raxd 2p seo
0) Patera +2 fond
where f(x) =h, sin db; sin 5°. This method can be extended
fo @ multi-machine system represented by a second order
model and an energy-type Lyapunov function can be
generated. But this generalisation becomes increasingly
difficult as the degree of detail in the teptesentation of
‘machines and associated control apparatus increases
. Aizerman’s Method
Aizerman’s method as applied to power systems is closely
related to that of Cartwright, A general quadratic form is
chosen to serve as a Lyapunov function, The coefficients in
the quadratic form are determined so that it will be a valid
Lyapunov function for the linearised system. The V function
thus generated is applied to the nonlinear system with the
restriction which applies to I to establish limits above the
linear approximation, which the nonlinearities may not
exceed. Since a V function generated for a linear system is
applied to a nonlinear model, the stability region will be very
small. Further, the assumption of a quadratic form for a mult
‘machine system is not an easy task. Hence this method is not
suitable for power system models which are highly nonlinear
D. Zubov's Method
Conventionally, Zubov’s method of constructing a
Lyapunov function involves solution of the following partial
differential equation for a system ¥ = F(X)
ye
i ex,
where V is the desired Lyapunov function and @ is a positive
definite or a semidefinite function of the state variables. The
closed form solution and a unique Lyapunov funetion can be
obtained only if @ suitable choice of this @ function can be
made. When the solution is not possible in closed form,
solution by series expansion of V is necessary.
‘Yu and Vongsuriya [10] applied this method to a basie
second order model of a single-machine infinite-bus system
with nonlinear damping coefficient and transient saliency, The
transcendental terms in the system equations were represented
by a truncated series and a positive definite function @ (8.00)
was used, The approximated boundary of the asymptotic
stability region of the post-fault system was given by the
l=)INDIAN INSTITUTE OF TECHNOLOGY. MADRAS 600 036 DEC.27-30
truneated Lyapunov function series. It has been
experienced that the approximate stability region does not
‘approach the true boundary monotonically with the increase of
the number of terms N in the truncated Lyapunov function
series. Further, the programming effort and computational
time requirement are considerable. The application of the
‘method to higher-order systems is difficult and tedious. To
‘overcome these difficulties, the conventional Zubov’s method
‘was further generalized and a stability boundary in closed
form was obtained by Prabhakara eral [11].
E. Stegov's Procedure
Hewit and Storey [12] used Szego's procedure to generate
‘a Lyapunov function. The method is suitable for the following,
system.
aan
sgh) +265)
The nonlinear funetions (x) and g(a) ean be represented
either as. polynomials or as. such, and the closed-form
Lyapunov function can be obtained, Prusty and Sharma [13]
applied this method to single-machine inginite-bus systems,
with nonlinear damping and transient saliency. The obtained
Lyapunov function was
Vent ean [toda jondes f° [aratenda
where ay(xy)= of +a 0524 40°) and
fils) = by sina, + 3°) ~e, sin 2(x, + 3°)~ Py?
However, this method is suitable only for a second-order
system,
F. Popov's Method
‘The Popov frequency domain criterion forthe stability of a
nonlinear system containing the nonlinearity in the frst and
third quadrants, is used to construct Lyapunov functions.
Using the Popov criterion as the necessary and sufficient
condition for the existence of the Lyapunov functions, two
‘methods, (I) based on Kalman’s construction procedure [14]
which is meant for single-argument nontinearity and further
modified by Desoer and Wu [15] with multi argument
nonlinearities and (2) based on the Moore-Anderson theorem
[16,17,18,19] have been developed to construct Lyapunov
functions for power systems. Pai and Murthy (20) applied
this method for a two-machine system, including the effect of
‘ransfer conductances in system models. Henner proved that
the generalized Popov criterion cannot be used to construct a
Lur'e type Lyapunoy function for a multi-machine power
system model if transfer conductances are included. Gudaru
[21] attempted to generalize the procedure of Pai and Murthy
{20] and generated a V function for the basic second order
modelof the multi-machine system — with transfer
conductances. Sastry [22], have shown that this generalized
V function is not a valid V function and that the Popov
method is inapplicable to multi-machine power system
‘models with transfer condutances taken into account. The
‘Moore-Anderson criterion is applicable only to systems with
single-argument nonlinearities. When the effect of flux decay
is also included in the system model, the system will involve
rmulticargument nonlinearities and Moore-Anderson’s,
‘theorem cannot be used directly. Kakimoto era! (23) applied
Pai’s correction which is applicable to systems with multi=
argument nonlinearities, along with Moore-Anderson’s
theorem.
G Lagrange’s State Function Method
In classical mechanics Lagrange’s state function is used as
‘an energy function for the system represented by simultaneous
second-order differential equations [24]. Since the swinging
behaviour of a power system can be represented by
simultaneous second-order differential equations, Lagrange”s
state funetion can be used to construct a Lyapunov function
for the power system. The detailed procedure is given in [25,
26]. The model to which this was applied includes salien
‘lux decay effect and the first order feedback-type voltage-
regulator,
H. Lagrange Charpit Method
The Lagrange-Charpit Method, which is a well known
technique for solving partial differential equations, is used to
construct the Lyapunov Function, The Lagrange-Charpit
method of constructing the Lyapunov function for @ nonlinear
autonomous system represented by X= /(X), f(0)=0
based on the solution of the following partial differential
equation
PUV,P)=PT F(X) +e(X)=0 where P=AV/AX and
4) is an arbitrary non-negative function whose opposite
sign may be the time derivative of the obtained Lyapunov
fiction V. The determination of the scalar function, V,
through the Lagrange-Charpit method is done in [27, 28}.
1 Variable Gradient Method
The Variable Gradient Method has been applied to both
single-machine infinite-bus systems and multi-machine
systems [29, 30,31, 32]. Its possible to include the effect of
flux decay, proportional plus derivative type automatic
voltage regulator and a second order speed governor in the
system model for a single-machine infnite-bus system. In the
case of multiemachine systems, the method has been applied
to generate the Lyapunov function for a basic second order
‘model and for a system which includes the effect of first order
governor, Energy type Lyapunov functions have been
generated. The Variable Gradient Method is a logical and
systematic one for generating a Lyapunov function. But the
main disadvantage of the method is that a particular choice of