An Optimal Design of Coordinated PI based PSS with

TCSC Controller using Modified Teaching Learning

Based Optimization

Bagepalli Sreenivas Theja, Anguluri Rajasekhar Ajith Abraham1,2

Department of Electrical Engineering 1
IT4Innovations, VSB-Technical University of Ostrava, Czech
National Institute of Technology-Warangal, India Republic
2
Email: [email protected] Machine Intelligence Research Labs (MIR Labs), WA, USA
[email protected] Email: [email protected]

Abstract—This paper deals with an interesting application of Transmission devices (FACTS), which can enhance the power
recently evolved Teaching Learning based Optimization (TLBO) system stability and power transfer capability. They are
algorithm in designing coordinated Proportional-Integral (PI) economical, fast acting and can improve the efficiency and
controller based Power System Stabilizer (PSS) for single security of power system [4]. Thyristor Controlled Series
machine infinite bus power system equipped with Thyristor Compensator is one of the first generation FACTS devices. It
Controlled Series Compensator (TCSC). As the design is for is economical and effective means of enhancing dynamic
coordinated system, traditional TLBO results in suboptimal stability of power system by quick and flexible means of
solution and hence we propose a modified TLBO method based adjusting line reactance. It assures better control over power
on the concept of opposition based learning for designing
flow, improvement of transient stability limits and fault
coordinated controllers. Computer simulations of the proposed
approach on various loading conditions reveal the superiority of
current limitation [5-9]. For the small signal stability studies
modified TLBO in designing coordinated controller for of Single Machine Infinite Bus system (SMIB) linear model of
enhancing the dynamic stability of power system. Philip-Heffron is considered. To avoid the destabilizing
interactions the tuning of TCSC controller is coordinated with
Keywords: PSS, TCSC, TLOBA, small signal stability, SMIB. PSS. To further enhance the dynamic stability, PI controllers
are incorporated along with TCSC and PSS controllers.
I. INTRODUCTION As the coordinated controllers consist of more parameters
to be selected judiciously for better performance of the power
Owing to the growing complexity of modern day power
system, this calls for real parameter optimization in n-
systems, they are often interconnected with weak tie lines.
dimensional hyperspace. To carry out this optimization task
Fast acting, high gain Automatic Voltage Regulator’s (AVR)
we chose Teaching Learning Based Optimization (TLBO)
are being employed to the synchronous generators to maintain
algorithm, a newly evolved optimization algorithm. TLBO
the distantly located, inter connected power systems at
draws its inspiration from knowledge sharing phenomenon
constant operating voltage [1]. Though AVRs can enhance the
between students and teacher in a classroom. To further
overall transient stability, they are responsible for low
enhance the performance of TLBO method in designing the
frequency generator rotor angle oscillations (0.1-3 Hz). They
coordinated controllers we propose a new variant TLBO
may further grow in magnitude affecting the small signal
method based on the concept of Opposition. It is referred as
stability, which is the ability of the power system to remain in
TLOBA i.e., teaching learning opposition-based algorithm.
synchronism when the small disturbances due to variations in
The rest of paper is organized as follows. Section II deals
generation and loads occur [2].
with mathematical modeling of power system considered. In
In order to produce positive damping on these small
Section III a brief outline of problem is discussed. Section IV
frequency oscillations, Power System Stabilizers (PSS) are
summarizes the proposed approach followed by design
employed. The purpose of PSS is to introduce supplementary
perspectives in Section V. In section VI we elucidated the
signals (derived from speed deviation signal Δω) in the performance of modified TLBO over various loading
feedback loop of voltage regulator. Design of effective PSS is conditions and at the end we provide some conclusions and
very difficult when the frequency of oscillations begun to vary future scope in Section VII.
over a wide range. Also PSS causes variations in voltage
profiles and their operation is relatively slow [3].
The recent advancements in the high power semiconductor
technology lead to the development of Flexible AC
 II. POWER SYSTEM MODELING B. PSS and Excitation system
The Single Machine connected to Infinite Bus through The conventional two-stage lead-lag Power System
transmission line with TCSC controller shown in Figure 1 is Stabilizer is considered in this study. IEEE Type-ST1A
being considered for small signal stability studies Excitation system is considered. The inputs to excitation
system are terminal voltage (VT), supplementary signal (Vs)
A. Generator Modeling from PSS and reference voltage (Vref). KA and TA are the gain
The 3rd order model consisting of the swing equation and and time constant of excitation system respectively.
the generator internal voltage equation describes the
generator. IEEE–ST1type excitation system is considered.

