Electrical Power and Energy Systems 89 (2017) 82–93

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

A high-precision real-time approach to calculate closest unstable

equilibrium points
Mehdi Zareian Jahromi, Shahram Montaser Kouhsari ⇑
Department of Electrical Engineering, Amirkabir University of Technology, Tehran, Iran

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, an innovative real-time approach is presented in order to compute the closest unstable
Received 31 July 2016 equilibrium points of system. The proposed approach estimates unstable equilibrium points of system
Received in revised form 23 December 2016 at the first time step of fault duration and there is no need to post the fault data. To such aim, a new con-
Accepted 21 January 2017
cept of equal area criteria is proposed in this paper which estimates the initial value of critical points of
the system. This value is used in order to calculate corrected kinetic energy and as a result the closest
unstable equilibrium point by taking into account the fault trajectory. Moreover, the details of power sys-
Keywords:
tem are considered to calculate unstable equilibrium point by utilizing network preserving model.
Unstable equilibrium point
Transient stability
Finally, several case studies have been conducted on IEEE 9 bus and the New-England 39 bus test systems
Equal area criterion to illustrate the benefits of the proposed approach. It is worth noting that considering structure preserv-
Corrected kinetic energy ing in modeling and at the same time simplicity in implementation and low computational burden are
the main salient features of the proposed approach. As a result, the proposed method is suitable for
real-time applications.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction decades [4]. Transient energy function methods which utilize spe-
cial form of Lyapanov function are an alternative approach to assess
1.1. Importance the system transient stability based on the system energy [5,6].
They prepared fast and efficient stability evaluation for diverse dis-
Power system transient stability assessment has become an turbances. However, lack of accuracy in the stability assessment as a
important issue in the large scale power systems. It’s real-time result of utilizing network reduction model, was a significant disad-
assessment is one of the most challenging problems in the control vantage of the conventional direct energy methods. Also, the meth-
center systems [1]. Evaluation of transient stability is one of the ods depend on recognition of unstable equilibrium point (UEP) of
most significant topics in power systems since its calculation is energy function that makes the transient stability assessment a
necessary, first to assess the system capability for tolerating the very difficult problem. In the last decades, numerous research stud-
large disturbance and second to propose the corrective instruc- ies have been carried out on both development of energy function to
tions. This is a non-linear problem in nature and it should be solved the different system components [7] and the advancement of meth-
in a large scale [2]. ods for identification the UEPs [8]. It is worth mentioning that the
Although the increase of computing power has resulted in a concept of controlling UEP [9] results in a feasible and less conven-
speed-up of online and real-time simulations, but the amount tional path to acknowledge stability of the given post fault state
and complication of simulations have also grown simultaneously. based on awareness of the fault trajectory. The significant disadvan-
The enhancement request for more detailed and complex models tage of the controlling UEP algorithm is the intrinsic problem of
can simply push any given computer to its limits [3]. direct recognition of the controlling UEP [10].
It should be noted that there are only a few works on contin-
1.2. Previous work gency screening taking no fault-on dynamics simulations into
account. Specifically, in [11] the closest UEP method has been
Direct energy methods were suggested in early 80s [4] and extracted and an algebraic formulation for estimating the critical
extended to the level of industrial deployments during the last four clearing time has been captured based on polynomial expansion
of the swing equations. But, considering latter assumption that
⇑ Corresponding author. dynamic related to the rotor angle in the fault duration is a con-
E-mail address: smontom@aut.ac.ir (S. Montaser Kouhsari). stant positive acceleration, makes this method unrealistic.

http://dx.doi.org/10.1016/j.ijepes.2017.01.011
0142-0615/Ó 2017 Elsevier Ltd. All rights reserved.
 M. Zareian Jahromi, S. Montaser Kouhsari / Electrical Power and Energy Systems 89 (2017) 82–93 83

Nomenclature

CCT critical clearing time xqi quadrature axis synchronous reactance

EAC equal area criterion Ei constant voltage behind direct axis transient reactance
NCEAC new concept of equal area criterion Bij network transfer admittance between bus i and bus j
UEP unstable equilibrium point Gij network transfer conductance between bus i and bus j
BCU based controlling unstable T mi time constant of AVR
TEF transient energy function li feedback gain of AVR
SDG severely disturbed group li a constant gain to adjust the location of the desired
LDG less disturbed group operating points
PM proposed method T 0doi direct axis transient open-circuit time constant
TDS time domain simulation T 0qoi quadrature axis transient open-circuit time constant
COI center of inertia E0di direct axis internal voltage magnitude at bus i
DT decision tree E0qi quadrature axis internal voltage magnitude at bus i
AVR automatic voltage regulation Pdk real power demand at load node k
di rotor angle of i-th generator Q dk ðV k Þ reactive power demand at load node k
dicrit critical rotor angle of i-th generator ILK \/k constant current injection at load node k
xi rotor speed of i-th generator Vi external generator voltage magnitude at bus i
x0 reference rotor speed hi external generator voltage angle at bus
Mi inertia momentum of i-th generator Vk voltage magnitude at load node k
Di damping constant of i-th generator hk voltage angle at load node k
Pmi input mechanical power of i-th generator ILk \/k constant current injection at load node k
Pei electrical power of i-th generator Efi excitation voltage magnitude
Zs series impedance of a transmission line 1; . . . ; n generator internal buses
ZL shunt admittance as a local load nþ1 infinite bus
x0di direct axis transient reactance n þ 2; . . . ; n þ m þ 1 load buses
x0qi quadrature axis transient reactance
xdi direct axis synchronous reactance

