Electrical Power and Energy Systems 73 (2015) 787–804

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

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

Automatic generation control of power system using a novel

quasi-oppositional harmony search algorithm
Chandan Kumar Shiva, G. Shankar, V. Mukherjee ⇑
Department of Electrical Engineering, Indian School of Mines, Dhanbad, Jharkhand, India

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

Article history: The present work approaches a novel quasi-oppositional harmony search (QOHS) algorithm, as an opti-
Received 27 May 2014 mization technique, for its optimum performance in the subject area of automatic generation control
Received in revised form 25 May 2015 (AGC) of power system. The proposed QOHS algorithm is applied with an aim to converge rapidly towards
Accepted 26 May 2015
the optimal solution(s) that houses both the characters of two guesses, i.e. opposite-point and
quasi-opposite point. The area of concern of this study is to discuss the multi-objective problems of an
interconnected power system for the benefits of AGC. The proposed QOHS algorithm is, individually,
Keywords:
applied to single-area, precede to two-area considering the non-linearity effects of governor dead band
Automatic generation control
Generation rate constraint
and generation rate constraint and, finally, extended to four-area power system showing the conse-
Governor dead band quences of multiple load disturbances. A case of robustness and stability analysis are also investigated
Interconnected power system for the studied two-area power system model. The control strategy, for the dynamic power system
Optimization model, is based on area control error. The simplicity of the structure and acceptability responses of the
Proportional–integral–derivative controller well-known proportional–integral–derivative controller enforces to implement as a controller in this
work. The comparative evaluation of the proposed QOHS algorithm is carried out by the way of compar-
ing the dynamic performances of the studied power system model with those offered by other algorithms
reported in the recent state-of-the-art literature. The simulation works, presented in the paper, reveal
that the proposed QOHS algorithm may be effectively utilized for the purpose of AGC study of power sys-
tem having varying degrees of complexities and non-linearities. Moreover, the proposed QOHS based
control strategy adopted in this work provides a robust and stable speed control mechanism.
Ó 2015 Elsevier Ltd. All rights reserved.

Introduction time-variant characteristics of power system, (b) an uncertainty

General
In power system dynamics, the control of generation and fre-
quency is the most important issue for its reliable operation [1].
The term frequency of a power system is dependent on active
power balance and, hence, a change in active power at one point
reflects the system operation at some other points. A considerable
drop in frequency may result in high magnetizing currents in ac
motors and transformers [2]. The other problem for interconnected
power system is power exchange among control areas which may
result in frequency deviation from its nominal one. Automatic gen-
eration control (AGC) performs a continuous real-time task to
adjust the power generation economically with active focus on fre-
quency control. The main issues regarding AGC are (a) the
⇑ Corresponding author. Tel.: +91 0326 2235644; fax: +91 0326 2296563.
E-mail addresses: chandankumarshiva@gmail.com (C.K. Shiva), gauri1983@g-
mail.com (G. Shankar), vivek_agamani@yahoo.com (V. Mukherjee).
in increase of load demand during the whole day and (c) genera-
tion losses which may change in operating point of power system
[3].
Literature review
An initial attempt in the area of AGC subject has been to control
the frequency deviation of power system via a flywheel governor of
synchronous machine [4]. This technique was not sufficient to con-
trol the frequency deviation at the instant application of load
demand. For the first time, it was apprehended that a supplemen-
tary action (i.e. load frequency controller) must be needed for
stable speed control mechanism. This concept constitutes the clas-
sical approaches to the AGC of power system. An early work with
the supplementary controller has been suggested by Cohn [5]
which was based on tie-line bias control strategy. The first classi-
fied optimal control theory has been initiated by Fosha and
Elgard [6]. In their work, they have proposed classical linear