Figure 1: Single machine infinite bus power system with TCSC [6] Figure2: PSS and IEEE Type-ST1A Excitation system [6]

The input signal to the PSS is (Δω) and output of PSS is a

The nonlinear model of SMIB system with TCSC is given
as below supplementary control signal (ΔVs) to excitation system. It
comprises a wash out block acting as high pass filter, with
δ•= ω0 (ω −1) ; (1)
time constant (Tw) high enough to allow signals associated
ω•= (Pm − Pe ) / M ; (2) with oscillations in input signal to pass unchanged. The lead-
lag compensation blocks produce a component of electrical
E•'q= (−E 'q − (x d − x 'd )i d + E fd ) / Tdo' ; (3)
torque in the direction of speed deviation (Δω). The gain (Kp)
E = (−E fd + K A (VR − VT + Vs )) / TA ;
•
fd
(4) determines the damping level.
'
E VB 2
V (X q − X ) '
C. Thyristor Controlled Series Compensator (TCSC)
Pe = q
sin δ − B d
sin 2δ ; (5)            TCSC   consists   of   three   main   components:   capacitor  
X d ∑1 2X d ∑1X q ∑1
bank   C,   bypass   inductor   L   and   bidirectional   thyristor   T1  
'
E Xd ∑ VB (X q − X 'd ) and  T2.  The  firing  angles  of  the  thyristors  are  controlled  to  
Eq = q
− cosδ ; (6)
X d ∑1 X d ∑1 adjust  the  TCSC  reactance.  
 
X q VB
VTd = sin δ (7)
X q ∑1
X Eff E 'q VBX 'd
VTd = + cosδ ; (8)
X d ∑1 X d ∑1
2
VT = (VTd 2
+ VTq) ; (9)  
Figure 3: TCSC configuration [6]
X Eff = X T + X L − X CF − X TCSC (α) ; X q ∑1 = X q + X Eff  
The  equivalent  capacitive  reactance  provided  by  TCSC  
X d ∑1 = X 'd + X Eff ; X d ∑ = X d + X Eff controller  as  function  of  firing  angle  is  given  as.
To obtain the Philip-Heffron’s model of Single Machine X C2 σ + sin σ
X TCSC (α) = X C −
Infinite Bus with TCSC controller, the system equations are XC − X P π (14)
to be linearized around an operating condition of Power 4X C2 cos 2 (σ / 2) (k tan(kσ / 2) − tan(σ / 2))
+
system. (X C − X P ) (k 2 −1) π
Δδ•= ω0 Δω (10) XC  =  Nominal  reactance  of  the  fixed  capacitor  C.  
'
Δω = [−K1Δδ − K 2 ΔE − K p Δσ − DΔω] / M
•
(11) XP  =  Inductive  reactance  of  inductor  L  connected  in  
q
                 parallel  with  C.  
•' '
ΔE = [−K 4 Δδ − K 3ΔE − K q Δσ + ΔE fd ] / T
q q
'
do
(12) σ = 2(π − α)  =  Conduction  angle  of  TCSC  controller.  
•
fd
'
ΔE = [−K A (K 5Δδ + K 6 ΔE + K v Δσ) − ΔE fd ] / TA
q
(13) XC
'
k= = Compensation ratio
K1 = ∂Pe / ∂δ, K 2 = ∂Pe / ∂E , K p = ∂Pe / ∂σ q
XP
K 4 = ∂E q / ∂δ, K 3 = ∂E q / ∂E 'q , K q = ∂E q / ∂σ
K 5 = ∂VT / ∂δ, K 6 = ∂VT / ∂E q ', K v = ∂VT / ∂σ
 III. PROBLEM FORMULATION an integral control action and eliminates the steady state
error. For TCSC controller u=Δσ and for PSS u=ΔVs.
A. TCSC Controller Transfer functions of PSS and TCSC controllers with PI
In   this   study   the   conventional   lead–lag   structure   has   controllers are
been   chosen   as   a   TCSC   controller.   The   TCSC   controller   ! sT $! 1+ sT $! 1+ sT $
block  representation  is  shown  in  Figure  4.  It  consists  of  a   u PSS = K P ## WP
&&## 1P
&&## 3P
&& G(s)                                (18)  
gain  block,  signal  wash  out  block  and  a  two  stage  lead-­‐lag   " 1+ sTWP %" 1+ sT2P %" 1+ sT4P %
phase  compensation  blocks.  These  blocks  serve  the  same   ! sT $! 1+ sT $! 1+ sT $
purpose   as   in   PSS.   The   phase   compensation   block   u TCSC = K T ## WT
&&## 1T
&&## 3T
&& G(s) (19)
" 1+ sT WT % "
1+ sT2T % "
1+ sT4T %
provides   the   appropriate   phase-­‐lead   characteristics   to  
compensate   for   the   phase   lag   between   input   and   the   Ki
output  signals.  The  signal  washout  block  serves  as  a  high-­‐ where G(s) = K p +
s
pass   filter.   Damping   level   can   be   adjusted   by   modulating   The input signal of the TCSC stabilizer is the speed
KT.   deviation Δω and the output is change in conduction
  angle Δσ. During steady state conditions XEff= XT
+XL−XTCSC(α ) and Δσ =0. During dynamic conditions the
0

