Electrical Power and Energy Systems 73 (2015) 632–643

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

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

Cuckoo Search algorithm based load frequency controller design

for nonlinear interconnected power system
A.Y. Abdelaziz a, E.S. Ali b,c,⇑
a
Electric Power and Machine Department, Faculty of Engineering, Ain Shams University, Cairo, Egypt
b
Electric Power and Machine Department, Faculty of Engineering, Zagazig University, Zagazig, Egypt
c
Currently, Electrical Department, Faculty of Engineering, Jazan University, KSA

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

Article history: A new optimization technique called Cuckoo Search (CS) algorithm for optimum tuning of PI controllers
Received 1 October 2014 for Load Frequency Control (LFC) is suggested in this paper. A time domain based-objective function is
Received in revised form 19 May 2015 established to robustly tune the parameters of PI-based LFC which is solved by the CS algorithm to attain
Accepted 26 May 2015
the most optimistic results. A three-area interconnected system is investigated as a test system under
various loading conditions where system nonlinearities are taken into account to confirm the effective-
ness of the suggested algorithm. Simulation results are introduced to show the enhanced performance of
Keywords:
the developed CS based controllers in comparison with Genetic Algorithm (GA), Particle Swarm
Cuckoo Search algorithm
Load Frequency Control
Optimization (PSO) and conventional integral controller. These results denote that the proposed con-
PI controller trollers offer better performance over others in terms of settling times and various indices.
Time delay Ó 2015 Elsevier Ltd. All rights reserved.
Generation Rate Constraint
Three area system

Introduction ANN suffers from the long training time, the selecting number of
layers and the number of neurons in each layer. Also, FLC requests
In the large scale power systems, LFC plays a serious role. The a hard work to catch the efficient signals and it needs fine tuning
LFC is aimed to ensure the frequency of each area and the inter area and simulation before operation.
tie line power within tolerable limits to deal with the fluctuation of Another approach is to use Evolutionary Algorithm (EA) tech-
load demands and system disturbances [1,2]. These important niques. EA is visualized to be very effective to deal with LFC prob-
functions are delegated to LFC due to the fact that a lem due to its ability to treat nonlinear objective functions. Among
well-designed power system should keep voltage and frequency the EA techniques, GA [24–29], PSO [30–33], Bacteria Foraging
in scheduled range while supplying an acceptable level of power [34–38], Artificial Bee Colony [39], and Ant Colony Optimization
quality [3,4]. [40] have attracted the attention in LFC controller design.
Several researches and techniques had been applied to the field Although these algorithms appear to be efficient for the design
of LFC during the last decades. Robust control [5–9], decentralized problem, they suffer from slow convergence problem in refined
control [10,11], linear quadratic problem [12], pole placement search stage, weak local search ability, which may lead them to
approach [13,14], variable structure control [15], and state feed- get trapped in local minimum solution. A new evolutionary com-
back [16], are used to deal with LFC problem design. However, putation algorithm, called CS algorithm has been presented by
these strategies have many problems which limit their applicabil- [41] and further formed newly by [42–44]. In addition, it is simple
ity. In an effort to overcome these problems, many researches have and population based stochastic optimization algorithm.
used artificial intelligence as Fuzzy Logic Controller (FLC) [17–21] Moreover, it requires less control parameters to be tuned. Also, it
and Artificial Neural Network (ANN) [21–23]. Although these is a compatible optimization tool for power system controller
methods are effective in dealing with the nonlinear characteristics design [45,46].
of the power system, they have their own problems. For example, This paper introduces a modern optimization algorithm called
CS for the optimum tuning of PI controller parameters in LFC prob-
lem. The motivation behind this research is to ensure and prove the
⇑ Corresponding author at: Electrical Department, Faculty of Engineering, Jazan
University, KSA.
robustness of CS based PI, and to enhance the performance of fre-
E-mail addresses: [email protected] (A.Y. Abdelaziz), ehabsali- quency deviation and tie line power under various loading condi-
[email protected] (E.S. Ali). tions in presence of system nonlinearities.

http://dx.doi.org/10.1016/j.ijepes.2015.05.050
0142-0615/Ó 2015 Elsevier Ltd. All rights reserved.
 A.Y. Abdelaziz, E.S. Ali / Electrical Power and Energy Systems 73 (2015) 632–643 633

