Expert Systems with Applications 37 (2010) 31713176

Contents lists available at ScienceDirect

Expert Systems with Applications

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

Vulnerability control of large scale interconnected power system using

neuro-fuzzy load shedding approach
Ahmed M.A. Haidar a,*, Azah Mohamed b, Aini Hussain b
a
b

Faculty of Electrical and Electronics Engineering, University Malaysia Pahang, Pahang, Malaysia
University Kebangsaan Malaysia, Selangor, Malaysia

a r t i c l e
Keywords:
Vulnerability control
Vulnerability index
Neuro-fuzzy
Load shedding

i n f o

a b s t r a c t
Vulnerability control is becoming an essential requirement for security of power systems in the new utility environment. It is a difcult task for system operator who under economic pressure may be reluctant
to take preventive action against harmful contingencies in order to guarantee providing continued service. For power systems which are operated closer to their stability limits, it is desirable to use load shedding as a form of vulnerability control strategy. This paper presents a neuro-fuzzy approach for
determining the amount of load to be shed in order to avoid a cascading outage. The objective is to
develop fast and accurate load shedding technique to control the vulnerability of power systems by
means of using a neuro-fuzzy controller. A case study is performed on the IEEE 300-bus test system so
as to validate the performance of neuro-fuzzy controller in determining the amount of load shed. Test
results prove that the neuro-fuzzy controller provides accurate and faster vulnerability control action.
2009 Elsevier Ltd. All rights reserved.

1. Introduction
The changing character of power systems with its increasing
dependence on the interconnected transmissions grid has introduced many challenges, of which the most pressing is the need
for signicant improvements in power system operational security
and control. Security of electricity supply networks has always
been a key issue in the development of the power industry. In recent years, the increasing development of supervisory control and
data acquisition systems and energy management systems, the
growing number of market participants and the development of
more complex market schemes have been increasingly reliant on
information technologies and sensing equipment. Power networks
are critically dependent on information and sensing equipment for
system reliability, operation, protection and maintenance. Therefore, in emergency and abnormal conditions, a power system operator has to deal with a large amount of data and apply the most
appropriate remedial control actions (Franco et al., 2001). However, due to emotional and psychological stress, an operator may
not be able to adequately respond to critical conditions and make
correct decisions. Mistakes can damage very expensive power
equipment or worse still lead to the major emergencies and catastrophic situations. Clearly, there is a strong need for automated
corrective procedures that can assist operators in vulnerability
control.
* Corresponding author. Tel.: +60 192704376; fax: +60 94242032.
E-mail address: ahaidar67@yahoo.com (A.M.A. Haidar).
0957-4174/$ - see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.eswa.2009.09.065

Several cascading failures and large area blackouts occurring in

recent years highlighted the need for vulnerability assessment of
power systems. Analysis of recent widespread outages demonstrates system vulnerabilities, such as blackout incidents that occurred in Tokyo in July 1987, Western Interconnection System in
the United States in 1996, the Brazilian power system in March
1999, Northeastern United States and Canada in August 2003 and
the latest blackout in Spain in July 2007. These vulnerabilities were
seen when a sequential series of so-called normal events quickly
diminished acceptable security limits and reliability margins intended to protect against multiple contingencies. Most of such
incidents are believed to be related to lack of vulnerability information about heavily stressed systems where large amounts of real
and reactive powers are transported over long transmission lines
and appropriate power sources are not available to maintain the
system.
Vulnerability control of power systems is very important so that
corrective measures can be taken for preventing deterioration in
service quality. Analysis of recent widespread outages demonstrates that blackouts rarely happened and are usually caused by
a sequence of low-probability disturbance which is generally not
expected by system operators. If fast control actions such as load
shedding and generation rejection are not taken proactively, the
system may cascade and separate into unplanned islands (Miroslav
et al., 2007). For power systems which are operated closer to their
stability limits, it is desirable to use load shedding when there is a
lack of adequate spinning reserve margin and a shortage of tie line
capacity. In the case of power decit in a power system, load