http://dx.doi.org/10.1016/j.ijepes.2015.05.048
0142-0615/Ó 2015 Elsevier Ltd. All rights reserved.
788 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804
Nomenclature
ACE area control error
Bi frequency bias constant of ith area (p.u.MW/Hz)
c portion (percentage) of power generated by the reheat
process in the total generated power
Di frequency dependency parameter of ith area
(p.u.MW/Hz)
Hi inertia constant of ith area (s)
i subscript referred to ith area
K pi power system gain constant of ith area (Hz/p.u.MW)
K ri reheat gain constant of ith area
Ri speed regulation parameter of ith area (Hz/p.u.MW)
control theory based on state variables and developed different
feedback theories for the design of AGC regulator.
Over the years, a number of evolutionary computational
intelligence-based techniques have been developed as well as
employed in the AGC study of power system. Pataya and Vichit
[7] have proposed genetic algorithm (GA) based intelligent con-
troller for AGC design strategy. Ghoshal [8] has proposed
GA/GA-SA based fuzzy AGC scheme for multi-area thermal gener-
ating system. Denna et al. [9] have presented an automatic defini-
tion of fuzzy rules based on Tabu search algorithm (TSA) for AGC
design. The concepts based on reinforcement learning have been
suggested by Imthias et al. [10] for AGC mechanism as a stochastic
multistage decision problem. In [11], particle swarm optimization
(PSO) algorithm has been proposed for optimization of proportiona
l-integral-derivative (PID) gains in a fuzzy based AGC system. An
application of TSA has been presented by Pathiya et al. in [12] for
AGC design of a two-area power system. The design of
multi-objective PID controller, optimized by PSO for AGC study of
power system, may be found in [13]. The utilization of PSO based
multi-stage fuzzy controller has been incorporated in [14] for the
AGC study. A modified GA based proportional-integral (PI) con-
troller has been appeared in [15] for interconnected power system
including the effects of generation rate constraint (GRC) and gover-
nor dead band. Nanda et al. [16] have proposed bacterial foraging
optimization algorithm (BFOA) for multi-area AGC system. The PI
and integral controller gains have been tuned by hybrid PSO for
deregulated power system in [17]. In [18], BFOA has been adopted
for the tuning of PI controller gains for the AGC study of two-area
power system. A novel gain scheduling control strategy, based on
craziness based PSO (CRAZYPSO), have been adopted by Gozde
and Taplamacioglu in [19] for AGC application. Gozde et al. [20]
have proposed artificial bee colony algorithm in AGC study of
reheat thermal power system. Shabani et al. [21] have adopted
imperialist competitive algorithm for the design of robust PID con-
troller for resistance to continuous load disturbances. An applica-
tion of non-dominating sorting GA for AGC study of multi-area
power system may be found in [22]. Panda et al. [23] have
employed BFOA-PSO for the AGC study of linear and non-linear
interconnected power system.
Literature survey reveals that a number of evolutionary opti-
mization techniques are being employed by the researcher pools
in the conventional vertical integrated utility model to study its
importance in AGC performance. These optimization techniques
have shown substantial improvement in AGC performance of
power system leaving behind some drawbacks those have been
further rectified by some other applied techniques. The possibility
of exploring some new hybrid optimization techniques is still pre-
vailing for the benefits of AGC performance, especially, by the way
of optimal tuning of controller gains. Therefore, it is justified that
T gi governor time constant of ith area (s)
T pi power system time constant of ith area (s)
T ri reheat time constant of ith area (s)
T ti turbine time constant of ith area (s)
Df i incremental change in frequency of ith area (Hz)
DP tiei incremental change in tie-line power flow of ith area
(p.u.MW)
DP di incremental change in load demand of ith area
(p.u.MW)
optimization techniques may be, significantly, explored in AGC
domain and might become an integral part for the design of AGC
controller.
Motivation behind the present work
Most of the solutions, surfaced in the literature so far for AGC
mechanism, have shown improved results. However, it does not
mean that these methods have not any limitations. During the evo-
lution of some new optimization techniques, a number of limita-
tions and problems have been found. The identified deficiencies
in GA are (a) the involvement of lots of crossover and mutation
operation in each iteration cycle, (b) taking more overall execution
time and (c) exhibiting premature convergence, i.e. trapping into
local minima [8,15]. PSO is developed through the simulation of
bird flocking in multi-dimensional search space. Empirical studies,
performed on PSO, indicate that even when the maximum velocity
and acceleration constant are correctly defined, the particles may
still diverge, i.e. go to infinity; a phenomena known as ‘‘explosion’’
of the swarm. Relay based identification technique needs extensive
mathematical calculations in order to obtain the optimal solution.
The final solution is approximated value rather than true optimal
one. The basis of calculation of this method is supported by
Laurent series expansion. BFOA is based on chemotactic movement
of virtual bacterium models, i.e. instituted by a trial solution of the
optimization problem. During the process of chemotactic, the per-
formance of BFOA depends on random search direction that may
lead to delay in reaching the global solutions. Also, the numbers
of BFOA parameters, as used for searching the total solution space,
are higher than GA and, hence, the possibility of trapping into local
minima is higher than GA.
However, harmony search (HS) algorithm (HSA) is a
derivative-free real-parameter optimization method [24]. HSA is
one of the most recent population based optimization method that
may be easily adopted for solving various kinds of engineering
optimization problems [25]. It generates a new vector after consid-
ering all of the existing vectors whereas GA considers only the two
parent vectors.
A few modified variants of HSA have been also proposed for
enhancing its solution accuracy and convergence rate. Mahdavi
et al. [26] have presented an improved HSA by introducing a strat-
egy to, dynamically, tune the key parameters. Omran and Mahdavi
[27] have proposed a global best HSA by borrowing the concept
from the swarm intelligence. Pan et al. [28] have proposed a
self-adaptive global best HSA for solving continuous optimization
problems. Banerjee et al. [29] have proposed opposition-based
HSA for engineering optimization problem with special emphasis
on reactive power compensation of an autonomous hybrid power
system model.
 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804
The present prospect explores the significances of HSA as an
optimization technique. This study is further extended to its uti-
lization in AGC domain as an optimization problem. Therefore,
the exploratory behaviors of HSA that makes it suitable in AGC
domain are as follows [24,25]. HSA
(a) imposes a unique dispersal and elimination technique that
finds fewer mathematical calculations, fast convergence
superiority and identifies the high performance region of
solution space,
(b) uses stochastic random search and, thus, any derivative
information is unnecessary and
(c) creates a new solution vector after considering all of the
existing solution vectors.
The present work introduces the concept of quasi-oppositional
based learning (QOBL) and its utilization in basic HSA with an
intention to increase its convergence speed in order to provide glo-
bal optimal solution. The idea of QOBL is blended with the basic
HSA and this new hybrid variant is classified as
quasi-oppositional HS (QOHS) algorithm. Therefore, this modified
variant of HSA (i.e. the proposed QOHS algorithm) is applied for
the optimization of PID controller gains of the studied power sys-
tem model. The significances of the proposed QOHS algorithm
are established by showing its consequences with the application
of step load perturbation (SLP). These load deviations affect the
system dynamic responses and, hence, distract the AGC perfor-
mance towards the optimal solution. The proposed QOHS algo-
rithm is applied to single-, two- and four-area power system
model to prove the scalability of this algorithm. The design prob-
lem of PID controller is formulated as an optimization problem
having time domain objective function in which the deviation in
frequency and tie line powers are involved.
Contribution of the present work
The main contributions of this paper are to:
(a) model the AGC dynamics of single-, two- and four-area
interconnected power system,
(b) explore a novel QOHS algorithm, as an optimizing tool, for
the designed power system model,
(c) study the effects of governor dead band and GRC for
two-area non-reheat power system,
(d) discuss the stability and robustness of two-area power
system,
(e) study the effects of multiple SLP at different time horizon for
four-area hydro-thermal power system,
(f) compare the system dynamic responses offered by the pro-
posed QOHS algorithm to those yielded by other algorithms
reported in the recent state-of-the-art literature and
(g) establish the optimizing capability of the proposed QOHS
algorithm as an optimization technique.
Organization of the paper
The rest of the paper is documented in the following headings.
The next section describes the dynamics of the designed AGC mod-
els of power system. In section ‘Mathematical problem formula-
tion’, mathematical problem of the present optimization task is
formulated. A brief description of the proposed QOHS algorithm
is presented in section ‘Proposed QOHS algorithm’. Simulation
based results of the present work are presented and discussed in
section ‘Simulated results and analysis’. Finally, the concluding
remarks are summarized in section ‘Conclusion’.
789
Dynamics of power system model
Design of PID controller
The PID controller is one of the most versatile controller which
is used in industry as a feedback loop controller. The favorable con-
ditions behind its selection in industry houses are its (a) simplicity
in structure and (b) reliability in operation. It has three modes of
operation those act simultaneously. A proportional controller has
the effect of reducing the rise time and makes the system
responses fast but, never be capable of reducing the steady state
error to zero. An integral controller, on the other hand, has the
effect of reducing the steady state error to zero but, at the same
time, distort the dynamic responses. A derivative controller
increases the stability, reduces the oscillation and improves the
dynamic responses at the cost of increase in noise and sensitivity.
The purpose of installing PID controller is to improve the
dynamic performances of the AGC system. From Fig. 1, it may be
noted that the proper design of PID controller requires determina-
tion of optimal values of its three parameters viz. proportional (K p ),
integral (K i ) and derivative (K d ) gains, connected in parallel.
Mathematical modeling
The dynamics of power system model suggest that power sys-
tem elements are non-linear in behavior and, hence, the actual
model of the power system may not be generated. However, the
AGC study is always associated with small change in load. Hence,
the power system model may be represented by a linear model.
For the AGC study, the power system model may be transformed
as a multivariable system using the state space matrix form, as pre-
sented in (1) and (2)
x_ ¼ Ax þ Bu þ Ld ð1Þ
y ¼ Cx ð2Þ
where x, u and d are the state, control and disturbance vectors,
respectively; A, B, C and L are matrices of real constant and y is
the output.
Fig. 2 represents the actual operation of AGC system. This figure
shows that the AGC mechanism is based on minimization of area
control error (ACE) involving both the frequency and tie-line power
deviation terms. In Fig. 2, f i and f 0i represent the generated and
nominal frequency, respectively, whereas DP ti and DP t0i represent
change in generated turbine power and nominal turbine power
outputs, in order.
AGC design of single-area reheats power system with and without
droop characteristic
The plant for a power system, generally, consists of governor
ðGg ðsÞÞ, turbine ðGt ðsÞÞ and power system ðGp ðsÞÞ block. The studied
Fig. 1. Structure of PID controller.
790 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804

Control strategy for single-area reheats power system

The AGC control strategy for isolated single-area power system,
mainly, depends on deviation in frequency. As shown in Fig. 3, the
frequency deviation ðDf Þ acts as the input signal of the feedback
control loop. The expression for ACE may be stated by (8).
ACE ¼ Df ð8Þ
In context of (8), the amount of control action taken by con-
troller ðuÞ may be stated by (9).
Z
d
u ¼ K p  ðACEÞ þ K i ðACEÞ þ K d ðACEÞ ð9Þ
dt

AGC model of two-area non-reheat power system

Fig. 2. Control structure for AGC of ith area.

The studied two-area power system with non-reheat turbine is
shown in Fig. 4 [23]. The difference between AGC of multi-area and
AGC model of an isolated single-area reheat thermal power system that of single-area is that not only the restoration of frequency of
is depicted in Fig. 3 [30]. The turbine model is a reheat type of sec- each area but also the net power interchange via tie-lines may also
ond order because steam has to pass through high pressure stage be restored to nominal values. So, a composite measure (called as
to low pressure stage with and without speed regulating parame- ACE) is taken as the variable of interest.
ter. The designed models for single-area reheat power system are
(a) reheat turbine without drop (RTWD) and (b) reheat turbine Control strategy for two-area non-reheat power system
with drop (RTD) characteristic. The transfer function model for The control configuration for AGC system depends upon the
Gg ðsÞ; Gt ðsÞ and Gp ðsÞ may be expressed by 3–5, in order. ACE term of both the areas. A good designing of AGC requires
ACE tends to be zero within the shortest period of time with min-
1 imum effect on damping. The expression for ACE of both the areas
Gg ðsÞ ¼ ð3Þ
1 þ sT g (area-1 and area-2) may be expressed by (10) and (11).
1 þ sT r c ACE1 ¼ B1 Df 1 þ DPtie ð10Þ
Gt ðsÞ ¼ ð4Þ
ð1 þ sT r Þ ð1 þ sT t Þ
ACE2 ¼ B2 Df 2  DPtie ð11Þ
Kp
Gp ðsÞ ¼ ð5Þ Corresponding to each controller input (ACE), the outputs (u1
1 þ sT p
and u2 ) may be stated by (12) and (13), respectively.
As shown in 3–5, the open loop transfer function (OLTF) with- Z
out droop characteristic ðPðsÞÞ may be stated by (6). d
u1 ¼ K p1  ðACE1 Þ þ K i1 ðACE1 Þ þ K d1
ðACE1 Þ ð12Þ
    dt
1 1 þ sT r c Kp Z
PðsÞ ¼ Gg ðsÞGt ðsÞGp ðsÞ ¼ d
1 þ sT g ð1 þ sT r Þð1 þ sT t Þ 1 þ sT p u2 ¼ K p2  ðACE2 Þ þ K i2 ðACE2 Þ þ K d2 ðACE2 Þ ð13Þ
dt
ð6Þ

