Multi-Objective GA Based PID Controller for
Load Frequency Control in Power Systems
M. A. Tammam** M. A. S. Aboelela*
* Department of Electrical Power and Machines, Faculty of Engineering, Cairo University, Egypt
**Invensys Process Systems, Cairo, Egypt
Correspondence Email: aboelelamagdy@yahoo.com
Abstract: This paper studies control of load
frequency in single area power system with PID
controller. In this study, PID parameters are
improved using the multi-objective genetic
algorithm technique. The proposed controller
compared with a conventional PID controllers tuned
by Ziegler-Nicholas technique, Particle Swarm
Optimization (PSO). The effectiveness of the
anticipated scheme is confirmed throw the
comparison of steady state response characteristics.
For this study, MATLAB-Simulink software is used.
Keywords: Load Frequency Control, Single Area
Power System, PID Controller, Multi-Objective
Genetic Algorithm
I. INTRODUCTION
Frequency is a key stability criterion in power
systems. To provide the stability, active power balance
and constant frequency are required. Frequency depends
on active power balance. If any change occurs in active
power demand/generation in power systems, frequency
cannot be held in its rated value. So oscillations
increase in both power and frequency. Thus, system
subjects to a serious instability problem. To improve the
stability of the power networks, it is necessary to design
a Load Frequency Control (LFC) system that controls
the power generation and active power [1].
Generally, ordinary LFC systems are designed with
Proportional-Integral (PI) controllers. However, since
the “I” control parameters are usually tuned; it is
incapable of obtaining good dynamic performance for
various load and system changes. Many studies have
been carried out in the past on this important issue in
power systems, which is the load frequency control. As
stated in some literature [2], some control strategies
have been suggested based on the conventional linear
control theory. These controllers may be inappropriate
in some operating conditions. This could be due to the
complexity of the power systems such as nonlinear load
characteristics and variable operating points.
M. A. Moustafa* A. E. A. Seif*
In this study, multi-objective genetic algorithm is
used to determine the parameters of a PID controller
according to the system dynamics. Adjusting the
maximum and minimum values of proportional (
 ),
integral (
 ) and integral (
 ) gains respectively, the
outputs of the system (voltage, frequency) could be
improved.
In this simulation study, a single area power system
is chosen and load frequency control of this system is
made by genetic based PID controller.
II. OVERVIEW ON GENETIC ALGORITHM
The Genetic Algorithm (GA) is an optimization and
search technique based on the principles of genetics and
natural selection. The GA allows a population
composed of many individuals to evolve under
specified selection rules to a state that maximizes the
“fitness” (i.e., minimizes the cost function) [3-5].
The GA begins, like any other optimization
algorithm, by defining the optimization variables, the
cost function, and the cost. It ends like other
optimization algorithms too, by testing for convergence;
a flow chart of the components of the GA is shown in
Error! Reference source not found..
Selecting Genetic Algorithm parameters like
mutation rate, and population size is very difficult due
to the many possible variations in the algorithm and
cost function. A Genetic Algorithm relies on random
number generators for creating the population, mating,
and mutation. A different random number seed
produces different results. In addition there are various
types of crossovers and mutations, as well as other
possibilities, like chromosome aging and Gray codes.
Comparing all the different options and averaging the
results to reduce random variations for a wide range of
cost functions is a daunting task. Plus the results may be
highly dependent on the cost function analyzed.
 Define cost function
Select GA parameters

Generate initial

Find cost for each

chromosome

Select mates

Crossover

Mutation

N
o Converge
nce Check

Done

Figure 1: Flow Chart of Genetic Algorithm

Figure 2: One Area Power Generation Model

 III. MULTI-OBJECTIVE GA
Multi objective formulations are realistic models for
many complex engineering optimization problems. As
soon as there are many (possibly conflicting) objectives
to be optimized simultaneously, there is no longer a
single optimal solution but rather a whole set of
possible solutions of equivalent quality [6-7].
  
