See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/266459484

Design of Decentralized Fuzzy Logic Load Frequency Controller

Article in International Journal of Intelligent Systems and Applications · March 2012

DOI: 10.5815/ijisa.2012.02.08

CITATIONS READS

22 10,671

2 authors:

Khaled Ellithy Khaled El-Metwally

Atlas a International Engineering Consultants Cairo University
39 PUBLICATIONS 471 CITATIONS 53 PUBLICATIONS 1,227 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Khaled El-Metwally on 27 April 2015.

The user has requested enhancement of the downloaded file.

I.J. Intelligent Systems and Applications, 2012, 2, 66-75
Published Online March 2012 in MECS (http://www.mecs-press.org/)
DOI: 10.5815/ijisa.2012.02.08

Design of Decentralized Fuzzy Logic Load

Frequency Controller
K. A. Ellithy
Department of Electrical Engineering, Qatar University, Qatar
Email: [email protected]

K.A. El-Metwally
Department of Electrical Engineering, Cairo University, Cairo, Egypt
Email: [email protected]

Abstract— This paper presents a novel approach for designing a maintain the interchange power between control areas at
decentralized controller for load frequency control of scheduled values by adjusting the MW output power of
interconnected power areas. The proposed fuzzy logic load the selected generators so as to accommodate changing in
frequency controller (FLFC) has been designed to improve the
load demands [4].
dynamic performance of the frequency and tie line power under
a sudden load change in the power areas. The effect of
generation rate constraint (GRC) for both areas has been Many investigations in the area of LFC problem have
considered in the controller design. The proposed FLFC been reported and a number of control strategies have
consists of two internal fuzzy logic controllers namely, the PD- been employed in the design of load frequency (LF)
like fuzzy logic controller and the PI-like fuzzy logic controller. controller in order to achieve better dynamic performance
The FLFC has been co-coordinated with the conventional [6-8, 9-15]. In recent years, fuzzy system applications
integral controller. Time-domain simulations using have received increasing attention in power system
MATALB/SIMULINK program has been performed to operation and control [16, 17, 19, 24]. Fuzzy logic based
demonstrate the effectiveness of the proposed FLFC. The
controllers have been suggested as an appropriate choice
simulation results show that the proposed FLFC can provide
good damping and reduce the overshoot even in the presence of
to control non-linear system [17,18, 20] and are being
the GRC. investigated as an alternative to conventional control. The
basic feature of fuzzy logic controllers (FLCs) is that the
control strategy can be simply expressed by a set of rules
Index Terms—Load Frequency Control, Generation Rate which describe the behavior of the controller using
Constraints, Conventional Integral Control, Fuzzy Logic linguistic terms. Proper control action is then inferred
Control. from this rule base. In additional, FLCs are relatively
easy to develop and simple to implement. In the design of
load frequency controller, it should be recognized that
there is a limit to the rate at which generating unit outputs
I. INTRODUCTION can be changed. This is particularly true of steam
Large-scale power systems are normally composed of generating units where mechanical and thermal stress are
control areas or regions representing coherent groups of limiting factor. The controller designed for unconstrained
generators. The various areas are interconnected through situation may not be suitable. The analysis performed in
tie lines. The tie lines are utilized for energy exchange [11, 21] indicate that with constraints imposed by
between areas and provide inter-area support in case of generation rate constraints (GRC), the dynamic responses
abnormal condition [1-5]. Area load-changes and of the power area experience larger overshoots and longer
abnormal conditions, such as outages of generation, leads settling times, compared to the case without considering
to mismatch in scheduled power interchanges between the GRC.
areas. These mismatches have to be corrected via
supplementary control. In recent years, large tie-line This paper presents a design for LF controller to
power fluctuations have been observed as a result of improve the dynamic performance of power areas under
increased system capacity and very close interconnection disturbances such as a sudden load changes. The
among power systems [1]. This observation suggests a generation rate constraint (GRC) has considered in the
strong need of establishing a more advanced load controller design. The paper is organized as follows:
frequency control (LFC) scheme. LFC is one of the major Section 2 provides the model of the power two-area.
requirements in providing reliable and quality operation Section 3 shows the designed fuzzy controllers. The
in interconnected power systems. The primary objectives effectiveness of the proposed controller on the dynamic
of the LFC in an interconnected power system are to performance is also contained in Section 4. The
regulate the frequency to the nominal value and to conclusions are given in Section 5.