3172

A.M.A. Haidar et al. / Expert Systems with Applications 37 (2010) 31713176

shedding schemes using relays are used to disconnect appropriate

amount of load and maintain system stability.
Different techniques have been proposed to solve the load shedding problem. The under frequency load shedding (UFLS) scheme is
a common practice used by utilities around the world for protection against generation and load mismatch. When the system frequency drops to some predened threshold, the UFLS relays start
to shed the load according to its pre-selected amount. UFLS relays
setting is based on some xed values obtained from off-line simulation results. Some utilities have implemented under voltage load
shedding (UVLS) to protect against voltage instability. This scheme
is generally implemented by using under voltage relays at specic
locations to disconnect load within a specic time, when local voltage drops below a pre-determined threshold. The use of voltage relays to implement UVLS for fast voltage collapse is not reliable
since the relays would need to operate within a few cycles to avoid
uncontrolled loss of load and cascading outages (Jeff, Abraham,
Richard, & Satish, 2004). In recent years, load shedding schemes
have been repackaged using breaker interlocks schemes integrated
with programmable logic controllers (PLC). In the PLC-based load
shedding scheme, load shedding is initiated based on the system
frequency deviations and/or other triggers. The circuit breaker tripping can be programmed based on the system loading, available
generation and other specic logics. However, this scheme is not
viable because it is limited to monitoring sections of a power system that are connected to the data acquisition system and also it
incurs increased hardware cost (Farrokh et al., 2005). Another
method used to solve the steady state load shedding scheme is
by using genetic algorithm in which the scheme is treated as an
optimization problem with the objective of minimizing the difference between the load and generation active and reactive powers
(Sanaye & Davarpanah, 2005).
The conventional load shedding techniques may not work as
desired in emergency conditions due to the complexity and size
of modern power systems. Therefore, alternative methods are required for solving certain difcult power problems where the conventional techniques have not achieved the desired speed and
accuracy. Such techniques are referred to as computational intelligent techniques using fuzzy logic, neural network or expert systems. Fuzzy logic has been applied for safety analysis of power
protection and automation system action (Manana, Toader, & Anatoli, 2004). The fuzzy expert system was proposed for voltage
instability control to calculate the optimum and minimum ratio
of load shedding (Sallam & Khafaga, 2002). In addition, a fuzzy logic stabilizer has been developed for stability control of a 1 kVA
laboratory scale model of power system (Saud, Adel, & Abdullaziz,
2005).
In this research work, an intelligent load shedding scheme is
proposed using neuro-fuzzy controller as a means for vulnerability control of large scaled interconnected power systems. The
neuro-fuzzy controller considers two inputs and one output in
which the inputs are the vulnerability index using power system
loss (PSL) and the bus voltage magnitudes whereas the output is
the amount of load to be shed for each contingency case. The
performance of the proposed neuro-fuzzy controller in load
shedding is investigated by comparing it with the fuzzy logic
controller.

2. Fuzzy logic and neuro-fuzzy

Fuzzy logic has emerged as a promising tool for solving complicated problems dealing with systems whose behavior is very complicated to model. The mathematical modeling of fuzzy concepts
was rst presented by Lot Zadeh in 1965 to describe mathematically, classes of objects that do not have precisely dened criteria

of membership (Hung, Nadipuran, Carol, & Elbert, 2003). Fuzzy set

theory provides an excellent means for representing uncertainty
due to vagueness in the available data or unknown behavior of a
system. It can imitate human control processes closely and provides a means to show an estimate of the system. A user need
not know the actual mathematical relationship between variables.
The attractive features of fuzzy logic control are that it simplies
design complexity, improves control performance and more robust
to imprecision in system parameters and operating condition
changes.
In fuzzy logic, a typical fuzzy inference system (FIS) is used to
map a given input to an output using fuzzy rules and membership
functions which are often chosen arbitrarily. Every FIS consists of
three main components, namely, fuzzication, fuzzy inference
mechanism and defuzzication. Fuzzication is the process of converting crisp statements into fuzzy statements by using membership function which determines how much does a certain
variable belong to a predened class. Commonly, membership
function is described by a simple geometric function such as triangular, trapezoidal and Gaussian function. The fuzzied statements
are evaluated by rules, which include the knowledge of the FIS that
leads to some conclusion. The rules are predened and depend
both on the nature of application and experience of the designer.
Typical rules in the FIS mechanism developed by Mamdani and
Tuskamoto given in Negnevitsky (2002) are in the form of,

IF x1 A1i AND x2 A2i THEN y Bi

Since the output does not depend directly on the input, Takagi
and Sugeno proposed the FIS mechanism as,

