Electric Power Systems Research 114 (2014) 49–59

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

Optimal power flow using Teaching-Learning-Based

Optimization technique
H.R.E.H. Bouchekara a,∗ , M.A. Abido b , M. Boucherma a
a
Constantine Electrical Engineering Laboratory, LEC, Department of Electrical Engineering, University of Constantine 1, 25000 Constantine, Algeria
b
Electrical Engineering Department, King Fahd University of Petroleum and Minerals, Dhahran 31261, Saudi Arabia

a r t i c l e i n f o a b s t r a c t

Article history: Teaching-Learning-Based Optimization is a rising star among metaheuristic techniques with highly com-
Received 13 September 2013 petitive performances. This technique is based on the influence of a teacher on learners. In this paper, the
Received in revised form 14 March 2014 Teaching-Learning-Based Optimization technique is used to solve the optimal power flow problem. In
Accepted 30 March 2014
order to show the effectiveness of the proposed method, it has been applied to the standard IEEE 30-bus
and IEEE 118-bus test systems for different objectives that reflect the performances of the power system.
Keywords:
Furthermore, the obtained results using the proposed technique have been compared to those obtained
Teaching-Learning-Based Optimization
using other techniques reported in the literature. The obtained results and the comparison with other
Metaheuristic
Optimal power flow
techniques indicate that the Teaching-Learning-Based Optimization technique provides effective and
Power system optimization robust high-quality solution when solving the optimal power flow problem with different complexities.
Voltage profile © 2014 Elsevier B.V. All rights reserved.
Voltage stability

1. Introduction
The optimal power flow (OPF) problem is the backbone tool for
power system operation [1,2]. The objective of the OPF problem
is to determine the optimal operating state of a power system by
optimizing a particular objective while satisfying certain operating
constraints [3].
The OPF has been studied for over half a century since the
pioneering work of Carpentier [1,2]. Earlier, many traditional
(deterministic) optimization techniques have been successfully
used, the most popular were: gradient based methods, Newton-
based method, simplex method, sequential linear programming,
sequential quadratic programming, and interior point methods. A
survey of the most commonly used conventional optimization algo-
rithms applied to solve the OPF problem is given in [4,5]. Although,
some of these deterministic techniques have excellent convergence
characteristics and many of them are widely used in the industry
however, they suffer from some shortcomings. Some of their draw-
backs are: they cannot guarantee global optimality i.e. they may
converge to local optima, they cannot readily handle binary or inte-
ger variables and finally they are developed with some theoretical
assumptions, such as convexity, differentiability, and continuity,
∗ Corresponding author. Tel.: +213 666605628; fax: +213 31908113.
E-mail address: bouchekara.houssem@gmail.com (H.R.E.H. Bouchekara).
among other things, which may not be suitable for the actual OPF
conditions [5,6].
Furthermore, the rapid development of recent computa-
tional intelligence tools have motivated significant research in
the area of non-deterministic that is, heuristic, optimization
methods to solve the OPF problem in the past two decades
[6]. Some of these techniques are: Ant Colony Optimization
(ACO), Artificial Neural Networks (ANN), Bacterial Foraging
Algorithms (BFA), Biogeography-Based Optimization (BBO), Black-
Hole-Based Optimization (BHBO), Chaos Optimization Algorithms
(COA), Differential Evolution (DE), Evolutionary Algorithms (EAs),
Electromagnetism-Like Mechanism (EM), Evolutionary Program-
ming (EP), Evolutionary Strategies (ES), Fuzzy Set Theory (FST),
Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Sim-
ulated Annealing (SA), Tabu Search (TS), Gravitational Search
Algorithm (GSA). These methods are known for: their capabilities
of finding global solutions and avoid to be trapped with local ones,
their ability of fast search of large solution spaces and their ability
to account for uncertainty in some parts of the power system. A
review of many of these optimization techniques applied to solve
the OPF problem is given in [6,7].
One of the recently developed optimization techniques is the
Teaching-Learning-Based Optimization (TLBO), which is a pop-
ulation based optimization technique inspired by passing on
knowledge within a classroom environment, where learners first
acquire knowledge from teacher and then from classmates [8,9].

http://dx.doi.org/10.1016/j.epsr.2014.03.032
0378-7796/© 2014 Elsevier B.V. All rights reserved.
50 H.R.E.H. Bouchekara et al. / Electric Power Systems Research 114 (2014) 49–59

The main objective of this paper is to apply the TLBO to solve the 2.3. Objective constraints
OPF problem. The performance of the proposed technique is sought
and tested on the standard IEEE 30-bus and IEEE 118-bus test sys- OPF constraints can be classified into equality and inequality
tems where the objective functions are: minimization of generation constraints, as detailed in the following sections.
fuel cost, voltage profile improvement, voltage stability enhance-
ment, voltage stability enhancement during contingency condition, 2.3.1. Equality constraints
piecewise quadratic fuel cost curve and fuel cost minimization of The equality constraints of the OPF reflect the physics of the
generators with valve-point loading. power system. The physics of the power system are represented by
The remainder of the paper is organized as follows. First, the the typical power flow equations. These equality constraints are as
OPF is mathematically formulated. Then, the TLBO is presented. follows.
Next, we apply the proposed TLBO to solve the OPF problem in
order to optimize the power system operating conditions. Finally, (a) Real power constraints
we conclude our paper with some remarks and points.

