Advanced Optimization of Single Area Power Generation System using Adaptive Fuzzy Logic and PI Control

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 08 | Aug -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 1789
Advanced Optimization of Single Area Power Generation
System Using Adaptive Fuzzy Logic and PI Control
,
1Graduate Student, Dept. of Electrical Engineering, Rochester Institute of Technology, Dubai, UAE
2Professor, Dept. of Electrical Engineering, Rochester Institute of Technology, Dubai, UAE
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - In this paper, the open loop single area power
generation system is modelled using state space
representation. The output response which is frequency
deviation at steady state is simulated using MATLAB. Then,
Proportional Integral (PI) controller combined with
Adaptive Fuzzy Logic (FL) controller is added to the system
to understand the effect of conventional and modern control
on system steady state output response. The performance of
the system steady state output response is measured in
terms of undershoot percentage, settling time, and steady
state error. Simulation of the controlled system shows that
PI controller combined with Adaptive FL controller are
considered the most efficient, reliable, and robust type of
controller in addressing power generation optimization
problem. The output response of the controlled system has
settling time of 2.5 second, zero steady state error, and
undershoot of 0.03%.
Key Words: Optimization, Single Area Power
Generation System, Adaptive Fuzzy Logic Control, PI
Control, Steady State Output Response, Frequency
Deviation
1 INTRODUCTION
An interconnected system called Automatic Generation
Control (AGC) consists of two sub-systems: Load
Frequency Control (LFC) and Automatic Voltage Regulator
(AVR). AVR is responsible to regulate the terminal voltage
and LFC is employed to control the system frequency. In
this paper, modelling and simulation of LFC is considered
for careful analysis since LFC is more sensitive to load
changes compared to AVR. There is only weak coupling
between the two sub-systems; hence, the overlap of load
frequency and excitation voltage is negligible and the two-
sub-systems can be analyzed independently. Figure 1
illustrates how AVR and LFC are interconnected in AGC
system [1].
Fig- 1: Two Main Sub-Systems of Automatic Generation
Control
Optimizing thermal power generation system will reduce
energy or fuel consumption. Fuel reduction of even a small
percentage will lead to large energy saving which results
into saving the environment [2]. Hence, many researchers
have been interested to solve optimization problem in
thermal power generation systems. There are many
papers on optimization of two and three area thermal and
solar power generation systems. This paper is focused
exclusively on optimization of single area thermal power
generation system.
2 OPEN LOOP ANALYSIS
Figure 2 shows SIMULINK generated block diagram
representation of an uncontrolled generating unit which
consists of a speed governor, a turbine, and a generator
[1].
In some generating units, no re-heat component is
available. Re-heat or feed water re-heat is used to pre-heat
the water that is delivered to the steam boiler. In this
paper, the considered model does not have re-heat
component.
For computational simplicity in optimization problem, the
case where the thermal power generation system consists
of a single boiler, a single turbine, and a single generator is
considered. In many real world power generation systems,
the generation unit consists of multiple boilers, steam
turbines, and generators. “Network Power Loss” is
referred to the loss of power from one generator to

another or from one turbine to another. This loss of power
is experienced in systems with multiple components of
same type [3].
Fig-2: Block Diagram of Uncontrolled Single Area Power
Generation System
The inputs of the system shown in Figure 2 are Δ
representing the change in speed generation by utility and
Δ representing the change in load by consumer also
known as disturbance. Since user has no control over load
changes, Δ is considered as the only input of the system.
Effect of Δ diminishes once a controller is added to the
system.
The output of LFC is Δf which represents the change or
variation in steady state frequency. The objective is to
have a constant output frequency which corresponds to Δf
being zero or very small. The value of Speed Regulation R
also known as Droop is the ratio of frequency deviation
(Δf) to change in power output of the generator.
The uncontrolled system shown in Figure 2 is modelled
using state space representation shown in Equation 1 and
2 where A is the state matrix, B is the input matrix, and C is
the output matrix. X(t) is a column vector representing the
state variables used in system modelling. .
̇( )=A x(t) + B Δ (1)
y(t)= C x(t) (2)
The system shown in Figure 2 has 3 integral blocks which
corresponds to 3 state variables. Therefore, state matrix A
must be of size 3x3. Since the system has only one input
which is Δ the input matrix B must be a column vector
of size 3x1. The input is taken as unit step function. The
output of the system is frequency deviation of the
generator which corresponds to the state variable as
shown in Figure 2. The output matrix C is a row vector of
size 1x3.
To obtain state space representation of the system,
following transfer functions are developed:
Generator: = (3)
Turbine: = (4)
Governor: = (5)
Inverse Laplace Transform of Equations 3-5 is taken in
order to derive the differential equations 6-8.
Generator: ̇ = -0.1 + 0.1 -0.1 Δ (6)
Turbine: ̇ = - ⁄ + ⁄ (7)
Governor: ̇ = -200 - 10 (8)
Following is the state space representation of the
uncontrolled system shown in Figure 2.
A=













