CHAPTER 2

Literature Review of Load Frequency Control

2.1 Introduction

There are volumes of research articles which have been appeared in the literature
regarding Automatic Generation Control (AGC)/LFC of single area/multi area power
system considering various control strategies. These are summarized comprehensively in
[IBR 06, KUM 96]. This section, specifically gives a brief literature review on LFC of
power system using intelligent control techniques. A number of investigations have been
done in the area of LFC [JAL 92, KUM 96]. The successful operation of interconnected
power systems requires the matching of total generation with total load demand and
associated system losses [MIL 71, NAN 78]. With time, the operating point of a power
system changes, and hence, these systems may experience deviations in nominal system
frequency and scheduled power exchanges to other areas, which may yield undesirable
effects [CON 54, MIL 71].

There are two variables of interest, namely, frequency and tie-line power exchanges
[FOS 70, HAN 02]. Their variations are weighted together by a linear combination to a
single variable called the Area Control Error (ACE) [KWA 75, NAG 05]. The AGC
problem has been augmented with the valuable research contributions from time to time,
like AGC regulator designs incorporating parameter variations/uncertainties, load
characteristics, excitation control, and parallel ac/dc transmission links [MAT 07]. The
microprocessor based AGC regulator, self-tuning regulator, and adaptive AGC regulator
designs have also been presented [MOO 72, PAN 89]. The small signal analysis is justified
for studying the system response for small perturbations. One important factor for reliable
supply is dependent on matching the electricity generation to the load demand and this
factor is dependent on number of power plants. There are different types of power plants
which supply reliable and good quality of electricity to their consumers. Power plants are
dependent on type of energy source. So the researchers are investigating different types of
power plant. Power plants can be thermal, hydro, wind, solar, nuclear and some other type

8
source of electricity generation [HAS 12, KAZ 02]. Researchers incorporated different
types of nonlinearities as in actual system to demonstrate the real problem in the generating
plant [GRE 96, TRI 84].

2.2 Approaches to realistic systems

However, the implementation of AGC strategy based on a linearized model on an

essentially nonlinear system does not necessarily ensure the stability of the system.
Considerable attention has been paid by researchers to consider the system nonlinearities to
develop a robust controller. It is shown in the literature that governor dead-band
nonlinearity tends to produce continuous oscillations in the area frequency and tie-line
power transient response which produces destabilizing effect on the system [TRI 84].

The rst attempt in the area of AGC has been to control the frequency of a power
system via the y wheel governor of the synchronous machine [KUM 98]. This technique
was subsequently found to be insufficient, and a supplementary control was included to the
governor with the help of a signal directly proportional to the frequency deviation plus its
integral [LIA 93, LIM 96, MIL 71, MOO 72, NAG 05]. This scheme constitutes the
classical approach to the AGC of power systems. Very early works in this important area
of AGC have been by Cohnetal [KUM 96]. This work was based on tie-line bias control
strategy. Quazza [QUA 77] illustrated non interactive control considering:

1. Non-interaction between frequency and tie-line powers controls

2. Each control area taking care of its own load variations.
Supplementary controllers were designed to regulate the ACE to zero effectively
[CHA 05, COH 57]. Later on, energy source dynamics were incorporated in AGC
regulator design [MIL 71, NAN 04]. The standard definitions of the terms associated with
the AGC of power systems were nalized in [IEEE 92]. Following that, suggestions for
dynamic modeling for LFC are discussed thoroughly in [KUN 04]. Based on the
experiences with actual implementation of AGC schemes, modifications to the definition
of ACE are suggested from time to time to cope with the changed power system
environment [GEG 95]. Since many presently regulated markets are likely to evolve in to a

9
hybrid scheme, and some deregulated markets are already of this type, the effects of
deregulation of the power industry on LFC have been addressed thoroughly [CHO 98,
NAN 04, SAB 08, SHE 84].

2.2.1 Incorporating system nonlinearities

Later on, the effect of Generation rate constraint (GRC) was included in these types
of studies, considering both continuous and discrete power system models [SCH 96].
Incorporating the dynamics of the energy source in AGC regulator design, Kwatny [Link]
[KWA 75] have proposed an optimal tracking approach to AGC, considering load to be the
output of the dynamic system. The small signal analysis is justified for studying the system
response for small perturbations [GRE 96]. However, the implementation of AGC strategy
based on a linearized model on an essentially nonlinear system does not necessarily ensure
the stability of the system [KUN 07]. Considerable attention has been paid by researchers
to consider the system nonlinearities [CHI 05, DAS 91]. Destabilizing effect of governor
dead-band nonlinearity on conventional the AGC system shown in [IBR 04]. It is shown
that governor dead-band nonlinearity tends to produce continuous oscillations in the area
frequency and tie-line power transient response [YOU 00].