Copyright © 2012 MECS I.J. Intelligent Systems and Applications, 2012, 2, 66-75
 Design of Decentralized Fuzzy Logic Load Frequency Controller 67

II. MODEL OF TWO POWER AREA ACE1 = ΔP12 + B1Δf1

(2)
Fig. 1 shows the block diagram representing the two ACE2 = ΔP21 + B2 Δf 2
area power system. This model includes the conventional
integral controller gains (K1, K2) and the two auxiliary The ΔP12 (ΔPtie) is the tie-line power deviation. The
(stabilizing) signals (Δu1, Δu2). The stabilizing signals nomenclature used and the parameters values of the two
will be generated by the proposed fuzzy logic load power area are given in the Appendix.
frequency controller (FLFC). Each power area has a
number of generators which are closely coupled together There exists a maximum on the rate of change of
so as to form a coherent group, i.e. all the generators power that can be generated by steam plants. The
respond in unison to changes in the load. Such a coherent constraints of the nonlinear characteristics of the turbine
area is called a control area in which the frequency is control should be considered in the load frequency
assumed to be the same throughout in static as well as controller design. If these constraints are not considered
dynamic situation [1, 4]. It is conveniently assumed that in the controller design, the power area is likely to chase
each control can be represented by an equivalent large monetary disturbance. The controller designed for
generator, governor and turbine system. The conventional the unconstrained situation may not be suitable. The
LF controller shown in Fig. 1 is based upon tie-line bias analysis performed in [21] indicate that with constraints
control, where each control area tends to reduce the Area imposed by generation rate constraints (GRC), the
Control Error (ACE) to zero. The ACE given in (1) for dynamic responses of the power area experience larger
each area consists of a linear combination of frequency overshoots and longer settling times, compared to the
and tie-line power deviation [2]. case without considering the GRC. Moreover, if the

ACEi = ∑ (ΔPij + γ i Δf )
n parameters of the controller are not chosen properly, the
(1) power area may have unacceptable dynamic response
j =1 (large overshoot and long settling time) or become
An overall satisfactory performance is achieved when γi is unstable. In the time-domain simulations presented in this
selected to be equal to the frequency bias factor of that paper, the linear model of turbine shown in Fig. 2 is
placed by a nonlinear model shown in Fig. 6. This
replacement is done to take into account the effect of
GRC that emulates the practical limit on the response of

1
B1
R1 ΔPL1
ACE1 _ Governor Turbine _ Generator & load
+ ΔPV1 ΔPm1 1 Δf1
K 1 1 Σ
Σ − 1 Σ
+ s 1 + sτ g 1 1 + sτ T 1 + _ 2 H 1 s + D1
+
+
Δ u1 ΔP12=ΔPtie T12 Σ
s _
Δ u2
_ +
+ ΔPV2 ΔPm2 Δf2
K 1 1 1
Σ − 2 Σ
1 + sτ T 2
Σ
2 H 2 s + D2
+ s _ 1 + sτ 2 + _
ACE2
1 ΔPL2
B2 R2

Fig. 1 Block diagram of Load frequency control (LFC) of two-area

area. So γ i = Bi = 1 + Di the turbine [11, 21].

Ri The system response having the integral controller
gains (K1, K2) usually satisfies the desired objectives of
Thus, the ACEs for a two-area are
the LFC (ensures zero frequency error in the steady state).
The only problem of conventional integral controller is

Copyright © 2012 MECS I.J. Intelligent Systems and Applications, 2012, 2, 66-75
68 Design of Decentralized Fuzzy Logic Load Frequency Controller

that the system response is less damped and the overshot frequency and tie-line power of the power areas under
is large. To solve this problem, another control signals disturbances. This improvement can be achieved by
(Δu1, Δu2) are added to the system in presence of integral introducing an auxiliary stabilizing signal (Δu1 & Δu2)
controller. These signals are derived from the proposed drive from the controller as shown in Fig. 1. The
fuzzy logic controller. objective of the proposed fuzzy logic based frequency
controller (FLFC) is to increase transient response (i.e.
improve the damping and reduce the overshoots) of the
frequency deviations for the unequal areas and also to
maintain the interchanged power (tie-line power) about
ΔPv 1 1 ΔPt=ΔPm
the schedule, thus that loads in each area are self
Σ
+ - τt s contained.

To achieve these two goals the FLFC is designed to

have two fuzzy logic controllers working in parallel. The
Fig. 2 Nonlinear Turbine model with GRC first is a PD-like fuzzy logic controller which uses the
frequency deviation, Δf, and the rate of the frequency
deviation Δ2f, in order to enhance the transients of the
III. DESIGN OF FUZZY LOGIC LOAD FREQUENCY frequency deviation of each area. The second is a PI-
CONTROLLER Like fuzzy logic controller which uses the frequency
A. Introduction to Fuzzy Logic Control deviation, Δf, and the tie power deviation, ΔPtie, to reset
the tie power to zero. The use of the ∆f and ∆f2 for a PD
Fuzzy logic-based control (FLC) has been become like is chosen basically to add a D component to the
an important methodology in control engineering and has convention I-controller and thus improving the ∆f
been rapidly gaining popularity among engineers during transient behavior. However, this component alone is not
the past few years [16-18, 20-25]. This increased enough to provide good ∆Ptie behavior especially at the
popularity can be attributed to the fact that fuzzy logic steady state value. This created the need to use a PI-like
provides a powerful vehicle that allows engineers to fuzzy controller using the ∆f and ∆Ptie signals to ensure
incorporate human reasoning in control algorithm. The zero power in the tie line after disturbance.
fuzzy logic is used in system control design because it
shortens the time for engineering development and The output control signal from the FLFC, Uf, is
sometimes in the case of highly complex systems is the injected to the summing point of the designated area as
only way to solve the problem. The basic configuration of shown in Fig. 5. Introducing fuzzy logic control means
a fuzzy-logic control is composed of four principle introducing a non-linear controller. In this case integral
components: a fuzzification, a knowledge base, a control alone will not provide a complete reset to the
inference engine, and defuzzification [20]. Figure 3 error. This is the reason for providing a rest term using
shows the basic configuration of a FLC. the fuzzy PI-Like controller. The PD-fuzzy controllers
have been applied alone but they didn’t give a good
response for the tie-line power. Each input or output of
both the PD and the PI like fuzzy controllers, (fuzzy
variable, X = {Δf, Δ2f, ΔPtie, Upd, UPI}), is assigned seven
linguistic fuzzy subsets varying from, Negative Big (NB)
to Positive Big (PB). Each fuzzy subset is associated with
a triangular membership function to form a set of seven
normalized and symmetrical triangular membership
functions for each fuzzy variable, see Fig. 4.