100200
3.0
1
3.0
1
0
01.01.0
B=










0
0
1.0
C= 001
Figure 3 shows the steady state frequency
deviation of the uncontrolled single area model illustrated
in Figure 2. The system has settling time of 3 seconds,
undershoot of 6% which corresponds to transient
frequency of -0.06HZ, and steady state error of -0.048HZ.
The system performance can definitely be improved
especially with steady state frequency deviation. Hence,
addition of a controller is required to control the system
output response.
Fig-3: Output Response of Uncontrolled Power Generation
System

3 INTRODUCTION TO FUZZY LOGIC
Many industrial systems such as power generation system
are time-variant and are influenced significantly by
external disturbances. These disturbances cause changes
in system performance. The issue of controlling and
optimizing a dynamic system can be addressed using
Fuzzy Logic (FL). FL has been applied to power plant
optimization problems in many different ways such as
optimal distribution planning, generator maintenance
scheduling, load forecasting, load management, and
generation dispatch problem [4].
Fuzzy Logic (FL) by Dr. Zadeh is able to provide a
systematic way for the application of uncertain and
indefinite models when precise definition or mathematical
representation of the system is unavailable [5]. Power
system is a stochastic system that is highly affected by
non-internal factors such as weather and change of
seasons. Modelling a stochastic and time varying system is
a very challenging task [6]. FL control is able to enhance
system performance without the need of mathematical
modeling of the system. It is enough to have only some
knowledge about the system and its behavior. This is
considered as the most important advantage of FL.
FL is strongly based on linguistic interpretation of the
system. It establishes linguistic rules called membership
rules to determine a systematic way of modelling the
power system. Membership rules or membership
functions are fundamental part of FL. Let X be a set of
objects whose elements are denoted by x. Membership in a
subset A of X is the membership function [7].
A = {(x, mA(x)), x ε X)} (9)
Fuzzy sets are functions that map a value that might be a
member of a set to a number between zero and one
indicating its actual degree of membership. Fuzzy sets
produce a membership curves.
4 DESIGN OF FUZZY LOGIC CONTROL
There exist two types of FL control:
1. Static Fuzzy Control: This controller is used when
structure and parameters of the FL controller are
fixed and do not change during real time operation
[6].
2. Adaptive Fuzzy Logic Control: This controller is used
when structure and parameters of FL controller
change during real time operation. This type of
controllers is more expensive to implement;
however, it results in better performance and less
mathematical information about the system is
needed [6].
The Objective of using Adaptive FL control in optimization
problem is to minimize or maximize an objective function
f(x) in the presence of uncertainties, unknown variations,
and constraints. Adaptive FL control is difficult to analyze
because it is time varying; however, it ensures more
desired performance in comparison to Static FL control.
Figure 4 shows block diagram of a FL controller which
consists of the following 4 components [6]:
1. Rule-Base: It holds knowledge in terms of set of
linguistic rules called fuzzy rules defined by the user.
Fuzzy rules are built using membership functions.
2. Inference Mechanism: It selects relevant rules at the
current time and decides what the output of the
controller should be. Output of the controller u(t) is
input of the plant. In power system, the plant is the
uncontrolled/open loop system.
3. Fuzzification: It converts controller’s input into
information that can be used in inference
mechanism.
4. Defuzzification: It converts the output of the
controller into values that can be used by the plant.
Fuzzification and defuzzification are inverse
processes.
Fig- 4: Fuzzy Logic Controller Block Diagram
From Figure 4 it can be observed that FL controller has
two inputs as shown below:
e(t) = r(t) – y(t) > ACE (10)
( ) ( )̇ > ̇ (11)
If reference input r(t) is zero, then inputs of FL controller
will be:
e(t) = – y(t) > ACE (12)