series compensation is modulated for effective damping of

system oscillations. The effective reactance in dynamic
conditions is: XEff= XT +XL−XTCSC(α), where σ=σ +Δσ & 0

σ=2(π−α), α and σ being initial value of firing & conduction

0 0

angle respectively.

Figure 4: Structure of TCSC controller [6] B. Objective Function

The design of coordinated controller is done based on
The   damping   torque   contributed   by   TCSC   can   be   minimizing the objective function considered such that power
considered  to  be  in  two  parts.  The  first  part  KP  referred  as   system oscillations after a disturbance are effectively damped
direct   damping   torque   and   is   directly   applied   to   electro   out so as to improve the stability. In this approach the
mechanical   oscillation   loop   of   the   generator.   The   second   objective function is formulated in such way rotor speed
part   comprises   of   both   Kq   and   KV   referred   as   indirect   deviation Δω is minimized and mathematically formulated as
damping   torque,   applied   through   the   field   channel   of   follows
t1
generator.   The   damping   torque   contributed   by   TCSC   2
controller  to  the  electromechanical  oscillation  loop  of  the   J = ∑ ∫ t #$Δω(t,X)%& dt (20)
generator  is:   0

ΔTD  =  TDω0Δω  ≅  KPKTKDΔω (15) In the above equations, Δω(t, X) denotes the rotor speed
The   transfer   functions   of   the   PSS   and   the   TCSC   deviation for a set of controller parameters X. Here X
controller  are  (8)  and  (9)  respectively:   represents the parameters to be optimized. The optimization
is carried in two phases, initially the 10 parameters
⎛ sTWP ⎞⎛ 1 + sT1P ⎞⎛ 1 + sT3 P ⎞ corresponding to both TCSC and PSS controller are been
u PSS = K P ⎜⎜ ⎟⎟⎜⎜ ⎟⎟⎜⎜ ⎟⎟    (16)  
tuned coordinately and in second phase by fixing the obtained
⎝ 1 + sTWP ⎠⎝ 1 + sT2 P ⎠⎝ 1 + sT4 P ⎠ parameters of TCSC and PSS controllers, the PI parameters
⎛ sTWT ⎞⎛ 1 + sT1T ⎞⎛ 1 + sT3T ⎞ Kp and Ki of both TCSC and PSS are tuned coordinately to
uTCSC = KT ⎜⎜ ⎟⎟⎜⎜ ⎟⎟⎜⎜ ⎟⎟ (17)   obtain optimum system response.
⎝ 1 + sTWT ⎠⎝ 1 + sT2T ⎠⎝ 1 + sT4T ⎠
In this structure, the washout time constants TWT and TWP
are usually pre-specified, TWT= TWP=5s. The controller gains
KT & KP and the time constants T1T , T2T, T3T, T4T, T1P , T2P,
T3P , T4P are to be determined.
Δω ⎛ sTw ⎞ ⎛ 1 + sT1 ⎞
Input K ⎜⎜ ⎟⎟ ⎜ ⎟
⎝ 1 + sTw ⎠ ⎝ 1 + sT 2 ⎠