Similarly, the OLTF with droop characteristic ðPðsÞÞ may be sta-
ted by (7).
Gg ðsÞGt ðsÞGp ðsÞ
PðsÞ ¼
1 þ Gg ðsÞGt ðsÞGp ðsÞ=R
K p ð1 þ sT r cÞ
¼ ð7Þ
ð1 þ sT g Þ ð1 þ sT t Þ ð1 þ sT r Þ ð1 þ sT p Þ þ K p ð1 þ sT r cÞ=R
It may be noted that T g , T t and T p play important roles in the
dynamics of governor, turbine and power system model. Also,
PðsÞ has non-oscillatory step response for all kinds of turbines
due to feedback control loop while the step response of PðsÞ are,
generally, unstable for hydro turbines.
AGC model of four-area hydro-thermal power system
In the present work, the tuning of PID controller is extended to
four-area hydro-thermal power system [31]. The simplified dia-
gram of four-area power system is shown in Fig. 5. The configura-
tion of its ith control area for AGC study is shown in Fig. 6. The
transfer function model for governor, reheat turbine and power
system dynamics are expressed in 3–5, in order. As shown in
Fig. 5, area-1, 2, 3 are identical reheat thermal power system hav-
ing OLTF stated in (6) while area-4 is a hydro turbine power
system.
The expression for OLTF of hydro-turbine power system with-
out droop characteristic may be stated by (14) [31].

Fig. 3. AGC model of isolated single-area reheat thermal power system.

 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804 791

Fig. 4. AGC model of two-area non-reheat thermal power system.

lead compensating device is cascaded to PID controller in order to

improve the system dynamic performances. The transfer function
model for lead compensator device may be stated by (16) [31].
1 þ 3:8787s
LðsÞ ¼ ð16Þ
1 þ 0:03919s
For ith area, the expression for ACE may be stated by (17).

ACEi ¼ Bi Df i þ DPtiei ð17Þ

According to (17), the controlled signal, i.e. output of controller
Fig. 5. A simplified diagram of four-area hydro-thermal power system [31]. ðui Þ may be expressed by (18)
ui ¼ K i ðsÞ  ðACEi Þ ð18Þ
GðsÞ ¼ Gg ðsÞGt ðsÞGp ðsÞ where K i ðsÞ is the installed PID controller in ith area, may be stated
    
1 1 þ 0:513s 1s 80 by (19) [32].
¼ ð14Þ
1 þ 48:7s 1 þ 10s 1 þ 0:5s 1 þ 13s K ii
K i ðsÞ ¼ K pi þ þ s K di ð19Þ
Similarly, the expression for OLTF of hydro-turbine power sys- s
tem with droop characteristic may be stated by (15) [31].
Mathematical problem formulation
0:00551 s2  0:005231 s þ 0:01074
GðsÞ ¼ ð15Þ Design of objective function
s4 þ 2:179 s3 þ 0:4008 s2 þ 0:01768 s þ 0:01085
The eigenvalues calculated from (15), suggest that the two An attempt is made in this paper to design the optimal QOHS
eigenvalues (0.0294 ± i0.143) with damping ratio ð0:202Þ makes based PID controller for the studied test systems. Therefore, for
hydro power system unstable. To make this system stable, a phase the design and performance evaluation of any controller, one or

Fig. 6. Block diagram of ith control area of multi-area power system [31].
792 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804
more objective functions may be chosen as the performance crite-
ria for the assessment of the utilized optimization technique. In
this regard, a specific objective cost function (named figure of
demerit (FOD)) is used as the fitness function for evaluating the fit-
ness of the proposed QOHS algorithm. This objective function
directly depends on transient performance characteristics such as
overshoots, undershoots, settling times and time derivative of the
frequency. In this paper, the proposed QOHS algorithm is tested
and verified on the basis of integral of square error (ISE) criteria.
The value of FOD, as employed in the present work, may be calcu-
lated by (20)
Z ts
ISE ¼ ðACE2i Þdt
0 ation [35]. Such an efficient search, for a perfect state of harmony,
where t s is the time range of simulation (in s).
Constrained QOHS algorithm
The constraints of the present optimization task are the gains of
the controllers which must be within some pre-specified limits.
These limits may be framed by (21)
9 then, harmony memory (HM) is initialized by filling it up with
K min
pi 6 K pi 6 K max
pi ; i ¼ 1; 2; 3; 4 >
>
= random solutions (each harmony is calculated using the evalu-
K min
ii 6 K ii 6 K max
ii ; i ¼ 1; 2; 3; 4 >
>
; Step (b) Harmony improvisation (creation of a new solution): Each
K min
di 6 K di 6 K max
di ; i ¼ 1; 2; 3; 4
where the superscripts min and max represent the minimum and
the maximum values of the respective variables.
Measure of performance
The present study is focusing on optimal tuning of controller
gains so that the power system dynamic responses should neither
be too fast nor be too slow. In this work, to study the dynamic
behavior of the system, mainly, rise time (T r ), settling time (T s ),
peak value (Mp ) and peak time (T p ) are calculated at the end of
the developed program. The values of these transient parameters
yield information about the speed of dynamic responses. This
information may help a designer to determine the speed and nat-
ure of the response does/does not degrade the performance of the
system.
In addition to the objective function, two more performance
indices, i.e. integral of time absolute error (ITAE) and integral of
time square error (ITSE) are also calculated with the help of opti-
mized PID gains as obtained by the proposed QOHS algorithm.
These two performance indices may be stated by (22) and (23),
respectively.
Z ts
ITAE ¼ jACEi j t dt
0
Z ts
ITSE ¼ ðACE2i Þ t dt
0
Proposed QOHS algorithm
HSA is a new derivative-free real-parameter optimization algo-
rithm which draws inspiration from the musical improvisation
process of searching for a perfect state of harmony (solution)
[24,33]. HSA imposes a fewer mathematical calculations, identifies
the high performance regions of the solution space and may be
applied in various aspects of optimization problems [25,34]. HSA
and its variants have been applied to a wide arena of real-life opti-
mization problems such as scheduling of multiple dam system,
load dispatch, ecological conservation, industrial operation, and
musical composition.
Basic HSA
As visualized in HSA, each musician corresponds to each deci-
sion variable; musical instrument’s pitch range corresponds to
decision variable’s range; musical harmony (at certain time)
responds to the solution vector at certain iteration and audience’s
corresponds to objective function. As musical harmony is impro-
ð20Þ vised time to time, solution vector gets improved iteration by iter-
is analogous to the procedure for finding the optimal solutions to
various engineering problems.
The optimization procedure of HSA may be sketched in three
main steps viz. initialization, harmony improvisation and selection.
These steps are stated below:
Step (a) Initialization: Program parameters are defined and,
ation of objective function).
ð21Þ
decision variable in the solution is decided in HSA and
Step (c) Selection: When a terminating condition is met, the best
harmony (solution) in the HM is updated. Otherwise, Step (a)
and Step (b) are repeated, sequentially.
Based on the work reported by Banerjee et al. [29], the compu-
tational procedure of the basic HSA may be summarized in
Algorithm 1. The notations used in Algorithm 1 are explained in
[29]. In Algorithm 1, HMS, HMCR, PAR, BW, NI and d are harmony
memory size, harmony memory consideration rate, pitch adjusting
rate, distance bandwidth, number of improvisations and total
number of search spaces (i.e. the dimension of the problem),
respectively. X 0i;j is an initial population and OX 0i;j is exactly oppo-
site of X 0i;j (for i ¼ 1; 2; . . . ; HMS and j ¼ 1; 2; . . . ; dÞ [29].
Algorithm 1: Pseudo code for HSA [29]
Step 1 Set the parameters: HMS; HMCR; PAR; BW; NI and d.
Step 2 Initialize the HM and calculate the objective function
value for each harmony vector.
Step 3 Improvise the HM filled with new harmony X new
vectors as follows:
for (j ¼ 0; j < d; j þ þÞ)
if ðr 1 < HMCRÞ then
ð22Þ xnew
j ¼ xaj % a 2 ð1; 2; . . . ; HMSÞ
if ðr2 < PARÞ then
ð23Þ xnew
j ¼ xnew
j  r3  BW % r 1 ; r 2 ; r 3 2 ½0; 1
end if
else
xnew
j ¼ paramin
j þ r  ðparamax
j  paramin
j Þ %
r 2 ½0; 1
end if
end for
Step 4 Update the HM as X worst ¼ X new if f ðX new Þ < f ðX worst Þ:
Step 5 If NI is completed, return the best harmony vector
X best in the HM; otherwise go back to Step 3.
 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804 793