NB
PGi − PDi − Vi Vj [Gij cos (ij ) + Bij sin (ij )] = 0 (6)
2. Optimal Power Flow formulation
j=1

As mentioned earlier, OPF is a power flow problem which gives (b) Reactive power constraints
the optimal settings of control variables for a given settings of load
by minimizing a predefined objective function such as the cost of

NB
QGi − QDi − Vi Vj [Gij sin (ij ) + Bij cos (ij )] = 0 (7)
active power generation and under the consideration of operating
j=1
limits of the system. The OPF problem can be formulated as a non-
linear constrained optimization problem as follows:
where  ij =  i −  j , NB is the number of buses, PG is the active power
Minimize J(x, u) (1) generation, QG is the reactive power generation, PD is the active
load demand, QD is the reactive load demand, Gij and Bij are the
Subject to g(x, u) = 0 (2) elements of the admittance matrix (Yij − Gij + j Bij ) representing the
conductance and susceptance between bus i and bus j, respectively.
and h(x, u) ≤ 0 (3)
2.3.2. Inequality constraints
where u, vector of independent variables or control variables; x,
The inequality constraints of the OPF reflect the limits on physi-
vector of dependent variables or state variables; J(x,u), objective
cal devices present in the power system as well as the limits created
function; g(x,u), set of equality constraints; h(x,u), set of inequality
to guarantee system security. These inequality constraints are as
constraints.
follows.
The control variables u and the state variables x of the OPF
problem are stated in (4) and (5), respectively.
(c) Generator constraints
For all generators including the slack: voltage, active and reac-
2.1. Control variables tive outputs ought to be restricted by their lower and upper
limits as follows:
These are the set of variables which can be modified to satisfy the
load flow equations. The set of control variables in the OPF problem VGmin ≤ VGi ≤ VGmax , i = 1, . . ., NG (8)
i i
formulation are:
PG , active power generation at the PV buses except at the slack PGmin ≤ PGi ≤ PGmax , i = 1, . . ., NG (9)
i i
bus; VG , voltage magnitude at PV buses; T, tap settings of trans-
QGmin ≤ QGi ≤ QGmax , i = 1, . . ., NG (10)
former; QC , shunt VAR compensation. i i

Hence, u can be expressed as: (d) Transformer constraints

Transformer tap settings ought to be restricted within their
uT = [PG2 · · ·PGNG , VG1 · · ·VGNG , QC1 · · ·QCNC , T1 · · ·TNT ] (4)
specified lower and upper limits as follows:
where NG, NT and NC are the number of generators, the number Timin ≤ Ti ≤ Timax , i = 1, . . ., NT (11)
of regulating transformers and the number of VAR compensators,
respectively. (e) Shunt VAR compensator constraints
Shunt VAR compensators must be restricted by their lower
and upper limits as follows:
2.2. State variables
QCmin ≤ QGCi ≤ QCmax , i = 1, . . ., NG (12)
i i
These are the set of variables which describe any unique state of
the system. The set of state variables for the OPF problem formula- (f) Security constraints
tion are: These contain the constraints of voltage magnitude at load
PG1 , active power output at slack bus; VL , voltage magnitude at buses and transmission line loadings. Voltage of each load bus
PQ buses, load buses; QG , reactive power output of all generator must be restricted within its lower and upper operating limits.
units; Sl , transmission line loading (or line flow). Line flow through each transmission line ought to be restricted
Hence, x can be expressed as: by its capacity limits. These constraints can be mathematically
formulated as follows:
xT = [PG1 , VL1 · · ·VLNL , QG1 · · ·QGNC , Sl1 · · ·Slnl ] (5)
VLmin ≤ VLi ≤ VLmax , i = 1, . . ., NL (13)
i i
where, NL, and nl are the number of load buses, and the number of
transmission lines, respectively. Sli ≤ Slmax , i = 1, . . ., nl (14)
i
 H.R.E.H. Bouchekara et al. / Electric Power Systems Research 114 (2014) 49–59 51

It is worth mentioning that control variables are self-
constrained. The inequality constraints of dependent variables
which contain load bus voltage magnitude; real power generation
output at slack bus, reactive power generation output and line load-
ing can be included into an objective function as quadratic penalty
terms. In these terms, a penalty factor multiplied with the square of
the disregard value of dependent variable is added to the objective
function and any unfeasible solution obtained is declined. Mathe-
matically, penalty function can be expressed as follows:
the students is assessed through the mean value of the population.
Moreover, the teacher puts effort to increase the mean of students
to a higher level, at which students will require another teacher of
better quality to teach them [8].
Remarkable results have been reported about TLBO outper-
forming ES, PSO, ABC, DE, and GEM on a number of constrained
benchmark functions and constrained mechanical design prob-
lems, TLBO discovers better or equal solutions much faster than
the abovementioned techniques as reported in [9].

2

NL
2
Jaug = J + P (PG1 = PGlim ) + v (VLi − VLlim ) + Q 3.2. TLBO algorithm
1 i
i=1

The TLBO algorithm is given in Algorithm 1. Like other


NG
2

nl
× (QGi − QGlim ) + S (Sli − Slmax )
2
(15) population-based algorithms, TLBO starts with an initialization
i i phase where a randomly generated population of candidate solu-
i=1 i=1
tions, are placed in the search space of the problem consisting of
where ␭P , ␭V , ␭Q and ␭S are penalty factors and xlim is the limit value n dimensions where each dimension is limited by an upper and a
of the dependent variable x. If x is higher than the upper limit, xlim lower bound. Then, the process of TLBO is divided into two parts
takes the value of this one, likewise if x is lower than the lower limit namely: the ‘Teacher Phase’ and the ‘Learner Phase’.
xlim takes the value of this limit hence:

xmax ; x > xmax 3.2.1. Teacher phase
lim
x = (16)
xmin ; x < xmin The knowledge of a class increases depending upon a good
teacher because he/she brings the level of his/her learners to his/her
level of knowledge. However, in actual life this is not always the
3. Teaching-Learning-Based Optimization (TLBO)
case because the level of learners depends on other factors like
their aptitudes and their efforts and commitment to learn. Thus, a
3.1. Overview
teacher can only increase the mean level of his/her learners [8]. The
procedure of how a teacher improves the mean level of learners is
Nature-inspired methods work on the principles of different
exposed in Algorithm 1 between lines (11–14).
natural phenomena and include, for example, Genetic Algorithm
(GA), Particle Swarm Optimization (PSO), Artificial Bee Colony
(ABC), Ant Colony Optimization (ACO), Harmony Search (HS), 3.2.2. Learner phase
and Grenade Explosion Method (GEM). However, in such and In this phase the learners improve their knowledge by inter-
many other techniques, the algorithm performance is affected acting with other learners i.e. between themselves. A learner i
by its specific control parameters [8]. This triggers the need interacts with another learner j randomly selected using randi as
for parameter-free optimization techniques where no algorithm shown in line (15). A learner learns something new i.e. increases his
parameters are required for the operation of the algorithm. knowledge if the second learner has more knowledge than him [8].
TLBO is an algorithm-specific parameter-less, nature-inspired The procedure for learner modification is given in lines (27–28).
metaheuristic method introduced recently by Rao and colleagues
[8]. In contrast with the other techniques, TLBO does not require Algorithm 1 (TLBO pseudocode.).
any algorithm-specific parameters and only requires such control- 1 n: dimension of the problem
ling parameters as population size and number of generations for 2 m: population size
3 MAXITER: maximum number of iterations
its operation [8]. Moreover, it is claimed that TLBO outperforms
4 Initialization()
some other widespread metaheuristics with regard to constrained 5 while ITER < MAXITER
benchmark functions, constrained mechanical design, and contin- 6 Elite ← SelectBest(P,Elite)
uous non-linear numerical optimization problems [9]. 7 for i = 1:m
Like other nature-inspired algorithms, TLBO uses a population 8 TF = round(1 + rand)
9 Xmean ← mean(Xi )
of solutions to evolve to the optimal solution [8]. It is inspired by the 10 Xteacher ← best(Xi )
process of knowledge, where learners first acquire knowledge from 11 Xnew,i = Xi + rand · (Xteacher − (TF · Xmean ))
a teacher and then between themselves [9]. By analogy to other 12 if f(Xnew,i ) < f(Xi )
population-based optimization techniques, in TLBO, the population 13 Xi ← Xnew,i
14 end if
is considered as a group or a class of learners, the design variables
15 j ← randi(m)
are analogous to subjects offered to learners and the learners’ result 16 if j =/ i
is analogous to the ‘fitness’, The teacher is considered as the best 17 if f(Xi ) < f(Xj )
solution obtained so far [8]. Therefore, TLBO works on the influence 18 Xnew,i = Xi + rand · (Xi − Xj )
of a teacher on learners and influence of learners on their col- 19 else
20 Xnew,i = Xi + rand · (Xj − Xi )
leagues. The influence is usually manifested by the learners’ results
21 end if
or outcomes. Better results of a class are typically represented 22 end if
by the students’ mean grade. In general, the teacher attempts 23 if f(Xnew,i ) < f(Xi )
to distribute knowledge among learners to increase their knowl- 24 Xi ← Xnew,i
25 end if
edge level and help them enhance their grades. Consequently, the
26 end for
teacher will increase the mean grade of the class according to his 27 P ← ReplaceWorstWithElite(P,Elite)
capability. However, despite the great effort made by the teacher, 28 P ← RemoveDuplicateIndividuals(P)
students will not only gain knowledge based on his teaching qual- 29 ITER = ITER +1
ity, but also on the quality of students sitting in the class. Quality of 30 end while
52 H.R.E.H. Bouchekara et al. / Electric Power Systems Research 114 (2014) 49–59

4. Application and results 870

860
In order to illustrate the effectiveness of the proposed TLBO tech-
nique, it has been tested on: the IEEE 30- and the IEEE 118-bus test
850
systems.
The developed program is written in MATLAB computing envi-
840

Cost ($/hr)
ronment and applied on a 2.20 GHz i7 personal computer with
8.00 GB-RAM. 830

4.1. IEEE 30-bus test system 820

The IEEE 30-bus test system has the following characteristics: 810
six generators at buses 1, 2, 5, 8, 11 and 13, four transformers with
off-nominal tap ratio at lines 6–9, 6–10, 4–12, and 28–27, 3) and 800

nine shunt VAR compensation buses at buses 10, 12, 15, 17, 20, 21,
23, 24 and 29 [10,11].
The line, bus data, generator data and the minimum and max-
imum limits for the control variables are given in [12] for Case 1
through Case 4 and in [13] for Cases 5 and Case 6. The proposed
technique has been applied to solve the OPF problem for several
cases with different objective functions.
4.1.1. Case 1: minimization of generation fuel cost
The minimization of the generation fuel cost is studied in this
case. Generally, the generation fuel cost curve is expressed by a
quadratic function, that is:
fi = ai + bi PGi + ci PG2 ($/h) (17)
i
where ai , bi and ci are the basic cost coefficient, the linear cost
coefficient and the quadratic cost coefficient of the ith generator,
respectively. The values of these coefficients are given in [12].
Therefore, the total fuel cost of all generator units, which repre-
sents the objective function for this case is expressed as follows:

NG
J= fi ($/h) (18)
i=1
where fi is the fuel cost of the ith generator given by (17).
The proposed technique has been run for Case 1 and the opti-
mal settings obtained are tabulated in Table 1. It appears that the
790
0 50 100 150 200 250 300 350 400 450 500
Iterations
Fig. 1. Fuel cost variation for Case 1.
total fuel generation cost is highly reduced compared to Case 0
(the initial case). Quantitatively, it is reduced from 901.9516 $/h to
799.0715 $/h which represents a percent cost reduction of 11.41%.
Furthermore, it is noteworthy to mention that Case 0 has some
voltage violations at bus 19 through bus 30. However, there are
no longer violations in the results obtained using TLBO. The trend
of minimizing the total generation fuel cost is sketched in Fig. 1. It
appears from this figure that the proposed method has an excellent
convergence speed. Moreover, the execution time of one iteration
is 0.92 s. The power flow through transmission lines of the consid-
ered system for this case is tabulated in Table 2 and a sketch of this
flow is given in Figs. 2 and 3.
4.1.2. Case 2: voltage profile improvement
One of the most important and significant safety and service
quality indices is bus voltage [12]. Though the minimization of total
generation fuel cost may give a feasible solution, the voltage profile
may not be acceptable. Therefore, the goal of this second case is the
minimization of the fuel cost along with the improvement of the
voltage profile by minimizing the voltage deviation of PQ buses

Power i-->k
100 Power k-->i

50
Power (MW)

-50

-100

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41

3
Losses
Losses (MW)

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41

Fig. 2. Active power flow and losses through transmission lines for Case 1.
 H.R.E.H. Bouchekara et al. / Electric Power Systems Research 114 (2014) 49–59 53

Table 1
Optimal settings of control variables.

Min Max Case 0 Case 1 Case 2 Case 3 Case 4 Case 5 Case 6

P1 50 200 99.2230 177.0578 176.7551 175.8077 173.9096 140.0000 194.1673

P2 20 80 80 48.6973 48.8437 48.4082 47.9080 55.0000 44.3449
P5 15 50 50 21.3044 21.5128 21.3735 21.3338 23.9338 20.6902
P8 10 35 20 21.0811 21.9212 22.0685 25.2421 35.0000 13.1553
P11 10 30 20 11.8843 12.2860 12.4877 12.8701 19.6291 10.0000
P13 12 40 20 12.0000 12.0007 12.0171 12.0023 16.9374 12.0000
V1 0.95 1.1 1.05 1.1000 1.0429 1.1000 1.1000 1.0500 1.0500
V2 0.95 1.1 1.04 1.0878 1.0259 1.0870 1.0868 1.0404 1.0287
V5 0.95 1.1 1.01 1.0617 1.0148 1.0830 1.1000 1.0150 1.0024
V8 0.95 1.1 1.01 1.0694 1.0071 1.0942 1.1000 1.0265 1.0163
V11 0.95 1.1 1.05 1.1000 1.0253 1.1000 1.0118 1.1000 1.0999
V13 0.95 1.1 1.05 1.1000 0.9876 1.1000 1.1000 1.1000 1.1000
T11(6 –9) 0.9 1.1 1.078 1.0447 1.0425 1.0155 0.9528 0.9810 1.0065
T12(6 –10) 0.9 1.1 1.069 0.9000 0.9000 0.9178 1.0999 0.9000 0.9073
T15(4 –12) 0.9 1.1 1.032 0.9863 0.9376 0.9820 1.0071 0.9557 0.9936
T36(28 –27) 0.9 1.1 1.068 0.9657 0.9702 0.9812 0.9819 0.9364 0.9485
QC10 0 5 0 0.0000 0.0000 4.9964 5.0000 0.0000 0.0000
QC12 0 5 0 0.0000 0.0000 4.9988 5.0000 0.0000 0.0000
QC15 0 5 0 0.0000 0.0000 5.0000 5.0000 0.0000 0.0000
QC17 0 5 0 0.0000 0.0000 5.0000 5.0000 0.0000 0.0000
QC20 0 5 0 0.0000 0.0000 5.0000 5.0000 0.0000 0.0000
QC21 0 5 0 0.0000 0.0000 5.0000 4.9999 0.0000 0.0000
QC23 0 5 0 0.0000 0.0000 4.9734 5.0000 0.0000 0.0000
QC24 0 5 0 0.0000 0.0000 5.0000 5.0000 0.0000 0.0000
QC29 0 5 0 0.0000 0.0000 4.9978 5.0000 0.0000 0.0000

Cost ($/h) – – 901.9516 799.0715 803.7871 799.9780 804.7369 647.9202 923.4147

Power loss (MW) – – 5.8219 8.6260 9.8641 8.76393 9.8231 7.1064 10.9599
Power loss (MVar) −4.6066 4.4629 11.6938 3.20629 11.4489 9.6327 22.6157
Voltage deviations – – 1.1496 1.8925 0.0945 2.14708 2.1632 1.4173 1.0282
Lmax – – 0.1723 0.1159 0.1369 0.11311 0.1150 0.1211 0.1253

from the unity 1.0. Thus, the objective function can be expressed as 
NL
follows: JVoltage deviations = |Vi − 1.0| (21)
i=1
J = Jcost + wJVoltage deviations (19)
and w is a weighting factor which has to be selected carefully. By
this section, a weight (an importance) is given to each one of the
where two terms of the objective function. In this study w is chosen as
100.

NG
The TLBO technique has been applied to search for the optimal
Jcost = fi (20) solution of Case 2. The variations of total fuel cost and voltage devia-
i=1 tions over the iterations are sketched in Fig. 4. These optimal results

20
Power i-->k
15 Power k-->i

10

5
Power (Mvar)

-5

-10

-15

-20
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41

5
Losses
Losses (Mvar)

-5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41

Fig. 3. Reactive power flow and losses through transmission lines for Case 1.
54 H.R.E.H. Bouchekara et al. / Electric Power Systems Research 114 (2014) 49–59

Table 2
Power flow through transmission lines for Case 1.