u Ki ⎛ 1 + sT3 ⎞
Kp + ⎜ ⎟
Output s ⎝ 1 + sT 4 ⎠

Figure 5: Structure of TCSC controller and PSS with PI controller

The proportional gain Kp provides a control action Figure 6: Phillips-Heffron model of SMIB with TCSC and PSS [6]
proportional to the error and reduces the rise time. The
integral gain Ki reduces the steady state error by performing
 IV. TEACHING LEARNING OPPOSITION BASED Pseudo code of Learner Phase
OPTIMIZATION ALGORITHM
For k = 1 to n
A. Teaching Learning Based Optimization Randomly select another learner Q, such that
Teaching Learning Based Optimization or simply TLBO X new
total−k,i
≠ X new
total−Q,i
is a new meta-heuristic [13-19] optimization algorithm
proposed by Rao et al [10]. TLBO can briefly classified into IF X new
total−k,i
< X new
total−Q,i

two phases i.e., (i) Teacher Phase and (ii) Learner Phase X new
j,k,i
= X new
j,k,i (
+ rand i X new
j,k,i
− X new
j,Q,i )
i. Teacher Phase ELSE
In this his phase a teacher tries to ameliorate the mean X new
j,k,i
= X new
j,k,i (
+ rand i X new
j,Q,i
− X new
j,k,i )
result of class in the subject taught by him/her based on level End IF
of knowledge and skill he/she had in that particular subject. End FOR
For any ith iteration, let us consider there are m number of
Accept X new if it gives a better function value.
subjects (design variables), n number of learners (population j,k,i

size, k=1, 2,..,n) and Tj,i be the mean result of the learners in
B. Teaching Learning Opposition Based Optimization
jth subject (where j=1,2,…m). However the best overall result
i. Opposition-Based Learning
Xtotal-k-best,i (considering all the subjects together) in a class of
Most of the evolutionary optimization methods start with
learners can be considered as result of best leaner k-best and
some initial solutions and usually start with random guesses.
the best learner identified is replaced by the teacher. As the
The computational time depends upon the distance between
teacher Xtotal-k-best,i will try to move mean Ti towards its own
initial guess and optimal solution. Hence if the guess is not in
level, an adaptive heuristic is used to update the solution and
the vicinity of optimal solution computation time may
is done according to the difference between the existing mean
increase. The chance of improving our convergence can be
result of each subject and the corresponding result of the
done by starting with a fitter solution by simultaneously
teacher for each subject is given by.
checking the opposite solution [11]. If x is a obtained solution
Difference _ mean j,k,i = rand i ( X j,k−best,i − TFTj,i ) (21) of given function then the opposite solution x’ can be
where TF is termed as teaching factor, which decides calculated as follows
whether the value of mean is to be changed or not. The value x' = a + b − x (23)
of TF can be either 1 or 2, which is decided randomly with where x ∈ R within an interval of [a,b]
equal probability and randi is a random number in the range
[0, 1]. Xj,k-best,i is the result of the teacher in subject j. ii. Opposition-based Optimization
Difference_Meanj,k,i, defined in Eqn (21) is used in updating Let P = {x1 ,x 2 ,...,x D } be a point in D-dimensional space,
the existing solution according to the following expression.
where x1 ,x 2 ,...,x D ∈ R and x i ∈ [a i ,bi ] ∀i ∈ {1,2,...,D} Now
X new
j,k,l
= X j,k,l + Difference _ Mean j,k,i (22)
the opposite point P' = {x1' ,x '2 ,...,x 'D } is defined as
where X new
j,k,i
and X j,k,i are the new and existing values
corresponding to jth subject of kth learner of ith iteration. A x 'i = a i + bi − x i (24)
greedy mechanism is performed between X new and X j,k,i , the Now, with above definition of opposite point the
j,k,i
opposition based optimization can be formulated as follows.
learner with better function value is retained.
ii. Learner Phase
()
Assuming f ⋅ is fitness function via which candidate fitness