According to achievement of Lyapanov family function intro- based on the corrected kinetic energy and also new concept of
duced in [6,12] authors in [13] proposed an approach to estimate equal area. The proposed approach in this paper, calculates UEP
critical clearing time based on bounding fault-on trajectory and which is applied for estimating critical clearing time and as a result
thereby the fault-cleared state. assessing power system transient stability. The general novelties
are described as follows:
1.3. Contribution and organization of the paper
(1) Despite of previous suggested methods in [3,6–9,12], pro-
As mentioned, parallel goals have been investigated in the pre- posed method in this paper estimates initial critical angle
vious published papers which is categorized into two groups: first, that is relatively very close to the exact UEP. To do so, new
methods that consider details and nonlinearity in power system concept of equal area criteria is proposed. This method is a
such as [4,14–18] and second, methods evaluate transient stability modification of EAC that determines initial critical angle
in real-time such as [19–24]. Unfortunately each group has serious from generator terminal regardless the effect of AVR and
drawbacks which are listed as follows: GOVERNOR.
The approaches suggested in the first group are pursuing tran- (2) While in [3,5,14,25], authors suggested methods based on
sient stability assessment taking details and nonlinearity into network reduction model, proposed methods in [2,4,10,16],
account. These approaches are not suitable for real-time or near have demonstrated that it is vital to consider network pre-
real-time transient stability assessment application because: serving structure. In this paper, proposed method utilizes
network preserving structure to take specifically the effect
1. These approaches have considerable computational cost. of dynamic controllers such as AVR and GOVERNOR into
2. They need to find the potential function of each element may be consideration. To do so, an innovative and uncomplicated
added into network. method based on corrected critical kinetic energy is pro-
3. They require post-fault data for transient stability assessment. posed, that consider the effect of AVR and GOVERNOR with-
out solving any complicated potential equations.
Since transient stability assessment is interested in large scale (3) To real-time assessment of transient stability and UEP, it is
power systems, as a result, the approaches in the first group are important to have a method with two salient features: (1)
not reliable for real-time analysis of large scale power networks. low computational burden and (2) independency form post
The approaches presented in the second group, first priority is fault data. We propose a method that has very low compu-
real-time assessing of the power system. Authors of these tational cost. Also our method is capable of calculating UEP
approaches have applied network reduction model in the problem in real-time, after first time step of fault occurrence, without
solving and as a result, the accuracy of the evaluation has been scarifying any details in system modeling and considering
decreased significantly. fault trajectory into account.
In this paper, an innovative hybrid approach to find closest 1. Low computation burden: For more clarification, as men-
unstable UEP is proposed which simultaneously considers two tioned before, proposed methods in [4,12,26,27] are basi-
aforementioned goals of the groups. The proposed approach is cally offline from computational burden point of view.
84 M. Zareian Jahromi, S. Montaser Kouhsari / Electrical Power and Energy Systems 89 (2017) 82–93

This is because they use potential function with many where,

equations to solve. from this point of view, this paper uses
1 0 1
network preserve structure, our method only uses kinetic P ei ¼ E V i sinðdi  hi Þ þ 0 E0di V i cosðdi  hi Þ
x0di qi xqi
energy and considers the effects of AVR and GOVERNOR
much simpler than the methods which use potential x0di  x0qi
þ V 2i sin½2ðdi  hi Þ ð2Þ
energy function [6,12,16,28]. 2x0di x0qi
2. Independency form post fault data: this independency is
much important than low computational burden, because 2. In order to model the exciter dynamics, a simplified exciter
it simply states that only fault data can be used in the model is utilized in one-axis generator model as follows [16]:
real-time assessment of transient stability and UEP calcu-
T v i E_ fi ¼ Efi  li ki V i cosðdi  hi Þ þ li ð3Þ
lation and not post fault data. From this point of view,
potential need post fault data and cannot assess the 3. In order to model the external generator bus, the power flow
power system transient stability in real-time. But this equations at each generator bus are expressed as follows [16]:
paper suggests an approach that calculates the UEP in
the first time step after fault occurrence while it takes X
nþ1
0¼ V i V j ðBij sinðdi  hi Þ þ Gij cosðdi  hi ÞÞ
AVR and GOVERNOR effects and also fault trajectory into j–i
consideration.
X
nþmþ1
þ V i V l ðBil sinðdi  hl Þ þ Gil cosðdi  hl ÞÞ ð4Þ
Combination of mentioned points in 1–3, are accountable for l¼nþ2
real-time issue of the transient stability assessment, because firstly,  x10 E0qi V i sinðdi  hi Þ  1
x0qi
E0di V i cosðdi  hi Þ
using the proposed approach there is no need to calculate post-fault di

x0di x0qi
data and secondly, as will be demonstrated in the simulation results,  2x0di x0qi
V 2i sin½2ðdi  hi Þ
the proposed approach will evaluate the system stability condition
in moment of the fault clearance easily, precisely and quickly.
In the next section, the network preserving model and its X
nþ1
0 ¼  V i V j ðBij cosðdi  hi Þ þ Gij sinðdi  hi ÞÞ
related equations are described as a network model of the system. j–i
Problem statement and solution procedure of this research is dis-
X
nþmþ1
cussed in Section 3 and simulation results are discussed in Sec-  V i V l ðBil cosðdi  hl Þ þ Gil sinðdi  hl ÞÞ
tion 4. Finally, conclusion is included in Section 5. l¼nþ2
ð5Þ
V 2i
þ x0di
 x10 E0qi V i cosðdi  hi Þ  x10 E0di V i sinðdi  hi Þ
2. Model considerations di qi

x0 x0qi
 2xdi0 x0
V 2i cos½2ðdi  hi Þ  1
As mentioned in the previous section, the main limitation of tran- di qi

sient function methods was the accuracy of the network model [2]. 4. In order to model the load bus, the load at each load bus is
Also, from the real-time applications point of view, calculation of assumed as a hybrid combination of three components. (1)
potential energy part of transient energy function was another con-
Constant power Pdk þ Q dk . (2) constant current injection
cern regard to transient stability assessment issues, because the
ILk \u and (3) frequency dependent load Dk h_ k . The power
k
potential energy function has many time-consuming computational
flow equations at each load bus are given by Eqs. (6), (7) [16]:
terms which needs the post fault data. Therefore, all of the previous
direct methods which use potential energy function to estimate UEP P dk þ V k ILK cosðhk  uk Þ  Dk h_ k
of system are not suitable for real-time and online applications.
X
nþmþ1
ð6Þ
¼ V k V j ðBkj sinðdk  hi Þ þ Gkj cosðdk  hi ÞÞ
2.1. Network preserving model j¼1

Network-preserving models were first suggested in 1980s [4]. Q dk þ V k ILK sinðhk  uk Þ