  
  (1)
A reasonable solution to a multi objective problem is
to investigate a set of solutions, each of which satisfies
the objectives at an acceptable level without being
dominated by any other solution.
Being a population based approach, GA are well
suited to solve multi-objective optimization problems.
A generic single-objective GA can be modified to find a
set of multiple non-dominated solutions in a single run.
The ability of GA to simultaneously search different
regions of a solution space makes it possible to find a
diverse set of solutions for difficult problems with non-
convex, discontinuous and multi-modal solutions
spaces. The cross over operator of GA may exploit
structures of good solutions with respect to different
objectives to create new non-dominated solutions [8-
10].
The goal of MOO is to find as many of these
solutions as possible. If reallocation of resources cannot
improve one cost without raising another cost, then the
solution is Pareto optimal. A Pareto GA returns a
population with many members on the Pareto front. The
  ,  (4)
population is ordered based on dominance . Several
different algorithms have been proposed and
successfully applied to various problems such as [5]:
Vector-Evaluated GA (VEGA), Multi Objective GA
(MOGA), A Non-Dominated Sorting GA (NSGA) and
Non-Dominated Sorting GA (NSGA II) which is used
in the proposed research.
IV. PROBLEM FORMULATION
Non-reheat type single area thermal generating
system represents by block diagram of closed loop
controlled system model. As shown in Figure 2,  is
the system frequency (Hz),  is regulation constant
(Hz/unit),  is speed governor time constant (sec), 
Where K ? , K A , K B are the PID controller parameters.
is turbine time constant (sec),  is inertia constant (s)
(LFC) using an assumed plant may not ensure the
stability of the overall real system.
For the single area non-reheat thermal system
considered in this study, the conventional Proportional
integral (PI) controller was replaced by a PID controller
with the following structure [13-14]:


Where
 is proportional gain,
 and
 are integral
and derivative time constants, respectively.
In this simulation, the objective is to minimize the
cost function. For this reason the objective function is
chosen as the Integral Square Error (ISE). The ISE
squares the error to remove negative error components.
  ∑   
 (2)
Where e(t) represents the deviation in the frequency ∆f.
The minimization fitness function becomes
!"#    $ % &. (  ) %  (3)
Where &. ( is the percentage overshoot,  is the
settling time (sec.), and $ and ) are positive real
numbers taken normally as 0.5 and 0.5.
The control signal for the conventional PID controller
in the sense of ISE only can be given in the following
equation.
*   +
  
*   -
  
 . /∑  
0 (5)

Based on these objectives; the multi-objective
optimization problem can be stated as:
1!#!2!3 !"#
Subjected to:

45 6
 6
478

45 6
 6
478 6

45 6
 6
478
and  is area parameter (Mw/Hz) [11-12]. Percentage of overshoot (P.O) and settling time (Ts)
are two more objective functions have been added to the
Basically, electric power system components are non- ISE performance index to define the multi-objective
linear; therefore a linearization around a nominal genetic algorithm problem.
operating point is usually performed to get a linear
system model which is used in the controller design The nominal system parameters in Figure 3 are:
process. The operating conditions of power systems are
continuously varying. Accordingly, the real plant
 1,   0.08 F,

 1, 
 0.3 F,
usually differs from the assumed one. Therefore,
classical algorithms to design a Load Frequency Control

H  120, H  20 F,   2.4
 Figure 3: Single Area Power System Simulink Model with Genetic based PID Controller

V. RESULTS AND DISCUSSIONS

By using multi-objective genetic algorithm technique performance in terms of overshoot and setting time [15-
in conjunction with equation (1)-(5), optimal controller 16].
parameters were obtained as shown in Table 1.
This shows the efficiency of the multi-objective
Table 1: PID Controller Parameters using Genetic genetic algorithm tuned PID controller over the
Algorithm Technique performance of the PSO based controller, the
PID parameters


 conventionally tuned PID controller (Ziegler-Nichols
method) and conventional PI controller.
values 3.8192 2.2784 4.0498

Performance of the multi-objective genetic algorithm

based PID controller was compared with the PSO based
self tuning PID controller developed in [6], the
conventionally tuned PID controller (Ziegler-Nichols
method), and conventional PI controller.

Table 2: Comparison between Proposed Genetic based

PID, PSO based PID, Ziegler-Nicholas Tuned PID &
Conventional PI Controller.
Controller Settling Peak
Time Overshoot
Proposed Genetic-PID 3.5 0.002400
PSO based PID 11.5 0.002662
Ziegler-Nichols PID 17.3 0.009763
Conventional PI 13.5 0.027350 Figure 4: Frequency Deviation of Single Area Power
System Compared to Conventional PID and PI
Figure 4 shows time response with the conventionally controllers
tuned PID controller (Ziegler-Nichols method) and the
conventional PI controller.

Figure 5 shows time response with PSO based PID

controller; System was simulated for 20 seconds with
step change of 0.01 p.u.