Line Line flow Line Line flow Line losses

From To MW Mvar From To MW Mvar MW Mvar

1 2 118.1446 −15.7977 2 1 −115.9046 16.1876 2.2401 0.3899

1 3 58.9123 −1.3724 3 1 −57.6153 1.8452 1.2969 0.4728
2 4 34.116 −3.0655 4 2 −33.555 0.4798 0.561 −2.5856
2 5 63.4673 0.1879 5 2 −61.8579 1.7447 1.6094 1.9326
2 6 45.3184 −3.837 6 2 −44.3088 2.5511 1.0096 −1.2859
3 4 55.2153 −3.0452 4 3 −54.8686 3.0689 0.3467 0.0237
4 6 49.7658 −4.1591 6 4 −49.5083 4.0228 0.2574 −0.1363
4 12 30.6313 −7.2253 12 4 −30.6313 9.4285 0 2.2032
5 7 −11.0378 6.0844 7 5 11.1089 −8.198 0.0711 −2.1135
6 7 34.1836 1.6218 7 6 −33.9089 −2.702 0.2747 −1.0802
6 8 11.3603 −5.088 8 6 −11.3445 4.1144 0.0158 −0.9736
6 9 18.6798 −2.2759 9 6 −18.6798 2.9206 0 0.6447
6 10 12.9337 −3.2421 10 6 −12.9337 4.1073 0 0.8652
6 28 16.0227 −0.213 28 6 −15.9847 −1.1336 0.038 −1.3466
8 28 2.425 −1.4359 28 8 −2.4211 −3.4315 0.0038 −4.8674
9 11 −11.8844 −13.3475 11 9 11.8844 13.9235 0 0.576
9 10 29.7656 −13.1558 10 9 −29.7656 14.166 0 1.0101
10 20 9.3463 −0.2655 20 10 −9.2771 0.4199 0.0692 0.1544
10 17 5.7452 1.9148 17 10 −5.7351 −1.8886 0.01 0.0262
10 21 15.7107 4.4805 21 10 −15.6322 −4.3115 0.0785 0.1689
10 22 7.5314 1.7869 22 10 −7.4945 −1.711 0.0368 0.0759
12 13 −12 −5.7784 13 12 12 5.9865 0 0.2081
12 14 7.4429 0.4532 14 12 −7.3856 −0.334 0.0573 0.1192
12 15 17.6056 −1.0625 15 12 −17.4331 1.4024 0.1725 0.3399
12 16 6.8101 0.8088 16 12 −6.7728 −0.7305 0.0372 0.0783
14 15 1.1211 −1.3069 15 14 −1.1155 1.3119 0.0056 0.005
15 18 5.6764 0.5257 18 15 −5.6468 −0.4652 0.0296 0.0605
15 23 4.7414 −0.6974 23 15 −4.7218 0.7369 0.0196 0.0395
16 17 3.2731 −1.0693 17 16 −3.2648 1.0886 0.0083 0.0194
18 19 2.4463 −0.4349 19 18 −2.4429 0.4418 0.0034 0.0069
19 20 −7.0575 −3.8419 20 19 7.0765 3.8799 0.019 0.0379
21 22 −1.8682 −1.8886 22 21 1.8689 1.89 0.0007 0.0014
22 24 5.6237 −0.181 24 22 −5.5925 0.2295 0.0312 0.0486
23 24 1.5212 1.4817 24 23 −1.5161 −1.4713 0.0051 0.0104
24 25 −1.589 −0.4559 25 24 1.5935 0.4637 0.0045 0.0078
25 26 3.5401 2.3564 26 25 −3.5006 −2.2973 0.0395 0.059
25 27 −5.1334 −2.8229 27 25 5.1656 2.8844 0.0322 0.0615
27 28 −17.7746 7.009 28 27 17.7746 −5.7913 0 1.2177
27 29 6.185 −0.3413 29 27 −6.1139 0.4755 0.071 0.1342
27 30 7.0548 0.8528 30 27 −6.9186 −0.5965 0.1362 0.2564
29 30 3.7139 1.3646 30 29 −3.6816 −1.3036 0.0323 0.061

are given in Table 1. The total generation fuel cost and voltage one of Case 1 as shown in Fig. 5. We can notice from this figure that
deviations are 803.7871 $/h and 0.0945 p.u. for this case compared the voltage profile has been greatly improved.
to 799.0715 $/h and 1.8925 p.u. for Case 1. Although the cost has
been increased by 0.59%, the voltage profile has been improved by 4.1.3. Case 3: voltage stability enhancement
95.01%. The system voltage profile in this case is compared to the Transmission systems are forced to operate very close to their
security limits, due to economical and environmental reasons.
One important characteristic of the power system is its ability to

840 1
Cost
1.1 Case 1
Voltage deviations
Case 2
1 p.u.
Voltage deviations (p.u.)
Voltage Profile (p.u.)

Voltage Profile (p.u.)