Quasi-opposition: a concept
Algorithm 2: Pseudo code for the calculation of
quasi-opposite number
Tizhoosh [36], originally, has introduced the opposition-based
learning (OBL) concept. It is used to accelerate the convergence M ¼ ðparamin þ paramax Þ=2
rate of different optimization methods in the field of computa- if (OX o < M)
tional intelligence. In OBL, current population as well as its oppo- QOX o ¼ M þ ðOX 0  MÞ  r1 % r 1 2 ½0; 1
site population is considered simultaneously in order to get else
better candidate solution. Researchers have proved that an oppo- QOX o ¼ OX 0 þ ðM  OX 0 Þ  r 1
site candidate solution has a better chance to be closer to the glo- end if-else
bal optimum solution [29,37] than a random candidate solution.
Some of the contributions of OBL in the soft computing field
include oppositional-based ant colony optimization [38], Quasi-opposite point
opposition-based gravitational search algorithm [39], The quasi-opposite point QOP ðQOP 1 ; QOP 2 ; . . . ; QOP i ; . . . ; QOP d Þ
oppositional-based harmony search algorithm [29], and so on. for d-dimensional search space is expressed by (29).
Regarding OBL utilization, the two important mathematical defini-  
paramin þ paramax
tions (such as opposite-number and its opposite-point) are stated QOP0i ¼ rand i i
; paramin
i þ para max
i  X i
2
in the next two sub-sections.
i ¼ 1; 2; . . . ; d ð29Þ
Opposite-number
The pseudo code for the calculation of quasi-opposite point is
Opposite-number may be defined as the mirror point of the
shown in Algorithm 3.
solution from the center of the search space (M). Let an initial pop-
ulation ðX 0 Þ be a real number ðRÞ in an interval ½paramin ; paramax .
Algorithm 3: Pseudo code for the calculation of
Then, the mathematical expression for M may be expressed by
quasi-opposite point
(24).
M j ¼ ðparamin
j þ paramax
j Þ=2
M ¼ ðparamin þ paramax Þ=2 ð24Þ
if (OX oj < Mj )
Its mirror point, i.e. opposite initial population ðOX 0 Þ is, mathemat- QOX oj ¼ M j þ ðOX 0j  M j Þ  r1 % r1 2 ½0; 1
ically, may be expressed by (25) else
QOX oj ¼ OX 0j þ ðM j  OX 0j Þ  r1
OX 0 ¼ paramin þ paramax  X 0 ð25Þ
end if-else
min max
where para and para are the extreme points of the search
space. Similarly, this definition may be extended to higher dimen-
sions [36].
Quasi-oppositional population initialization
Opposite-point
Even when there is not a prior knowledge about the solution(s),
If X ðX 1 ; X 2 ; . . . ; X d Þ is a point in d-dimensional search space
better starting candidate solution may be obtained by utilizing
where fX 1 ; X 2 ; . . . ; X d g 2 R. Then, the center of search space for this
quasi-oppositional concept. The proposed scheme (QOBL) may be
point ðMj Þ may be stated by (26).
stated in the following sequences:

M j ¼ ðparamin
j þ paramax
j Þ=2 ð26Þ (a) initialize the opposite population ðOX 0 Þ from random popu-
lation space ðN p Þ [39],
Its opposite-point OX ðOX 1 ; OX 2 ; . . . ; OX d Þ may be defined by (b) calculate quasi-opposite population ðQOX 0 Þ using (28) and
(27). (c) select the N p fittest individuals from f OX 0 [ QOX 0 g as initial
population.
OX j ¼ paramin
j þ paramax
j  Xj j ¼ 1; 2; . . . ; d ð27Þ
Quasi-oppositional generation jumping
The extensive work as done on OBL, utilized in HSA, proved that The evolutionary process may be forced to jump to a new candi-
the projection of quasi-concept in OBL (QOBL) is more likely to be date solution which is, ideally, better than the current one. Based on
closer to solution than the OBL [29,37]. Regarding QOBL, the the jumping probability (J r , jumping rate), a new population is gen-
description of two important definitions (such as quasi-opposite erated, then the quasi-opposite population is calculated and the
number and quasi-opposite point as used in QOBL) are defined in HMS fittest HM are selected from the union of the current population
the next two sub-sections. and the quasi-opposite population (refer section ‘Quasi-opposite
point’). But, in the calculation of quasi-oppositional generation
Quasi-opposite number jumping, the opposite of each variable and middle point are, dynam-
Quasi-opposite number may be defined as the number between ically, calculated. That is, the maximum and minimum value of each
h min i
para þparamax variable in the current population is used to calculate the
the center of the search space 2
and the
  middle-to-opposite-points instead of using variable’s pre-defined
min max
opposite-number para þ para  X . Quasi-opposite number interval boundaries.
ðQOX 0 Þ may, mathematically, be expressed by (28).
 
paramin þ paramax Proposed QOHS
QOX 0 ¼ rand ; paramin þ paramax  X 0 ð28Þ
2
In this work, HSA is improvised by employing the QOBL concept
The pseudo code for the calculation of quasi-opposite number is in order to enhance its harmony (solution) and to accelerate the
given in Algorithm 2. convergence speed [29]. The pseudo code, as used in the proposed
794 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804

QOHS algorithm, is presented in Algorithm 4. Here, X 0i;j is an initial and the quasi-opposite point of X 0i;j , respectively, (for
population whereas OX 0i;j and QOX 0i;j are, exactly, the mirror-image i ¼ 1; 2; . . . ; HMS and j ¼ 1; 2; . . . ; dÞ [29].

Algorithm 4: Pseudo code for the proposed QOHS algorithm

Step 1 Set the parameters: HMS; HMCR; PARmin ; PARmax ; BW min ; BW max and NI:
Step 2 Initialize the HM with X 0i;j :
Step 3 % Quasi-oppositional HM initialization
for (i ¼ 0; i < HMS; i þ þÞ
for (j ¼ 0; j < d; j þ þÞ
OX 0i;j ¼ paramin max
i;j þ parai;j  X 0i;j ; % OX 0 : Opposite of initial X 0
M i;j ¼ ðparamin max
i;j þ parai;j Þ=2;
if (OX 0i;j < Mi;j )
QOX 0i;j ¼ Mi;j þ ðOX 0i;j  M i;j Þ  r 1 ; % r 1 2 ½0; 1
end if
else
QOX 0i;j ¼ OX 0i;j þ ðM i;j  OX 0i;j Þ  r1 ;
end else
end for
end for
% End of quasi-oppositional HM initialization.
Select HMS fittest individuals from set of fX 0i;j ; QOX 0i;j g as initial HM; HM being the matrix of fittest X vectors
Step 4 Improvise a new harmony X new as follows:
Update PARðgnÞ and BWðgnÞ .
for (i ¼ 0; i < HMS; i þ þÞ
for (j ¼ 0; j < d; j þ þÞ)
if ðr 2 < HMCRÞ then
X new
i;j ¼ X ai;j ; % a 2 ð1; 2; . . . ; HMSÞ
if ðr3 < PARðgnÞÞ then
X new
i;j ¼ X new i;j  r 4  BWðgnÞ; % r 2 ; r 3 ; r 4 2 ½0; 1
end if
else
X new
i;j ¼ paramin max
i;j þ r 5  ðparai;j  paramin
i;j Þ; % r5 2 ½0; 1
end else
end if
end for
end for
Step 5 Update the HM as X worst ¼ X new if f ðX new Þ < f ðX worst Þ
Step 6 Quasi-opposition based generation jumping
if ðr 6 < J r Þ % r 6 2 ½0; 1, J r : Jumping rate
for (i ¼ 0; i < HMS; i þ þÞ
for (j ¼ 0; j < n; j þ þÞ
OX i;j ¼ paramin max
i;j ðgnÞ þ parai;j ðgnÞ  X i;j ;
% paramin
i;j ðgnÞ: minimum value of jth variable of ith parameter in the current generation (gn)
%paramax
i;j ðgnÞ: maximum value of jth variable of ith parameter in the current generation (gn)
M i;j ¼ fparamin max
i;j ðgnÞ þ parai;j ðgnÞg=2
if (X i;j < Mi;j )
QOX 0i;j ¼ Mi;j þ ðOX 0i;j  M i;j Þ  r7 ; % r7 2 ½0; 1
end if
else
QOX 0i;j ¼ OX 0i;j þ ðM i;j  OX 0i;j Þ  r 7 ;
end else
end for
end for
end if
Select HMS fittest HM from the set of fX i;j ; QOX i;j g as current HM.
% End of quasi-oppositional generation jumping.
If NI is completed, return the best harmony vector X best in the HM; otherwise go back to Step 4.
 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804 795

Table 1 QOHS algorithm is going to be tested for the studied two cases. The
Optimized controller gains of the studied optimization techniques for single-area two cases considered for this example are:
reheat power system.