The structure of network is preserved; as a result the related Lya-
X
nþmþ1
ð7Þ
punov function is called ‘‘structure preserving” or ‘‘topological” ¼ V k V j ðBkj cosðdk  hi Þ þ Gkj sinðdk  hi ÞÞ
energy function. Due to the lack of reduction in network models, j¼1
the identity of all the nodes is not spoiled. Recently, a model has
been suggested in [16] for preserving model which is used in this
study. The dynamics expression which exists in the model can be 2.2. Transient energy function
described as follows:
First, the infinite bus n + 1 is assumed to be reference bus, buses The most recent energy function for network preserving model
1, . . ., n are generator buses and buses n + 2, . . ., n + m + 1 are load was expressed in [16]. According to [29], TEF consists of two terms
buses. of the potential energy and kinetic energy. As described in [16], the
potential energy function has many complex computational terms
1. In order to model the internal generator bus, a two-axis gener- which their calculation is very difficult and time consuming. How-
ator model considering first-order exciter system is selected for ever, the kinetic energy term is simple function of rotor angle and
each generator [16]. inertia. In order to clarify the importance of the issue, Fig. 1 shows
the potential energy function of a system with two generators.
d_ i ¼ xi More details are provided in [4]. As shown in Fig. 1, the potential
Mi x _ i ¼ Di xi þ Pmi  Pei energy function of this system with two generators has many local
xdi x0di and global extremum points where fault trajectory move to one of
T 0doi E_ 0qi ¼  xxdi0 E0qi þ x0di
V i cosðdi  hi Þ þ Efi ð1Þ
di these extremum points based on the fault location. Finding UEP
xqi x0qi
T 0qoi E_ 0di ¼  xqi0 E0di þ between these extremums which is used to assess the transient
x
x0qi
V i sinðdi  hi Þ
qi stability of power system is the most important part of the direct
 M. Zareian Jahromi, S. Montaser Kouhsari / Electrical Power and Energy Systems 89 (2017) 82–93 85

Potential energy function 1

-1

-2

-3

-4
5
4
3
2
1 4 5
0 3
-1 1 2
-2 -1 0
-3 -2
Rotor angle of Second machine (Rad) -4
-5 -4 -3
-5 Rotor angle of first machine (Rad)

Fig. 1. Potential energy function of system with two generators.

method approaches. On the other hand, finding UEPs of multi Vn

V g ∠θ g
machine power system using potential energy function, is very dif- E ∠δ
ficult, complex, and time consuming because there are many non- jX d' ZL
linear fault trajectories and minimization of potential energy
function needs post fault data.

3. Proposed approach
G
Zs
An appropriate method for finding UEP of the system must be
able to estimate a suitable value of the UEP at the first time step
of the fault simulation which takes details of power system into
consideration. From transient stability studies point of view, the
critical clearing angle of generators is UEP of system when a distur-
Fig. 2. Single line diagram used in studying proposed concept of EAC.
bance occurs in system. In fact, UEP is the point where the kinetic
energy of generators during fault and potential energy of genera- considering second order model of synchronous generator, the elec-
tors in post fault duration are equal [4]. Therefore, when a fault tric output power generator (Pe ) is calculated as follows [31]:
occurs in the system, the minimum potential energy of system in
stable state is equal to the maximum kinetic energy that system 1
Pe ¼ EV g sinðd  hg Þ ð8Þ
can be achieve to be stable. According to [30], the potential energy x0d
|fflfflffl{zfflfflffl}
function of power system consists of potential energy function of P max
each generator of system; hence, potential energy of system is
minimum when the potential energy of each generator is mini- where (E\d) is the internal voltage bus and (V g \hg ) is external volt-
mum. On the other hand, the minimum potential energy of each age bus. In NCEAC, each generator is modeled by the internal volt-
generator for transient stability is equal to maximum kinetic age (E\d) which transmits electrical power to the power network
energy of each generator that can be achieved to be stable. As which is modeled with terminal voltage (V g \hg ). Therefore, the pro-
shown in Fig. 1, there are different minimum points in the poten- posed approach considers each generator of network as a single
tial energy function for different fault locations. Therefore, there generator (E < d) which is connected to an equivalent generator
are different fault trajectories and estimation of minimum point with terminal voltage (V g \hg ). As shown in Fig. 3, when a fault
by the potential energy function is very difficult and it needs post occurs in the system, P max is changed and it obviously depends on
fault data. A generator obtains the maximum kinetic energy when the fault location and the electrical fault trajectory.
a three phase fault occurs in terminal of generator, but this energy The initial fault dependent UEP can be calculated as follows
is different when a fault occurs in the other points of system and [31]:
also initial acceleration of the generators (x _ ) are different based Z dc Z dmax

on the fault location since maximum output electrical power is ðPm  Pmax2 sinðd  hÞdd ¼ ðPmax1 sinðd  hÞ  Pm Þdd ð9Þ
d0 dc
changed because of different fault location. Therefore, UEP depends
on the fault trajectory in the system. Therefore, the dc ¼ dc ¼ UEPinitial which is calculated by solving (9) is
a suitable initial value of UEP ¼ dc when a three phase short circuit
3.1. New concept of equal area criterion occurs in a point very close to the generator. According to (9), it is
concluded that the effects of automatic voltage regulator (AVR)
The proposed method presents a new concept of equal area cri- and governor are not considered in this calculation. As mentioned
terion (NCEAC) to estimate an appropriate approximation for initial in the contribution section, the goal of proposing NCEAC is fast
value of UEP in multi machine power systems. As shown in Fig. 2, and reasonable estimating of UEP (dinitial
c ) initial value.
86 M. Zareian Jahromi, S. Montaser Kouhsari / Electrical Power and Energy Systems 89 (2017) 82–93

Pe
Pe1 : Pre & Post fault duration

Pmax1

PM Pe 2 : during fault

Pmax 2

0 δ0 δc δ max π δ − θg
Fig. 3. EAC concept in P  d graph for multi machine power system.