Nomenclature

f the system frequency in Hz, K min max

PPi ; K PPi the lower and the upper limit of proportional gain of
i subscript referring to area (i = 1, 2, 3) area i
Ri the regulation constant (Hz/p.u MW) for area i, K min max
IIi ; K IIi the lower and the upper limit of Integral gain of area i
T gi the speed governor time constant in second for area i Pa the probability to abandon a nest
T ti the turbine time constant in second for area i, n number of nests
T ri the reheat time constant in second for area i, xti the current solution
K ri the p.u megawatt rating of high pressure stage for area xtþ1
i
a new solution
i
Tw the hydro turbine time constant, List of abbreviations
K d , K p , K i the electric governor derivative, proportional and LFC Load Frequency Control
integral gains, respectively GA Genetic Algorithm
T Pi , K Pi the time constant and gain of power system respec- PSO Particle Swarm Optimization
tively for area i CS Cuckoo Search
DPtiei the difference between the actual tie-line power and FLC Fuzzy Logic Controller
scheduled one ANN Artificial Neural Network
B the biasing factor in pu MW/Hz ANFIS adaptive neuro-fuzzy inference system
K PPi , K IIi the gains of PI controller of area i PI proportional plus integral
N the number of area in power systems GRC Generation Rate Constraint
tsim the simulation time in second ACE Area Control Error
t time in second IAE the integral of absolute value of the error
T ij synchronizing coefficient ITAE the integral of the time multiplied absolute value of the
J objective function error
Ui the control signal of area i ISE the integral of square error
Ki the controller of area i ITSE the integral of time multiply square error

Three area power system [17]. The detailed designed model of 3 area hydro-thermal power
system for load frequency control is shown in Fig. 1. The thermal
The system under study consists of 3 areas of equal sizes. Areas plant has a single stage reheat steam turbine and the hydro plant
1 and 2 are reheat thermal systems and area 3 is a hydro system is equipped with an electric governor. In this model, nonlinearities

Generation Rate Constraint

Generation Rate Constraint

Generation Rate Constraint

Fig. 1. A three area model.

634 A.Y. Abdelaziz, E.S. Ali / Electrical Power and Energy Systems 73 (2015) 632–643

case [47]. Sometimes, it can reach to several seconds in conven-

Start
tional systems [4]. The system parameters are given in appendix.
The transfer functions of different blocks used in power system
Initialization of CS parameters model are given below [17,21]:
The transfer function of the hydraulic turbine is

Initial generation of host nest population T w S þ 1

: ð1Þ
0:5T w S þ 1
Transfer function of the hydraulic governor is
Evaluation of fitness function for the generated
host nest population K d S2 þ K p S þ K i
2
: ð2Þ
K d S þ ðK p þ f =RÞS þ K i
Move all cuckoos
Modification of nest position using Lévy towards better Transfer function of the governor is
flight equation environment
1
: ð3Þ
TgS þ 1
Evaluate the new solution
Transfer function of the steam turbine is
K rTrS þ 1
No : ð4Þ
Check the condition is TrS þ 1
satisfied? Transfer function of the Re-heater is
Yes 1
; ð5Þ
Print the result TtS þ 1
and transfer function of the Generator is

Stop
KP
: ð6Þ
TPS þ 1
Fig. 2. Flowchart of Cuckoos Search optimization algorithm. For the ith area, the Area Control Error (ACE) signal made by fre-
quency and tie-line power variations is stated by:

are represented in Generation Rate Constraint (GRC) and commu- ACEi ¼ B  Df i þ DPtiei : ð7Þ
nication time delay. GRC illustrates the limitation on the genera-
tion rate of change in the output generated power due to the Optimization model of load frequency control system
limitation of thermal and mechanical movements [4], for thermal
stations it is taken to be 0.1 Pu Mw per minute [40]. On the other For the three area considered, the conventional integral con-
hand, communication time delay is introduced by many signal pro- troller was exchanged by a PI one with the given equation:
cessing and data exchanging operations. In this paper, a time delay
between the control center and each unit is stimulated. Time delay K IIi
K i ðSÞ ¼ K PPi þ : ð8Þ
is in the order of 100 ms and can extend to 250 ms at the worst S

0.02
CS
GA
0.01 PSO
Integral

0
Change in f1

-0.01

-0.02

-0.03

-0.04

-0.05
0 10 20 30 40 50 60 70 80 90 100
Time in second

Fig. 3. Change in f 1 for case 1.