2.3 Control techniques

A lot of control techniques are proposed by the researches in there pioneer work to
design LFC controllers. The controllers are based on:

Classical control techniques

LQR based controlling techniques
Proportional, Derivative, Integral controlling techniques
Soft computing techniques/Artificial intelligence (AI) techniques
Fuzzy logic based techniques
Neural network based techniques
Genetic Algorithm based techniques
Particle Swarm based techniques

10
 Other hybrid techniques

Description of load frequency control techniques are described by different researchers.

2.3.1 Classical control techniques

The pioneering work by a number of control engineers, namely Bode, Nyquist, and
Black, has established links between the frequency response of a control system and its
closed-loop transient performance in the time domain [OGA 02]. The investigations
carried out using classical control approaches reveal that it will result in relatively large
overshoots and transient frequency deviation [CHA 05, KUN 04, MOM 08, WOO 96].
Moreover, the settling time of the system frequency deviation is comparatively long and is
of the order of 1020s [VIN 98].The AGC regulator design techniques using modern
optimal control theory enable the power engineers to design an optimal control system with
respect to given performance criterion [PAN 89]. Fosha and Elgerd [FOSA 70] were the
rst to present their pioneering work on optimal AGC regulator design using this concept.
A two area interconnected power system consisting of two identical power plants of
nonreheat thermal turbines was considered for investigations [CON 54, QUA 77].

Adaptive control is a control methodology of controller to adapt with parameter

variations. Adaptive control strategy is found to be very effective in improving the stability
and transient response of the system. There is a difference between robust control and
adaptive control technique. Robust control technique guarantees that control laws for
parameter variations cant change for a given boundary. While adaptive control law will be
change with changing conditions. Ross proposed an error adaptive computer control
(EACC) technique which gave new and different degrees of freedom for automatic control.
It is a logical computer technique which monitors the control error signal and makes
logical decisions on how much, if any, control action should be taken depending on the
characteristics of the error signal [ROS 66]. Control action may be reduced to any
judiciously chosen minimum by adjusting the EACC to discriminate against certain classes
of disturbance, e.g., statistical, deterministic, and periodic. Fast control action may still be
taken for large and/or sustained disturbances. Tripathi [Link] proposed an adaptive regulator
which is implemented on a microprocessor. They suggested that application of self-tuning

11
control for load-frequency regulators is suitable for alternative integral control, for load-
frequency control [TRI 84]. Vaj [Link] presents an adaptive regulator which uses the a priori
known information and satisfies the multi-objective character of the control of the
Hungarian power system [VAJ 85]. Pan [Link] uses a PI adaptation to satisfy the hyper
stability condition for taking care of the parameter changes of the system. Only the
available information on the states and output are required for the control [PAN 89]. Sho
[Link] proposed a multiarea adaptive load frequency control (LFC) based on the self-tuning
regulator (STR) for a comprehensive automatic generator control simulator (AGCS) [SHO
93]. Liaw [Link] proposed controller uses a signal synthesis adaptation approach such that
the performance of the controlled system is insensitive to the changing load and system
parameter variations [LIA 94]. It is designed such that only the outputs of the controlled
plant are fed back to the controller. A reduced reference model is used to simplify the
design, making the controller easier to implement. Self tuning controllers are also a part of
adaptive control scheme [SHE 84].

The self tuning controllers are optimized the running parameter of the system for
fulfillment of objective function which could be maximized or minimized [YAM 89].
Generally the error is minimized and the efficiency is maximized. Lee [Link] used a control
methodology where extended least squares technique is used for estimation of parameters
[LEE 91]. The area generator corrective control is computed by a self-tuning algorithm
derived from the minimization of a generalized cost function, which is defined in terms of
the system output and the weighting on the control effort.

[Link] LQR based controlling techniques

Optimal control design for the linear systems with quadratic performance (Linear
Quadratic Regulator (LQR) has been established [IBR 09, JHO 96, KAK 09, RAY 01].
The purpose of optimal regulator design is to determine the optimal control rule. There are
several competing objectives that need to be simultaneously satisfied (system step
response, rise time, overshoot, disturbance rejection, or integral absolute error). These
objectives are imbedded in systems Eigen values that are measures of system stability and

12
robustness. [Ray 09] proposed a combination of Matching conditions and Lyapunov
stability theory is adopted to implement a robust stabilizing controller.

Yang [Link] suggested a controller which is based on structured singular values

(SSVs) and each local area load-frequency controller can be designed independently [YAN
96, YAN 02]. The robust stability condition for the overall system can be easily stated to
achieve a sufficient interaction margin and a sufficient gain and phase margin defined in
classical feedback theory during each independent design.

Ko [Link] published that LQR problem needs to be reformulated for finding a

common Lyapunov function for the set of considered linear systems [KO 08]. This is
accomplished by representing the underlying control optimization problem in terms of a
system of linear-matrix-inequality (LMI) constraints. The solution of LMI equations
involves a form of quadratic Lyapunov function that not only gives the stability property of
the controlled system but can also be used for achieving certain performance specifications
[WAN 93].