In the course of time a learner may interact randomly with is measured and according to the above given definitions of
other learners with the help of communications, discussions, P and P' , if f (P' ) ≥ f (P) then the point P can be replaced
etc. If a leaner interacts with other learner who has more
with P' ; hence, the point and its opposite point are evaluated
knowledge than him or her, he/she tries to learn new things simultaneously in order to go with the fitter one.
and tries to increase his/her knowledge. For a class of n
learners the learning phenomenon of this phase is expressed iii. Proposed Algorithm
with following pseudo code. Opposition scheme discussed above is applied two times
for the proposed TLOBO method at starting of teaching phase
and learning phase respectively. Once the algorithm has
started with random initial population simultaneously
opposite population are also calculated and then best n values
are picked up (based on the fitness value) and then passed in
to the teacher phase. Similarly before entering in to the
learning phase opposite population is evaluated and the best n
values are passed in to the learning phase and the rest is same
as that of TLBO. This is continued till the termination
0.16
criterion is reached. PSO - PSS
Instead of using predefined interval boundaries 0.14 PSO tuned TCSC-PSS
TLOBA tuned TCSC-PSS
[a i ,bi ] here we used the minimum and maximum values TLOBA tuned PI basedTCSC-PSS
0.12
( [a min ,b max ] ) of each dimension in current population to

rotor angle deviation

j j
0.1
calculate the opposite population. This type of opposition
helps the learners to get good information and it is computed 0.08

as: 0.06
OPi, j = a min
j
+ b max
j
− Pi, j (25)
0.04
where Pi,j is the jth vector of the ith learner in the
0.02
population. OPi,j is the opposite position of Pi,j ; a min
j
and
0
b max
j
are the minimum and maximum values of the jth 0 1 2 3 4
time (sec)
5 6 7

dimension in current population respectively.

Figure 7(a): Rotor Angle Deviation: Heavy loaded
V. DESIGN OF COORDINATED PI CONTROLLER BASED PSS -4
x 10
WITH TCSC CONTROLLER 8
PSO-PSS
A. Parameters of power system considered 6 PSO tuned TCSC-PSS
TLOBA tuned TCSC-PSS
For the small signal stability analysis of single machine TLOBA tuned PI based TCSC-PSS
infinite bus the design of the system and system data is taken 4
from [6]. As the optimization is to be carried out in a bounded
Speed Deviation

search we had used the following ranges for different 2

parameters in our design and they are recorded in Table I. The

parameters being considered for tuning were KT, KP, T1T, T2T, 0

T3T, T4T, T1P, T2P, T3P, T4P and PI controller parameters (Kp, Ki)
of both TCSC and PSS controllers. (the parameters with -2

subscript T indicates they belong to TCSC controller and that

of P indicates they belong to PSS Control. The ranges over -4

which these parameters tuned as per standards are 30< KP, KT

-6
<80 & 0.1 < T1T, T3T, T1P, T3P < 0.6 & 0.02 < T2T, T4T, T2P, T4P 0 2 4 6 8 10
< 0.4 & 0 < Kp < 50, 0 < Ki < 10. Time (sec)

Figure 7(b): Speed Deviation: Heavy loaded

B. Parameters of TLOBA Algorithm
0.16
The objective function considered consists of total 14-D PSO-PSS
i.e., 10-D for PSS-TCSC followed by 4-D optimization of 0.14 PSO tuned TCSC-PSS
two PI controllers. Hence termination criteria of 200 TLOBA tuned TCSC-PSS
0.12 TLOBA tuned PI based TCSC-PSS
Functional Evaluations (NFEs) are considered to get optimum
design. The population has been judiciously chosen to be 10.
rotor angle deviation