 A.Y. Abdelaziz, E.S. Ali / Electrical Power and Energy Systems 73 (2015) 632–643 635

The control signal is specified by equation: The regular ranges of the optimized gains are [2 to 2] as
obtained in [35–37] and the objective function are calculated
U i ðSÞ ¼ K i ðSÞACEi ðSÞ: ð9Þ
under disturbance of 1% in all areas.
The sum of time multiple absolute errors in ACE is considered as
performance index, hence J can be outlined as: Cuckoo Search algorithm
N Z
X tsim
J¼ tjACEi jdt; ð10Þ The breeding behavior of cuckoo species and the basic items of
i¼1 0
this algorithm are discussed below.
To enhance the system response in terms of the settling time
and overshoots, it is necessary to minify equation (10). The design Cuckoo breeding behavior
process can be formed as the following constrained optimization
problem. Lessen J subject to: CS is a metaheuristic search technique that has been introduced
by Yang and Deb in 2010 [42]. The algorithm is inspired by obligate
K min
PPi 6 K PPi 6 K max
PPi ; brood parasitism of cuckoo species by laying their eggs in the nests
ð11Þ
K min
IIi 6 max
K IIi 6 K IIi : of other host birds. At the proper minute, the hen cuckoo flies
down to the host’s nest, ejects one egg out of the nest, lays an

0.02
CS

GA

0.01 PSO

Integral

0
Change in f3

-0.01

-0.02

-0.03

-0.04

-0.05
0 10 20 30 40 50 60 70 80 90 100
Time in second

Fig. 4. Change in f 3 for case 1.

0.005

-0.005
Change in ACE1

-0.01

-0.015

-0.02

CS
-0.025 GA
PSO
Integral
-0.03
0 10 20 30 40 50 60 70 80 90 100
Time in second

Fig. 5. Change in ACE1 for case 1.

636 A.Y. Abdelaziz, E.S. Ali / Electrical Power and Energy Systems 73 (2015) 632–643

egg and flies off. This procedure takes about 10 s. A female may from a probability density function which has a power law tail.
visit up to 50 nests during a breeding season. The host birds may This procedure demonstrates the optimal random search pattern
detect that the eggs are not their own and either throw them away and is found in nature [45]. When generating a new egg, a Lévy
or abandon the nest. This has resulted in the evolution of cuckoo flight is done starting at the location of a randomly chosen egg,
eggs which mimic the eggs of local host birds [43]. Moreover, the if the objective function value at these new coordinates is better
timing of egg laying of some species is amazing. Parasitic cuckoos than another chosen egg then that egg is shifted to this new loca-
select a nest where the host bird just laid its own eggs. In general, tion. One of the advantages of CS over other optimization algo-
the cuckoo eggs hatch slightly earlier than their host eggs. It dis- rithms are that only one parameter, the fraction of nests to
lodges all host progeny from host nests. It is a larger bird than its abandon P a ; requires to be inspected. The use of Lévy flights as
hosts, and needs to monopolize the food supplied by the parents. the search method means that the CS can detect each optimum
The chick will roll the other eggs out of the nest by pushing them in a design space and it has been proved in comparison with
with its back. If the host’s eggs hatch before the cuckoos, the other algorithms [46].
cuckoo chick will push the other chicks out of the nest in an iden-
tical manner. Cuckoo Search implementation

Lévy flights To apply this search as an optimization algorithm, three pre-

sented approximation rules are adopted as below [45,46]:
The use of Lévy flights for local and global searching is a vital
portion of the CS [44]. The Lévy flight process is a random walk 1. Cuckoos select random nest for laying their eggs. Artificial
that is characterized by a series of instantaneous jumps elected cuckoo can lay only one egg at the time.

0.015
CS
GA
0.01 PSO
Integral

0.005
Change in ACE3

-0.005

-0.01

-0.015

-0.02
0 10 20 30 40 50 60 70 80 90 100
Time in second

Fig. 6. Change in ACE3 for case 1.

x 10-3
8
CS

7 GA
PSO
Integral
6

5
Change in P2

-1

-2
0 10 20 30 40 50 60 70 80 90 100
Time in second

Fig. 7. Change in P2 for case 1.