NB NM NS Z PS PM PB

The fuzzifier maps the input crisp values into fuzzy

variables using normalized membership functions and
input gains. The fuzzy logic inference engine then infers θi
the proper control action based on the available rule-base. Xmin Xrange Xmax
The fuzzy control action is translated to the proper crisp
value through the defuzzifier using normalized Fig. 4 Fuzzy variable, Xi, Seven membership functions
membership functions and output gains.
B. Fuzzy logic load frequency controller design
A proper design of fuzzy logic based frequency
controller can improve the transient response of the

Copyright © 2012 MECS I.J. Intelligent Systems and Applications, 2012, 2, 66-75
 Design of Decentralized Fuzzy Logic Load Frequency Controller 69

The Xmax and Xmin represent maximum and the achieved by integrating the output of a PD-Like
minimum variation of the input and output signals. These controller [18]. Thus the final configuration of the
values are selected based on simulation information. The proposed FLFC is shown in Fig. 5. A symmetrical fuzzy
selected value of Xmax and Xmin was based on the rule set is used to describe both the PD and the PI like
performed dynamic simulations under different system fuzzy controller behaviors as shown in Table 1 and 2. The
operating conditions including different load changes to fuzzy rules have been selected based on experience
obtain a robust modulated signal (output signal of the gained form different simulations.
designed FLFC controller) which improve the dynamic
performance (reducing the overshoot and improving the Table 1: PD-Fuzzy Logic Rules
damping) of the system under diffent operating
conditions and system parameters change. The range of Δ2f
each fuzzy variable is normalized between –1 to 1 by Δf
introducing a scaling factor to represent the actual signal NB NM NS Z PS PM PB
so that: NB NB NB NB NB NM NS Z
NM NB NB NM NM NS Z PS
The PD-Like fuzzy controller initial gains are given by NS NB NM NM NS Z PS PM
[17, 18]:
1 1 Z NM NM NS Z PS PM PM
K Δf = ; K Δ2 f = 2 ; K upd = U pd max (3) PS NM NS Z PS PM PM PB
Δf max Δ f max
Similar definition applies in the case of the PI-like fuzzy PM NS Z PS PM PM PB PB
controller PB Z PS PM PB PB PB PB
1 1
K Δf = ; KPtie = ; K upi = U pi max (4)
Δf max ΔPtie max Table 2: PD-Fuzzy Logic Rules

Where KΔf, KPtie , Kupi are the gains of frequency ΔPtie

deviation, tie power deviation, and the output control Δf
signal respectively. While, Δfmax, ΔPtiemax, Upimax are the NB NM NS Z PS PM PB
maximum deviation in frequency, tie-line derivative and NB NB NB NB NB NM NS Z
max control limit respectively. These gains are actually
scaling factors to adapt input signal to the normalized NM NB NB NM NM NS Z PS
fuzzy membership functions. The reason of using the NS NB NM NM NS Z PS PM
inverse of the signal expected maximum value is to
Z NM NM NS Z PS PM PM
provide full coverage of the membership functions range
and thus utilizing the complete rule base. The values of PS NM NS Z PS PM PM PB
maximum variation of the input and output signals (∆fmax, PM NS Z PS PM PM PB PB
∆fmax, ∆Ptiemax) can be easily found from the system
simulations under disturbances [7]. These initial gains PB Z PS PM PB PB PB PB
can be further tuned slightly to achieve better
performance. The PI-like fuzzy control action can be

1 KΔf
Δf Normalized Kupd
2 2
KΔ f FLC
2
Δ f
PD-Like fuzzy
+ 1
+
ΔUf
KΔf′
Normalized
Kupi +
FLC +
3 KPtie
ΔPtie 1
z
Delay
PI-Like fuzzy

FLFC

Fig. 5 The structure of the proposed FLFC

Copyright © 2012 MECS I.J. Intelligent Systems and Applications, 2012, 2, 66-75
70 Design of Decentralized Fuzzy Logic Load Frequency Controller

Each entity in the table represents a fuzzy rule of the from

"if antecedent then consequence", e.g. the shaded rule in
wi = μ NB (Δf )xμ PB Δ2 f ( ) (7)

Table 1 is
where μ(x) is the membership value of the fuzzy variable,
if Δf is NB and Δ2 f is PB then U pd is Z (5)
x.