IF x1 A1i AND x2 A2i THEN y f x1 ; x2

whereby the output is a function, f which depends on the input. In

case of a constant function, the system is known as zeroth order
system or Sugeno system. When f is a linear function, the system
is of rst order and it is known as TakagiSugeno system. When
applying these systems using constant or linear function, it is possible to use optimization techniques to nd the best parameters
to t data instead of attempting to do it heuristically. The Takagi
SugenoKang method of fuzzy inference is similar to the Mamdani
method in many respects. In fact, in the rst two parts of the fuzzy
inference process, fuzzifying the inputs and applying the fuzzy
operators are exactly the same. Finally, defuzzication is performed
whereby the applied rules are transformed back into crisp consequents. In the literature, there are many different approaches for
defuzzication.
Neuro-fuzzy system is a combination of neural network and
fuzzy logic in which it combines the learning and adapting abilities
of neural networks with the fuzzy interpretation of fuzzy logic system. An example of a neuro-fuzzy system is the adaptive neural
network based fuzzy inference system (ANFIS) which combines
the TakagiSugeno FIS with neural network. The ANFIS denes ve
layers which perform the function of fuzzication of the input values, aggregation of membership degree, evaluation of the bases,
normalization of the aggregated membership degree and evaluation of function output values. A typical ANFIS structure with ve
layers and two inputs, each with two membership functions is
shown in Fig. 1. The ve layers of the ANFIS are connected by
weights. The rst layer is the input layer which receives input data
that are mapped into membership functions so as to determine the
membership of a given input. The second layer of neurons represents association between input and output, by means of fuzzy
rules. In the third layer, the output are normalized and then passed
to the fourth layer. The output data are mapped in the fourth layer
to give output membership function based on the pre-determined
fuzzy rules. The outputs are summed in the fth layer to give a
single valued output. The ANFIS has the constraint that it only

3173

A.M.A. Haidar et al. / Expert Systems with Applications 37 (2010) 31713176

4. Load shedding using fuzzy logic and neuro-fuzzy controllers

Fig. 1. ANFIS Structure with 2 inputs, 1 output, and 2 membership functions for
each input.

supports the Sugeno-type systems of rst or zeroth order. The system can only be designed as a single output system and the system
must be of unity weights for each rule.
3. Load shedding concept
During steady state operation of a power system, the power balance is always maintained. However, such balance of power may
be disturbed by sudden changes in load or loss of generation. If
at any stage, it is found that the current operating state of a power
system is insecure where some or all of the system constraints are
violated, then fast corrective actions need to be taken so as to bring
the system back to a secure operating state. Initially, generators are
re-dispatched optimally in a manner so as to satisfy the generation
security constraints. If re-scheduling of generators fails to provide
a feasible solution for secure operation, load shedding is restored.
It is noted that load shedding is only used as a last resort so as
to avoid a complete system failure (Dash, Mishra, & Panda, 2000).
Assume that a power system is operating in an insecure state
with apparent power loads, Sdi, i = 1,2, . . ., n. After applying load
shedding, the system moves into a secure state with loads,
Ssdi ; i 1; 2; . . . ; n. Thus, the loads after load shedding will become,

Ssdi 6 Sdi

The total system loads after load shedding are also required to
satisfy the following equality constraints,

X
X

Psdi
Q sdi

P Gi  PLoss
X

Q Gi  Q Loss

4a
4b

where, PGi, QGi, active and reactive power generation of ith generator
unit, respectively; PLoss, QLoss, active and reactive total system losses,
respectively.
The objective of solving the load shedding problem is to minimize the difference between the generated power at base case
and the generated power at contingency case. The difference can
be expressed mathematically as:

