Improved Particle Swarm Optimization Based

Load Frequency Control In

A Single Area Power System
Saumya Kr. Gautam, Nakul Goyal
Department of Electrical Engineering, IT-BHU
Varanasi,India
[email protected]
[email protected]

Abstract— In this paper, an improved particle swarm complexity of the power systems such as nonlinear load
optimization (IPSO) based load frequency control (LFC) of a characteristics and variable operating points. According to
single area power system is presented. Although the Particle some authors, variable structure control [4] maintains stability
Swarm Optimization approaches have several advantages, they of system frequency. However, this method needs some
can still have drawbacks like local optimal trapping due to
information for system states, which are very difficult to know
premature convergence(i.e. exploration problem). This paper
proposes an improved PSO framework adopting a crossover completely. Also, the growing needs of complex and huge
operation scheme to increase exploitation capability of PSO.The modern power systems require optimal and flexible operation
study has been realized for control of a single area of them. The dynamic and static properties of the system must
interconnected power system with IPSO optimized self-tuning be well known to design an efficient controller. On the other
PID controller. The comparison between a conventional hand, to handle such a complex system is quite complicated
Proportional-Integral (PI) controller and the proposed PSO [5]. According to [6], conventional PID control schemes will
based controller showed that the proposed controller can not reach a high of control performances. Since the dynamic
generate a better transient response for a step load change. For behavior even for a reduced mathematical model of a power
this application, MATLAB-Simulink software is used.
system is usually nonlinear, time-variant and governed by
strong cross-couplings of the input variables, special care has
Keywords— Load Frequency Control, A single Area Power to be taken for the design of the controllers. For this reason,
system, Particle Swarm Optimization, Transient Response, PID recently, a lot of artificial intelligence based robust controllers
controllers such as genetic algorithm, tabu search algorithm, fuzzy logic
and neural networks are used for PID controller parameters
I. INTRODUCTION
tuning in LFC by authors [7, 8, 16, 17]. Since, Particle Swarm
The dynamic behavior of many power systems and resulted Optimization algorithm is an optimization method that finds
in industrial loads heavily depends on disturbances and in the best parameters for controller in the uncertainty area of
particular on changes in the operating point [1]. Load controller parameters and obtained controller is an optimal
frequency control in power systems is very important in order controller, it has been used in almost all sectors of industry
to supply reliable electric power with good quality. The goal and science. One of those areas is the load frequency control
of the LFC is to maintain stable system frequency which has as shown in [10].
zero steady state errors, and to provide load sharing between In this study, it is used to determine the parameters of a PID
areas in a multi area interconnected power system. In addition, controller according to the system dynamics changing with
the power system should fulfill the proposed dispatch daily period. In addition, for different applications, this
conditions. Power systems are divided into control areas method does not require a certain model [3]. In the integral
connected with tie-lines. All generators are supposed to controller, if the integral gain, Ki, is very high, undesirable
constitute a coherent group in each control area. From the and unacceptable large overshoots will be occurred [3].
experiments on the power system, it can be seen that each area However, adjusting the maximum and minimum values of
needs its system frequency to be controlled [2]. proportional (Kp) and integral (Ki) gains respectively, the
Generally, ordinary LFC systems are designed with outputs of the system (voltage, frequency) can be improved.
Proportional-Integral (PI) controllers. However, since the “I” In this simulation study, Kp is made equal a regulation
control parameters are usually tuned, it is incapable of constant “R” to obtain robustness, and it is shown that the
obtaining good dynamic performance for various load and overshoots and settling times with the proposed Improved
system change scenarios. Many studies have been carried out PSO tuned PID controller are better than the outputs of the
in the past about the load frequency control. In literature, other controllers.
some control strategies have been suggested based on the
conventional linear control theory [3]. These controllers may
be unsuitable in some operating conditions due to the

978-1-4244-9074-5/10/$26.00 ©2010 IEEE 2010 Annual IEEE India Conference (INDICON)