( ) ( )̇ = ( )̇ > ̇ (13)
To create a FL controller, following steps must to be taken
[7]:
1. Define the controller inputs:
Error = set point – process output
Error change = current error – last error
2. Define the controller output:
Output = controller output – plant input
1. Create membership functions:
Membership functions are developed based on
designer’s knowledge and experience about the
system. Membership functions are used to define
fuzzy rules.
2. Create fuzzy rules:
Fuzzy rules are defined using IF-THEN relationships.
They need to be manually tuned or adjusted in order
to obtain the desired system response.
3. Simulate the results:
SIMULINK can be used to simulate the steady state
output response.
The inputs of FL control shown in Equation 12 and 13 can
be classified into membership functions. In this paper, the
inputs are classified into 7 membership functions:
NB: Negative Big, NM: Negative Medium, NS: Negative
Small, ZZ: Zero, PS: Positive Small, MP: Positive Medium,
PB: Positive Big. These 7 membership functions lead to 49
fuzzy rules as shown in Table 1.
Membership functions must be symmetrical and each
membership function overlaps with the adjacent functions
by 50%. Membership functions are normalized in the
interval [-L, L] which is symmetric around zero [6].
The two inputs are combined together using AND
operation. Table 1 is constructed based on experience and
knowledge known about power generation systems.
Table-1: Fuzzy Logic Membership Rules
Fuzzy Inference System (FIS) in MATLAB is used to design
a FL controller based on the fuzzy rules defined in Table 1.
The controller output is the input of the plant. Centeroid
method is used to defuzzificate the values. The range of
each membership function is defined based on human’s
experience and knowledge about power generation
system. There are various types of membership functions
used in FIS such as triangular, trapezoidal, PI-curve, bell-
shaped, and S-curved [8]. In this paper, triangular
membership functions are used.
5 FEEDBACK ANALYSIS
To have a stable system after implementation of FL
controller, controllability and observability are very
important factors. Implementation of FL controller
guarantees a closed loop globally stable system if the
corresponding open loop system is controllable,
observable, and stable [6]. Hence, the system shown in
Figure 2 which has order of 3 is checked for the above
conditions:
1. The system is controllable. The rank of
controllability matrix is 3.
2. The system is observable. The rank of observability
matrix is 3.
3. The system is stable since all the three poles lie on
the left half plane. The poles are -10.8290 + 0.0000i,
-1.3022 + 2.1837i, -1.3022 - 2.1837i.
( )̇
AND NB NM NS ZZ PS PM PB
e(t)
NB NB NB NB NB NM NS ZZ
NM NB NM NM NM NS ZZ PS
NS NM NS NS NS ZZ PS PM
Z NB NM NS ZZ PS PM PB
PS NM NS ZZ PS PS PS PM
PM NS ZZ PS PM PM PM PS
PB ZZ PS PM PB PB PB PB