The H control strategy was used many scientists and researchers. The advantage of
H control technique is to use this technique in multi variable system and also where cross
compiling between the channel exits [AL 07, RAH 98]. Although the H control problem
can be regarded as robustness against exogenous signal uncertainty, in the case when
parameter uncertainty appears in the plant modeling, robust behavior on H performance
as well as stability cannot be guaranteed by standard H control. The necessary and
sufficient conditions for the existence of robust H controller design with circular pole
constraints are derived in terms of a linear matrix inequality (LMI) in order to contruct a
desired controller. Kan [Link] proposed a H load-frequency control for interconnected
power systems with circular pole constraints [KAN 05]. In their study of robust H control
problem whose the objective is to design controllers such that the closed-loop system is
stable and the H norm of a specified closed-loop transfer function is minimized. Cimen
[Link] designed a robust load frequency controller based on H design techniques applied
MIMO system [CIM 00]. Al-Tamirni proposed that H load-frequency controller for a
power system generator will be designed without the knowledge of the system model. The

13
idea is to solve for an action dependent value function of the zero-sum game instead of
solving for the state dependent value function which satisfies a corresponding game
algebraic Riccati equation (GARE). Mandour [Link] proposed an algorithm was the H
control using iteration. This algorithm will use the loop shifting two Riccati formula of
H control [MAN 10].The disadvantage of H approach is that this technique includes
higher level of mathematical analysis and system nonlinearity cannot be well handled.

[Link] Proportional, Derivative, Integral controlling techniques

Among various types of load frequency controller, the PI controller is most widely
used to speed-governing system for LFC scheme [CHA 07, JHO 96, NAN 04]. An
advantage of the PI control technique is to reduce the steady-state error to zero by feeding
the errors in the past forward to the plant [CHA 98, PAT 90, SHO 93].

The most of proposed techniques were based on the classical proportional and
integral (PI) or proportional, integral and derivative (PID) controllers. Its use is not only
for their simplicities, but also due to its success in a large number of industrial applications
[LIM 96, SAB 08, VIN 98]. For PID controller design, overshoot/undershoot and settling
time are used as objective function for multi-objective optimization in LFC problem [KAR
02]. One of the developments in the field of modern control theory is in the direction of its
application in optimal control allows the power engineers to cope with the problems arising
out of due to the complex structure of power systems [SON 97]. The development of
design techniques for load frequency control of a power system in the last few years is very
significant.

AGC regulator designs are based on adaptive control schemes [RUB 94]. The AGC
regulator design techniques using modern control system theory enable the power
engineers to design the optimal control system with respect to given performance criteria.
Most of the researchers have taken the dynamic system equations for the development of
adaptive controller, to cope up the change in process parameters and unmodelled process
dynamics [TAL 99]. Many researchers have carried out research work on adaptive
automatic generation control scheme which has been reported in the literature. The main

14
contribution of the proposed controller is to enhance the controlled performance of the
conventional PID controller by adding a self-tuning on the existing conventional PID
controller. The conventional PID controller with online self-tuning pre compensation
shows superior performance than conventional PID controller [PAN 89].

The emerging techniques of artificial intelligence (AI) have a common feature, i.e.
they had the ability to process the complex information [CHI 96, GAH 94, HOL 75, RER
03]. In AI technique all the short circuit analysis in three-phase is carried out offline, and
the fault is located online within short time. Different AI tools were successfully employed
for all power system purposes such as Expert Systems, Artificial Neural Network (ANN),
Fuzzy Logic and genetic algorithm realizing distinctive performances over the
conventional ones [CHA 09, DAS 03, KAZ 02]. Among the various AI based techniques,
fuzzy logic approach is observed to be applicable and attractive for dealing with complex
and ill-defined problems which may be impossible or too expensive with conventional
methods [CHA 07, CHA 98, CHO 98].

2.3.2 Soft computing techniques/Artificial intelligence (AI) techniques

The emerging techniques of artificial intelligence (AI) have a common feature, i.e.
they had the ability to process the complex information. In AI technique all the short
circuit analysis in three-phase is carried out offline, and the fault is located online within
short time. Different AI tools were successfully applied for all power system purposes such
as Expert Systems, Artificial Neural Network (ANN), Fuzzy Logic and genetic algorithm
realizing distinctive performances over the conventional ones [ZEY 02]. Among the
various AI based techniques, fuzzy logic approach is observed to be applicable and
attractive for dealing with complex and ill-defined problems which may be impossible or
too expensive with conventional methods.