 A.Y. Abdelaziz, E.S. Ali / Electrical Power and Energy Systems 73 (2015) 632–643 637

0.1
CS
GA
PSO
0.05 Integral

0
Change in f1

-0.05

-0.1

-0.15

-0.2
0 10 20 30 40 50 60 70 80 90 100
Time in second

Fig. 8. Change in f 1 for case 2.

0.1
CS
GA
PSO
0.05 Integral

0
Change in f3

-0.05

-0.1

-0.15

-0.2
0 10 20 30 40 50 60 70 80 90 100
Time in second

Fig. 9. Change in f 3 for case 2.

2. Elitist selection process is applied, so only the eggs with highest where a > 0 is the step size which should be proportional to the
quality are passed to the next generation. scales of the optimization problem. The product  intends entry-
3. The number of available hosts nests is specified and a host wise walk during multiplications. Lévy flights supply a random
can reveal a foreigner egg with a probability Pa 2 ½0; 1. If walk while their random steps are drawn from a Lévy distribution
cuckoo egg is detected by the host, it may be thrown away, for large steps.
or the host may leave its own nest and commit it to the
cuckoo intruder.
vy  u ¼ tk ; ð1 < k < 3Þ;
Le ð13Þ
The last assumption can be approximated by a factor Pa of the n
nests being exchanged by new nests. The quality of a solution is which has an infinite variance with an infinite mean. Here the
relative to the objective function. Based on the previous rules, steps form a random walk process with a power law step-length
the flowchart of CS is displayed in Fig. 2. The parameters of CS distribution with a heavy tail. Some of the new solutions should
are shown in appendix. Other forms of fitness can be determined be generated by Lévy walk round the best solution; this will
in GA [48,49] and PSO [50,51]. accelerate the local search. However, a substantial fraction of
When generating a new solution xtþ1
i for the ith cuckoo, a Lévy the new solutions should be generated by far field randomization
flights is carried out and whose places should be far enough from the current best
solution, this will ensure the system will not be trapped in a local
xitþ1 ¼ xti þ a  Le
vyðkÞ; ð12Þ optimum.
638 A.Y. Abdelaziz, E.S. Ali / Electrical Power and Energy Systems 73 (2015) 632–643

Results and simulations conventional integral controller are suffered from high settling
time and unwanted oscillations. Moreover, the developed con-
Various comparative cases are tested in this section to prove the trollers are more successful in enhancing the damping feature of
efficiency of the developed CS algorithm for optimizing controller power system compared with PSO and GA. Thus, power system
parameters. The proposed CS methodology, PSO and GA are pro- oscillations are efficiently attenuated with the implementation of
grammed in MATLAB 7.1 with I5, CPU 2.53 GHz and 4.0 GB RAM. the developed controllers. Hence, CS provides more satisfactory
The obtained results are the best for all algorithms depending on results than other controllers.
value of J. The convergence times for CS, PSO and GA are 23.3,
28.7 and 40.2 s respectively. Case 2: step increase in all areas

Case 1: step increase in DPD1 In this case, a 1% step increment in demand of all areas simul-
taneously is applied as the second operating point. The signals of
A 1% step increment in demand of the first area (DPD1 ) is the closed loop system are introduced in Figs. 8–12. It is clear from
employed as the first operating point. The system responses are these Figures, that the system oscillations are poorly damped for
shown in Figs. 3–7. In these Figures, the responses with conventional controller. Also compared with PSO and GA, the

0.03
CS

0.02 GA
PSO
Integral
0.01

0
Change in ACE1

-0.01

-0.02

-0.03

-0.04

-0.05

-0.06

-0.07
0 10 20 30 40 50 60 70 80 90 100
Time in second

Fig. 10. Change in ACE1 for case 2.

0.04
CS
GA
PSO
0.02 Integral

0
Change in ACE3

-0.02

-0.04

-0.06

-0.08
0 10 20 30 40 50 60 70 80 90 100
Time in second

Fig. 11. Change in ACE3 for case 2.

 A.Y. Abdelaziz, E.S. Ali / Electrical Power and Energy Systems 73 (2015) 632–643 639

developed controllers have a lower settling time and system 10%. It is clear that the system is stable with the developed con-
response is driven back to zero speedily. Moreover, the seniority trollers. In addition, the designed controllers have the ability of
of the suggested algorithm in tuning controllers is demonstrated. adding adequate damping to the system oscillatory modes under
various conditions and the robustness of these controllers is
Case 3: parameter variation believed.