3.2. Stable or unstable cases (calculating critical kinetic energy) 3.3. Calculating critical angle

There are two different conditions when a three phase short cir- After calculation of Kecritical considering stable or unstable cases
cuit occurs in terminals of a generator and it remains until (d ¼ dc ): conditions, the high precision value of UEP of each generator with-
A. Stable case and B. Unstable case [32]. According to the red curve out considering the fault trajectory effect for a fault very close to
(Stable Case) in Fig. 4, stable state occurs in the system when the the generator is estimated by following equations:
fault is cleared at dc , and the minimum kinetic energy of the gen-
erator becomes zero at the post fault duration. rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
In this state, once again, a three phase short circuit fault is 1 2KeCritical
KeCritical ¼ M x2 ! xCritical ¼ ð12Þ
assumed at generator terminals which are removed at 2   M
(dc ¼ p  d0 ). According to this hypothetically fault scenario, if 1 1 M xCritical
x_ ¼ Pm ! x ¼ Pm t ! t Critical ¼ ð13Þ
the kinetic energy of generator becomes zero after clearing the M M Pm
fault, therefore the high precision value of UEP is equal to p  d0 dd
for a fault very close to the generator and critical kinetic energy ¼ x ! dn ¼ xn  t n þ d0 ! dcritical ¼ xcritical  tcritical þ d0 ð14Þ
dt qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
is calculated as follows: rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2KeCritical
2KeCritical M M 2KeCritical
Kecrite ¼ Pm ðp  d0  d0 Þ ¼ P m ðp  2d0 Þ ð10Þ dCritical ¼  þ d0 ¼ þ d0 ¼ d ð15Þ
M Pm Pm
Otherwise, if the minimum kinetic energy of system is greater than
zero after clearing the fault, the critical kinetic energy is obtained
from the unstable case. 3.4. Effects of AVR and governor on UEP
Black curve in Fig. 4 shows unstable behavior of generator for
the mentioned fault. In this case, the minimum kinetic energy of Due to severe faults in the terminal of generator, terminal volt-
generator is greater than zero after clearing the fault at dc . In other age changes significantly when AVR has maximum contributions
words, the minimum kinetic energy of generator in post fault dura- in the control of voltage terminal. Also, rotor angle is accelerated
tion is the amount of energy that generator has gained in fault as a result of severe faults near the generator terminal which
duration for the instability. Thus, high precision value of critical causes the maximum contribution of governor in the control of
kinetic energy of generator for a fault very close to the generator the generator accelerations. Therefore, the estimation of initial
will be obtained as follows: UEP which is dependent on the fault trajectory has maximum cal-
culation error with high perecisin value of UEP for the faults near
Kecritical ¼ Kemax  Kemin ð11Þ the generator. This latter error can be estimated as follows:
Kinetic Energy

K e,max

Unstable Case

K e, crit K e,max K e,min K e,max

K e,crit
Three Phase short K e,min Stable Case
circuit duration

K e,min

** *
c c

Fig. 4. Kinetic energy behavior during stable or unstable condition.

 M. Zareian Jahromi, S. Montaser Kouhsari / Electrical Power and Energy Systems 89 (2017) 82–93 87

According to [33], utilizing Taylor expansions, rotor angle (di ðtÞ), 4. Results and discussion
and rotor angular speed (xi ðtÞ) of generators, can be approximated
as follows [33]: As discussed in the proposed method section, an efficient sim-
2 3 ulation method for real-time assessment of transient stability must
ð1Þ t ð2Þ ð3Þ t
di ðtÞ ¼ di ð0Þ þ di ð0Þt þ di ð0Þ: þ di ð0Þ þ    ð16Þ have two inseparable features: low computational cost and ability
2! 3!
to estimate transient stability condition of the system at first time
t2 t3
xi ðtÞ ¼ dð1Þ ð2Þ ð3Þ
i ð0Þ þ di ð0Þt þ di ð0Þ:
ð4Þ
þ di ð0Þ þ    ð17Þ step of simulation without any data from subsequent time steps. In
2! 3! order to evaluate the proposed method from the two above-
It is worth noting that in (16), di ð0Þ is the rotor angle before fault mentioned aspects, several case studies have been considered
ðmÞ and studied.
occurrence and di is the m-th order derivatives of rotor angle at
t ¼ 0 ½Se. It has been proved in [33] that the odd coefficients of
(16) are equal to zero in the fault duration. Moreover, the first 4.1. Case1: IEEE 9 bus test system
two non-zero terms of (16) have a sufficient accuracy to estimate
the rotor angle for the times less than 0.3 of fault durations [33]. In order to verify the performance and benefits of the proposed
According to this explanation, it is concluded that (16) and (17) method, IEEE 9 bus test system which includes 3 generators and 9
can be rewritten as follows: lines has been utilized [34]. In this research, the synchronous ref-
erence frame is employed and it is assumed that the speed and
ð2Þ t2
di ðtÞ ¼ di ð0Þ þ di ð0Þ ð18Þ rotor angle of each generator are compared with speed and rotor
2! angle of generator 1 which is considered to be reference angle
ð2Þ
xi ðtÞ ¼ di ð0Þt ð19Þ (slack). Also, it is assumed that IEEE 9 bus is simulated with time
_ i ðtÞ is calculated as follows:
Therefore, by using (19), x step of 0.001[s].
Table 1 shows obtained UEP of the system by using the pro-
x_ i ðtÞ ¼ dð2Þ
i ð0Þ ¼ Constant ð20Þ posed method and time domain simulation method when a three
phase short circuit fault occurs at a point very close to each bus.
_ i ðtÞ is equal to a constant value
As it can be seen in (20), x According to Table 1, the estimated critical angle by the proposed
ð2Þ ð2Þ
(di ð0Þ ¼ cte) at the first time step of fault duration where di is approach is very accurate and results are close to those obtained
dependent on the fault location and it can be obtained as follows: through the time domain simulation method. It should be noted
1 that by comparing required CPU times for the proposed method
ð2Þ
di ð0Þ ¼ ½Pmi  Pei jt¼0 ð21Þ and time domain simulation, it is clear that the proposed method
Mi
is faster than the time domain simulation method since the pro-
According to (21), x _ i ðtÞ is maximum when a three phase short circuit posed approach estimates the critical angle of generators at first
fault occurs at the voltage terminal of generator since Pei ¼ 0 in this time step of fault duration and it doesn’t need any post fault data.
state and x _ i ðtÞ has a less value for other fault locations depending on Also, Fig. 6 shows UEPs of system and potential energy function of
the electrical distance or proximity of the fault. Hence, the calcula- system for different fault locations. As can be seen in Fig. 6, speci-
tion of initial UEP has maximum error with high precision value of fied points are local minimum of potential energy function of sys-
UEP when a three phase short circuit occurs in a point very close to tem which is estimated by the proposed approach in this paper
the generator because AVR and governor control have a maximum without considering the potential energy function and the post
effect in this state and the initial UEP has less error with the high pre- fault data.
cision value of UEP for the other fault locations. Now, suppose that
initial UEP of generators dc ¼ UEPinitial for a fault very close to the gen- 4.1.1. Scenario 1 (Fault at bus 6)
erator is obtained by solving (9) and d c is obtained for a generator Fig. 7a shows the rotor angle of generators when a three phase
using the algorithm which is described in Section 3.3. Then, it is con- short circuit fault is happened at bus 6. This fault is cleared when
cluded that the maximum calculation error of the critical angle by the rotor angle of generator 2 is equal to 88.20 (Deg). As it can be
taking into account effects of AVR and governor is given by: seen in this fig, both generators are stable because clearing angle is
e ¼ d 
c  dc ð22Þ less than the critical angle of both generators. On the other hand,
Fig. 7b shows the rotor angle of generators when the fault is
Therefore, according to (20), the calculation error by considering removed at the moment that rotor angle of generator 2 is equal
effects of AVR and governor for the other fault locations can be esti- to 96.91 (Deg) and angle of generator 3 is equal to 101.6 (Deg).
mated by (23): According to Fig. 7b, it is obvious that the generator 2 is unstable
  and generator 3 is stable because angle of clearing the fault is less
x_ 
e ¼ e ð23Þ than estimated critical angle for GEN3 and it is greater than esti-
x_
mated critical angle for GEN2.
In this equation x_ is rotor acceleration at first time step of fault Finally, Fig. 7c shows that both generators are unstable when
duration when a three fault short circuit occurs in a point near the mentioned fault is cleared at the moment in which rotor angle
the generator and x _  is rotor acceleration at first time step of fault of GEN2 is equal to 101.2 (Deg) and angle of GEN3 is equal to 115.4
duration when a fault occurs in the locations of network. In the next (Deg), because the angle of clearing the fault is greater than both of
step, once again (9) is solved for the intended fault location to esti- critical angle of generators.
mate new initial critical angle (dinitial;new
c ) relevant to the intended
fault location. Finally, the high precision value of critical angle of 4.1.2. Scenario 2 (Fault at bus 9)
each generator dependent on fault location is as follows: In this case, a three phase short circuit fault occurs at bus 9 and
it is removed when rotor angle of GEN2 is equal to 56.13 (Deg) and
dc ¼ dinitial;new
c þ e ¼ UEP ð24Þ
rotor angle of GEN3 is equal to 96.62 (Deg). As can be seen in
Therefore, the potential energy of the system is minimum for these Fig. 8a, both generators are stable because according to Table 1
values of UEPs. Fig. 5 shows the calculation procedure of UEPs for angle of clearing the fault is less than the critical angle of GEN2
system. and GEN3. Fig. 8b shows that both generators are unstable when
88 M. Zareian Jahromi, S. Montaser Kouhsari / Electrical Power and Energy Systems 89 (2017) 82–93