[Link] Fuzzy logic based techniques

A fuzzy logic based intelligent controller is designed to facilitate the smooth

operation and less oscillatory when system is subjected to a sudden load change [ANA 09,
ARA 09, GEG 95]. Fuzzy controller is based on a logical system called fuzzy logic which

15
is much closer in spirit to human thinking and natural language than classical logical
systems [CHA 07, HA 98, IND 95, SIJ 02, XIU 06]. Fuzzy set theory and fuzzy logic
establish the rules of a nonlinear mapping [CHA 09]. The main goal of LFC in
interconnected power systems is to protect the balance between production and
consumption [EI 08, ISM 92]. Because of the complexity and multi-variable conditions of
the power system, conventional control methods may not give satisfactory solutions [MAT
06, MOO 01]. On the other hand, their robustness and reliability make fuzzy controllers
useful in solving a wide range of control problems [MEN 03, NAN 04, TAL 99]. Load
frequency control in two area system using fuzzy logic algorithm is found to be suitable
[SON 97, SON 98, XIU 06]. But the fix fuzzy rule expert systems have some drawbacks as
[CHA 07]:

1. It is difficult to acquire knowledge

2. There is no adaptability and hence for dynamic time varying system, it is unable to
perform well due to change in system.

To overcome these drawbacks, many researchers used adaptive technique with

fuzzy logic to control the dynamics. Talaq proposed an adaptive fuzzy gain scheduling
scheme for conventional PI and optimal load frequency controllers. Comparison
between fixed gain and adaptive fuzzy gain controllers were shown. In this controller a
sugeno type fuzzy inference system is proposed [TAL 99]. This controller performs better
for linear and nonlinear system both. They also discussed that adaptive fuzzy control
methodology also takes lesser operating time because here lesser need of training patterns.
Trabelsi [Link] proposed an adaptive T-S fuzzy logic controller where fuzzy clustering is
based for fuzzy clustering [TRA 04]. Osyal [Link] proposed a Dynamical Fuzzy Network
(DFN) for LFC. A DFN is that which contains dynamical elements such as delayers or
integrators in their processing units is used in the adaptive controller design for load
frequency control [OYS 05]. Adaptation is based on the DFN training that is adjusting
parameters of DFN for load frequency control in power systems with Broyden-Fletcher-
Golfarb-Shanno (BFGS) gradient algorithm. This is done by minimizing the cost

16
functional of load frequency errors. The cost gradients with respect to the network
parameters are calculated by adjoint sensitivity.

EI-Metwelly presented an adaptive fuzzy logic control approach for designing a

decentralized controller for load frequency control of interconnected power areas [EI 08].
The T-S fuzzy-logic configuration is adapted to implement the FLFC. For adaptation EI-
Metwelly choose the variable structure algorithm to implement our estimator since it leads
to a robust performance and an efficient numerical implementation. Pra [Link] published a
fuzzy logic based novel method of quenching transients of load frequency of a single area
power system [PRA 08]. Prakash [Link] investigates the impact of addition of slider gain
with fuzzy logic controller for load frequency control of interconnected thermal-thermal
and thermal-hydro interconnected power systems [PRA 09]. Ramesh [Link] designed a fuzzy
logic controller which is worked for an application of HVDC link to stabilize the
frequency oscillation in a parallel ACDC interconnected power systems [RAM 10].
Ozkop [Link] presented a load frequency control for four area power systems to use fuzzy
gain scheduling of PI controller [OZK 10].

[UMR 10] proposed a new intelligent control technique is based on polar fuzzy
sets. The polar fuzzy sets were first introduced in 1990 by Hadipriono and Sun [CHA 09].
Polar fuzzy sets differ from standard fuzzy sets only in their Polar fuzzy sets are defined on
a universe of angle and hence repeat shapes every 2 radian [CHA 07]. Polar fuzzy logic
controller performs to improve the stability and dynamic performance of the power system
[ORT 95].

[Link] Neural network based techniques

In the last few years, considerable progress has been made in application of ANNs.
This is because, unlike expert systems, neural networks do not relay on a knowledge (Rule)
base, but look for and identify patterns, given appropriate design and training [AHA 02,
CHA 01, ISH 01, KAZ 02, ZEY 02]. Artificial Neural Network was not only applied for
location of faults in transmission line [CHA 99] but also in related field such as, fault
diagnosis in power systems, high impedance fault detection, fault location in sub-station, to

17
mention a few. A brief comparison of various analytical techniques with ANNs for fault
location in transmission system is provided in [IBR 05]. Load frequency control using
artificial neural network is described in [DEM 01], this paper presents a suitable neural
network based solution for load frequency control problem in a deregulated environment.
The proposed control system ensures that the selected generator companies will
automatically track the load changes to keep the system frequency and tie line power
interchanges close to specified values. In the proposed control structure, it is assumed that
each control area has its own generation, distribution and transmission network, which
distribution company is responsible to track areas load frequency of different and
honoring tie line power exchange contracts with neighbors by securing as much
transmission and generation capacity as needed.