0.1

VI. SIMULATIONS AND RESULTS 0.08

In this context we considered three different loading 0.06

conditions and they are as follows:
i. Nominal Loading: Pe=1.0, Qe=0.303. 0.04
ii. Light Loading: Pe=0.3, Qe=0.015 and system inertia
reduces by 25%. 0.02

iii. Heavy Loading: Pe=1.01, Qe=0.1 and total line

0
reactance increases by 30%. 0 1 2 3 4 5 6 7
For a step change of 5% in input (Pm), in Figures 7-9 the time (sec)

responses obtained for nominal, heavy and light loaded Figure 8(a): Rotor Angle Deviation: Light loaded
systems are depicted in terms of speed deviation and rotor
angle deviations. Figure 8 shows the convergence
characteristics of TLOBA progressing towards optimum
values without and with PI controllers. Table 3 shows the
time domain indices values for different loading conditions in
terms of peak value and settling time.
 x 10
-4
shoot and low settling time of speed deviation response.
10
PSO - PSS However the peak over shoot value for the rotor angle
PSO tuned TCSC-PSS deviation response with proposed controller is a bit high
TLOBA tuned TCSC-PSS when compared to case without coordinated tuned PI
TLOBA tuned PI based TCSC-PSS
controllers for TCSC and PSS controllers.
5
Similarly Figures 8(a) and 8 (b) shows the speed deviation
speed deviation

and rotor angle deviation responses for lightly loaded system.

As expected proposed PI based TLOBA tuned coordinated
TCSC-PSS controller has shown less settling times for both
0 speed and rotor angle deviation responses. Unlike heavy
loaded system, this system enriched with coordinated
controllers gave less the less peak over shoots for both rotor
angle and speed deviation responses. Figures 9(a) and 9(b)
-5
depicts the speed deviation and rotor angle deviations for the
0 1 2 3 4 5 6 7 nominal loaded system. PI based TCSC-PSS controller tuned
time (sec)
with TLOBA algorithm has shown relatively less settling
Figure 8(b): Rotor Angle Deviation: Light loaded times and less peak over shoot values for both speed and rotor
angle deviation responses.
0.14
PSO-PSS -4
x 10
0.12 PSO tuned TCSC-PSS 2.8
TLOBA tuned TCSC-PSS NOMINAL SYSTEM
TLOBA tuned PI based TCSC-PSS LIGHT LOADED SYSTEM
0.1 2.6 HEAVY LOADED SYSTEM
rotor angle deviation

Objective Function (In Log scale)

0.08
2.4

0.06
2.2
0.04

2
0.02

1.8
0
0 2 4 6 8 10
time (sec)
1.6
Figure 9(a): Rotor Angle Deviation: Nominal loaded 0 50 100 150 200
Number of Function Evaluations

x 10
-4
Figures 10(a): Convergence of TLOBA towards minimum: without
10
PI Controller
PSO-PSS -4
x 10
8 PSO tuned TCSC-PSS 1.6
TLOBA tuned TCSC-PSS NOMINAL SYSTEM
6 TLOBA tuned PI based TCSC-PSS LIGHT LOADED SYSTEM
1.4 HEAVY LOADED SYSTEM
Objective Function (In Log scale)

4
speed deviation

1.2
2

0 1

-2
0.8
-4

0.6
-6
0 1 2 3 4 5 6 7
time (sec)
0.4
Figure 9(b): Speed Deviation: Nominal loaded 0 50 100 150 200
Number of Function Evaluations

Figures 7(a) and 7(b) show the rotor angle deviation and Figure 10(b): Convergence of TLOBA towards minimum: without
the speed deviation responses respectively for the heavy PI Controller
loaded system. From Figures 7(a-b) and the time domain
indices recorded in Table 3 it is clear that proposed PI
controller based TCSC-PSS has produced less peak over
 Table 1: Parametric Values Obtained for coordinated TCSC- PSS Using TLOBA and obj func. Minimization values

Parameter Nominal Heavy Light

Loading Loading Loading
KTCSC 30.0000 30.0000 30.0000
T1T 0.4226 0.4185 0.1000
T2T 0.2778 0.2394 0.1912
T3T 0.3428 0.1000 0.2532
T4T 0.3940 0.2067 0.2568
KPSS 30.0000 30.0000 30.0000
T1P 0.3076 0.2198 0.2482
T2P 0.1960 0.3138 0.2733
T3P 0.1021 0.2163 0.1811
T4P 0.3921 0.3353 0.1678
Obj Fun without PI 1.7684e-04 1.9699e-04 1.9932-03
mean(std) (1.300e-07) (4.394e-07) (1.125e-08)