1.05

820 0.5

800 0 0.95
50 100 150 200 250 300 350 400 450 500 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
PQ Bus Number PQ Bus Number

Fig. 4. Variations of fuel cost and voltage deviations for Case 2. Fig. 5. System voltage profile improvement.
 H.R.E.H. Bouchekara et al. / Electric Power Systems Research 114 (2014) 49–59 55

conserve constantly acceptable bus voltage at each node under 920 0.145
Cost
normal operating conditions, after load increase, following system L
900 max 0.14
configuration changes or when the system is being subjected to a
disturbance. The non-optimized control variables may lead to pro-
880 0.135
gressive and uncontrollable drop in voltage resulting in an eventual

Voltage Profile (p.u.)

wide-spread voltage collapse [14]. As a result, voltage stability has
860 0.13
become an issue of paramount importance to electric utility and it

L max
is attracting more and more attention. Out of the different meth- 840 0.125
ods to assess the voltage stability margin, Kessel and Glavitch [15]
have developed a voltage stability index (L-index) based on the fea- 820 0.12
sibility of power flow equations for each node. The L-index of a bus
indicates the proximity of voltage collapse condition of that bus. 800 0.115

It varies between 0 and 1 corresponding to no load and voltage

collapse, respectively. 780
50 100 150 200 250 300 350 400 450
0.11
500
For a power system with NB, NG and NL buses representing the PQ Bus Number
total number of buses, the total number of PV buses and the total
number of PQ buses, respectively, we can separate buses into two Fig. 6. Variations of fuel cost and stability index for Case 3.
parts: PV buses at the head and PQ buses at the tail as follows:
      
IL VL YLL YLG VL After applying the TLBO technique, it appears from Table 1 that
= [Ybus ] = (22) the value of Lmax is significantly reduced in this case compared to
IG VG YGL YGG VG Case 1 and Case 2. Consequently, distance from collapse point is
increased. Fig. 6 shows the variations of fuel cost and the stability
where YLL , YLG , YGL and YGG are sub matrix of Ybus .
index during the optimization process.
The following hybrid system of equations can be written:
      
VL IL HLL HLG IL 4.1.4. Case 4: voltage stability enhancement during contingency
= [H] = (23)
IG VG HGL HGG VG condition
A new case study where the voltage stability is enhanced during
where H matrix is generated by the partial inversion of Ybus , HLL , a contingency condition has been proposed in [12]. The contingency
HLG , HGL and HGG are sub matrix of H, VG , IG , VL and IL are voltage state is simulated as outage of lines (2–6). Since the voltage stability
and current vector of PV buses and PQ buses, respectively. is a major concern, the objective function for this case is chosen as
The matrix H is given by: in Case 3.
  The proposed TLBO technique has been run to search for the
ZLL −ZLL YLG optimal solutions for this case. These solutions are displayed in
−1
[H] = ZLL = YLL (24)
YGL ZLL YGG − YGL ZLL YLG Table 1. It is evident that the proposed TLBO is very effective in
optimizing the system and enhancing the voltage stability in the
Therefore, the L-index denoted by Lj of bus j is represented as contingency. Furthermore, the comparison of the results obtained
follows: for this case using TLBO with those presented in [12] reveals that
 
  NG  the Lmax is reduced from 0.1347 to 0.1150 and the cost is reduced
 Vi  from 810.2661 $/h to 804.7369 $/h.
Lj = 1 − HLGji  j = 1, 2, . . ., NL (25)
 Vj 
i=1
4.1.5. Case 5: piecewise quadratic fuel cost curve
For stable situations the condition Lj ≤ l must not be violated for
In practice, many thermal generating units may be supplied with
any of the buses j.
multiple fuel sources like coil, natural gas and oil. Therefore, the
Hence, a global indicator Lmax describing the stability of the
fuel cost functions (curves) can be divided as piecewise quadratic
complete subsystem is given by [15]:
cost functions for various fuel types [12]. The problem in this case
Lmax = max(Lj ) j = 1, 2, . . ., NL (26) becomes a non-convex optimization problem with discontinuous
values at each boundary. It is obvious that this problem has many
The lower Lmax , the more the system is stable.
local optimal solutions and therefore, traditional solution meth-
Enhancing voltage stability can be achieved through minimizing
ods -if they are used- are susceptible to be trapped on these local
the voltage stability indicator L-index values at every bus of the
optimal solutions [12]. In this case, the fuel cost curve is expressed
system and consequently the global power system Lmax [12].
by a piecewise quadratic function therefore, fi can be expressed as
In order to minimize total the fuel cost and enhance the voltage
follows:
stability in order to move the system far from the voltage collapse
point, the objective function chosen in this case is a linear combi- fi = aik + bik Pi + cik Pi2 min
if Pik max
≤ Pi ≤ Pik (30)
nation of total fuel cost and Lmax . Hence, the objective function can
be expressed as follows: where k is the fuel option.
In this study, generators of buses 1 and 2 have two fuel options

NG
i.e. k = 2. The fuel cost coefficients of these generators are given in
JCost = fi (27)
[12] while, the fuel cost coefficients of remaining generators are
i=1 the same as in Case 1. The upper limit of voltage magnitude at bus
JVoltage stability enhancement = Lmax (28) 1 is 1.05 and no VAR compensation buses are considered for fair
comparison with the results reported in [12]. Thus the objective
J = JCost + wJVoltage stability enhancement (29)
function can be given by:
where w is a weighting factor set at 6000 in this study as in [12]
J = JCostGenerators 1 and 3 + JCostRemaining generators (31)
and [16].
56 H.R.E.H. Bouchekara et al. / Electric Power Systems Research 114 (2014) 49–59

Table 3
Comparison of the simulation results for Case 1.