Generally, in the all applications, the learning algorithms cause the adjustment of
the connection weights so that the controlled system gives a desired response. Bevrani [Link]
proposed a new approach based on artificial Flexible Neural Networks (FNNs) to design of
load frequency controller for a large scale power system in a deregulated environment. In
this approach, the power system is considered as a collection of separate control areas
under the bilateral Load Frequency Control (LFC) scheme [BEV 06]. In their work, in
order to greatest response and better performance, we have proposed FNN-based load
frequency controller with dynamic neurons that have wide ranges of variation. Sabahi [Link]
were developed a modified dynamic neural network (MDNN) controller in two-area power
system for generating electricity with good quality [SAB 07]. Also, in this paper to achieve
the sensitivity of power systems model, neural network emulator used for identification the
power system. In PI controller, the nonlinearities of the system are not accounted for and
they are incapable of gaining good dynamical performance for a wide range of operating
conditions in a multi-area power system [AHA 06]. A strategy for solving this problem
because of the distributed nature of a multi-area power system is presented by using a
multi-agent reinforcement learning (MARL) approach [DAN 10]. It consists of two agents
in each power area; the estimator agent provides the area control error (ACE) signal based
on the frequency bias (b) estimation and the controller agent uses reinforcement learning to
control the power system in which genetic algorithm optimization is used to tune its

18
parameters. This method does not depend on any knowledge of the system and it admits
considerable exibility in dening the control objective. Also, by nding the ACE signal
based on estimation the LFC performance is improved and by using the MARL parallel,
computation is realized, leading to a high degree of scalability. Sundaram [Link] presented
an Artificial Neural Network (ANN) which is applied to self tune the parameters of PID
controller [SUN 11]. Multi area system had been considered for simulation of the
proposed self tuning ANN based PID controller.

After the evolution of soft computing tools, many researchers are tried to find better
output. So they tried newer techniques that were based on Genetic Algorithms and Particle
Swarm. These are also called optimization techniques.

[Link] Genetic Algorithm based techniques

A newly intelligent control technique is the genetic algorithms based load frequency
controller. Genetic algorithms (GA) are global search techniques, based on the operations observed
in natural selection and genetics [ABD 95, GHO 04, JUA 02]. They operate on a population of
current approximations [AL 00, MAT 02]. The individuals initially drawn at random, from which
improvement is sought. Individuals are encoded as strings (chromosomes) constructed over some
particular alphabet, e.g., the binary alphabet {0, 1}, so that chromosomes values are uniquely
mapped onto the decision variable domain. Once the decision variable domain representation of the
current population is calculated, individual performance is assumed according to the objective
function which characterizes the problem to be solved. It is also possible to use the variable
parameters directly to represent the chromosomes in the GA solution.

At the reproduction stage, a fitness value is derived from the raw individual
performance measure given by the objective function, and used to bias the selection
process. Highly fit individuals will have increasing opportunities to pass on genetically
important material to successive generations. In this way, the genetic algorithms search
from many points in the search space at once and yet continually narrow the focus of the
search to the areas of the observed best performance. The selection individuals are then
modified through the application of genetic operators, in order to obtain the next
generation. Genetic operators manipulate the characters (genes) that constitute the

19
chromosomes directly, following the assumption that certain genes code, on average, for
fitter individuals than other genes. Genetic operators can be divided into three main
categories reproduction, cross over and mutation.

1. Reproduction: Selects the fittest individuals in the current population to be used in

generating the next population.

2. Cross over: Causes pairs, or larger groups of individuals to exchange genetic

information with one another.

3. Mutation: Causes individual genetic representations to be changed according to some

probabilistic rule.

Genetic algorithms are more likely to converge to global optimal than conventional
optimization techniques, since they search from a population of points, and are based on
probabilistic transition rules. Conventional optimization techniques are ordinarily based on
deterministic hill climbing methods, which, by definition, will only find local optima.
Genetic algorithms can also tolerate discontinuities and noisy function evaluations.

In the study of [PRA 08], the optimal values of the parameters 1 (or) 2 and 1

(or) 2 which minimize an array of different performance indices are easily and accurately
computed using a genetic algorithm.

Al-Hamouz proposed the variable structure controller (VSC) feedback gains

selection by GA. The proposed method provides an optimal and systematic way of
feedback gains selection in the VSC compared to trial and error methods. The application
of the proposed method to the LFC problem reveals that not only the system performance
is highly improved but also the control effort is dramatically reduced as compared to
previous methods [AL 00]. Using GA and by proper choice of the performance index to be
evaluated, the designer can achieve a tradeoff between the frequency deviation and the
control effort.

20
 Padhy [Link] select the fitness function on integral of the square of the area control
error (ISACE) and integral of the time multiplied absolute value of the area control error
(ITAACE) have been utilized for his genetic algorithm tool. By selecting one of the
criterions at a time the fitness function is minimized by using the genetic algorithm. By
minimizing the fitness function we get the optimal parameters of either the integral
controller or PID controller [PAD 08].