Cases Methods K p (ve) K i (ve)
Case 1(a): RTWD Relay based [30] 3.6317 3.302
QOHS 4.90 1.7262
Case 1(b): RTD Relay based [30] 6.164 1.930
QOHS 4.90 2.10
Simulated results and analysis
In this section, the effectiveness of the proposed QOHS algo-
rithm is tested through three kinds of decentralized power system.
The first power system, shown in Fig. 3, is a single-area reheats test
system explained in section ‘AGC design of single-area reheats
power system with and without droop characteristics’. The second
one, shown in Fig. 4, is a two-area non-reheat power system dis-
cussed in section ‘AGC model of two-area non-reheat power sys-
tem’ while the third one is a four-area hydro-thermal power
system (refer Figs. 5 and 6) which is elaborated in section ‘AGC
model of four-area hydro-thermal power system’. The QOHS based
simulation results of the studied power system models are dis-
cussed and compared to GA [23,40], hBFOA-PSO [23], BFOA
[23,40], PSO [23], ZN [40] and relay based identification techniques
[30]. In the analysis of AGC work, the type of system, nature of area
load demand and location of load connection are mentioned in
their respective studied examples. The relevant system parameters
are given in Appendix Section while the symbols used are depicted
in Nomenclature Section. The numbers of iteration cycle and pop-
ulation size are chosen as 100 and 60, respectively, for all the stud-
ied cases. The simulation results are carried out on an Intel, core-i3
processor CPU of 2.40 GHz and 4 GB RAM computer in the MATLAB
7.10.0499 (R2010a) environment. The major observations of the
present work are documented below. The Results of interest are
bold faced in their respective tables.
K d (ve) Case 1(a): single-area RTWD characteristic.
1.75 Case 1(b): single-area RTD characteristic.
2.10
1.164 The results, as yielded by the proposed QOHS algorithm, are
1.40 compared to those reported by Padhan and Majhi in [30]. The
model parameters for this example are presented in Section A.1
of Appendix Section [30]. For the simulation work, SLP of
0.01 p.u. is applied at t = 1 s for both the two cases.
Case 1(a): Single-area RTWD characteristic
The optimized controller gains, the transient specifications
(such as T r , T s ; M p and T p ), FOD value and the studied two perfor-
mance indices (such as ITAE and ITSE) are shown in Tables 1 and
2, respectively. From the Table 2, it is clear that the proposed
QOHS based FOD value is smaller as compared to relay based
method reported in [30]. This demonstrates that the system with
the proposed QOHS technique provides more optimal controller
gains as compared to the relay based approach. Improvements of
99.82%, 26.21%, 72.5% and 23.5% are also observed for T r , T s , M p
and T p , in order, with the utilization of the proposed QOHS opti-
mized PID controller gains. Examining the comparative dynamic
response of frequency deviation, as presented in Fig. 7(a), it is
evident that the proposed QOHS method yields much better
dynamic response than the relay based method. It may also be
observed that oscillations are attenuated and the system offers
better AGC performance. Hence, the proposed QOHS based con-
troller gains enhance the system stability and improve the damp-
ing characteristics of the power system. Fig. 7(b) portrays faster
convergence profile of FOD value, as yielded by the proposed
QOHS method.

Example 1: Single-area power system Case 1(b): Single-area RTD characteristic

This case is an extension of Case 1(a) in which drooping phe-
The model of an isolated single-area power system, shown in nomena is added to the single-area reheat turbine system. The
Fig. 3, is considered as Example 1. The performance of the proposed model parameters are the same as those of Case 1(a). In order

Table 2
Comparative transient parameter specifications of the studied optimization techniques for single-area reheat power system.

Cases Methods Df ISE (103) ITAE (104) ITSE (105)

3
T r (s) T s (s) M p (%) (10 ) T p (s)

Case 1(a): RTWD Relay based [30] 1.03  105 6.95 200.00 1.51 5.2 335.2 3.35
QOHS 0.0058 9.42 55.00 1.15 4.7 283.1 3.40
Case 1(b): RTD Relay based [30] 0.03 4.33 73.00 1.22 6.14 237.4 3.62
QOHS 3.6  103 8.26 69.00 1.20 5.28 207.3 2.89

Fig. 7. For single-area (RTWD characteristic) test system: (a) comparative frequency deviation and (b) QOHS based convergence profile of FOD value.
796 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804

Fig. 8. For single-area (RTD characteristic) test system: (a) comparative frequency deviation and (b) QOHS based convergence profile of FOD value.

Fig. 9. Comparative dynamic responses of the studied optimization techniques for two-area non-reheat turbine power system with 0.1 p.u. SLP in area-1 at t = 0 s: (a) Df 1 , (b)
Df 2 and (c) DP tie .

to get the satisfactory closed loop response, the optimized con-
troller gains are tabulated in Table 1. It may be noted from
Table 2 that the proposed QOHS method provides more efficient
solution in controlling the damping and oscillation characteristics
of this system when all the values like FOD, ITAE and ITSE are
considered simultaneously. Moreover, significant improvements
of 88%, 47.5%, 5.4% and 1.63% are observed for T r , T s , M p and
T p , in order, with the use of the proposed QOHS method. The
comparative dynamic response is presented in Fig. 8(a). It may
be observed that the system dynamic performance is improved
with the use of the proposed QOHS method. The convergence
profile of FOD value, as offered by the proposed QOHS algorithm
and presented in Fig. 8(b), shows promising convergence
characteristics.
Example 2: Two-area power system
The model for the two-area power system, shown in Fig. 4, is
considered as Example 2. The model parameters for this example
are presented in Section A.2 of the Appendix Section [23]. Types
and area of application of load demand for the studied cases of this
example are mentioned in their respective cases. The three cases
considered for this example are:
Case 2(a): two-area non-reheat power system.
Case 2(b): two-area non-reheat power system with governor
dead band.
Case 2(c): two-area non-reheat system with GRC ðaÞ limit of
±0.05.
 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804 797

Table 3 The system eigenvalues for the different approaches are pro-
Optimized controller gains of the studied optimization techniques for two-area non- vided in Table 5. It may be observed from this table that the pro-
reheat power system.
posed QOHS method provides eigenvalues far away from the
Methods K p (ve) K i (ve) K d (ve) imaginary axis (of s-plane) as compared to other methods. It may
Area-1 Area-2 Area-1 Area-2 Area-1 Area-2 be understood in this way that the decay in transient phenomena
GA [23] 0.2346 0.2346 0.2662 0.2662 – –
may be faster with the proposed QOHS algorithm as compared to
hBFOA–PSO [23] 0.4383 0.4383 0.3349 0.3349 – – other techniques. Hence, compared to other methods reported in
BFOA [23] 0.3317 0.3317 0.471 0.471 – – the recent state-of-the art literature, the proposed QOHS provides
PSO [23] 0.3597 0.3597 0.4756 0.4756 – – better system stability, improves the nature of dynamic responses
QOHS 4.0 1.6662 3.9156 0.4925 1.6928 0.0011
and the oscillations are attenuated.
For entry ‘‘–’’ means not applicable.