Method Cost Method description References

TLBO 799.0715 Teaching-Learning-Based Optimization

GSA 798.6751 Gravitational Search Algorithm [18]
DSA 799.0943 Differential Search Algorithm [19]
BBO 799.1116 Biogeography-Based Optimization [16]
DE 799.2891 Differential Evolution [12]
SA 799.45 Simulated Annealing [20]
AGAPOP 799.8441 Adaptive Genetic Algorithm with Adjusting Population Size [21]
BHBO 799.9217 Black-Hole-Based Optimization [22]
EM 800.078 Electromagnetism-Like Mechanism [23]
EADHDE 800.1579 Genetic Evolving Ant Direction HDE [24]
EADDE 800.2041 Evolving Ant Direction Differential Evolution [25]
PSO 800.41 Particle Swarm Optimization [10]
FPSO 800.72 Fuzzy Particle swarm optimization [26]
IGA 800.805 Improved Genetic Algorithms [27]
PSO 800.96 Particle swarm optimization [26]
GAF 801.21 Fuzzy genetic algorithm [26]
ICA 801.843 Imperialist Competitive Algorithm [28]
EGA 802.06 Enhanced Genetic Algorithm [29]
TS 802.2900 Tabu search [30]
MDE 802.376 Modified Differential Evolution Algorithm [31]
IEP 802.465 Improved Evolutionary Programming [32]
EP 802.62 Evolutionary Programming [17]
RGA 804.02 Refined Genetic Algorithm [33]
GM 804.853 Gradient Method [11]
GA 805.94 Genetic algorithm [33]

where 4.1.6. Case 6: fuel cost minimization of generators with

valve-point loading

2 In practice, the fuel cost function of generating units has non-
JCostGenerators 1 and 2 = aik + bik Pi + cik Pi2 if min
Pik max
≤ Pi ≤ Pik differentiable points due to the valve-loading point effect and
i=1 change of fuels. In order to take into account the valve-loading
effects of units, a recurring rectifying sinusoidal term is added to
for fuel optionk; k = 1 and 2 (32)
the basic quadratic cost function, as follows [3]:

and fi = ai = bi Pi + ci Pi2 + |di × sin (ei × (Pimin − Pi ))| (34)

where di and ei are the coefficients that represent the valve-point

NG loading effects.
JCostRemaining generators = ai + bi Pi + ci Pi2 (33) In this case, generators of buses 1 and 2 are considered to have
i=3 the valve-point loading effects on their characteristics. Therefore,
the objective function can be represented by:
The optimal settings of control variables found using the J = JCostGenerators 1 and 2 + JcostRemaining generators (35)
proposed TLBO technique are given in Table 1.The fuel cost mini-
mization study has resulted in 647.9202 $/h. The variation of the where
fuel cost i.e. the objective function over iterations in this case is 
2
depicted in Fig. 7. JCostGenerators 1 and 2 = ai + bi Pi + ci Pi2 + |di × sin (ei × (Pimin − Pi ))|
i=2
(36)

760 and

NG

740 JCostRemaining generators = ai + bi Pi + ci Pi2 (37)

i=3

720 Fuel cost coefficients of generators 1 and 2 are given in [17] and
those of the remaining generators have the same values as in Case
Cost ($/hr)

1.
700
The obtained optimal settings of control variables when the
proposed technique is run, are given in Table 1 and the fuel cost
680 variation during the optimization process is sketched in Fig. 8.

4.1.7. Comparative study

660
Under the same conditions i.e. control variables limits, con-
straints and system data, the results obtained using the TLBO
640
0 50 100 150 200 250 300 350 400 450 500 technique reported in this paper are compared to some other
Iterations techniques reported in the literature as shown in Tables 3–7,
respectively. It appears from these tables that, the proposed tech-
Fig. 7. Fuel cost variation for Case 5. nique outperforms many techniques used to solve different OPF
 H.R.E.H. Bouchekara et al. / Electric Power Systems Research 114 (2014) 49–59 57

1040 Table 7
Comparison of the simulation results for Case 6.

1020 Methods Cost

TLBO 923.4147
DSA 923.4573
1000
BBO 919.7647
Cost ($/hr)

EP 919.8900
FPSO 923.5400
980
PSO 923.7200
GSA 929.7240
960 MDE 930.7930
IEP 953.5730

940
problems because the results obtained using TLBO are either bet-
ter or comparable to those obtained using other techniques. This
920
0 50 100 150 200 250 300 350 400 450 500 highlights its ability to find better quality solution.
Iterations Moreover, the proposed technique has been compared with
three mathematical non-linear programming methods for the solu-
Fig. 8. Fuel cost variation for Case 6.
tion of OPF [34]. Specifically, the Interior Point (IP), the Sequential
Quadratic Programming (SQP) and the Active Set (AS) methods
Table 4 have been applied to Case 1 through Case 6. The results of the pro-
Comparison of the simulation results for Case 2 where VD stands for voltage posed TLBO are compared with those of IP, SQP and AS techniques
deviation. as given in Table 8. The superiority of the proposed TLBO method
Methods Cost VD Objective function is clearly shown as it gives better solutions. Furthermore, in Case
5 the three mathematical non-linear programming methods have
TLBO 803.7871 0.0945 813.2355
DSA 803.8274 0.0977 813.5961 converged to local optima i.e. they are trapped on local optima.
GSA 804.3148 0.0933 813.6417
BBO 804.9982 0.1020 815.1982 4.1.8. Robustness study
PSO 806.3800 0.0891 815.2900
In order assess the robustness and the effectiveness of the pro-
EM 804.2600 0.1270 816.9600
BHBO 804.5975 0.1262 817.2140 posed technique a statistical study has been carried out. In this
DE 805.2619 0.1357 818.8319 study the TLBO has been run 50 times with different initial solu-
tions for three selected cases namely: Case 1, Case 5 and Case 6. The
best, the mean, the median, the worst and the standard deviation
Table 5 (noted as SD) obtained are displayed in Table 9. We can notice that
Comparison of the simulation results for Case 3. the best, the mean, the median and the worst values of objective
Method Cost Lmax Objective function functions after 50 trials are very close. This is also shown by the
low values of the standard deviations calculated. This study reveals
TLBO 799.9780 0.1131 1478.6611
BBO 805.7252 0.1104 1468.1252
the effectiveness of the proposed technique and its ability to reach
DSA 799.8895 0.1140 1484.1108 either to optimum value or very near to it in every trial. It is worth
GSA 806.6013 0.1162 1504.0833 mentioning that similar statistical analysis can be done for the other
BHBO 803.5193 0.1168 1504.4869 cases.
EM 805.0580 0.1167 1505.0420
DE 807.5272 0.1219 1538.9272
PSO 801.1600 0.1246 1548.7600 4.2. IEEE 118-bus test system