A parameter variation test is utilized to affirm the effectiveness Case 4: step response with different times delay
of the suggested CS algorithm in tuning controllers. Figs. 13–16
show the responses of frequency and ACE deviation of the second The communication time delay is added here, as a second
and third area respectively with variation in all time constants by source of nonlinearities, to investigate its effect on the closed loop

x 10-3
3
CS
2.5 GA
PSO
2 Integral

1.5

1
Change in P3

0.5

-0.5

-1

-1.5

-2

-2.5
0 10 20 30 40 50 60 70 80 90 100
Time in second

Fig. 12. Change in P3 for case 2.

0.02

-0.02

-0.04

-0.06
Change in f1

-0.08

-0.1

-0.12

-0.14

Normal
-0.16 Increase by 10%
Decrease by 10%

-0.18
0 5 10 15 20 25 30 35 40 45 50
Time in second

Fig. 13. Change in f 1 for case 3.

640 A.Y. Abdelaziz, E.S. Ali / Electrical Power and Energy Systems 73 (2015) 632–643

0.02

-0.02

-0.04

-0.06
Change in f3

-0.08

-0.1

-0.12

-0.14
Normal
-0.16 Increase by 10%
Decrease by 10%

-0.18
0 5 10 15 20 25 30 35 40 45 50
Time in second

Fig. 14. Change in f 3 for case 3.

0.01

-0.01

-0.02
Change in ACE2

-0.03

-0.04

-0.05

Normal
-0.06
Increased by 10%
Decrease by 10%
-0.07
0 10 20 30 40 50 60
Time in second

Fig. 15. Change in ACE2 for case 3.

system of LFC. The performance of the system with different times 3 Z

X 100

delay is offered in Figs. 17 and 18. The simulation results detect ITAE ¼ tðjDACEi jÞdt; ð15Þ
i¼1 0
that, the developed controllers provide adequate performance
and they compensate the time delay which donates a more 3 Z
X 100
enhancement in system performance. ISE ¼ ðDACEi Þ2 dt; ð16Þ
i¼1 0

Performance indices and robustness

3 Z
X 100
ITSE ¼ tðDACEi Þ2 dt ð17Þ
To prove the superiority of the developed controllers, some 0
i¼1
indices like IAE, ITAE, ISE and ITSE are being used as:
It is remarkable that the lower the value of these indices is,
3 Z
X 100 the better the system response in terms of time domain charac-
IAE ¼ ðjDACEi jÞdt; ð14Þ teristics. Parameters of controllers are recorded in Table 1 while
0
i¼1 the values of various indices are illustrated in Table 2. It can be
 A.Y. Abdelaziz, E.S. Ali / Electrical Power and Energy Systems 73 (2015) 632–643 641

0.02
Normal
0.01 Increase by 10%
Decrease by 10%
0

Change in ACE3 -0.01

-0.02

-0.03

-0.04

-0.05

-0.06

-0.07

-0.08
0 10 20 30 40 50 60
Time in second

Fig. 16. Change in ACE3 for case 3.

-0.02

-0.04

-0.06
Change in f1

-0.08

-0.1

-0.12

-0.14 Without time delay

Time delay=0.1 sec
Time delay=0.3 sec
-0.16 Time delay= 0.6 sec

0 5 10 15 20 25 30 35 40
Time in second

Fig. 17. Change in f 1 for case 4.

seen that the values of these system performance with the pro- Conclusions
posed controllers are smaller compared with those of PSO, GA
and conventional controller. This demonstrates that the overshoot CS algorithm is suggested in this paper to tune the param-
and settling time are largely minimized by applying the proposed eters of PI controllers for LFC problem. An integral time abso-
CS algorithm. lute error of the ACE for all areas is chosen as the objective
function to enhance the system response in terms of the set-
Merits of CS over others tling time and overshoots. The major contributions of this
paper are:
The settling times of disperse states for each area under differ-
ent operating conditions are shown in Table 3. It is clear that the 1. Establishment of the dynamic model for 3 area power system
settling time associate with CS is the smallest one compared with considering GRC and communication time delay as nonlineari-
PSO, GA and conventional integral control. Moreover, the values of ties sources with LFC based CS to assure the superiority of the
these times are smaller than those obtained in [17,21]. developed controllers over PSO, GA and conventional inte-
Consequently, CS based controllers provide better performance gral controller throughout different disturbances for various
than other controllers. Therefore, the proposed controllers signals.
approach using CS are more accurate and quicker than the other 2. The robustness of the developed controllers are confirmed
schemes for complex dynamical system. through parameter variations.
642 A.Y. Abdelaziz, E.S. Ali / Electrical Power and Energy Systems 73 (2015) 632–643