Robustness analysis of two-area non-reheat power

Case 2(a): Two-area non-reheat power system
The optimal controller gains and transient response parameters
(based on Fig. 9) are presented in Tables 3 and 4, respectively. It is
clear from Table 4 that all the values of ISE, ITAE and ITSE are lesser
as compared to those offered by the other reported techniques like
GA [23], hBFOA-PSO [23], BFOA [23] and PSO [23]. For the simula-
tion work, SLP of 0.1 p.u. is applied at t = 0 s in area-1. Critical anal-
ysis of the dynamic responses reveal that the performance of the
system is improved by using the proposed QOHS based optimized
controller gains. The same is reflected in the comparative dynamic
responses (Df 1 , Df 2 and DP tie ) as offered by the different optimiza-
tion techniques presented in Fig. 9. Moreover, the proposed
method outperforms and outlasts GA [23], hBFOA-PSO [23], BFOA
[23] and PSO [23] in sense of damping, the magnitude of over-
shoot/undershoot and settling time. Hence, the proposed QOHS
algorithm is a better optimization tool in AGC performance
analysis.
Stability analysis of two-area non-reheat power system. The eigen-
value analysis using the state-space approach is a feasible method
to show the stability of the designed power system. The condition
for stability analysis is that the system is absolutely stable if the
entire eigenvalues lie on the left side of the imaginary axis (of
s-plane) and vice versa. Moreover, with this eigen theory,
small-signal analysis behavior may also be studied. For the stabil-
ity analysis, an incremental change in load demand is applied in
the form of SLP. The purposes of using the step load are that (a)
it is bounded input signal, (b) the system dynamic behavior may
be studied and (c) the stability analysis is possible.
system. Robustness analysis is carried out to study the uncertainty
in dynamic behavior at nominal condition to wide change in some
of the important parameters of the power system. The aim of this
study is to find out the range of parameters so that the power sys-
tem responses are remain under-damped. Over-damped and criti-
cally damped systems are, normally, avoided due to its sluggish in
nature and large time constant.
In this paper, robustness analysis is carried out by varying the
operating load condition and governor time constant in the range
of ±50% by applying SLP of 0.1 p.u. in area-1 with nominal system
parameters. The profiles of robustness analysis of two-area
non-reheat power system are presented in Figs. 10 and 11. It
may be inferred from these figures that the proposed QOHS based
optimized PID controller gains provides a robust and stable speed
control mechanism. The superiority of the proposed QOHS method
is verified and power system oscillations are effectively alleviated.
Moreover, the tuned value of the controller gains obtained at the
nominal loading with nominal parameters need not be reset for
wide change in the system loading.
Case 2(b): Two-area non-reheat power system with governor dead
band
The study is further extended by introducing non-linearity in
the power system dynamics by incorporating governor dead band.
It is defined as the rate of change of total magnitude of a sustained
speed when there is no change in valve position. There is a specific
time, called as dead band, during which no change in valve position
is observed. It creates continuous oscillations in the system fre-
quency and tie-line power flow profiles [19,23]. The property of
the governor dead band is that, it produces sustained sinusoidal

Table 4
Comparative transient parameter specifications of the studied optimization techniques for two-area non-reheat power system.

Methods Df 1 Df 2 DP tie ISE ITAE ITSE

T r (s) (106) T s (s) M p (%) T p (s) T r (s) T s s) Mp (%) T p (s) T r (s) (105) T s (s) M p (%) T p (s)

GA [23] 4.16 8.22 0.23 0.67 0.052 9.83 0.18 1.37 200.00 10.15 0.06 1.08 0.062 4.69 0.12
hBFOA- PSO [23] 3.35 5.05 0.25 0.68 0.0606 4.39 0.21 1.58 1.13 4.80 0.08 1.18 0.046 3.34 0.08
BFOA [23] 0.850 5.32 0.24 0.68 0.034 6.22 0.19 1.47 1.14 5.72 0.07 1.07 0.040 1.53 0.05
PSO [23] 1.37 5.46 0.24 0.68 0.047 6.31 0.19 1.47 0.017 4.95 0.07 1.07 0.039 1.55 0.04
QOHS 0.0217 2.55 0.04 0.13 0.0062 3.82 0.02 0.89 25.00 4.43 0.01 0.62 0.0011 0.12 0.00079

Table 5
Calculated eigen values of the studied optimization techniques for two-area non-reheat power system.

GA [23] hBFOA-PSO [23] BFOA [23] PSO [23] QOHS

33.5439 33.4873 33.5177 33.5105 34.5866
33.5400 33.7178 33.5114 33.5036 33.7213
0.6943 ± i3.0838 0.6607 ± i3.648 0.6332 ± i2.86 0.6433 ± i2.81 0.2017 ± i6.4226
1.3727 ± i1.825 1.3500 ± i1.7778 0.9472 ± i1.466 0.8704 ± i1.40 0.5800 ± i3.5543
1.3609 0.9796 ± i0.4958 0.9663 ± i0.50 1.4722 0.3411
0.4312 0.2475 1.3109 0.9598 ± i0.561 1.6150 ± i0.3565
0.4233 33.4873 33.5177 34.5866
798 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804

Fig. 10. Robustness analysis of two-area non-reheat power system with change in governor time constant: (a) Df 1 , (b) Df 2 and (c) DP tie .

Fig. 11. Robustness analysis of two-area non-reheat power system with change in rated load condition: (a) Df 1 , (b) Df 2 and (c) DP tie .

Table 6
Optimized controller gains of the studied optimization techniques for two-area non-reheat power system with governor dead band.

Methods K p (ve) K i (ve) K d (ve)

Area-1 Area-2 Area-1 Area-2 Area-1 Area-2

ITAE:hBFOA-PSO [23] 0.5484 0.5484 0.2277 0.2277 – –
ISE:hBFOA-PSO [23] 0.2186 0.2186 0.3673 0.3673 – –
ISE:CRAZYPSO [23] 0.2514 0.2514 0.2491 0.2491 – –
QOHS 1.00 1.00 1.00 0.9894 0.642 0.3339

For entry ‘‘–’’ means not applicable.

oscillation of natural period of approximately T 0 ¼ 2 s. A different
approach is adopted to describe the governor dead band function
in order to introduce the effect of non-linearity in real-time power
system application. After Fourier series expansion of the function
and neglecting the fourth and higher order terms, the transfer
function of governor with non-linearity effect may be given by
(26) [19].
0:8  ð0:2=PÞs
Gg ¼
1 þ sT g
The relevant system parameters for this test case are presented
in Section A.3 of Appendix section [23]. A SLP of 0.01 p.u. is applied
in area-1 at t = 0 s. The proposed QOHS optimized controller gains
are featured in Table 6 along with those reported in the literatures
like ITAE:hBFOA-PSO [23], ISE:hBFOA-PSO [23] and ISE:CRAZYPSO
[23]. The comparative studies of transient parameters along with
fitness function of the studied optimization techniques are pre-
sented in Table 7. Table 7 records the sharp improvement in ISE,
ITAE and ITSE values with the proposed QOHS method as com-
ð26Þ
pared to other techniques. Critical analysis of the dynamic

Table 7
Comparative transient parameter specifications of the studied optimization techniques for two-area non-reheat power system with governor dead band.

Methods Df 1 Df 2 DP tie ISE ITAE ITSE (103)

T r (s) T s (s) Mp (%) T p (s) T r (s) T s (s) M p (%) T p (s) T r (s) T s (s) Mp (%) T p (s)

ITAE:hBFOA-PSO [23] 0.00015 10.05 0.05 1.27 0.04 15.04 0.02 3.09 0.02 18.9 0.0054 2.42 0.0015 1.23 4.13
ISE:hBFOA-PSO [23] 0.000075 10.56 0.04 1.15 0.02 14.32 0.01 2.08 0.02 17.81 0.0034 2.08 0.0015 0.56 2.40
ISE:CRAZYPSO [23] 0.00082 11.09 0.04 1.15 0.08 17.99 0.01 2.61 0.02 22.07 0.0037 2.10 0.001 0.863 3.20
QOHS 0.0027 9.28 0.02 0.05 1.65 13.01 0.0035 1.28 .000047 16.71 0.001 2.29 0.0002 0.16 2.6
 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804 799

Fig. 12. Comparative dynamic responses of the studied optimization techniques for two-area non-reheat turbine system with governor dead band with 0.01 p.u. SLP in area-1
at t ¼ 0 s: (a) Df 1 , (b) Df 2 and (c) DP tie .

responses, offered by the studied optimization techniques, is of the system is, significantly, improved with the proposed QOHS
shown in Fig. 12. This figure reveals that the dynamic performance based approach. It may also be observed that the effects of dead
band may be effectively alleviated with the proposed QOHS
technique.

Case 2(c): Two-area non-reheat power system with GRC limit of ±0.05
For the power system, GRC limits the rate of generating unit
output which affects the AGC performance. This is, especially, true
Fig. 13. Dynamic structure of non-linear turbine model with GRC [40].
for thermal system where mechanical and thermal stresses are the
limiting factors for both the governor and turbine unit. The
Table 8 non-linear turbine model with GRC effect is shown in Fig. 13
Optimized controller gains of the studied optimization techniques for two-area power [40]. The nominal parameters for this case study are shown in
system with GRC. Section A.4 of Appendix Section [40]. For the simulation work,
Algorithms K p (ve) K i (ve) K d (ve) SLP of 0.05 p.u. is applied in area-1 at t = 0 s.
Area-1 Area-2 Area-1 Area-2 Area-1 Area-2
The proposed QOHS based optimized controller gains are fea-
tured in Table 8 along with Ziegler–Nichols (ZN) [40] method, GA
ZN [40] 0.5865 0.5865 0.5100 0.5100 0.1686 0.1686
GA [40] 0.0955 0.0955 0.4712 0.4712 0.0679 0.0679
[40] and BFOA [40]. The comparative dynamic response profiles
BFOA [40] 0.317 0.317 0.4873 0.4873 0.2506 0.2506 are presented in Table 9. It may be noted that the improvements
QOHS 1.00 0.1869 0.7511 0.0606 0.6654 0.0171 in ISE, ITAE and ITSE values have recorded with the proposed
QOHS method as compared to the reported algorithms (refer

Table 9
Comparative transient parameter specifications of the studied optimization techniques for two-area power system with GRC.