Table 2: Parametric Values Obtainedof coordinated PI controller TCSC- PSS Using TLOBA and obj func. Minimization values
Parameter Nominal System Heavy loaded System Light Loaded System

KP_TCSC 4.3709 5.2020 2.6232

Ki_TCSC 10.0000 10.0000 10.0000
KP_PSS 0.6785 3.6136 4.8863
Ki_PSS 10.0000 10.0000 10.0000
Obj Fun with mean 4.2233e-05 5.0702e-05 5.29012e-05
(standard deviation) (1.136e-07) (2.4243e-07) (9.71640e-09)

Table 3. Settling times and peak values for various loading conditions

Settling time Ts(sec) Peak value

Loading Response PSO- PSO- TLOBA- TLOBA-PI- PSO-PSS PSO-TCSC- TLOBA- TLOBA-PI-
condition PSS TCSC-PSS TCSC-PSS TCSC-PSS PSS TCSC-PSS TCSC-PSS
Speed
Nominal deviation
8.234 5.65 3.65 2.67 9.58e-04 3.095e-04 2.725e-04 1.717e-04
system Rotor angle
deviation
9.452 6.212 4.215 3.172 0.1256 0.0848 0.0767 0.0765
Speed
Heavy deviation
8.23 6.16 3.31 2.68 7.816e-04 6.87e-04 4.09e-04 2.814e-04
loaded Rotor angle
system deviation
8.76 7.59 5.66 4.07 0.1422 0.1035 0.0909 0.0922
Speed
Light deviation 7.76 4.56 3.92 3.16 8.07e-04 6.258e-04 4.41e-04 2.84e-04
loaded Rotor angle
system deviation 8.17 5.45 5.17 2.83 0.146 0.0975 - 0.0916