Defuzzification is achieved using the center of The net control action (Uf) of the FLFC is the sum of the
gravity method [20]. In general, if the control output individual PD and PI-like fuzzy controller outputs (Upd,
membership functions centroids are represented by Upi) thus:
θ1,…,θM. Thus, for M rules, the crisp output of the
controller is: U f = U pd + U pi (8)
M

∑wθ i i
(6)
=θ ζ
T
u= i =1
M
IV. COMPUTER SIMULATION RESULTS
∑w
i =1
i

The block diagram of the simulated model of the two

Where area system is given in Fig. 6. The system is controlled by
using: a) conventional integral controller; b) propose
ζ = [ζ 1 ....ζ i ....ζ M ] FLFC. The effects GRC for both areas have been
included in the simulations. The GRC is taken into
w . account by adding a limiter to turbine as shown in Fig. 6.
ζ i = M
i
The values ± 0.2 are considered for the limits. In addition,
∑ w k to prevent the excessive control action, a limiter is also
k =1 added to the integral control part. The simulations have
been performed for the following two cases:
The strength of the ith is wi and it is calculated based on
interpreting the ‘and’ conjunction as a product of the
membership values corresponding to the measured values
of Δ f and Δ f. For example, the rule strength of the
2

shaded rule in table 1 is given by (7)

B1

20.6

20
1/R1

dPL1
I1

0.3
1 1 1 1 ΔPm1 1
s 0.2s+1 0.5 s 10s+0.6
k1 T1 GRC1 Turbine1 Inertia, Load1
Governor1 Sum

df

u d2f du/dt dw1

Ptie
PI PD FLC 1

T12
400,800,5, PD
1 f1-f2
2
s
dP-tie I3
df
Mux

df

u d2f du/dt

Ptie dw2
PI PD FLC 2

I2

0.3
1 1 1 1 ΔPm2 1
s 0.3s+1 0.6 s 8s+0.9
k2 T2 GRC2 Turbine2 Inertia, Load2
Governor2
Sum1

1/R2
16
B2
dPL2
16.9

Fig. 6 Simulink Block diagram of LFC of two-area including GRC

Copyright © 2012 MECS I.J. Intelligent Systems and Applications, 2012, 2, 66-75
 Design of Decentralized Fuzzy Logic Load Frequency Controller 71

Case 1: Simulations result with conventional integral

controller Δf1 ΔPc1 ΔPm1 ΔPv1 ΔP12 Δf2 ΔPc2 ΔPm2 ΔPv2
⎡ ⎤
⎢ λ1 0 . 1075 0 . 1624 0 . 7266 0 . 0028 0 . 0006 0 . 0000 0 . 0000 0 . 0000 0 . 0000 ⎥
In this paper, the integral controller gains (K1, K2) ⎢ ⎥
⎢ ⎥
have been determined using eigenvalue analysis. ⎢λ2 0 . 0000 0 . 0000 0 . 0000 0 . 0000 0 . 0018 0 . 1414 0 . 2281 0 . 6235 0 . 0052 ⎥
⎢ ⎥
Satisfactory system transient response can be achieved ⎢ λ3 0 . 4198 0 . 3091 0 . 1483 0 . 0336 0 . 0259 0 . 0283 0 . 0199 0 . 0126 0 . 0025 ⎥
⎢ ⎥
by improving the relative damping (damping factor) of ⎢ ⎥
⎢ λ4 0 . 4198 0 . 3091 0 . 1483 0 . 0336 0 . 0259 0 . 0283 0 . 0199 0 . 0126 0 . 0025 ⎥
the dominant eigenvalue modes associated with the ⎢ ⎥
⎢ λ5 0 . 0345 ⎥
frequency deviations. The common index for system ⎢
0 . 0449 0 . 0319 0 . 0151 0 . 0034 0 . 0383 0 . 3918 0 . 2719 0 . 1681
⎥
damping is the relative damping of the dominant modes. ⎢λ 0 . 0449 0 . 0319 0 . 0151 0 . 0034 0 . 0383 0 . 3918 0 . 2719 0 . 1681 0 . 0345 ⎥
⎢ 6 ⎥
The relative damping of a complex eigenvalues λi = σi ± ⎢
⎢ λ7 0 . 0488 0 . 0660 0 . 0250 0 . 1214 0 . 2880 0 . 0510 0 . 0921 0 . 0436
⎥
0 . 2641 ⎥
jωi is given by (9). ⎢
⎢λ
⎥
0 . 0488 0 . 0660 0 . 0250 0 . 1214 0 . 2880 0 . 0510 0 . 0921 0 . 0436 0 . 2641 ⎥
−σi (9)
⎢ 8 ⎥
ξi = ⎢ ⎥