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on September 21,2021 at 20:08:29 UTC from IEEE Xplore. Restrictions apply.
 II. LOAD FREQUENCY CONTROL In a physical n-dimensional search space, the position and
velocity of individual i are represented as the vectors Xi =
The objectives of the LFC are to maintain reasonably ( xi,…, xin) and Vi = (vi,….vin) in the PSO algorithm. Let
uniform frequency, to divide the load between generators, and Pbesti = ( xi1pbest,….xinpbest ) and Gbesti = ( xigbest,…xngbest ) be
to control the tie-line interchange schedules. Basically, single the best position of individual i and its neighbors’ best
area power system consists of a governor, a turbine and a position so far, respectively. Using the information, the
generator with feedback of regulation constant. System also updated velocity of individual i is modified under the
includes step load change input to the generator. This work following equation in the PSO algorithm:
mainly related with the controller unit of a single area power
system. Simple block diagram of a single area power system Vik+1 = ωVik + c1r1×(Pbestik – Xik) + c2r2× (Gbestik – Xik)
with the controller is shown in Figure 1. (2)
Where,
Vik : velocity of individual i at iteration k,
ω : inertia weight parameter,
c1, c2 : acceleration coefficients,
r1, r2 : random numbers between 0 and 1,
Xik : position of individual i at iteration k,
Pbestik : best position of individual i until iteration k,
Gbestik : best position of the group until iteration k.

In this velocity updating process, the values of parameters

Fig.1. A single area power system with the controllers (ΔPL = 0.01)
such as ω, c1 and c2 should be determined in advance. In
So far, PID controllers have widely been used in process general, the weight ω is set according to the following
control. With simple structure, they yet can effectively control equation [9]:
various large industrial processes. There are many tuning ω = ωmax – (ωmax–ωmin)×iter/Itermax (3)
approaches for these controllers, but each has own where,
disadvantages or limitations. As a result, the design of PID ωmax ,ωmin initial and final weights,
controllers still remains a remarkable challenge for Itermax maximum iteration number,
researchers. In simple words, the PID controller is used to Iter current iteration number.
improve the dynamic response as well as reduce or eliminate
the steady-state error. The derivative term normally adds a Each individual moves from the current position to the next
finite zero to the open loop plant transfer function and can one by the modified velocity in (2) using the following
improve the transient response in most cases. The integral equation:
term adds a pole at origin resulting in increasing the system Xik+1=Xik +Vik+1 (4)
type and therefore reducing the steady-state error. Fig. 2 shows the concept of the searching mechanism of
Furthermore, this controller is often regarded as an almost PSO using the modified velocity and position of individual i
robust controller. As a result, they may also control uncertain based on (2) and (4) if the values of ω, c1, c2, r1, r2 are 1.
processes. The well-known PID controller transfer function is
as follows :

+ + ×s (1)

III. OVERVIEW OF PARTICLE SWARM OPTIMIZATION

Kennedy and Eberhart developed a PSO algorithm based
on the behavior of individuals (i.e., particles or agents) of a
swarm [11]. Its roots are in zoologist’s modeling of the Fig. 2
movement of individuals within a group. It has been noticed
that members of the group seem to share information among IV. IMPROVED PARTICLE SWARM OPTIMIZATION WITH
them, a fact that leads to increased efficiency of the group [12]. CROSSOVER OPERATION
The PSO algorithm searches in parallel using a group of In order to increase the diversity of a population, the
particles. Each particle corresponds to a candidate solution to crossover operation is newly introduced to the PSO
the problem. A particle moves toward the optimum based on
mechanism[14], thereby can effectively explore and exploit
its present velocity, its previous experience, and the
promising regions in a search space. The position of particle i,
experience of its neighbors. In an -dimensional search space,
Xi = (xi1,....xin) , obtained in (4) is mixed with Pbesti to
the position and velocity of particle are represented as vectors. generate a trial vector X^i = (x^i1..., x^in) as follows :