Finally Figures 10(a) and 10 (b) shows the convergence

characteristics of TLOBA algorithm towards optimum values Various simulations for different loading conditions have
without and with PI controllers respectively. The parameters been explored, and results validates the superior performance
obtained via optimal tuning are recorded in Tables 1 and 2. of the proposed system when tuned optimally.
From both the Tables it is further evident that objective Our future research will be focusing on implementing
function value is less for the PI controller case. fractional order controllers via a multi-objective frame work
for multi-machine systems.
VII. CONCLUSIONS
In this paper, a new intelligent method of designing the ACKNOWLEDGEMENTS
coordinated PI controller based TCSC-PSS, tuned with This research work is supported in the framework of the
TLOBA algorithm using Philip-Heffron’s model for SMIB IT4 Innovations Centre of Excellence project, reg. no.
CZ.1.05/1.1.00/02.0070 by operational programme ‘Research
has been proposed. To show the efficacy of proposed
and Development for Innovations’ funded by the Structural
approach we also compared our method with PSO, and also
Funds of the European Union and state budget of the Czech
systems not involving PI controller.
Republic, EU.
 REFERENCES [10] Ali Taleb Al-Awami. “Design of robust PSS and Facts-based
Controllers for Stability Enhacement of Power Systems”, PHD Thesis,
[1] Y.L. Abdel-Magid, M. Bettayeb, M.M. Dawoud, “Simultaneous 2004.
stabilization of power systems using genetic algorithms” in IEE
Proceedings Generation Transmission Distribution, vol. 144, No. 1, [11] R.V. Rao, V.J. Savsani and D.P. Vakharia, “Teaching-Learning-Based
Optimization: A novel optimization method for continuous non-linear
January 1997, pp. 39-44
large scale problems”, Information Sciences, 183 (1), 2012, pp. 1-15.
[2] Y.L. Abdel-Magid, M. Bettayeb, M.M. Dawoud, “Simultaneous
stabilization of power systems using genetic algorithms” in IEE [12] M. Shokri, H.R. Tizhoosh, M.Kamel, “Opposition-based Q(λ)
algorithm” In Proc. IEEE World Congress on Comput, Intell,
Proceedings Generation Transmission Distribution, vol. 144, No. 1,
January 1997, pp. 39-44 Vancouver, BC, Canada, 2006, pp. 646-653.
[3] Vikal, R.; Goyal, G.; , "TCSC Controller Design Using Global [13] H. Izakian, A. Abraham, V. Snasel, Comparison of Heuristics for
Scheduling Independent Tasks on Heterogeneous Distributed
Optimization for Stability Analysis of Single Machine Infinite-Bus
Power System," 15th International Conference on Intelligent System Environments, The 2009 IEEE International Workshop on HPC and
Grid Applications (IWHGA2009), China, IEEE Press, USA, ISBN 978-
Applications to Power Systems, 2009. ISAP '0, pp.1-7.
0-7695-3605-7, pp. 8-12, 2009.
[4] Ali Taleb Al-Awami. Design of robust PSS and Facts-based Controllers
[14] F. Xhafa, E. Alba, B. Dorronsoro, B. Duran and A. Abraham, Efficient
for Stability Enhacement of Power Systems, PHD Thesis, 2004K.
Batch Job Scheduling in Grids Using Cellular Memetic Algorithms,
Elissa, “Title of paper if known,” unpublished.
Studies in Computational Intelligence, Springer Verlag, Germany,
[5] Magaji, N.; Mustafa, M.W.; , "TCSC Damping controller design based ISBN: 978-3-540-69260-7, pp. 273-299, 2008.
on Self-learning fuzzy controller Techniques," 5th International Power
[15] S. Das, A. Biswas, S. Dasgupta and A. Abraham, Bacterial Foraging
Engineering and Optimization Conference (PEOCO), 2011 pp.387-
Optimization Algorithm: Theoretical Foundations, Analysis, and
391, 6-7 June 2011.
Applications, Foundations of Computational Intelligence Volume 3:
[6] Tehrani, F.M.; Shahgholian, G.; Pourghassem, H.; , "Dynamic study Global Optimization, Studies in Computational Intelligence, Springer
and stability analyze of damping cohefision and reactance in TCSC Verlag, Germany, ISBN: 978-3-642-01084-2, pp. 23-55, 2009.
controller connected on optimization SMIB system," 2011 IEEE 3rd
International Confernce on Communication Software and Networks [16] H. Liu, A. Abraham and M. Clerc, Chaotic Dynamic Characteristics in
Swarm Intelligence, Applied Soft Computing Journal, Elsevier Science,
(ICCSN), pp.270-274.
Volume 7, Issue 3, pp. 1019-1026, 2007.
[7] Panda, S; Padhy, N.P.: “Coordinated Design of TCSC Controller and
PSS Employing Particle Swarm Optimization technique,” [17] F. Xhafa and A. Abraham, Meta-heuristics for Grid Scheduling
Problems, Metaheuristics for Scheduling: Distributed Computing
International   Journal   of   Computer   and   Information   Engineering  
2007. Environments, Studies in Computational Intelligence, Springer Verlag,
Germany, ISBN: 978-3-540-69260-7, pp. 1-37, 2008.
[8] Narne, R.; Panda, P.C.; Therattil, J.P.; , "Transient stability
[18] R. Thangaraj, M. Pant, A. Abraham and P. Bouvry, Particle Swarm
enhancement of SMIB system using PSS and TCSC-based controllers,"
IEEE Ninth International Conference on Power Electronics and Drive Optimization: Hybridization Perspectives and Experimental
Illustrations, Applied Maths and Computation, Elsevier Science,
Systems (PEDS), 2011, pp.214-218.
Netherlands, Volume 217, No. 1, pp. 5208-5226, 2011.
[9] Panda, S; Patel R. N; Padhy N. P; “Power System Stability
Improvement by TCSC Controller Employing a Multi- Objective [19] H. Liu, A. Abraham, O.K. Choi and S.H. Moon, Variable
Neighborhood Particle Swarm Optimization for Multi-objective
Genetic Algorithm Approach”, International Journal of Electrical and
Computer Engineering 1:8 2006. Flexible Job-shop Scheduling Problems, The Sixth International
Conference on Simulated Evolution And Learning (SEAL06), China,
Springer Verlag, Germany, Lecture Notes in Computer Science,
T.D.Wang et al. (Eds.): SEAL 2006, LNCS 4247, pp. 197-204, 2006.