( )
⎢ λ9 0 . 0480 0 . 0026 0 . 0010 0 . 6371 0 . 0000 0 . 0206 0 . 0007 0 . 0003 00 . 2987 ⎥
⎣ ⎦
σ i2 + ω i2
From the participation factors, it can be observed that
where ωi characterizes the frequency and σi the damping the largest participation factor 0.4198 is associated with
behavior of oscillations. the eigenvalue modes λ3,4 while the largest participation
factor 0.3918 is associated with the modes λ5,6. Figures
Participation factors technique [4, 26], can be used 7 and 8 show the system frequency response with
to associate state-variables (frequency deviation Δfi = 0, different values of integral controller gains K1 = K2 =
i=1,2) with the specific modes (dominant eigenvalues K1,,2 under 20% and 1% changes in the load demand at
modes λ3,4 & λ5,6). The participation factors Pki is defined area 2. From Fig. 7 and 8, it can be seen that the gains
as : K1 = K2 = K1,,2=0.3 gives the best damping response.
vik wki From Table 3 and the time response shown in Figs. 7 and
Pki = n
(10) 8, it can be seen that the higher gains of integral
∑vi =1
ik wki controller can lead the system to instability. The results
have shown that the system becomes unstable at K1,,2≥
0.74. Optimization technique presented in [12] can also
where vik and wik are the kth entries of the right and left be applied to determine the optimal integral controller
eigenvector associated with λi. The Pki provide a measure gains.
of the contribution of kth state variable to the ith
eigenvalue and so can be used to correlate the Case2: Simulations Result with the proposed fuzzy logic
contribution of each state variable to each eigenvalue frequency controller
mode The system eigenvalues are listed in Table 3. The
table clearly show that the system is stable at 0.3 gains From the system dynamic response shown in Figs.7
setting (all eigenvalues having negative real part) while and 8, it can be seen that the conventional integral
the system is unstable at 0.8 gains setting. controller provide zero frequency deviations (Δfi=0,
i=1,2) in the steady-state but it exhibits poor dynamic
Table 3: System Eigenvalues with integral controller performance; i.e. it exhibits large overshoot and less
Eigenvalues Stable Unstable damping [1, 9]. In order to improve frequency damping
λi (K1 =K2=0.3) (K1 =K2=0.8) (i.e. ensures zero frequency deviations in shortest time
possible) and to reduce the frequency overshoot
λ1 -5.8468 -5.788 whenever there is any disturbance. the fuzzy logic
λ2 -4.2717 -4.1894 frequency controller (FLFC) is proposed. To
demonstrate the effectiveness of the proposed FLFC
λ3,4 -0.3768 ± 1.7234i * -0.0912 ± 1.7122i compare to the conventional integral controller, time-
domain simulations of the two area power system have
λ5,6 -0.2231+ 1.5992i ** 0.0287 ± 1.5766i
been performed under disturbances (1% and 5% load
λ7,8 -0.2537 + 0.0484i -0.2216, -0.8494 changes). Similar responses were obtained for load
disturbance applied to area 2 as shown in Figs. 9-12.
λ9 -0.3468 -0.9992 Figures 9-12 show the system dynamic response for
*ξ3,4 of stable mode λ3,4=21.3592%
both 1% and 5% load change. From Figs. 9-12, it can be
**ξ5,6 of stable mode λ5,6= 13.8169% seen that the dynamic response of the frequency and tie-
line deviations is less damped (large settling time) and
The Participation factors have been calculated and the large overshoot with only the integral controller for both
eigenvalues modes associated with the frequency disturbances. On the other hand the dynamic response of
deviations have been identified. The calculated the frequency and tie-line deviations is well damped and
participation factors are given as : the overshoot (%OS) is reduced with the proposed
FLFC controller for all disturbances considered. The

Copyright © 2012 MECS I.J. Intelligent Systems and Applications, 2012, 2, 66-75
72 Design of Decentralized Fuzzy Logic Load Frequency Controller

improvements in %OS are shown in Figs. 14 and 15. The disturbance matrix Γ equals to
This proves that the designed decentralized FLFC ⎡ 1 ⎤
T

controller is more effective controller in reducing ⎢−2H 0 0 0 0 0 0 0 0 ⎥

Γ=⎢ 1 ⎥
system overshoot and decreasing the settling time. The ⎢ 0 1
0 0 0 0 − 0 0 0 ⎥
proposed controller still yields good dynamic ⎢
⎣ 2H 2 ⎥
⎦
performance for both areas when the GRC is included.

B. Optimal load frequency controller

Case3: Simulations results with the optimal load
frequency controller The optimal controller is design to minimize the
quadratic performance index of the following form :
The proposed FLFC has also been compared with the
optimal controller determined by the LQR technique ∞
[9,10] and it is found that the FLFC the proposed FLFC J = ∫ (Δx T QΔx + Δu T RΔu )dt (12)
can achieve better dynamic performance than the optimal
0
controller as shown in Fig. 13. The state space model of
the two-area under study and the designed optimal state-
Subject to the dynamic system equation in (11), Q is a
feedback gains determined by LQR are given as follows:
positive semi-definite matrix and R is a positive definite
matrix. The optimal gain vector is given by :
A. State-variable model

The state-space equations of the linear time-invariant of K = R −1 B T P (13)

the two power areas:
Where P is determined by solving the following Riccati
• equation :
Δx = AΔx + BΔu + ΓΔPL (11)
PA + AT P − PBR−1 BT P + Q = 0 (14)
Δy = CΔx
The calculated optimal state-feedback gains are given as :
Where:

[
Δx= Δf1 ΔPc1 ΔPm1 ΔPv1 ΔP12 Δf2 ΔPc2 ΔPm2 ΔPv2 ]
T
is the ⎡
⎢ 86.8631
⎢
2.5342 0.7289 − 20.0053 1.6851 − 41.3502 − 1.2733 − 0.2936 12.1153 ⎥
⎤
⎥
K =⎢
state variables vector. ⎢
⎥
⎥
⎢ − 32.1644 − 0.7891 − 0.1957 7.7316 − 0.6917 71.1184 3.0647 0.9946 − 17.8207 ⎥
⎣ ⎦

Δu = [Δu1 Δu2]T is the input vector .