DS SCC  SBC

PSL

SCCL SIL SLD

Pn

SBCL

i1 SLGO;i W G;i

Pm

i1 SLLO;i W L;i

where, SBCL, apparent power loss at base case; SCCL, apparent power
loss at contingency case; SIL, increase in total load; SLD, amount of
load disconnected; SLGO,i, loss of generated MVA due to generator
outage; SLLO,i, loss of transported MVA due to line outage; WG,i,
weight of individual generator power output; WL,i, weight of individual line power inuence; n, number of generators; m, number
of lines.
The selection of inputs to the controller is an important design
consideration and therefore for power system vulnerability control, the PSL and voltage magnitudes are selected as input variables
for the load shedding controllers. Load shedding based on the proposed fuzzy logic and neuro-fuzzy controllers are implemented
using the Matlab version 7 on an Intel Pentium 2.13 GHz with
2.48 Gb of RAM.
4.1. Load shedding using fuzzy logic controller

where, SCC, generated apparent power at contingency case; SBC, generated apparent power at base case.
The difference between the generated power at contingency
case and the generated power at base case, DS, will give the
amount of load to be shed. By shedding such amount of load, a
power system can be said to remain in a secure state. This statement implies that,

DS Scontroller

Conventional controllers are derived from control theory techniques based on mathematical models of a process. These controllers are characterized with design procedures and usually have
simple structures to yield satisfying results. However, in a number
of cases, when parameter variations take place or when there is no
simple mathematical model, fuzzy logic based control systems
have shown superior performance to those obtained by conventional control algorithms. The proposed controller for load shedding is designed using two different controllers. In the rst
controller, the rules of the FIS are dened based on knowledge of
the power system taking into account engineering judgment. In
the second controller, the FIS is expanded to a neuro-fuzzy system
to dene the rules using neural network without prior knowledge
about the power system.
To implement the proposed load shedding scheme for vulnerability control of power systems, rstly, base case simulation is carried out on a power system so as to analyze the system behavior at
the base case condition. The next step is to analyze the system
behavior when subjected to credible system contingencies such
as line outage (LO), generator outage (GO), load increase (LI) and
disconnection of loads (DL). The power ow simulation outputs
for the contingency cases are used to check the voltage limits
and the transmission line thermal limits. For each contingency,
the vulnerability index based on power system loss (PSL) is then
calculated in which PSL considers total system loss, generation loss
due to generation outage, power line loss due to line outage, increase in total load and amount of load disconnected. The rational
for considering PSL is due to the fact that losses in a power transmission system are a function of not only the system load but also
of the generation (Ahmed, Mohamed, & Hussain, 2007). PSL is then
used as the input in to the load shedding controller and it is given
by,

where, Scontroller is the amount of load shed estimated by a

controller.

The proposed fuzzy logic controller (FLC) for load shedding of a

large scale power system has two inputs and one output. The
power system loss (PSL) and the lowest voltage magnitude (V) in
a power system are considered as the inputs and the amount of
load shed (LS) is considered as the output of the FLC. Here, the inputs of FLC are fuzzied according to its membership functions as
shown in Fig. 2. For each input variable, ve linguistic variables are
dened as more vulnerable (MV), vulnerable (V), more alert (MA),
alert (A) and invulnerable (I). However, for the output variable, LS,
ve linguistic variables are dened as high (H), medium (M), low
(L), very low (VL) and very very low (VVL). The values of the input

1 MV

MA

Degree of membership

A.M.A. Haidar et al. / Expert Systems with Applications 37 (2010) 31713176

Degree of membership

3174

0.5
0
0

0.2

0.4

0.6

0.8

1 MV

MA

0
0.7

0.75

VVL V L
1

0.8
0.85
voltage
L

0
0.1

0.9

0.95

0.4

0.5

0.6

0.7

0.8

0.9

1
0.8
0.6
0.4
0.2
0
0.65

0.7

0.75

0.8

0.85

0.9

0.95

0.5
Fig. 3. Initial membership functions of the input variables.

0
0

10

20
Load Shedding

30

4.2. Load shedding using neuro-fuzzy logic controller

40

linguistic variables are in the range of 01 per unit whereas the

values of the output linguistic variables are in the range of 0
40 MVA p.u. These values can be readjusted and the form of the
membership functions can be changed in order to get satisfactory
results. After assigning the input and output fuzzy sets, mapping
each of the possible ve input fuzzy values of PSL and voltage magnitudes to the ve output fuzzy values are carried out through a
rule base to obtain a fuzzy associative memory as shown in Table 1.
Using the two inputs with ve linguistic variables for each input,
there will be a maximum of 25 decision rules for the FLC. Every entity in the table represents a rule. An example of a rule relating two
inputs and one output dened in the form of a logic, is given as, IF
(PSL is MV) and (V is MV) THEN (LS is H).
The inference mechanism is used to compute the FLC output
membership grades. There are several defuzzication methods,
but probably the most popular one is the centroid technique. It
nds the point where a vertical line would slice the aggregate set
into two equal masses, mathematically, in the centre of gravity.
By means of defuzzication (Negnevitsky, 2002), the nal output
of FLC is the estimated amount of load shed which is given by,

Pn
lsi llsi
SFLC Pi1
n
i1 llsi

where, l(lsi) denotes the output membership grade for the ith rule.