0.01

-0.01

-0.02
Change in ACE1

-0.03

-0.04

-0.05

-0.06
Without time delay
-0.07 Time delay=0.1 sec
Time delay=0.3 sec
Time delay=0.6 sec
-0.08
0 5 10 15 20 25 30 35 40 45 50
Time in second

Fig. 18. Change in ACE1 for case 4.

Table 1
computational time and gets trapped in local minimum solution.
Gains for different algorithms.
Also, PSO suffers from weak local search ability and algorithm
K PP1 K II1 K PP2 K II2 K PP3 K II3 may lead to possible entrapment in local minimum solutions.
CS 0.2674 0.2219 0.2194 0.1841 0.1818 0.0681 4. The capability of the developed controllers to compensate the
PSO 0.299 0.232 0.234 0.168 0.192 0.088 communication time delay and preserve its satisfactory perfor-
GA 0.2198 0.1997 0.1986 0.1587 0.1876 0.1976 mance is demonstrated.
5. The effectiveness of the developed controllers in terms of vari-
ous indices and settling time is proved.
Table 2
Values of performance indices.

IAE ITAE ISE ITSE

Appendix

CS 4.1453 28.9463 0.4882 2.7914

(a) The parameters of system under study are given below [17]:
PSO 4.2616 31.2114 0.4892 2.8225
GA 5.0199 44.7268 0.5297 3.5032 T P1 = T P2 = T P3 = 20 s; T t1 = T t2 = 0.3 s; T r1 = T r2 = 10 s; T 12 =
Integral 6.462 91.7536 0.5726 4.276 T 31 = T 23 = 0.545 p.u; T g1 = T g2 = 0.08 s; K P1 =K P2 =K P3 =
120 Hz/p.u MW; B1 = B2 = B3 = 0.425 p.u MW/Hz; a12 =a31
=a23 =-1; R1 =R2 =R3 =2.4 Hz/p.u MW; T W = 1 s; K r1 =K r2 =0.5
P.u MW; f =60 Hz; K d =4.0; K i = 5.0; K p =1.0.
3. CS outperforms PSO, GA in solving LFC problem due to only one (b) CS parameters: Max generation = 100; Number of nests = 50;
parameter required to fine-tune. Thus, it has the priority to reach P a =0.25.
the optimal solution rapidly. On the other hand, GA deals with a (c) GA parameters: Max generation = 100; Population size = 50;
population of solutions, thus leading to the disadvantage of Crossover probabilities = 0.75; Mutation probabilities = 0.1.
requiring a great number of function evaluations, large (d) PSO parameters: Max generation = 100; Population in
swarm = 50; C1 = C2 = 2; x = 0.9.

Table 3
Settling time of each variable for different controllers and operating conditions.

Area 1 demand change All areas change

Df 1 Df 3 DACE2 DP 2 Df 1 Df 3 DACE2 DP 3
CS 17.94 18.14 17.26 20.79 25.89 26.29 17.26 38.4
PSO 18.41 18.87 17.68 21.46 26.11 26.74 18.36 42.18
GA 21.49 31.88 20.71 46.75 27.71 27.49 20.71 47.88
Integral 48.97 50.84 69.88 59.8 56.15 63.29 69.94 68.74
ANFIS [17] 36.9 62.7 62.7 55.2 31.6 32.4 62.7 66.3
ANN [21] – – – – 35 45 – 45
Fuzzy [21] – – – – 50 50 – 50
 A.Y. Abdelaziz, E.S. Ali / Electrical Power and Energy Systems 73 (2015) 632–643 643