Figure 6 shows the time domain performance of the

system under the proposed multi-objective genetic
algorithm based PID controller with step change of 0.05
p.u.

At the simulation, the multi-objective genetic

algorithm was run for 1000 generations with a
population size of 100; As seen in the time response, the
genetic algorithm tuned controller gives better Figure 5: Frequency Deviation of Single Area Power
System Compared to PSO-PID controller
 algorithms: A tutorial". Reliability Engineeringand
System Safety , 91 (992–1007), 992-1007 (2006).
[7] 5Ivo F.Sbalzarini, Sibylle Muller and Petros
Koumoutsakos. Multi Objective Optimization
Using Evolutionary Algorithms. Center for
Turbulence Research Proceedings of the Summer
Program (2000).
[8] Mehdi Nikzad, R. Hemmati, S. A. Farahani and S.
M. Boroujeni." Comparison of Artificial
Intelligence Methods for Load Frequency Control
Problem". Australian Journal of Basic and Applied
Sciences , 4910-4921 (2010).
[9] M. LÜY, İ. K. Load Freqency Control In A Single
Area Power System By Artificial Neural Network
(ANN). University Of Pitesti, Electronic And
Figure 6: Frequency Deviation of Single Area Power Computers Science, Scientific Bulletin, No. 8,
Vol.2, , ISSN-1453-1119 (2008).
System with ∆LM = 0.05 p.u
[10] Haluk GÖZDE1, M. C. TAPLAMACIOĞLU, İ.
KOCAARSLAN, E. ÇAM. "Particle Swarm
VI. CONCLUSION
Optimization Based Load Frequency Control in A
Single Area Power System". University Of Pitesti –
In this proposed study, a new multi-objective genetic
Electronics And Computers Science, Scientific
algorithm based PID has been investigated for
Bulletin, No. 8, Vol. 2 , 1453–1119 (2008).
automatic load frequency control of a single area power
[11] Elgerd, O. I. Electric Energy Systems Theory.
system. For this purpose, first, more adaptive tuning
London: McGrawhill Book Company (1983).
mechanism for the PID controller parameters is
[12] P.Kundur. Power System Stability And Control.
obtained. It has been shown that the proposed control
New York: McGraw-Hill (1994).
algorithm is effective and provides significant
[13] Burns, S. Advanced Control Engineering.
improvement in system performance. Therefore, the
Plymouth - UK: A Division of Reed Educational
proposed multi-objective genetic algorithm based PID
and Profession Puplishing LTD., ISPN
controller is recommended to generate good quality and
0750651008 (2001).
reliable electric energy. In addition, the proposed
[14] Richard Bishop, C. D. Modern Control Systems.
controller is very simple and easy to implement since it
Colorado, USA: Prentice Hall (2008).
does not require many information about system
[15] Yildiz, C., Yilmaz, A., & Bayrak, M. "Genetic
parameters. Comparison of the proposed multi-
Algorithm Based PI Controller For Load Frequency
objective genetic algorithm based PID controller with
Control In Power Systems". Proceedings of 5th
conventional PID controllers was presented.
International Symposium on Intelligent
Manufacturing Systems, (pp. 1202-1210).
Bibliography
Kahramanmaraş, Turkey (2006).
[16] Mohamed A. M. M. Tammam, "Multi objective
[1] Bevrani, H. Robust Power System Frequency
genetic algorithm controller’s Tuning for load
Control. Brisbane, Australia: Springer Science +
frequency control in Power systems", Unpublished
Business Media, LLC (2009).
M. Sc. Thesis, Cairo University (2011).
[2] H. Shayeghi a, H. S. "Load Frequency Control
Strategies A State Of The Art Survey for The
Researcher". Energy Conversion and Management
Journal , 344-353 (2009).
[3] R. L. Haupt. Practical Genetic Algorithms - Second
Edition. Hoboken, New Jersey: A John Wiley &
Sons, Inc., Publication (2004).
[4] Darrell Whitley. A Genetic Algorithm Tutorial.
Colorado - USA: Computer Science Department,
Colorado State University (2005).
[5] Kalyanmoy Deb, Samir Agrawal, Amrit Pratap and
T. Meyarivan. A Fast Elitist Non-Dominated
Sorting Genetic Algorithm for Multi-Objective
Optimization :NSGA-II. Kanpur, India: Indian
Institute of Technology Kanpur, (KanGAL Report
No.200001) (2000).
[6] 4Abdullah Konak, D. W. Coit and Alice E. Smith.
"Multi-objective optimization using genetic