Start

Estimate initial critical angle of each

generators for the fault at their terminal using
equal area criterion

Putting fault at terminal of each generator

Applying stable
Stable ?
Yes case

No
Applying
unstable case

Putting fault at arbitrary location of system

Calculate initial critical angle of each

generator for intended fault location

Calculate error data and estimate high

precision value of UEPs of System
Fig. 5. Procedure of stable or unstable case algorithm.

Table 1
Critical angle of IEEE 9 bus test system.

Fault point Proposed method (PM) Time-domain simulation (TDS) Error %

 SM PM 
G2 G3 CPU½S  10 2 G2 G3 CPU[s] dc dc 
 dSM   100
c

dc (Deg) dc (Deg) dc (Deg) dc (Deg) G2 G3

BUS1 112.4 115.21 0.1327 114.3 116.4 21.140 1.66 1.022
BUS2 83.44 32.31 0.1012 80.66 31.21 3.44 3.52
BUS3 51.32 110.3 0.0992 49.29 108.7 4.11 1.47
BUS4 89.67 90.28 0.1119 91.02 91.44 1.48 1.26
BUS5 109.2 85.43 0.1317 113.5 84.1 3.78 1.58
BUS6 93.3 112.7 0.1239 91.2 111.7 2.30 0.895
BUS7 92.9 48.9 0.0928 96.21 51.09 3.44 4.28
BUS8 107.44 89.23 0.1017 105.5 88.14 1.83 1.23
BUS9 61.72 105.8 0.0865 59.96 103.6 2.93 2.12

clearing angle of the fault is equal to 64.37 (Deg) for GEN2 and The estimated UEP is compared with time-domain simulation
107.9 (Deg) for GEN3. The reason of generator instability in Fig. 8 method as listed in Tables 2 and 3. It should be noted that it is
is that clearing angle is greater than the estimated critical angle assumed that time step of simulation is 0.001[s] where rotor angle
of both generators. Fig. 8 results clearly emphasizes on the ability of generator 2 is used as reference angle to study of this simulation.
of the proposed method in order to estimate the critical angle According to Table 2 it can be seen that the estimated UEP of the
(UEP ¼ dc ) in both online and real time applications. system by PM and time domain simulation method are very close.
In the following, by comparing of their elapsed time, it is clear that
4.2. Case 2: IEEE 39 bus test system (10-machine New-England power the proposed method of this paper is faster than time domain sim-
system) ulation method. Moreover, Table 3 validates ability of the PM in
estimation of UEPs of system in first time step of fault duration.
Schematic of IEEE 39 bus with 10 generators, 45 lines has been In the next step, four scenarios have been study for validation of
shown in [35]; also, lines and generators data are listed in this ref- results in Table 2 where details of the scenarios are provided in
erence. Like the IEEE9- bus section, in order to estimate UEP of gen- Table 4. According to Figs. 9 and 10 and Table 2, the PM estimates
erators and system by the proposed method of this paper, a three the UEPs of system with high precision. Because they remain stable
phase short circuit fault occurs at a point very close to each bus. when angles of clearing the fault for the generators are less than
 M. Zareian Jahromi, S. Montaser Kouhsari / Electrical Power and Energy Systems 89 (2017) 82–93 89

CP : Critical Point

CP for fault at bus 2 CP for fault at bus 4 CP for fault at bus3

CP for fault at bus 8 CP for fault at bus 6

CP for fault at bus 7

CP for fault at bus 5 CP for fault at bus 9

CP for fault at bus 1

Potential Energy Function

0
0

50

-5 100
0
20 40 60 80 150 Rotor Angle of Generator 2 (Deg)
100 120
140
Rotor Angle of Generator 3 (Deg) 160 180 200

Fig. 6. UEPs of IEEE 9bus test system for different fault location.