References [26] Ramakrishna K, Bhatti T. Sampled-data automatic load frequency control of a

single area power system with multi-source power generation. Electric Power
Compon Syst 2007;35(8):955–80.
[1] Elgerd O. Electric energy systems theory: an introduction. 2nd ed. New
[27] Bensenouci A. Optimal design of discrete output feedback control using
Delhi: Tata McGraw-Hill; 1983.
genetic algorithm for a multi area power system. J King Abdulaziz Univ Eng Sci
[2] Saadat H. Power system analysis. International edition. McGraw-Hill; 1999.
2009;20(2):3–26.
[3] Wood A, Wollenberg B. Power generation, operation and control. 2nd ed. John
[28] Daneshfar F, Bevrani H. Multiobjective design of load frequency control using
Wiley & Sons; 1984.
genetic algorithms. Int J Electr Power Energy Syst 2012;42:257–63.
[4] Bevrani H. Robust power system frequency control. New York: Springer; 2009.
[29] Golpîra H, Bevrani H. A framework for economic load frequency control design
[5] Mandour M, Ali E, Lotfy M. Robust load frequency controller design via genetic
using modified multi-objective genetic algorithm. Electr Power Compon Syst
algorithm and H1. Modern Electric Power Systems, Wroclaw, Poland MEPS’10
2014;42(8):788–97.
– Paper P16, 2010.
[30] Yousuf M, Al-Duwaish H, Al-Hamouz Z. PSO based single and two
[6] A. Bensenouci and A. Abdel Ghany, ‘‘Performance Analysis and Comparative
interconnected area predictive automatic generation control. WSEAS Trans
Study of LMI-Based Iterative PID Load-Frequency Controllers of a Single-Area
Syst Control 2010;5(8):677–90.
Power System’’, WSEAS Transactions on Power Systems, Vol. 5, No. 2, 2010, pp.
[31] Gozde H, Taplamacioglu M. Automatic generation control application with
85–97.
craziness based particle swarm optimization in a thermal power system. Int J
[7] Dey R, Ghosh S, Ray G, Rakshit A. H1 load frequency control of interconnected
Electr Power Energy Syst 2011;33:8–16.
power systems with communication delays. Int J Electr Power Energy Syst
[32] Kumari N, Jha A. Effect of generation rate constraint on load frequency control
2012;42:672–84.
of multi area interconnected thermal systems. J Electr Electron Eng Res
[8] Toulabi M, Shiroei M, Ranjbar A. Robust analysis and design of power system
2013;5(3):44–9.
load frequency control using the Kharitonov’s theorem. Int J Electr Power
[33] Sudha K, Vakula V, Shanthi R. PSO based design of robust controller for two
Energy Syst 2014;55:51–8.
area load frequency control with nonlinearities. Int J Eng Sci Technol
[9] Maher R, Mohammed I, Ibraheem I. Polynomial based H1 robust governor for
2010;2(5):1311–24.
load frequency control in steam turbine power systems. Int J Electr Power
[34] Saikia L, Nanda J, Mishra S. Performance comparison of several classical
Energy Syst 2014;57:311–7.
controllers in AGC for multi-area interconnected thermal system. Int J Electr
[10] Chidambaram IA, Velusami S. Design of decentralized biased controllers for
Power Energy Syst 2011;33:394–401.
load-frequency control of interconnected power systems. Electr Power
[35] Ali E, Abd-Elazim S. Bacteria foraging optimization algorithm based load
Compon Syst 2005;33(12):1313–31.
frequency controller for interconnected power system. Int J Electr Power
[11] Yazdizadeh A, Ramezani M, Hamedrahmat E. Decentralized load frequency
Energy Syst 2011;33(3):633–8.
control using a new robust optimal MISO PID controller. Int J Electr Power
[36] Ali E, Abd-Elazim S. BFOA based design of PID controller for two area load
Energy Syst 2012;35:57–65.
frequency control with nonlinearities. Int J Electr Power Energy Syst
[12] Hassan M, Abouelsoud A, Soliman H. Constrained load-frequency control.
2013;51:224–31.
Electr Power Compon Syst 2008;36(3):266–79.
[37] Panda S, Mohanty B, Hota P. Hybrid BFOA-PSO algorithm for automatic
[13] Sivaramakrishnan A, Hariharan M, Srisailam M. Design of variable structure
generation control of linear and nonlinear interconnected power systems. Int J
load frequency controller using pole assignment technique. Int J Control
Appl Soft Comput 2013;13:4718–30.