Nikzad [Link] proposed that the calculation of the performance index for each of the
individuals in the current population, to do this the system must be simulated to obtain the
value of the performance index [NIK 10]. Prasanth [Link] designed GA controller which
consists of two crisp inputs namely deviation of frequency and the other is derivative of
frequency deviation. In this work, they investigate the optimum adjustment of the classical
AGC using GA and performance indices, namely the integral of time multiplied absolute
value of the error (ITAE) [PRA 08].

Shankar [Link] introduced the FACTS based decentralized controller for load
frequency control of two area interconnected hydrothermal power system considering the
combined effect of RFB ( Redox Flow Batteries ) and TCPS ( Thyristor Control Phase
Shifter ) as a FACTS device which is incorporated with tie-line power flow of the system.
The proposed controller is design using genetic algorithm based integral controller in
which Integral Square Error (ISE) criterion is consider for the optimization of the system
error [SHA 12]. The proposed controller gives better transient responses and helps in better
stabilizing frequency response as well as improves the tie line power flow of the system.

Aditya [Link] proposed that Optimum gain settings of different types of controllers
are obtained using the genetic algorithm (GA) for a two area hydro power system. Aditya
minimized an objective function for obtaining the optimum gain settings of integral (I), PI,
ID, and PID controllers using a genetic algorithm. It was found that an integral controller
gives unstable dynamic responses and a PI controller gives highly oscillatory dynamic
responses for two area hydro power systems because of the special characteristics of a
hydro system [ADI 10].

21
 Panda [Link] proposed GA along with decomposition technique as developed has
been used to obtain the optimum megawatt frequency control of multi-area electric energy
systems. Time-domain simulations are conducted with trapezoidal integration along with
decomposition technique. The following two modifications have been proposed

Modification in parent selection

Modification in crossover mechanism

In this proposed algorithm, the following modifications have been proposed with an
intuition to have better trade-off between exploration of unknown solution space and
exploitation of already known knowledge of solution to find the global optimum in less
number of generations. In this work, one point crossover also called Holland crossover is
adopted with a probability Pc [0.6, 0.95] with modifications in exchange of chromosomal
materials [PAN 10].

IBRAHEEM [Link] is used to get the optimal AGC feedback gains with the help of
GA. A penalty function based strategy is used to satisfy transient response specifications of
system frequency and power flow deviations and consequently, area control error (ACE) is
minimized to zero. This proposed AGC scheme is tested on a two-area interconnected
power system consisting of thermal power plants with reheat turbines. The area
interconnections considered are with AC link only and AC link in parallel with DC link.
The response plots achieved with proposed scheme are compared those obtained with
optimal AGC regulators designed using Linear Quadratic Regulator (LQR) concept [IBR
09]. The investigation of the system dynamic responses under load disturbance conditions
reveal that proposed GA based AGC scheme yields appreciably better results as compared
to those obtained with LQR concept based optimal AGC scheme.

Arivoli [Link] investigates the performance of the two-area identical, thermal

reheat systems interconnected with AC-DC tie-lines using Mutual Aid Criterion
(MAC) with Genetic algorithm (GA).The optimum proportional controller feedback
gain Kp is obtained by plotting the cost curve for various values of Kp against the cost
function of area [ARI 11].

22
 Bhongade [Link] has been tuned PID controller according to GA based performance
indices. Their developed model also includes the Superconducting Magnetic Energy
Storage (SMES) units to inject or absorb the active power of an interconnected power
system. The functioning of Genetic Algorithm based PID controller has been tested on a
39-bus New England system and 75-bus Indian power system network [BHO 11]. The
results of controller have been compared with those obtained by using the Least Square
Minimization method.

Mandour [Link] proposed three different robust controllers. The first is based on H
control design in order to obtain robustness against uncertainties. The second controller is a
reduced model to the H controller because the first one is very complex for practical
implementation. Genetic algorithm (GA) is used in the third controller to optimize
proportional integral differential (PID) controller parameters. Results show that the PID
controller has the best performance in damping oscillations, minimizing overshoot by a
great amount and reaching zero value in about 5 seconds [MAN 10].

Juang [Link] proposed that to reduce the fuzzy system design effort and the number
of fuzzy rules, the fuzzy system is designed automatically by genetic algorithms. To
improve the design performance, a new genetic algorithm using elitist strategy combined
with similarity measure on relatives between individuals is proposed. This way, we can
reduce the de-sign effort and find a better fuzzy rule set [JUA 02].

A newly intelligent control technique is the design of a fuzzy system by

evolutionary algorithms has been proposed [JUA 04]. In this work, they apply the idea of
evolutionary fuzzy systems to the LFC problem. During control a fuzzy system issued to
decide adaptively the proper proportional and integral gains of a PI controller according the
area-control error and its change [ADI 03, CLE 02, COO 01].