150
Rotor Angle (Deg)

100 Gen2 clearing Angle=101.6

Gen3 Clearing Angle=88.2
50
(a)
0

-50
300
Rotor Angle (Deg)

200 Gen2 Clearing Angle=106.4

Gen3
100
(b)
Clearing Angle=96.91

0
Rotor Angle (Deg)

200
Gen2
Gen3 Clearing Angle=115.4
100 (c)
clearing Angle=101.2

0
Starting time of fault=5 (Se)

-100
0 1 2 3 4 5 6 7 8 9 10
Time (Se)

Fig. 7. Rotor angle of generators when fault occurs at bus 6 at 5.00[Se] (Scenario 1).

their estimated UEP by PM in Table 2 and they become unstable generator 9 becomes unstable after clearing the fault. Because
when angles of clearing the fault for generators are greater than angle of clearing the fault for Gen9 is greater than the estimated
their estimated UEP by PM. For instance, the PM forecasts that value for UEP by PM in Table 2. Also in this state, angles of clearing
all generators become unstable when the fault occurs at bus 24 the fault for the other generators are less than their estimated UEP
and it is removed when rotor angle of generators are greater than by PM in Table 2. Therefore, these generators remain stable after
their estimated UEP by PM in Table 2. clearing the fault.
Fig. 9 shows the rotor angle of generators when a three phase In the following, the same scenarios have been done for the
short circuit fault occurs at bus 24. The fault is cleared when rotor other clearing angles of the fault and they are shown in Figs. 11
angle of generator 1, 8, 9 and 10 are equal to 52.41°, 15.14°, and 12, respectively. According to these figures and Table 2, the
6.20° and 55.21° respectively. As it can be seen in this figure, PM estimates the UEPs of system very accurately. Because they
all generators are stable since their clearing angles are less than remain stable when angles of clearing the fault for the generators
their estimated UEP by PM in Table 2. are less than their estimated UEP by PM in Table 2 and they
In the next scenario, the fault is removed when rotor angle of become unstable when angles of clearing the fault for generators
generators 1, 8, 9 and 10 are equal to 57.32°, 15.21°, 6.62° are greater than their estimated UEP by PM. For instance, as it
and 58.31° respectively. In this state, According to Fig. 10, only can be seen in Fig. 11, the generators 3, 4, 5, 6 and 7 are stable
90 M. Zareian Jahromi, S. Montaser Kouhsari / Electrical Power and Energy Systems 89 (2017) 82–93

150
Gen2

Rotor Angle (Deg)

100 Gen3 Clearing Angle=96.62

50 Clearing Angle=56.13
(a)
0

-50

-100

Gen2
Rotor Angle (Deg)

200 Gen3

Clearing Angle=107.9
100 (b)
Clearing Angle=64.37

0
Starting time of fault=5 (Se)

-100
0 1 2 3 4 5 6 7 8 9 10
Time (Se)

Fig. 8. Rotor angle of generators when fault occurs at bus 9 at 5.00[Se] (Scenario 2).

Table 2
UEP of IEEE-39 system by proposed method (PM) and time-domain simulation method (TDS).

Fault G1 G3 G4 G5 G6 G7 G8 G9 G10
point
UEP [Deg] UEP [Deg] UEP [Deg] UEP [Deg] UEP [Deg] UEP [Deg] UEP [Deg] UEP [Deg] UEP [Deg]
(Bus#)
PM TDS PM TDS PM TDS PM TDS PM TDS PM TDS PM TDS PM TDS PM TDS
2 62.45 61.88 29.23 30.21 36.45 36.24 21.5 22.12 34 34.18 33.5 32.94 37.84 37.14 23.76 24.07 70.77 71.16
10 89.5 89.31 12.21 11.98 39.25 38.92 31.6 31.54 36.82 36.72 35.54 36.08 41.2 40.92 28.62 27.88 18.71 19.31
12 99.42 100.1 5.38 4.78 45.68 46.04 35.77 34.21 43.06 42.87 42.07 41.98 59.18 58.98 41.36 42.54 89.52 88.95
24 57.5 57.4 10.2 9.97 18.1 18.2 20.55 20.92 20.72 20.18 27.57 26.72 15.4 15.3 6.57 6.38 58.5 58.4
30 118 116 12.4 13.2 32.7 33.5 17.2 16.8 28.1 29.2 27.9 26.5 58.5 59.6 20.8 21.2 10.1 9.26
33 74.4 73.5 12.1 11.6 28.4 27.3 20.8 19.5 28.3 27.2 26.3 26.8 36.4 35.7 21.9 22.72 75.7 73.8
37 70.0 68.7 16.3 15.8 33.8 32.4 23.4 21.8 32.7 34.4 31.4 32.7 38.0 40.0 24.9 25.8 73.1 75.2

Table 3
Comparison of TDS and PM for IEEE 39-bus test system.

Method Fault scenario Applying Removing UEP estimated UEP estimated Simulation Simulation Maximum calculation
Faulted Bus# fault-time [Se] fault-time [Se] time [Se] CPU-time [Se] time [Se] CPU-time [Se] error between PM and
TDS%
TDS 2 0 0.17 8.00 27.899 10 34.874 3.24
PM 0.001 0.008194 10 8.194
TDS 10 0 0.52 7.61 26.168 10 34.387 2.65
PM 0.001 0.008421 10 8.421
TDS 12 0 0.45 6.52 22.833 10 35.021 4.56
PM 0.001 0.008673 10 8.673
TSM 24 0 0.28 7.44 25.282 10 33.982 3.18
PM 0.001 0.009071 10 9.071
TDS 30 0 0.82 9.03 31.208 10 34.561 1.72
PM 0.001 0.00814 10 8.142
TDS 33 0 0.22 8.32 28.223 10 33.921 2.57
PM 0.001 0.00797 10 7.975
TDS 37 0 0.21 7.21 24.980 10 34.6471 2.79
PM 0.001 0.00827 10 8.278

Table 4
Different scenarios for fault at bus 24 (Section 4.2).

Scenario number Fault clearing angle (Degree)

G1 G3 G4 G5 G6 G7 G8 G9 G10
1 52.41 7.35 15.82 17.26 18.17 24.94 15.14 6.20 55.21
2 57.32 8.05 16.92 18.41 19.13 25.84 15.21 6.62 58.31
3 57.59 8.70 17.70 19.20 19.60 26.65 15.52 6.71 58.62
4 58.2 10.5 19.2 22.5 23.2 28.1 15.94 6.81 59.2
 M. Zareian Jahromi, S. Montaser Kouhsari / Electrical Power and Energy Systems 89 (2017) 82–93 91

G1
100 Clearing Angle= 6.2o G3
G4
Clearing Angel = - 15.14o G5
G6
50
G7
G8
Rotor Angle [Deg]

G9
0 G10

-50

-100

Clearing Angle = -52.41o Clearing Angle = -55.21o

-150
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time [Se]

Fig. 9. Rotor angle of generators for scenario 1.