1984;40:487–98.
[38] Saini R, Gupta R, Parmar G. Optimization of LFC using bacteria foraging
[14] Ehsan H, Yazdizadeh A. Decentralized self-tuning pole placement controller
optimization algorithm. Int J Emerg Technol Comput Appl Sci 2013;3(2):133–8.
for load frequency control in KHOZESTAN Area. In: 3rd IEEE multi-Conference
[39] Naidu K, Mokhlis H, Bakar A. Multiobjective optimization using weighted sum
on System and Control, St. Petersburg, Russia; July 8–10, 2009. p. 571–5.
artificial bee colony algorithm for load frequency control. Int J Electr Power
[15] Vrdoljak K, Peri’c N, Petrovi’c I. Sliding mode based load-frequency control in
Energy Syst 2014;55:657–67.
power systems. Electric Power Syst Res 2010;80:514–27.
[40] Omar M, Soliman M, Abdel Ghany A, Bendary F. Optimal tuning of PID
[16] Rahmani M, Sadati N. Two-level optimal load-frequency control for multi-area
controllers for hydrothermal load frequency control using ant colony
power systems. Int J Electr Power Energy Syst 2013;53:540–7.
optimization. Int J Electr Eng Inf 2013;5(3):348–60.
[17] Khuntia S, Panda S. Simulation study for automatic generation control of a
[41] Yang X. Engineering optimization: an introduction with metaheuristic
multi-area power system by ANFIS approach. Int J Appl Soft Comput
applications. John Wiley and Sons; 2010.
2012;12:333–41.
[42] Yang X, Deb S. Engineering optimization by Cuckoo Search. Int J Math Model
[18] Farhangi R, Boroushaki M, Hosseini S. Load frequency control of
Numer Optimization 2010;1(4):330–43.
interconnected power system using emotional learning-based intelligent
[43] Yang X, Deb S. Cuckoo Search via Lévy Flights. In: Proceedings of world
controller. Int J Electr Power Energy Syst 2012;36:76–83.
congress on nature & biologically inspired computing (NaBIC 2009), USA 2009.
[19] Sudha K, Raju Y, Sekhar A. Fuzzy C-means clustering for robust decentralized
p. 210–4.
load frequency control of interconnected power system with generation rate
[44] Yang X, Deb S. Multiobjective Cuckoo Search for design optimization. Comput
constraint. Int J Electr Power Energy Syst 2012;37:58–66.
Oper Res 2013;40:1616–24.
[20] Sudha K, Santhi R. Load frequency control of an interconnected reheat thermal
[45] Basu M, Chowdhury A. Cuckoo Search algorithm for economic dispatch. Energy
system using type-2 fuzzy system including SMES units. Int J Electr Power
2013;60:99–108.
Energy Syst 2012;43:1383–92.
[46] El-Fergany A, Abdelaziz A. Capacitor allocations in radial distribution networks
[21] Prakash S, Sinha S. Load frequency control of three area interconnected hydro-
using Cuckoo Search algorithm. IET Gener, Transm Distrib 2014;8(2):223–32.
thermal reheat power system using artificial intelligence and PI controllers. Int
[47] Wu H, Tsakalis K, Heydt G. Evaluation of time delay effects to wide area power
J Eng Sci Technol 2011;4(1):23–37.
system stabilizer design. IEEE Trans Power Syst 2004;19(4):1935–41.
[22] Saikia L, Mishra S, Sinha N, Nanda J. Automatic generation control of a multi
[48] Sivanandam SN, Deepa SN. Introduction to genetic algorithms. Springer; 2008.
area hydrothermal system using reinforced learning Neural Network
[49] Houck C, Joines J, Kay M. A genetic algorithm for function optimization: a
Controller. Int J Electr Power Energy Syst 2011;33:1101–8.
matlab implementation. NCSU-IE TR 1995;95–05.
[23] Patel N, Jain B. Automatic generation control of three area power systems
[50] Poli R, Kennedy J, Blackwell T. Particle swarm optimization. Swarm Intell
using ANN controllers. Int J Comput Eng Res 2013;3(6):1–7.
2007;1(1):33–57.
[24] Abdel-Magid Y, Dawoud M. Optimal AGC tuning with genetic algorithms. Int J
[51] Oshaba AS, Ali ES, Abd-Elazim SM. MPPT control design of PV system supplied
Electr Power Syst Res 1997;38:231–8.
SRM using BAT search algorithm. Sustain Energy, Grids Networks
[25] Aditya S, Das D. Design of load frequency controllers using genetic algorithm
2015;2C:51–60.
for two area interconnected hydro power system. Electr Power Compon Syst
2003;31(1):81–94.