Du et. Al developed an online algorithm for Fuzzy Logic Controller (FLC) with
Genetic Algorithm (GA).The complex AGC control system s are separated into individual
single-input single-output (SISO) system . This SISO cascade loop models consider the
fast load changes and slow plant utility response as different disturbances. The

23
mathematical model for Multi-Area AGC system used in their paper is a four-area
interconnected power system model with reheat nonlinearity effect of the steam turbine
and upper and lower constraints for generation rate nonlinearity of hydro turbine [Du 06].

Abdennour proposes an adaptive optimal gain scheduling approach to the Load

Frequency Control (LFC) problem. The obtained gains by GA are used to train an
Adaptive Networkbased Fuzzy Inference System (ANFIS) to provide a general mapping
between the operating conditions and the optimal control gains. Both approaches (GA and
ANFIS) are compared. This paper published that the ANFIS gives better result than GA
[ABD 02].

No doubt that the GA gives better results to previous techniques but it has some problems
as:

To find the optimal solution for complex high dimensional, multimodal problems
often requires very expensive fitness function evaluations.
Genetic algorithm does not handle well with complexity. That is, where the number
of elements which are exposed to mutation is large there is often an exponential
increase in search space size.
GA may have a tendency to converge towards local optima or even arbitrary points
rather than the global optimum of the problem.
Operating on dynamic data sets is difficult, as genomes begin to converge early on
towards solutions which may no longer be valid for later data.
GAs cannot effectively solve problems in which the only fitness measure is a single
right/wrong measure (like decision problems), as there is no way to converge on the
solution (no hill to climb).

For specific optimization problems and problem instances, other optimization

algorithms may find better solutions than genetic algorithms (given the same amount of
computation time) as swarm intelligence (e.g.: ant colony optimization, particle swarm
optimization).

24
[Link] Particle Swarm optimization techniques

Many researchers gave different variants of PSO to solve the problem of LFC. To
improve the performance of particle-swarm optimization (PSO) hybrid methods [GHO 04,
KEN 95, KEN 97, QI 06] are proposed. PSO is a relatively new evolutionary computation
technique [ADH 07, ALR 06, ASA 04, KOK 08].

PSO is initialized by a population of random solutions and potential solution is

assigned a randomized velocity [CLE 99, DOR 08, GUA 07]. The potential solutions,
called particles, are then flown through the problem space [HIS 06, HO 05]. Each particle
keeps track of its coordinates in the problem space, which are associated with the best
solution or fitness achieved so far [LIP 03, MA 06, SHI 98]. The fitness value is stored and
it is called as pbest and globaly best value is called gbest. Thus at each time step, the
particle changes its velocity and moves toward its pbest and gbest; this is global version of
PSO [RAN 04, SID 07, VLA 04].

Gozde [Link] and Nikzad et. al are proposed swarm optimization based optimal
proportional-plus-integral (PI) controller. The design is determined an optimization
problem and a novel cost function is derived for increasing the performance of
convergence to the solution [NIK 11]. To optimize the parameters of the cost functions and
the PI-controller, the craziness based particle swarm optimization (CRAZYPSO) algorithm
is used [GOZ 09]. Jayanthi [Link] is proposed a controller for investigating with spinning
reserves such as Superconducting Magnetic Energy Storage device (SMES) and Gas
Turbine units. PI controller gains are tuned using an EAPSO technique to find best
parameter for the tuning of controller [ISM 12, JAY 11, OMA 07, TAH 08].

Gautam [Link] proposed Improved-PSO optimized self-tuning PID controller [BOR

11, DAS 11, GAU 10, GOZ 08, SHA 09, SOU 10]. Improved-PSO framework adopting
a crossover operation scheme to increase exploitation capability of PSO.

To ease the design effort and improve the performance of the controller, design of
the Fuzzy-PI (FPI) controller by hybridizing a genetic algorithm and particle-swarm
optimization, called FPIHGAPSO, is proposed [EBE 01, GHO 03, HAN 02, SWA 08].

25
FPIHGAPSO is based on the hybrid of the genetic algorithm and particle-swarm
optimization [JUA 02, JUA 04, JUA 04, KRI 02, NAK 03, ROY 06, SHA 10, TSA 04]. In
FPIHGAPSO, elites in the population of GAs are enhanced by particle-swarm
optimization and these enhanced elites are selected as parents for cross over and mutation
operations.