2010 Annual IEEE India Conference (INDICON)

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on September 21,2021 at 20:08:29 UTC from IEEE Xplore. Restrictions apply.
 X^ijk+1 = xijk+1 , if rij < CR In this simulation, the objective is to minimize the error and
= xijp,k , otherwise (5) the maximum overshoot. For this reason the objective
function is chosen as the Integral Square Error (ISE). The ISE
For j=1,2,….n, where rij is a uniformly distributed random squares the error to remove negative error components [13].
number between [0,1], and CR is the crossover rate in the T
range of [0,1]. When the value of CR becomes one, there is no ISE = ∫ e 2 (t )dt
0
crossover like in the conventional PSO. If the value of CR is
zero, the position will always have the crossover operation
similar to the GA mechanism.A proper crossover rate CR can
be determined by empirical studies to improve the diversity of
a population. Fig. 3 gives an example of the crossover
mechanism for an individual i.
The trial vector X^ijk+1 is used to update the Pbesti and
Gbesti at iteration k+1 using the greedy selection. Pbestik+1 is
set to X^ijk+1 if the fitness value of X^ijk+1 is better than that of
Pbestik . The developed crossover operation is applied for the
improvement of Pbesti while the PSO evolution process of
each particle is conserved by (4).
Fig. 4. A single area power system with proposed IPSO-PID controller.

VI. SIMULATION RESULTS

The ordinary single area power system parameters
consisting of the speed governor, turbine and generator are
given in Table 1. Here the governor free operation is assumed
and load demand ΔPL = 0.01 .
TABLE I

Description Parameter Value

Governer Gain Kh 1
Governer Time Th 80e-3
Fig.3. Illustration of crossover operation
Constant
V. IMPLEMENTATION OF IPSO FOR PID TUNING IN LOAD Turbine Gain Kt 1
FREQUENCY CONTROL
Turbine Time Tt 0.3
constant
A. Model with proposed PID controller Load Model Gain Ky 120
Load Time constant Ty 20
The framework of PSO based self-tuning PID controller is
depicted as Figure 3. To find the optimum parameters (Kp, Ki,
At the simulation, the population size is taken 10. c1 and c2
Kd) of PID controller, PSO program should search in 3-
dimensional search space. In an ordinary load frequency constants are taken 2. ωmax = 0.9 and ωmin = 0.4 . CR is
control systems, since a regulation constant R is used as Kp taken to be 0.6. For conventional PI controllers Ki is taken as
parameter in PID controller, especially I (integral) controller 0.65 and R = 2.4 .The values of PID Parameters as obtained
is used in LFC systems. At the proposed system, Kp is made by IPSO optimization :
equal regulation constant R, and ID (integral-derivative) Kp = R = 3.935;
controller is used in LFC system. Thus, for robustness, Ki = 8.1472;
regulation constant is tuned according to load and system Kd = 1.5761;
changes, too. With the optimized parameters based on PSO TABLE III
algorithm, the proposed PID controller of the LFC can achieve
Controllers Settling Maximum
optimal properties. The block diagram of a single area power Times(sec) Overshoot(Hz)
system with this controller is shown in Figure 4. Conventional PI 13.5 sec 0.0264
controller
B. Performance index for IPSO
Proposed IPSO-PID 8 sec 0.003330
During the simulation study, error signal which is required controller
for the controller is transferred to PSO software with error.mat
component. All positions of particles on each dimension are Simulation results for the single area power system are
clamped in limits which are specified by the user, and the shown in Table 2. As can be observed, the settling time and
velocities are clamped to the range [ Vmax Vmin ]. overshoots(around 10 times) with the proposed PSO-PID