Fuzzy logic system needs rules to be dened rst, but one may
have no knowledge about a power system for the formation of
rules. Therefore, automated parameters tuned by a neural network
embedded inside a fuzzy system can replace the need for prior
knowledge about a power system. In this study, a TakagiSugeno
FIS is adapted to the ANFIS as it is more effective for system identication. For load shedding estimation using neuro-fuzzy logic
controller (NFLC), the ANFIS output is associated with the amount
of load shed in MVA p.u. Similar to the FLC, the inputs to the NFLC
are the PSL and lowest voltage magnitude. Fig. 3 shows the initial
membership functions of the input variables in which the membership function parameters are selected to satisfy the desired output. A multilayer feed forward neural network trained by using the
back propagation algorithm is used to adjust the membership
function parameters according to the inputoutput characteristic
of the training patterns. The neural network computation time depends on the number of rules, which on the other hand depend
euphonically on the number of the membership function and inputs. The parameters associated with the membership function
can change through the training process. The adjustment of these
parameters is facilitated by a gradient vector, which provides a
measure of how fuzzy inference system models the input/output
relations. Once the gradient vector is obtained, any of the optimization routines can be applied to adjust the parameters so as to reduce an error usually dened by the sum of the squared difference
between actual and desired outputs.
The overall output of the ANFIS is the estimated amount of load
shed by NFLC which calculates the sum of outputs of all defuzzication neurons and is given by,

SNFLC

n
X

l i ki0 ki1 x1 ki2 x2

i1

Table 1
Fuzzy decision table.

MV
V
MA
A
I

0.3

Input variable V

Fig. 2. Membership functions for PSL, V and LS.

PSL

0.2

Input variable PSL

0.5

0.65

0.5

Degree of membership

Degree of membership

Degree of membership

PSL

Volt.
MV

MA

H
H
M
M
L

H
M
M
L
L

M
M
L
L
VL

M
L
L
VL
VL

L
L
VL
VVL
VVL

 i,
where, ki0, ki1 and ki2, sets of consequent parameters of rule i; l
normalized ring strength.
After modeling the ANFIS, it is validated with a set of testing
data and checking data set. The reason for using a checking data
set for model validation is to avoid the point in the training process
when the model begins to over t the training data. In principle,
the model error for checking data set tends to decrease during
training until the point when over tting begins. This is when

3175

35

0.5

0
0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Load Shedding

Degree of membership Degree of membership

A.M.A. Haidar et al. / Expert Systems with Applications 37 (2010) 31713176

Input variable PSL

1

30
25
20
15
10
5
1

0.5

0.9

0
0.65

0.8
0.8

0.7

0.75

0.8

0.85

0.9

0.95

voltage

Input variable V

0.7

0.2
0

0.4

0.6

PSL

Output variable LS

Fig. 5. Generated FIS output of FLC.

Training data
ANFIS output

50

5. Simulation results and discussion

0
-50
1

0.9

0.8

0.7

0.2

0.4

0.8

0.6

Input variable PSL

Input variable V

Fig. 4. Final membership functions of the input variables.

the model error for the checking data suddenly increases. When
checking data are presented to the ANFIS model, it is considered
that the model has parameters associated with minimum error.

For load shedding estimation using NFLC, the ANFIS output is

associated with the amount of load shed in MVA p.u. In the ANFIS
implementation, 70% of the data set is used for training, 20% for
checking and 10% for testing. Each ANFIS is trained for 150 iterations because accurate result can be achieved in 150 iterations.
The ANFIS is generated using 15 Gaussian membership functions
for each input. The number of inputs and the number of membership functions determine the number of fuzzy rules and therefore
the training time. After training the ANFIS, the nal membership
functions of the input variables and the output variable which is
the amount of load shed are obtained as shown in Fig. 4.