300
G1
o
Clearing Angle= -15.21 G3
250
G4
200 G5
o
Clearing Angle = 6.62 G6
150 G7
o
Clearing Angle= -58.31 G8
Rotor Angle [Deg]

100 G9
G10
50

-50

-100

-150
o
Clearing Angle =-57.32
-200
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time [Se]

Fig. 10. Rotor angle of generators for scenario 2.

when their clearing angles are 8.70°, 17.70°, 19.20°, 19.6° and Also, the required CPU-time for each fault scenario demonstrates
26.65° respectively whereas according to Fig. 11 these generators that the PM is faster than real-time because of its low computa-
become unstable when clearing angle of generators 3, 4, 5, 6 and tional cost. But, CPU-times of TDS method are greater than real-
7 are 10.5°, 19.2°, 22.5°, 23.2° and 28.1° respectively. Therefore, time. Also, as it can be seen in the column of ‘‘UEP Estimated
it is concluded that the generators have become unstable only for a CPU-time” in Table 3, the CPU-time of TDS method is more greater
few degrees clearing angel greater than estimated UEP by PM than the entire simulation time (10.0[Se]) whereas, the CPU-time
which emphasizes on the accuracy of PM to estimate UEPs of for the PM is even less than 0.001[Se]. On the other hand, the esti-
system. mated UEPs by the PM are very close to the simulation results of
The summary of simulation results are presented in Table 3. time-domain simulation method. According to the last column of
Simulation results in this table shows that the PM estimates UEP Table 3, it can be seen that maximum calculation error between
of the system at each fault scenario during the first 0.001[Se] PM and TDS is less than 5% where the error is calculated as follows:
!
because time step of real-time simulation is assumed 0.001[Se].
jUEP PM  UEP TDS j
Hence it can be said, the PM has the ability of estimating the Error ð%Þ ¼ max i i
 100 : i ¼ 1; 2; 3; . . . ; N
UEP of the system at first time step of real-time simulation without UEPTDS
i

any subsequent data. While the time domain method determines ð25Þ
the condition of the system and generators during the post-fault. where N is number of generators and i = 1, 2, 3, . . ., N.
92 M. Zareian Jahromi, S. Montaser Kouhsari / Electrical Power and Energy Systems 89 (2017) 82–93

300
G1
Clearing Angle = -15.52 o G3
G4
200 G5
Clearing Angle = 6.71 o
G6
G7
100 G8
Rotor Angle [Deg]

G9
G10
0

-100

-200
Clearing Angle= -58.62 o
Clearing Angle =-57.59 o
-300
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time [Se]

Fig. 11. Rotor angle of generators for scenario 3.

300
o
G1
Clearing Angle of G1=-58.2
G3
o
200 Clearing Angle of G3=-10.5
G4
o
Clearing Angle of G4=-19.2 G5
100 o G6
Clearing Angle of G5=22.5
Rotor Angle [Deg]

o
G7
Clearing Angle of G6=23.2
0 G8
o
Clearing Angle of G7=28.1 G9
-100 Clearing Angle of G8=-15.94
o G10
o
Clearing Angle of G9=6.81

-200 Clearing Angle of G10=-59.2

o

-300

-400
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time [Se]

Fig. 12. Rotor angle of generators for scenario 4.

Therefore, simulation results show ability of the PM to fast and 2. The PM calculates the UEP independent from post fault data as a
real-time estimation of UEPs of system along with high accuracy in result of indirect calculation of the potential terms in transient
a real-time simulation. energy function. As it can be seen in Table 3, the PM estimates
Observing Cases 1 and 2, it can be concluded that: UEPs of the system at the first step after the fault is cleared. But
in TDS, takes significant time to detect instability.
1. The PM method estimates the UEP with negligible error with 3. According to Tables 2 and 3, the PM has negligible computa-
very significant speed. The PM calculates high precision cor- tional time compared with TDS method. As a result of indirect
rected kinetic energy which takes AVR and GOVERNOR effects calculation of potential term, the computational burden is
into account. As a result, the estimated UEP have minimum reduced significantly. Also it is obvious that the nature of the
deviation compared with TDS method. As shown in Figs. 11 PM is in a way that the simulation time of the PM is much smal-
and 12, the PM is able to perfectly response to the 0.1 degree ler than TDS and also always below total simulation time (in the
deviation which may cause some generators become unstable case 2: 10 s.) as a result, the PM is technically suitable for real-
in comparison with Figs. 11 and 12. time and online applications.
 M. Zareian Jahromi, S. Montaser Kouhsari / Electrical Power and Energy Systems 89 (2017) 82–93 93