Methods Df 1 Df 2 DP tie ISE ITAE ITSE

T r (s) (105) T s (s) Mp (%) T p (s) T r (s) T s (s) Mp (%) T p (s) T r (s) T s (s) Mp (%) T p (s)

ZN [40] 4.00 6.76 0.11 0.64 0.02 0.01 0.6 0.01 0.004 8.66 0.03 1.00 0.04 0.60 0.06
GA [40] 0.06 6.11 0.10 0.66 0.01 0.01 0.5 0.01 0.001 6.83 0.02 1.02 0.03 0.55 0.04
BFOA [40] 0.40 4.36 0.11 0.61 0.31 0.01 0.4 0.01 1.42 6.53 0.03 0.98 0.02 0.49 0.02
QOHS 2.00 3.33 0.098 0.58 0.0001 0.01 0.2 0.01 0.001 3.30 0.02 0.949 0.01 0.28 0.01
800 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804

Fig. 14. Comparative dynamic responses of the studied optimization techniques for two-area non-reheat turbine system with GRC having 0.05 p.u. SLP in area-1 at t ¼ 0 s: (a)
Df 1 , (b) Df 2 and (c) DP tie .

Table 10
Comparative optimized controller gains of the two studied optimization techniques for four-area power system for Case 3(a).

Methods K p (ve) K i (ve) K d (ve)

Area-1 Area-2 Area-3 Area-4 Area-1 Area-2 Area-3 Area-4 Area-1 Area-2 Area-3 Area-4
IMC [31] 2.445 2.445 2.445 0.145 1.584 1.584 1.584 0.021 1.526 1.526 1.526 18.77
QOHS 8.954 7.74 8.33 1.316 4.44 8.02 8.61 0.616 3.50 8.799 1.43 8.75

Table 11
Comparative transient parameter specifications of the two studied optimization techniques for four-area power system for Case 3(a).

Parameters IMC [31] QOHS

T r (s) T s (s) Mp (%) (103) T p (s) T r (s) T s (s) Mp (%) (104) T p (s)
5
Df 1 2.53  10 11.14 16.0 0.48 0.0001 10.86 95.00 0.27
Df 2 0.04 24.82 3.20 1.58 0.04 29.45 8.70 1.43
Df 3 0.04 24.82 3.20 1.58 0.05 28.73 11.00 0.91
Df 4 0.08 24.32 3.20 3.54 0.24 26.65 9.74 3.72
DP tie12 0.02 29.35 0.88 1.47 0.02 30.41 3.17 1.47
DP tie31 0.02 29.35 0.88 1.47 0.02 45.77 3.11 1.51
DP tie41 0.06 24.04 1.10 1.79 0.12 55.29 3.75 1.60
ISE 2.56  104 4.26  105
ITAE 0.5294 0.5098
ITSE 3.8  104 6.80  105

Table 9). The comparative dynamic responses offered by the stud-
ied optimization techniques are shown in Fig. 14. It is clearly
revealed that the AGC performance of the system is improved with
the proposed QOHS based approach (refer Fig. 14).
Example 3: Four-area hydro-thermal power system
The simplified diagram of a four-area hydro-thermal power sys-
tem [31], presented in Fig. 5, is considered as Example 3. The block
diagram of its ith control area is shown in Fig. 6. The relevant sys-
tem parameters are presented in Section A.5 of Appendix section
[31]. The results yielded by the proposed QOHS algorithm are com-
pared to those reported by Tan in [31]. In [31], internal model con-
trol (IMC) technique is adopted. The two cases considered for this
example are:
Case 3(a): application of SLP of 0.01 p.u. at t ¼ 0 sec in area-1.
Case 3(b): application of SLP of 0.01 p.u. at t ¼ 0 sec in area-1
followed by SLP of 0.01 p.u. at t ¼ 100 s in area-3.
 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804 801

Fig. 15. Comparative dynamic responses of the studied optimization techniques for four-area power system with 0.01 p.u. SLP in area-1 at t ¼ 0 s: (a) Df 1 , (b) Df 2 , (c) Df 3 , (d)
Df 4 , (e) DP tie12 , (f) DP tie23 , (g) DP tie31 and (h) DP tie41 .

Case 3(a): Application of SLP of 0.01 p.u. at t ¼ 0 s in area-1 Tan in [31]. The comparative dynamic responses are presented in
The proposed QOHS based optimized controller gains for this Table 11. A significant improvement in the value of maximum over-
test case is tabulated in Table 10 along with those reported by shoot is observed from this table. Also, sharp improvements in the
802 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804

fitness function (FOD or ISE) as well as the two performance indices Fig. 15) clearly reveal that the frequency and the tie-line power
values (like ITAE and ITSE) are observed. A close analysis of the deviations are driven to almost zero by the proposed QOHS method
dynamic responses, offered by the comparative two methods, (refer with significant improvement in dynamic performance.