ΔPL = [ΔPL1 ΔPL2 ] is the disturbance to the system.

Δf1 Response during 20% Load Change at Area 2
0.015 ___ ___
K1,2 = 0.3 K1,2 = 0.5 ___ K1,2 = 0.7 ___
K1,2 = 0.8

⎡ − D1 1 −1 ⎤ 0.01
⎢ 2H 0 0 0 0 0 0 ⎥
2H1 2H1
⎢ 1
⎥
−1 1
⎢ 0 0 0 0 0 0 0 ⎥
⎢ τ T1 τT1 ⎥
⎢ −1 −1 1 ⎥ 0.005
⎢ 0 0 0 0 0 0 ⎥
⎢ R1τ g 1 τ g1 τ g1 ⎥
⎢ − K 1 B1 0 0 0 − K1 0 0 0 0 ⎥
⎢ ⎥
A = ⎢ T12 0 0 0 0 − T12 0 0 0 ⎥ 0
⎢ 1 − D2 1 ⎥
⎢ 0 0 0 0
2H 2 2H 2 2H 2
0 0 ⎥
⎢ ⎥
−1 1
⎢ 0 0 0 0 0 0 0 ⎥
⎢ τT2 τT2 ⎥ -0.005
⎢ −1 −1 1 ⎥
⎢ 0 0 0 0 0 0 ⎥
⎢ R 2τ g 2 τ g2 τ g2 ⎥
⎢ 0
⎣ 0 0 0 K2 − K 2 B2 0 0 0 ⎥⎦
-0.01
0 5 10 15 20 25
Time [sec]
T
⎡ 1 ⎤
⎢0 0 0 0 0 0 0 0⎥ Fig. 7 Response of area 1 frequency deviation ∆f1due to 20%
τ load disturbance at area 2 at different integral controller
B = ⎢ G1
⎥
⎢0 1
0 0 0 0 0 0 0⎥ gains.
⎢⎣ τ G 2
⎥⎦

Copyright © 2012 MECS I.J. Intelligent Systems and Applications, 2012, 2, 66-75
 Design of Decentralized Fuzzy Logic Load Frequency Controller 73

Δf2 Response during 20% Load Change at Area 2 x 10

-3
Δf2 Response during 5% Load Change at Area 2
0.04 ___ ___ ___ ___ 2
K1,2 = 0.3 K1,2 = 0.5 K1,2 = 0.7 K1,2 = 0.8
With FLFC
0.03 Without FLFC
1

0.02
0

0.01
-1
0
-2
-0.01

-3
-0.02

-4
-0.03

-0.04 -5
0 5 10 15 20 25 0 5 10 15 20 25
Time [sec] Time [sec]

Fig. 8 Response of area 2 frequency deviation ∆f2 due to Fig. 11 Response of area 2 frequency deviation ∆f2 due to 5%
20% load disturbance at area 2 at different integral load dsiturbance at area 2 with GRC
controller gains

-4 -3
x 10 Δf2 Response during 1% Load Change at Area 2 x 10 Δptie Response during 5% Load Change at Area 2
4 14
With FLFC With FLFC
Without FLFC 12 Without FLFC
2

10
0

8
-2

6
-4
4

-6
2

-8 0

-10 -2
0 5 10 15 20 25 0 5 10 15 20 25
Time [sec] Time [sec]
Fig. 9 Response of area 2 frequency deviation ∆f2 due to1% load Fig. 12 Response tie‐line power deviation ∆Ptie due to 5% load
dsiturbance at area 2 with GRC disturbance at area 2 with GRC

x 10
-3
Δptie Response during 1% Load Change at Area 2 Δf2 Response during 20% load change in Area 2
2.5 0.01
With FLFC
Without FLFC
2 0.005

1.5 0

1 -0.005
With FLFC
Without FLFC
0.5 -0.01 With Optimal

0 -0.015

-0.5 -0.02
0 5 10 15 20 25 0 5 10 15 20 25
Time [sec] Time (sec)
Fig. 10Response
Fig. 10 Responsetie‐line
tie-line power
power deviation
deviation ∆Ptie ∆P
duetietodue
1% to 1%
load Fig.Fig.
13 13
Response of area 2 frequency deviation ∆f2 ∆f
due to 20% load
Response of area 2 frequency deviation 2 due to 20%
loaddisturbance
disturbance at at area 2 with
GRCGRC
area 2 with loaddsiturbance
disturbanceatatarea
area22