[Link] Other hybrid techniques

Some researchers used other optimization technique to control the system

frequency and the inter area tie power as near to the scheduled values as possible. GA is
less sensitive to local minimum as compared to the conventional approach. GA
manipulates the representation of potential solution, rather than the solutions itself. The
premature convergence of GA degrades its efficiency and reduces the search capability. To
overcome this problem a more recent and powerful computational intelligence technique
bacterial foraging (BF) is available in which the number of parameters that are used for
searching the total solution space is much higher compared to those in GA. Social foraging
behavior of Escherichia coli bacteria is used by Nanda [Link] [NAN 09] and Ali [Link] [ALI
10]. The Bacterial Foraging Optimization Algorithm (BFOA), as it is called now, is
currently gaining popularity in the community of researchers, for its effectiveness in
solving certain difficult real world optimization problems. The foraging strategy of
Escherichia coli bacteria present in human intestine can be explained by four processes,
namely chemotaxis, swarming, reproduction, and elimination dispersal. BFOA due to its
unique dispersal and elimination technique can find favorable regions when the population
involved is small. These unique features of the algorithms overcome the premature
convergence problem and enhance the search capability. Hence, it is suitable optimization
tool for power system controllers. In [NAN 09], Comparison of convergence
characteristics of BF, GA, and classical approach revealed that the BF algorithm is quite
faster in optimization, leading to reduction in computational burden and giving rise to
minimal computer resource utilization.

Pothiya [Link] focused on a new optimization technique of a fuzzy logic based

proportional integral (FLPI) load frequency controller by the multiple Tabu search (MTS)

26
algorithm [POT 06]. The Tabu search (TS) algorithm is an iterative search that starts from
some initial feasible solution and attempts to determine a better solution in the manner of a
hill-climbing algorithm. The TS algorithm has a flexible memory in which to maintain the
information about the past step of the search and uses it to create and exploit the better
solutions. The main two components of the TS algorithm are the Tabu list (TL) restrictions
and the aspiration criterion (AC). The proposed technique for designing a FLPI controller
helps us save time when compared to those from conventional trial and error design
procedures. Another benefit of this approach is that it does not require experts for the
design of the fuzzy logic controller.

Vrdoljak [Link] presented a new discrete-time sliding mode controller for load-
frequency control (LFC) in control areas (CAs) of a power system. Design of the discrete-
time sliding mode controller for LFC with desired behavior is accomplished by using a
genetic algorithm. GA used for the purpose of finding optimal sliding mode algorithm
parameters [VRD 10]. Dong [Link] presented a new approach which is based on active
disturbance rejection control (ADRC). Estimating and mitigating the total effect of various
uncertainties in real time [DON 10]. Ganpathy [Link] proposed a new approach to the design
of decentralized controllers, using Multi-Objective Evolutionary Algorithm (MOEA), for
load frequency control of interconnected power systems with AC-DC parallel tie-lines.
Steady state Multi-Objective Evolutionary Algorithm is based on -dominance concept
[GAN 10].

Chatterjee published a concept of Dual mode control is applied in the PI

controller, such that the proportional mode is made active when the rate of change of the
error is sufficiently larger than a specified limit otherwise switched to the integral mode.
A digital simulation is used in conjunction with the Hooke-Jeeves optimization technique
to determine the optimum parameters of the PI controller [CHA 11]. Their design method
for a robust controller was based on Quantitative Feedback Theory (QFT). In QFT one of
the main objectives is to design a simple, low order controller with minimum bandwidth.
To indicate the effectiveness of proposed technique, this method has been compared with a
PI controller optimized by Genetic Algorithm or Particle Swarm Optimization [SAK 11].

27
 Paramasivam [Link] developed a sophisticated application of Redox Flow
Batteries (RFB) coordinated with Unified Power Flow Controller (UPFC) for the
improvement of Load Frequency Control of a multi- unit multi- area power system.
The UPFC offers an effective means to enhance improvement in the power transfer
capability of the tie-line. The main application of UPFC is to stabilize the frequency
oscillations of the inter-area mode in the interconnected power system by the
dynamic control of tie-line power flow. The Artificial Bee Colony algorithm is used to
optimize the parameters of UPFC and the cost function of the two area power
system along with the integral controller [PAR 11]. Soheilirad [Link] proposed a new
evolutionary computing method based on imperialist competitive algorithm (ICA), which
was formulated by Gargari and Lucas, is used for tuning the parameters of a PID controller
[SOH 12].

Optimal gain tuning of PI controllers for various case studies for the LFC
problem is proposed and obtained using Artificial Bee Colony (ABC) algorithm by [JAY
12]. The Artificial Bee Colony algorithm which was introduced in 2005 by Karaboga,
is used as an optimization search simulates the intelligent foraging behavior of a honey
bee swarm. Compared with the usual algorithms, the major advantage of ABC
algorithm lays in that it conducts both global search and local search in each
iteration and as a result the probability of finding the optimal parameters is
significantly increased, which efficiently avoid local optimum to a large extent. Jayanthi
[Link] were tested this controller with SMES units. Superconducting energy storage
systems (SMES) represent a fascinating prospective FACTS technology as they can
generate / absorb active and reactive power in rapid response to power system
requirements.

*******

28