Table 2
Results of load shedding using NFLC.
Contingency
Case

Amount of Load Shed Estimated by The number of weak buses

NFLC (MVA p.u)

LO-L93
LO-177
LO-182
LO-242
LO-305
TO-393
TO-397
TO-404

9.14
5.56
6.45
2.78
2.73
3.69
3.18
13.64

47,43,44,113
159,157,122,121,120,118,117,115
159,157,122,121,118,117,115
9038,9033,9032,9031
225,224,223,192
9042,9038,9037,9036,9035,9033, 9032,9031
9042,9038,9035,9033,9031
9072,9071,9052,9044,9043,9042, 9041, 9038, 9036,9035,9033, 9032,9031,9004,9003,139

TO-406
GO-84
GO-213
GO-233
GO-236

2.73
7.47
3.85
5.95
17.07

9042,9038,9035,9033,9032,9031,15
9042,9038,9035,9033,9032,9031
9038,9033,9031
9038,9033,9031
9072,9071,9052,9044,9043,9041,9038,9037,9036,9035,9034,9033,9032,9031,9004,9003,562,53,5,47

GO-241

14.68

GO-7001
GO-7002
GO-7061
GO-7017
GO-7024
GO-7166
LI-2.6%

1.88
3.69
3.95
2.73
2.61
3.13
13.40

LI-3.1%

18.24

GO-143,185

24.13*

9042,9071,9052,9044,9043,9042,9041,9038,9037,
9036,9035,9033,9032,9031,9034,9007,9004,9003,52
9038,9033,9032,9031
9042,9038,9037,9036,9035,9033,9032, 9031
9038,9033,9031,7061,61,59,58
9042,9038,9035,9033,9032,9031,9017,17,15
9038,9033,9031
9035,9033,9032,9031
9071,9052,9043,9042,9041,9038,9037,9036,903,
9033,9032,9031,9004
9072,9071,9052,9044,9043,9042,9041,9038,903,
9036,9035,9034,9033,9032,9031,9007,9004,900,52
9072,9071,9052,9044,9043,9042,9041,9038,903, 9036,9035,9034,9033,9032,9031,9007,9006,900,
9004,9003,9001,53,52,51,40,38,37,33, 178
9071,9052,9044,9043,9042,9041,9038,9037,903,
9035,9034,9033,9032,9031,9007,9006,9005,900,
9003,9001,55,3,52,51,48,43,41,40,38,7,33
225,224,223,162
157,122,121,118,117,115

TO-400 & LO-219 & DL-167 20.14

LO-101,305
LO-182,272

5.33
7.35

3176

A.M.A. Haidar et al. / Expert Systems with Applications 37 (2010) 31713176

Absolute Error of NFLC and FLC

0.35

FLC
NFLC

0.3
0.25
0.2
0.15
0.1
0.05
0

50

100

150

about the amount of load to be shed from a system during vulnerable conditions.
The proposed load shedding technique can be implemented in a
practical power system by providing information such as amount
of load to be shed to system operators. In a dispatch center, information from the network is arriving through the SCADA and is
passed to the energy management system (EMS) software. The
EMS should be able to detect potential insecure condition of power
systems. If an alert state is detected, operators expect to get suggestion from the EMS application, namely in terms of the most
adequate control actions to be taken. In this approach, NFLC is
the basic tool used to provide, in a fast way, that information to
the operators about the amount of load to be shed from the system.
In this method it is not necessary to wait for detection of frequency
or voltage drop and disconnect some loads by some delay.