Copyright © 2012 MECS I.J. Intelligent Systems and Applications, 2012, 2, 66-75
 74 Design of Decentralized Fuzzy Logic Load Frequency Controller

-3
x 10 ΔP-tie response during 20% load change in Area 2
20 to enhance the transients of the frequency deviation of
each area while the PI-fuzzy logic controller uses
WithWith
FLFCFLFC deviations of the frequency deviation and the tie line
15
WithoutFLFC
Without FLFC power to enhance the transient of the tie line power
WithOptimal
With Controller
OptimalController deviation and keep power in the tie line to the scheduled
value. The computer simulations results show that the
proposed FLFC is more effective means for improving
10
the dynamic performance of the two power area
compared to the conventional integral controller. The
proposed FLFC controller still achieves good dynamic
5
performance when the GRC is considered and ensures
the stability of power areas for all load demand changes.
Moreover, the proposed controller type is relatively
0 simple and suitable for practical on-line implementation.

-5
REFERENCES
0 5 10 15 20 25

Fig. 14 Response of area 2 frequency deviation ∆Ptie due to 20% load [1] Fosha, C.E. and Elgerd, O.I., The megawatt-frequency
dsiturbance at area 2 control problem: a New Approach via Optimal Control
Theory, IEEE Trans. Power Apparatus and Systems, 1970,
89(4): 563-577.
[2] Kothari, D.P. and Nagrath, I.J. Modern Power System
Analysis, 1st Edition, McGraw-Hill, New York, 2008.
[3] Kundur P. Power System Stability & Control, McGraw-
Hill, New York, 1994.
[4] Wood, A. Power Generation, Operation and Control, 2nd
Edition, Prentice-Hall. New York, 1996.
[5] Shayeghi H, Shayanfar HA, Jalili A. Load frequency
control strategies: a state of-the-art survey for the
researcher. Energy Conversion and Management, 2009,
50(2): 344–353
[6] Tan W. Decentralized load frequency controller analysis
Fig. 15 Percent overshoot of area 2 frequency deviation Δf1
and tuning for multi-area power systems, Energy
under load disturbance at area 2
Conversion and Management , 2011,52(5): 2015–2023.
[7] Al-Badi, M., Awlad Thani, A. Al-Omeiri, B. and Ellithy,
%Overshoot of Δf2 at differnt load change at area 2 K.A. Comparative Study of Load Frequency Controller
2
1.8 Design for Interconnected Power Systems, The Journal
1.6 for Scientific Research-Science and Technology, Sultan
1.4
1.2 Qaboos University,m 2002 , 7: 81-90.
%OS

1
0.8
[8] Chan, W.C. and Hsu, Y.Y., Automatic Generation
0.6 Control of Interconnected Power Systems Using Variable
0.4
0.2
Structure Controller. IEE Proc. on Generation,
0 Transmission and Distribution Pt. C, 1981: 269-279.
1% 2% 4% 6% 8% 10% 12% 14% 16% 18% 20%
[9] Feliach, A. , Optimal Decentralized Load Frequency
ΔPL
Control, IEEE Trans. on Power Systems, 1987, 2: 379-
Wthout FLFC With FLFC 386.
[10] Hiyama, T., Design of Decentralized Load Frequency
Fig. 16 Percent overshoot of area 1 frequency deviation Δf2 Regulators for Interconnected Power Systems, IEE Proc.,
under load disturbance at area 2 Pt C, 1982, 129 (1): 17-23.
[11] Hiyama, T. Optimization of Discrete-Type Load
Frequency Regulator Considering Generation Rate
V. CONCLUSIONS Constraints, Proc. IEE, Pt. C, Generation, Transmission
and Distribution, 1982, 129 (6): 285-289.
A decentralized fuzzy logic frequency controller [12] Kothari, M.L., Nanda, J., Kothari, D.P. and Das, D.,
(FLFC) coordinated with conventional integral Discrete-Mode Automatic Generation Control of Two-
controller has been proposed to damp out the deviations Area Reheat Thermal System with New Area Control
of the frequency and tie-line power for the Error, IEEE Transactions on Power Systems, 1989, 4(2):
interconnected power areas and also to keep the 730-738.
[13] Oysal, Y. A Comparative Study of Adaptive Load
interchanged power at the scheduled value. The
Frequency Controller Designs in a Power System with
proposed FLFC incorporate two fuzzy logic controllers Dynamic Neural Network Models, Energy Conversion
a PD-fuzzy logic controller and the PI-fuzzy logic and Management 2005, 46 (15-16): 2656-2668.
controller. The PD-fuzzy controller uses the frequency [14] Yang, T.C., Ding. and Yu, H. Decentralized power
deviation and the rate of the frequency deviation in order system load frequency control beyond the limit of