2010 Annual IEEE India Conference (INDICON)

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on September 21,2021 at 20:08:29 UTC from IEEE Xplore. Restrictions apply.
controller are much shorter than that of with the conventional VIII. REFERENCES
PI controller. Therefore, the proposed PSO-PID controller
provides better performance than conventional PI controller for [1] Unbehauen, H., Kocaarslan, I., Experimental Modelling and Adaptive
the single area power system. Power Control of a 750 MW Once-Through Boiler, Proceedings of 11th
IFAC World Congress, Tallin, SU, Vol. 4, pp. 226-231, 13-17 August
1990.
[2] C.S., Chang, Weihui Fu, Area load frequency control using fuzzy gain
scheduling of PI controllers, Electric Power systems Research, 42, pp.
145-152, 1997.
[3] A.Kumar, O.P.Malik, G.S.Hope, Variablestructure- system control
applied to AGC of an interconnected power system, IEE Proceedings,
Vol. 132, Pt. C, No. 1, pp. 23-29, January 1985.
[4] Z.M.Ai-Hamouz, Y.L.Abdel-Magid, Variablestructure load frequency
controllers for multi area power systems, Int. J. Electr. Power Energy
Syst., 15 (5), pp. 23-29, 1993.
[5] Unbehauen, H., Kocaarslan, I.,Experimental Modelling and Simulation
of a Power Plant, Proceedings of European Simulation Multi
Conference, Nürnberg, Germany, pp. 474 - 478, 10-13 June 1990.
[6] Unbehauen, H., Keuchel, U., Kocaarslan, I., Real- Time Adaptive
Control of Electrical Power and Enthalpy for a 750 MW Once-
Through Boiler, Proceedings of IEE International Control Conference
91, Edinburg, Scotland, Vol.1, pp. 42-47, 25-28 March 1991.
[7] Shayeghi, H., Jalili, A., Shayanfar, H.A., Robust Modified GA Based
Multi-Stage Fuzzy LFC, Elsevier Energy Conversion and Management
Fig.5 Transient Response with PI controller 48, 1656–1670, 2007
[8] Al-Hamouz, Z., Al-Musabi, N., Al-Duwaish, H., A Tabu Search
Approach For The Design of Variable Structure Load Frequency
Controller Incorporating Model Nonlinearities, Journal of Electrıcal
Engıneerıng, Vol. 58, No. 5, 264–270, 2007.
[9] J. B. Park, K. S. Lee, J. R. Shin, and K. Y. Lee, “A particle swarm
optimization for economic dispatch with nonsmooth cost functions,”
IEEE Trans. on Power Systems, Vol. 20, No. 1, pp. 34-42, Feb. 2005.
[10] Taher, S.A., Hemati, R., Abdolalipour, A., Tabie, S.H., Optimal
Decentralized Load Frequency Control Using HPSO Algorithms in
Deregulated Power Systems, American Journal of Applied Sciences 5
(9): 1167-1174, 2008.
[11] J. Kennedy and R. C. Eberhart, “Particle swarm optimization,”
Proceedings of IEEE International Conference on Neural Networks
(ICNN’95), Vol. IV, pp. 1942-1948, Perth, Australia, 1995.
[12] J. Kennedy and R. C. Eberhart, Swarm Intelligence, San Francisco, CA:
Morgan Kaufmann Publishers, 2001
[13] J. H. Ahn, S. Choi, and J. H. Oh, “A new way of PCA: Integrated-
squared-error minimization,” IEEE Trans. Neural Networks, 2003,
submitted.
Fig.6 Transient Response with IPSO-PID controller [14] J. B. Park, Y. W. Jeong, Joong-Rin Shin, K. Y. Lee “An improed
Particle Swarm Optimization for Non-Convex Economic Load
VII. CONCLUSIONS Dispatch Problems” IEEE Trans. On Power Systems, Vol.25,
no.1 ,February 2010.
[15] Y. L. Abdel-Magid and M. M. Dawoud, “Optimal AGC tuning with
In this study, a new particle swarm optimized LFC has the genetic algorithms,” Elect. Power Syst. Res., vol. 38, no. 3, pp.
been investigated for automatic load frequency control of a 231–238, 1996.
single area power systems. For this purpose, first, to obtain [16] J. Talaq and F. Al-Basri, ‘‘Adaptive fuzzy gain scheduling for load
frequency control’’, IEEE Trans. Power Syst., vol. 14, no. 1, pp. 145---
more adaptive tuning mechanism for the PID controller 150, Feb. 1999.
parameters and sensitivity of the system is increased. It has [17] C. S. Chang and W. H. Fu, “Area load frequency control using fuzzy
been shown that the proposed control algorithm is effective gain scheduling of PI controllers,’’ Elect. Power Syst. Res., vol. 42, no.
2, pp. 145---152, 1997.
and provides significant improvement in system performance.
Therefore, the proposed IPSO-PID controller is recommended
to generate good quality and reliable electric energy. In
addition, the proposed controller is very simple and easy to
implement since it does not require many information about
system parameters. Two area power systems operation will be
investigated in future. In addition, comparison of the proposed
IPSO-PID controller with PID tuning by Differential
Evolution and Genetic Algorithm will be subject to the future
work. Further research on checking the robustness of multi-
area systems and on decentralized PID tuning considering the
tie-line interchange power is under progress.

2010 Annual IEEE India Conference (INDICON)

Authorized licensed use limited to: Rochester Institute of Technology. Downloaded on September 21,2021 at 20:08:29 UTC from IEEE Xplore. Restrictions apply.