FL controllers are reliable and PI controllers are robust.
Combination of the two types of controllers can result in a
reliable, efficient, and robust controller design. Figure 5 is
the block diagram representation of the feedback single
area system generated in SIMULINK. The Adaptive FL and
PI controller are combined together in parallel to improve
the system behavior. This controller is called Adaptive FL
and PI controller. The system shown in Figure 5 has only
one input Δ and one output Δ .
Fig-5: Block Diagram Representation of Feedback Single Area Generating Unit
Parameters of the PI controller have been tuned carefully
to ensure performance improvement. Table 2 shows the
parameters of the PI controller implemented in Figure 5.
Equation 14 shows the transfer function of a PI controller
where U(S) is the controller output, E(S) is the controller
input, is the controller proportional constant, and is
the controller integral constant.
( )
( )
= (14)
Table-2: PI Controller Parameters for Feedback Single
Area Generating Unit Combined with Adaptive FL
Controller
Proportional
Constant ( )
Integral
Constant ( )
Integral Gain
( )
-0.25 -3.5 3.5
6 RESULT
Figure 6 shows the steady state frequency response of the
controlled system after implementation of Adaptive FL
controller described in Table 1 combined with PI
controller described in Table 2.
Reliability of FL controller and robustness of PI controller
are combined together to construct a well behaved
controlled system. As shown in Figure 6, the system
settling time is reduced to 2.5 seconds and the steady state
error is completely removed; this is the effect of integral
controller. The undershoot percentage is about 0.03%.
This is a well behaved system since all the parameters
have been improved significantly.
The primary objective of having controller in a power
generation system is to eliminate or minimize the steady
state frequency deviation. In power generation system,
followings are considered as standard performance
specifications of a well-behaved system:
1. Steady state frequency error should not be more
than 0.01HZ.
2. Settling time should be less than 3 seconds.
3. The maximum overshot/undershoot should not
be more than 6% which corresponds to transient
frequency of 0.06HZ.
Fig-6: Steady State of Feedback Single Area Model after
Implementation of FL Control

7 CONCLUSION
Adaptive FL controller and a suitable PI controller have
been combined to improve the system performance of a
single area power generation system. The membership
functions for Adaptive FL controller and the parameters of
PI controller have been tuned to ensure the specifications
are met.
Adaptive FL controller is robust, reliable, and most
commonly used in solving optimization problems. Table 3
compares the performance factors of uncontrolled vs
controlled single area system.
Table-3: Comparison of Uncontrolled vs. Controlled
Power Generation System
Settling
Time
(Sec)
Steady State
Error (HZ)
Undershoot
(%)
Uncontrolled 3 -0.048 6
Controlled 2.5 0 0.03
REFERENCES
[1] Dr. A. Ismail, Professor of Electrical Engineering at
Rochester Institute of Technology, Dubai, UAE .
Advances in Power Generation System Course
(2017).
[2] P. Pechtl. Integrated Thermal Power and
Desalination Plant Optimization. PowerGen
Middle East, Paper No.110 (2003).
[3] J. Zhu. Optimization of Power System Operation.
Second Edition. New Jersey, USA: Wiley (2015).
[4] K. Tomsovic. Fuzzy Systems Application to Power
Systems. Washington State University. Pullman,
WA. USA.
[5] E. Cam, I. Kocaarslan. Load Frequency Control in
Two Area Power Systems Using Fuzzy Logic
Controller. Kirikkale University. Turkey (2004).
[6] K. M. Passino, S. Yurkovich. Fuzzy Control.
California, USA: Addison Wesley Longman Inc
(1998).
[7] K. Tomsovic, M.Y.Chow. Tutorial on Fuzzy Logic
Applications in Power System. IEEE-PES Winter
Meeting in Singapore (2000).
[8] S.N. Sivanandam, S. Sumathi, S.N. Deepa.
Introduction to Fuzzy Logic using MATLAB.
Springer (2007).
BIOGRAPHIES
Nastaran Naghshineh is a
graduate of Electronics
Engineering from Simon Fraser
University, British Columbia,
Canada. Currently, she is
completing her Master’s degree
in Electrical Engineering,
specializing in Control System, at
Rochester Institute of
Technology (RIT) Dubai, UAE.
Email: nxn5534@rit.edu
Dr. Abdulla Ismail obtained his
B.Ss (’80), M.Sc. (’83), and Ph.D.
(’86) degrees, in Electrical
Engineering from the University
of Arizona, U.S.A. Currently, he is
a full professor of Electrical
Engineering and assistant to the
President at Rochester Institute
of Technology (RIT) Dubai, UAE.
Email: axicad@rit.edu

Advanced Optimization of Single Area Power Generation System using Adaptive Fuzzy Logic and PI Control

More Related Content

What's hot (19)

Similar to Advanced Optimization of Single Area Power Generation System using Adaptive Fuzzy Logic and PI Control (20)

More from IRJET Journal (20)

Recently uploaded (20)

Advanced Optimization of Single Area Power Generation System using Adaptive Fuzzy Logic and PI Control