Copyright © 2012 MECS I.J. Intelligent Systems and Applications, 2012, 2, 66-75
 Design of Decentralized Fuzzy Logic Load Frequency Controller 75

diagonal dominance, Electric Power Energy System, ΔPL load disturbance

2002, 24(3): 173–184. µ(x) membership value of the fuzzy variable x
[15] Yang, T., Cimen, H. and Zhu, O. Decentralized Load-
Frequency Controller Design Based on Structured
Singular Values,” IEE Proc. on Generation,
Transmission and Distribution, 1998, 145(1):.7-13. Parameters of the two area system (in 1000 MVA base)
[16] Ellithy, K.A. and Al-Naamany, A. A Hybrid Neuro- R1= 5%, R2 = 6.25%, D1 = 0.6, D2 = 0.9, H1 = 5 sec,
Fuzzy SVC Stabilizers for Power System Damping
Improvement in the Presence of Load Parameters H2 = 4 sec, τ G1 = 0.2 sec, τ G 2 = 0.3 sec, τ T 1 = 0.5 sec,
Uncertainty, International Journal of Electric Power τ T 2 = 0.6 sec ,T12 = 2 pu, K1=K2=0.3
Systems Research, 2000, 56(3): 211-223.
[17] El-Metwally, K.A. and Malik, O. Fuzzy logic power
system stabilizer, IEE Proc. Generation, Transmission,
and Distribution, 1995, 142(3): 277-281. Khaled A. Ellithy (SM’83, M’88,
[18] El-Metwally, K.A., A Fuzzy Logic Based PID for Power SM’00) received the MSc. and PhD.
System Stabilization, Journal of Electric machines and degrees from West Virginia
Power Systems , 2001, 29(7): 659-668. University, Morgantown, WV in 1983
[19] Momoh, J. Ma, X Tomsovic, K. Overview and literature and 1987, respectively. Dr. Ellithy is
survey of fuzzy set theory in power systems, IEEE Trans. now associate professor in Department
on Power Systems, 1995, 10(3): 1676 – 1690.
of Electrical Engineering at Qatar
[20] Passino, K. and Yurkovickh, S., H. Fuzzy Control, University and he was the power
Addison Wesley Longman, Inc., 1998. systems chair for Qatar Electricity & Water Company
[21] Moon, Y.H., Lee, H.S., Sung, J.G., and Shin, K.B. (QEWC). Dr. Khaled is an Honored Associate
Extended Integral Control for Load Frequency Control Research Fellow, Brunel University, UK. Dr. Khaled
with the Consideration of Generation-rate Constraints, has previously taught in ECE Department at West
Electrical Power and Energy Systems , 2002, 24(4): 263- Virginia University, USA; Sultan Qaboos University,
269. Oman; United Arab Emirates University, UAE; and
[22] İlhan K. and Ertuğrul. C. Fuzzy logic controller in Cairo University-Alfayum branch. He has supervised
interconnected electrical power systems for load- several MSc and PhD students in the area of power
frequency control, International Journal of Electrical system analysis. He has published several papers in
Power & Energy Systems, 2005, 27(8): 542-549
IEEE Transactions and international journals in
[23] Tang, K.S, Man K.F., Chen G. and kwong S., An
Optimal Fuzzy PID Controller, IEEE Trans. on electrical power engineering. He has been awarded the
Industrial Electronics , 2001, 48(4) 757-765. prizes of the best applied research papers at 2000 and
[24] Pothiya, S., Ngamroo, I., Runggeratigul, S. and 2010 GCC-Cigré power conferences. He received the
Tantaswadi, P. Design of Optimal Fuzzy Logic Based PI IEEE award of Outstanding Branch Counselor and
Controller using Multiple Tabu Search Algorithm for Advisory Nominee. He is the counselor of IEEE
Load Frequency Control, Journal of Control, student branch at Qatar University. He has carried out
Automation, and Systems, 2006, 4(2): 155-164. several consultants and research projects for power
[25] Talaq, J. and Albasri F. Adaptive Fuzzy Gain utilities and industrials. His current research interests
Scheduling for Load Frequency Control, IEEE Trans. on
include power systems dynamic & control, power
Power Systems, 1999, 14(1): 145-151.
[26] Perez-Arriaga, I.J., Verghese, G.C. Schweppe, F.C. systems quality and influence of extremely low
Selective Modal Analysis with Applications to Electric frequency electric and magnetic fields on humans.
Power Systems. Part I: Heuristic Introduction. Part II:
The Dynamic Stability Problem, IEEE Trans. On Power Khaled A. El-Metwally (M’95)
Apparatus and Systems, 1982, 101: 3117-3134. received his B.Sc. and M.Sc. from
Cairo University 1987 and 1990,
APPENDIX respectively. In 1994 he received his
Nomenclature PhD from University of Calgary,
Canada. He is an Associate Professor
ACE area control error in Cairo University, Egypt and currently in leave to
B frequency bias factor Sultan Qaboos University, Oman. His research interests
D frequency sensitive load coefficient include intelligent control systems, optimization,
FLFC fuzzy logic frequency controller modeling, PLC systems, SCADA systems and power
GRC generating rate constraint systems stability and control. Dr. Khaled EL-Metwally
H inertia constant of generating unit is the Middle East regional editor of the International
KI integral control gain Journal of Modeling Identification and Control
LQR linear quadratic regulator (IJMIC).
R droop characteristic
τg speed governor time constant
τT turbine time constant
Δf frequency deviation
ΔPtie incremental change in tie-line power

Copyright © 2012 MECS I.J. Intelligent Systems and Applications, 2012, 2, 66-75

View publication stats