Contingency Cases

6. Conclusion
Fig. 6. The accuracy of NFLC and FLC in load shedding.

Typical fuzzy inference mechanisms developed by Mamdani is

used for the FLC in which the output does not depend directly on
the input. The FIS maps a given input to an output based on the
fuzzy rules and membership functions, which are often chosen
arbitrarily. A more smooth membership function known as the
Gaussian function is used. Here, the center of gravity defuzzication method is applied to convert the fuzzy output into a crisp value in which the output of FLC is the estimated amount of load shed
as shown in Fig. 5.
The results of load shedding using the NFLC for various contingency cases are summarized as shown in Table 2. From the table, it is shown that for each contingency case, the amount of
load to be shed is determined by NFLC in terms of per unit
MVA. An important factor to consider in load shedding is to
determine the location where load is to be shed. Therefore, in
determining the load shedding location, the weak buses are identied by considering buses with low voltage magnitudes in the
range of (0.910.68) per unit. At these weak buses as indicated
in Table 2, it is noted that the voltage magnitudes are violated.
After determining the weak buses, it is considered that load
shedding is to be applied only at these buses. The determination
of the weak buses can assist system operators in determining the
appropriate load shedding location. Referring to Table 2, the critical contingency that causes greater amount of load shed is due
to multiple outages of generators 143 & 185, where the amount
of load shed is 24.13 MVA p.u which is equal to 9.67% of the total
generation.
To evaluate the accuracy of the NFLC in determining the amount
of load to be shed, it is compared with the FLC in terms of absolute
errors as shown in Fig. 6. In this case, the absolute error is the difference between the amounts of load shed determined by NFLC or
FLC and the actual amount of load shed obtained from contingency
and base case simulations. From the gure, it is seen that the accuracy of NFLC is high compared to the accuracy of FLC in which the
calculated average absolute errors of NFLC and FLC are 0.03 and
0.14, respectively. The average absolute error of NFLC is within tolerable limits as compared to that of FLC. Based on these results, it
can be concluded that the NFLC is more accurate than the FLC in
determining the amount of load shed. Hence, the NFLC can be used
as a tool to provide fast information to power system operators

This paper presented the outcome of research from the application of neuro-fuzzy technique for vulnerability control of large
scale interconnected power systems. To counteract the problem
of operating and managing large scale interconnected power systems in vulnerable condition, a new load shedding scheme is
developed by means of using neuro-fuzzy controller to determine
the optimal amount of load to be shed so that a power system
can remain in a secure condition. Existing techniques used for vulnerability control indicate that there is a need to obtain accurate
results using faster computational methods, particularly for real
time applications. The neuro-fuzzy technique proposed in this paper overcome these difculties by providing fast, cost effective and
accurate method which can be implemented for real time applications. The results proved that the neuro-fuzzy controller gives better performance than the fuzzy controller in determining the
amount of load shed.
References
Ahmed, H., Mohamed, A., & Hussain, A. (2007). Vulnerability assessment of a large
sized power system using a new index based on power system loss. European
Journal of Scientic Research, 17(1), 6172.
Dash, K., Mishra, S., & Panda, G. (2000). A radial basis function neural network
controller for UPFC. IEEE Transaction on Power System, 15, 12931299.
Farrokh, S., Dai, J., Shervin, S., Jacques, T., Hugo, C., Tanuj, K., et al. (2005). An
intelligent load shedding (ILS) system application in a large industrial facility. In
Proceeding of IEEE industry application conference.
Franco, C., Bruno, D., Pierangelo, A., Stefano, M., Andrea, M., Federico, S., et al. (2001).
Operation and management of the electric system for industrial plants: an
expert system prototype for load shedding operator assistance. Transaction on
Industry Application, 37(3), 701708.
Hung, T., Nadipuran, P., Carol, W., & Elbert, W. (2003). A rst course in fuzzy and
neural control. Chapman and Hall/CRC.
Jeff, M., Abraham, E., Richard, C., & Satish, R. (2004). Design of an under voltage load
shedding scheme. IEEE Power Engineering Society General Meeting.
Manana, D., Toader, M., & Anatoli, P. (2004). Fuzzy logic in power system
performability. In Second IEEE international conference on intelligent system.
Miroslav, B., Gagnon, J., Gomes, P., Walter, L., Chen-Ching, L., Vahid, M., et al. (2007).
Defense plan against extreme contingencies. CIGRE TF C2.02.24 Summary for
Electra.
Negnevitsky, M. (2002). Articial intelligence a guide to intelligent system (1st ed.).
Addison Wesley.
Sallam, A., & Khafaga, A. (2002). Fuzzy expert system using load shedding for
voltage instability control. In Proceeding of IEEE conference on power engineering.
Sanaye, M., & Davarpanah, M. (2005). A new adaptive multidimensional load
shedding scheme using genetic algorithm. In IEEE proceeding of the CCECE/CCGEI,
Saskatoon.
Saud A., Adel, A., & Abdullaziz, M. (2005). Hardware implementation of a fuzzy logic
stabilizer on a laboratory scale power system. Electric power systems research.
Berlin: Springer.