5. Conclusion [9] Chiang H-D, Wu FF, Varaiya PP. A BCU method for direct analysis of power
system transient stability. Power Syst, IEEE Trans 1994;9(3):1194–208.
[10] Chiang H-D, Li H, Tong J, Tada Y. On-line transient stability screening of a
Since the calculation of the potential energy function in multi practical 14,500-bus power system: methodology and evaluations. High
machine large power systems is very complex, estimation of criti- Performance Computing in Power and Energy Systems. Springer; 2013. p.
335–58.
cal angle (UEP) is very difficult and time consuming. Taking the
[11] Roberts L, Champneys A, Bell K, di Bernardo M. An algebraic metric for
power system nonlinearity and network preserving model into parametric stability analysis of power systems. arXiv preprint
account, calculation of UEPs becomes more complicated transient arXiv:1503.07914; 2015.
[12] Yousefian R, Kamalasadan S. A Lyapunov function based optimal hybrid power
stability assessment. This paper presents a novel approach to esti-
system controller for improved transient stability. Electr. Power Syst. Res.
mate the critical angles of the multi machine power systems in 2016;137:6–15.
which there is no need to post the fault data during the calcula- [13] Vu TL, Araifi SA, Elmoursi M, Turitsyn K. Simulation-free estimation of critical
tions. In order to achieve this purpose, the proposed approach uses clearing time. arXiv preprint arXiv:1506.07152; 2015.
[14] Chen L, Min Y, Xu F, Wang K-P. A continuation-based method to compute the
a hybrid method based on new concept of the equal area criterion relevant unstable equilibrium points for power system transient stability
and corrected kinetic energy that takes fault trajectory into analysis. Power Syst, IEEE Trans 2009;24(1):165–72.
account. The proposed method essentially was based on hybrid off- [15] Liu J-H, Chu C-C. Wide-area measurement-based voltage stability indicators by
modified coupled single-port models. Power Syst, IEEE Trans 2014;29
line and online calculations. In the offline calculation utilizing the (2):756–64.
proposed NCEAC, relatively accurate initial estimation of UEP [16] Chu C-C, Chiang H-D. Constructing analytical energy functions for network-
regardless of the fault trajectory was obtained. This initial point preserving power system models. Circ, Syst Signal Process 2005;24(4):363–83.
[17] Zhang R, Xu Y, Dong ZY, Wong KP. Post-disturbance transient stability
is corrected utilizing critical kinetic energy to include the effect assessment of power systems by a self-adaptive intelligent system. Gener,
of AVR and GOVERNOR. In the online calculation, once more by Transm & Distrib, IET 2015;9(3):296–305.
using the proposed NCEAC, relatively accurate initial estimation [18] Paul A, Senroy N. Critical clearing time estimation using synchrophasor data-
based equivalent dynamic model. Gener, Transm & Distrib, IET 2015;9
of UEP considering the fault trajectory was captured. Utilizing off-
(7):609–14.
line calculation, the inaccuracy of the latter UEP is compensated by [19] Savulescu S. Real-time stability in power systems: techniques for early
corrected critical angle obtained from offline calculation. The aim detection of the risk of blackout. Springer; 2014.
[20] Jayasekara B, Annakkkage U. Transient security assessment using multivariate
of the compensation is to include the effects of the AVR and GOV-
polynomial approximation. Electr Pow Syst Res 2007;77(5):704–11.
ERNOR considering the fault trajectory. Simulation results show [21] Amjady N, Banihashemi S. Transient stability prediction of power systems by a
the effectiveness and ability of the proposed method in estimating new synchronism status index and hybrid classifier. Gener, Transm & Distrib,
the critical angles of system (UEP) at first time step of the fault IET 2010;4(4):509–18.
[22] Deng H, Zhao J, Wu X, Men K. Real time transient instability detection based on
duration with the great accuracy and low computational cost. trajectory characteristics and transient energy. In: Power and Energy Society
Since the proposed approach in this paper can estimate the critical General Meeting, 2012 IEEE, 2012 p. 1–7.
angle of each generator of system before clearing the fault while [23] Cepeda JC, Rueda JL, Colomé DG, Echeverría DE. Real-time transient stability
assessment based on centre-of-inertia estimation from phasor measurement
considering all details of the system, it is an efficient, the accurate unit records. Gener, Transm & Distrib, IET 2014;8(8):1363–76.
and simple method with the low computational cost and practical [24] Xu Y, Dong ZY, Meng K, Zhang R, Wong KP. Real-time transient stability
approach for both online and real-time applications. assessment model using extreme learning machine. Gener, Transm & Distrib,
IET 2011;5(3):314–22.
[25] You D, Wang K, Ye L, Wu J, Huang R. Transient stability assessment of power
system using support vector machine with generator combinatorial
References trajectories inputs. Int J Electr Power Energy Syst 2013;44(1):318–25.
[26] Vu TL, Turitsyn K. Lyapunov functions family approach to transient stability
[1] Dou C-X, Yang J, Li X, Gui T, Bi Y. Decentralized coordinated control for large assessment. IEEE Trans Power Syst 2016;31(2):1269–77.
power system based on transient stability assessment. Int J Electr Power [27] Rahmatian M, Chen YC, Palizban A, Moshref A, Dunford WG. Transient stability
Energy Syst 2013;46:153–62. assessment via decision trees and multivariate adaptive regression splines.
[2] Karami A, Esmaili S. Transient stability assessment of power systems described Electr. Power Syst Res 2017;142:320–8.
with detailed models using neural networks. Int J Electr Power Energy Syst [28] Anghel M, Milano F, Papachristodoulou A. Algorithmic construction of
2013;45(1):279–92. Lyapunov functions for power system stability analysis. IEEE Trans Circuits
[3] Sharifian A, Sharifian S. A new power system transient stability assessment Syst I Regul Pap 2013;60(9):2533–46.
method based on Type-2 fuzzy neural network estimation. Int J Electr Power [29] Guruprasada Rau V, Krishnamoorthy N. Construction of improved Lyapunov
Energy Syst 2015;64:71–87. functions and extension of transient stability regions for power systems. Int J
[4] Chiang H-D. Direct methods for stability analysis of electric power systems: Electr Power Energy Syst 1989;11(1):65–9.
theoretical foundation, BCU methodologies, and applications. John Wiley & [30] Fang D-Z, Chung T, Zhang Y, Song W. Transient stability limit conditions
Sons; 2011. analysis using a corrected transient energy function approach. Power Syst,
[5] Wadduwage DP, Wu CQ, Annakkage U. Power system transient stability IEEE Trans 2000;15(2):804–10.
analysis via the concept of Lyapunov Exponents. Electric Power Syst Res [31] Saadat H. Power system analysis. Tata McGraw Hill Publishing Company, New
2013;104:183–92. Delhi; 2002, Bharathidasan Engineering College; 2015.
[6] Verdejo H, Kliemann W, Vargas L. Application of linear stability via Lyapunov [32] Vaahedi E, Mansour Y, Chang A, Corns B, Tse E. Enhanced ‘‘Second Kick”
exponents in high dimensional electrical power systems. Int J Electr Power methods for on-line dynamic security assessment. Power Syst, IEEE Trans
Energy Syst 2015;64:1141–6. 1996;11(4):1976–82.
[7] Pai A. Energy function analysis for power system stability. Springer Science & [33] Anderson P, Fouad A, Happ H. Power system control and stability. Syst, Man
Business Media; 2012. Cybern, IEEE Trans 1979;9(2). 103-103.
[8] Sulistiawati IB, Priyadi A, Qudsi OA, Soeprijanto A, Yorino N. Critical clearing [34] Wahab NIA, Mohamed A, Hussain A. An improved method in transient stability
time prediction within various loads for transient stability assessment by assessment of a power system using probabilistic neural network. J Appl Sci
means of the extreme learning machine method. Int J Electr Power Energy Syst Res 2007;3(11):1267–74.
2016;77:345–52. [35] Saha TK, Mithulananthan N. Test System Report; 2011.