Fig. 16. Comparative dynamic responses of the studied optimization techniques for four-area power system with 0.01 p.u. SLP at t ¼ 0 s in area-1 followed by 0.01 p.u. SLP at
t ¼ 100 s to area-3: (a) Df 1 , (b) Df 2 , (c) Df 3 , (d) Df 4 , (e) DP tie12 , (f) DP tie23 , (g) DP tie31 and (h) DP tie41 .
 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804
Case 3(b): Application of SLP of 0.01 p.u. at t ¼ 0 s in area-1 followed
by SLP of 0.01 p.u. at t ¼ 100 s in area-3
The comparative system dynamic responses of two methods
(viz. the responses offered by IMC reported in [31] and to those
offered by the proposed QOHS algorithm) are shown in Fig. 16.
Critical analysis of the dynamic responses offered by these two
competitive methods reveal that the system performance with
the proposed QOHS method is improved. It may also be noted that
the designed controller gains, offered by the proposed QOHS algo-
rithm, are robust and perform satisfactorily with the application of
multiple load disturbances. System regains its normal operating
condition with multiple SLP at different time interval.
Conclusion
In this paper, AGC performance analysis of single- and
multi-area power system is performed by using a novel optimiza-
tion technique. The proposed QOHS algorithm, based on
quasi-oppositional concept, is blended with the basic HSA to
enhance the AGC performance. The performance investigation of
the proposed QOHS algorithm starts the AGC study of a linear
single-area power system model. Then, it is extended to two-area
power system with and without non-linearity effect whose stabil-
ity and robustness studies are also carried out. And, finally, this
concept is further extended to a four-area hydro-thermal intercon-
nected power system having multiple SLP at different time loca-
tion. The potential benefits yielded by the proposed QOHS
algorithm are compared to other techniques for similar power sys-
tem configuration. Time-domain simulation of the investigated
power system model reveals that the proposed QOHS algorithm
of the present work shows better tuning capability and is quite
robust and stable in comparison to other similar population based
optimization algorithms reported in the recent state-of-the-art lit-
erature. The proposed QOHS algorithm may be applied by the
future researchers to solve some other engineering optimization
tasks.
Appendix A
A.1. Nominal parameters of single-area power system with reheat
turbine [30]
f ¼ 60 Hz; K p ¼ 120 Hz/p.u.MW; T p ¼ 20 s; T t ¼ 0:3 s; T g ¼
0:08 s; R ¼ 2:4 Hz/p.u.; T r ¼ 2:4 sec; c ¼ 0:35.
A.2. Nominal parameters of two-area system with non-reheat turbine
[23]
Pr ¼ 2000 MW (rating); f ¼ 60 Hz; B1 ¼ B2 ¼ 0:425 p.u.MW/Hz;
R1 ¼ R2 ¼ 2:4 Hz/p.u; T g1 ¼ T g2 ¼ 0:03 s; T t1 ¼ T t2 ¼ 0:3 s;
K p1 ¼ K p2 ¼ 120 Hz/p.u.MW; T p1 ¼ T p2 ¼ 20 s; T 12 ¼ 0:545 p.u.;
a1 ¼ a2 ¼ 1.
A.3. Nominal parameters of two-area power system with governor
dead band [23]
f ¼ 60 Hz; B1 ¼ B2 ¼ 0:425 p.u.MW/Hz; R1 ¼ R2 ¼ 2:4 Hz/p.u;
T g1 ¼ T g2 ¼ 0:03 s; T t1 ¼ T t2 ¼ 0:3 s; K p1 ¼ K p2 ¼ 120 Hz/p.u.MW;
T p1 ¼ T p2 ¼ 20 s; T 12 ¼ 0:0707 p.u.; a1 ¼ a2 ¼ 1.
A.4. Nominal parameters of two-area power system with GRC [40]
f ¼ 60 Hz; T p1 ¼ T p2 ¼ 20 s; T t1 ¼ T t2 ¼ 0:3 s; T 12 ¼ 0:545 p.u.;
T g1 ¼ T g2 ¼ 0:08 s; K p1 ¼ K p2 ¼ 120 Hz/p.u.MW; a1 ¼ a2 ¼ 1;
R1 ¼ R2 ¼ 2:4 Hz/p.u.MW; B1 ¼ B2 ¼ 0:425 p.u.MW/Hz.
803
A.5. Nominal parameters of four-area power system [31]
For reheat turbines power system: f ¼ 60 Hz; T gi ¼ 0:08 s;
T ti ¼ 0:3 s; T pi ¼ 20 s; K pi ¼ 120 Hz/p.u.MW; Ri ¼ 2:4 Hz/p.u.;
Bi ¼ 0:425 p.u.MW/Hz; T 12 ¼ T 13 ¼ T 23 ¼ 0:707 p.u.; i ¼ 1; 2; 3.
For hydro-turbine power system: f ¼ 60 Hz; K p ¼ 80 Hz/
p.u.MW; T p ¼ 13 s; T w ¼ 1 s; R ¼ 2:4 Hz/p.u., B4 ¼ 0:425 p.u.MW/
Hz; T 14 ¼ 0:0707 p.u.
References
[1] Wood AJ, Wollenberg BF. Power generation operation and control. New
York: John Wiley & Sons; 1996.
[2] Kundur P. Power system stability and control. New Delhi: Tata McGraw Hill;
2008.
[3] Bevrani H. Intelligent automatic generation control. New York: Springer; 2009.
[4] Kumar IP, Kothari DP. Recent philosophies of automatic generation control
strategies in power systems. IEEE Trans Power Syst 2005;20(1):346–57.
[5] Cohn N. Some aspects of tie-line bias control on interconnected power
systems. Am Inst Elect Eng Trans 1957;75:1415–36.
[6] Fosha CE, Elgerd OI. The megawatt frequency control problem: a new approach
via optimal control theory. IEEE Trans Power Appl Syst 1970;89(4):563–77.
[7] Pataya D, Vichit A. Genetic algorithms based on an intelligent controller.
Expert Syst with Appl 1996;10(3–4):465–70.
[8] Ghoshal SP. Application of GA/GA-SA based fuzzy automatic generation control
of a multi-area thermal generating system. Electr Power Syst Res
2004;70(2):115–27.
[9] Denna M, Mauri G, Zanaboni AM. Learning fuzzy rules with Tabu search-an
application to control. IEEE Trans Fuzzy Syst 1999;7(2):295–318.
[10] Ahamed TP Imthias, Nagendra Rao PS, Sastry PS. A reinforcement learning
approach to automatic generation control. Electr Power Syst Res
2002;63(1):9–26.
[11] Ghoshal SP. Optimization of PID gains by particle swarm optimization in fuzzy
based automatic generation control. Electr Power Syst Res 2004;72:203–12.
[12] Pothiya S, Ngamroo I, Runggeratigul S, Tantaswadi P. Design of optimal fuzzy
logic based PI controller using multiple Tabu search algorithm for load
frequency control. Int J Contr Autom Syst 2006;4(2):155–64.
[13] Sabahi K, Sharifi A, Aliyari M, Teshnehlab M, Aliasghary M. Load frequency
controller in interconnected power system using multi-objective PID
controller. J Appl Sci 2008;8(20):3676–82.
[14] Shayeghi H, Jalili A, Shayanfar HA. Multi stage fuzzy load frequency control
using PSO. Int J Energy Convers Manage 2008;49(10):2570–80.
[15] Ramesh S, Krishnan A. Modified genetic algorithm based load frequency
controller for interconnected power system. Int J Electr Power Eng
2009;3(1):26–30.
[16] Nanda J, Mishra S, Saikia LC. Maiden application of bacterial foraging based
optimization technique in multi area automatic generation control. IEEE Trans
Power Syst 2009;24(2):602–9.
[17] Taher S, Hematti R, Abdolalipour A, Tabei SH. Optimal decentralized load
frequency control using HPSO algorithms in deregulated power systems. Am J
Appl Sci 2008;5(9):1167–74.
[18] Ali ES, Abd-Elazim SM. Bacteria foraging optimization algorithm based load
frequency controller for interconnected power system. Int J Electr Power
Energy Syst 2011;33(3):633–8.
[19] Gozde H, Taplamacioglu MC. Automatic generation control application with
craziness based particle swarm optimization in a thermal power system. Int J
Elect Power Energy Syst 2011;33:8–16.
[20] Gozde H, Taplamacioglu MC, Kocaarslan I. Comparative performance analysis
of artificial bee colony algorithm in automatic generation control for
interconnected reheat thermal power system. Electr Power Syst Res
2012;42:167–78.
[21] Shabani H, Vahidi B, Ebrahimpour M. A robust PID controller based on
imperialist competitive algorithm for load-frequency control of power system.
ISA Trans 2013;52:88–98.
[22] Panda S, Yegireddy NR. Automatic generation control of multi-area power
system using multi-objective non-dominated sorting algorithm-II. Int J Electr
Power Energy Syst 2013;53:54–63.
[23] Panda S, Mohanty B, Hota PK. Hybrid BFOA-PSO algorithm for automatic
generation control of linear and non-linear interconnected power systems.
Appl Soft Comput 2013;13:4718–30.
[24] Geem ZW, Kim JH, Loganathan GV. A new heuristic optimization algorithm:
harmony search. Simulations 2001;76(2):60–8.
[25] Das S, Mukhopadhyay A, Roy A, Abraham A, Panigrahi BK. Exploratory power
of the harmony search algorithm: analysis and improvements for global
numerical optimization. IEEE Trans Syst 2011;41(1):89–106.
[26] Mahdavi M, Fesanghary M, Damangir E. An improved harmony search
algorithm for solving optimization problems. Appl Math Comput
2007;188:1567–79.
[27] Omran MGH, Mahdavi M. Global-best harmony search. Appl Math Comput
2008;198:643–56.
[28] Pan QK, Suganthan PN, Tasgetiren MF, Liang JJ. A self-adaptive global best
harmony search algorithm for continuous optimization problems. Appl Math
Comput 2010;216:830–48.
804 C.K. Shiva et al. / Electrical Power and Energy Systems 73 (2015) 787–804
[29] Banerjee A, Mukherjee V, Ghoshal SP. An opposition-based harmony search
algorithm for engineering optimization problems. Ain Shams Eng J
2014;5(1):85–101.
[30] Padhan DG, Majhi S. A new control scheme for PID load frequency controller of
single-area and multi-area power systems. ISA Trans 2013;52:242–51.
[31] Tan W. Unified tuning of PID load frequency controller for power systems via
IMC. IEEE Trans Power syst 2010;25(1):341–50.
[32] Vagia M. PID controller design approaches-theory, tuning and application to
frontier areas. Intech; 2012.
[33] Geem ZW. Recent advances in harmony search. New York: Springer; 2010.
[34] Lee KS, Geem ZW. A new meta-heuristic algorithm for continuous engineering
optimization: harmony search theory and practice. Comp Meth Appl Mech Eng
2005;194(36–38):2902–33.
[35] Yang XS. Music-inspired harmony search algorithm-theory and
applications. Springer; 2009.
[36] Tizhoosh HR. Opposition-based learning: a new scheme for machine
intelligence. In: Proceedings of the international conference on computation
of intelligence on modeling control and automation, vol. 1; 2005. p. 695–701.
[37] Rahnamayan S, Tizhoosh HR, Salama MMA. Opposition versus randomness in
soft computing techniques. Appl Soft Comput 2008;8(2):906–18.
[38] Haiping M, Xieyong R, Baogen J. Oppositional ant colony optimization
algorithm and its application to fault monitoring. In: Proceedings of the 29th
Chinese control conference; 2010. p. 3895–8.
[39] Shaw B, Mukherjee V, Ghoshal SP. Solution of reactive power dispatch of
power systems by an opposition-based gravitational search algorithm. Int J
Electr Power Energy Syst 2014;55:29–40.
[40] Ali ES, Abd-Elazim SM. BFOA based design of PID controller for two area load
frequency control with nonlinearities. Int J Electr Power Energy Syst
2013;51:224–31.