To demonstrate the scalability and suitability of the proposed

Table 6 TLBO technique for large-scale power systems, it has been applied
Comparison of the simulation results for Case 5. to the IEEE 118-bus test system. The system has 54 generators, 186
branches, 9 transformers, 2 reactors, and 12 capacitors. More details
Method Cost
about this system can be found in [35].
TLBO 647.9202
GSA 646.8481
PSO 647.6900 4.2.1. Case 7: minimization of generation fuel cost
BBO 647.7437 The objective of this case is as in Case 1 that is, the minimization
EP 647.7900 of the total fuel cost of all generator units. The execution time of
MDE 647.8460
one iteration for this test system is 3.77 s. The obtained results using
DSA 647.9215
IEP 649.3120
the proposed TLBO technique for Case 7 are tabulated in Table 10
DE 650.8224 and compared with those of GA, PSO and DSA. It is obvious that
EM 650.9371 the proposed technique outperforms the abovementioned methods
and gives better results. This confirms the potential of the proposed

Table 8
Comparison of the obtained results with those obtained using mathematical programming.

Method Case 1 Case 2 Case 3 Case 4 Case 5 Case 6

TLBO 799.0715 Cost = 803.7871VD = 0.0945 Cost = 799.9780Lmax = 0.1131 Cost = 804.7369Lmax = 0.1150 647.9202 923.4147
IP 799.0718 Cost = 803.8182VD = 0.1025 800.0225 Lmax = 0.1131 Cost = 804.7376Lmax = 0.1150 777.2738 923.4231
SQP 799.0718 Cost = 803.8106VD = 0.0942 800.0225 Lmax = 0.1131 Cost = 804.7377Lmax = 0.1150 737.6695 923.4160
AS 799.0718 Cost = 803.8118VD = 0.0942 800.0224 Lmax = 0.1131 Cost = 804.7376Lmax = 0.1150 736.3832 923.4160
58 H.R.E.H. Bouchekara et al. / Electric Power Systems Research 114 (2014) 49–59

Table 9 Table 10 (Continued)

Efficiency of the proposed technique after 50 trials for Case 1, Case 5 and Case 6.
GA PSO DSA TLBO
Objective function
V27 0.9596 1.0286 1.0183 1.0106
Best Mean Median Worst SD V31 0.9558 1.0132 1.0065 1.0019
V32 0.9954 1.0221 1.0151 1.0077
Case 1 799.071 799.072 799.072 799.072 0.000105 V34 1.0258 1.0289 1.0090 1.0121
Case 5 647.920 648.018 647.979 648.294 0.099571 V36 1.0103 1.0266 1.0073 1.0105
Case 6 923.415 923.419 923.415 923.561 0.020533 V40 1.0115 1.0201 0.9985 1.0014
V42 1.0256 1.0196 1.0005 1.0036
Table 10 V46 1.0244 1.0164 1.0167 1.0219
Optimal settings of control variables for the IEEE 118-bus test system. V49 1.0004 1.0357 1.0296 1.0347
V54 0.9943 1.0118 0.9982 1.0075
GA PSO DSA TLBO
V55 0.9993 1.0107 0.9972 1.0067
P1 1.3675 22.4593 26.8362 27.4218 V56 0.9882 1.0123 0.9980 1.0070
P4 35.5585 0.0953 0.0269 0.0076 V59 0.9857 1.0243 1.0176 1.0205
P6 4.8962 0.0605 0.1713 0.0748 V61 1.0105 1.0330 1.0263 1.0192
P8 4.6750 0.0004 0.0321 0.0006 V62 0.9896 1.0282 1.0242 1.0190
P10 353.3066 401.1231 398.4004 399.4280 V65 0.9935 1.0592 1.0600 1.0591
P12 95.7364 86.4173 85.3337 85.6184 V66 0.9927 1.0491 1.0441 1.0461
P15 45.1681 23.6916 21.4802 18.2351 V69 1.0241 1.0444 1.0574 1.0600
P18 56.0939 14.1255 12.9615 11.1616 V70 1.0072 1.0107 1.0271 1.0325
P19 13.3277 18.7619 20.8774 23.0465 V72 0.9690 1.0224 1.0240 1.0218
P24 8.1091 0.6752 0.0267 0.0125 V73 0.9984 1.0174 1.0240 1.0288
P25 158.7755 193.5494 192.9606 193.6707 V74 1.0253 1.0039 1.0196 1.0256
P26 234.7256 283.2726 278.7981 279.7224 V76 0.9991 0.9980 1.0167 1.0255
P27 26.8677 5.2022 13.0039 15.8441 V77 1.0096 1.0171 1.0353 1.0446
P31 15.6471 7.1953 7.3372 7.2543 V80 1.0224 1.0324 1.0469 1.0570
P32 12.6737 16.7919 16.1215 13.4763 V85 1.0214 1.0222 1.0472 1.0546
P34 3.6202 1.2614 2.4835 2.5040 V87 1.0349 0.9995 1.0478 1.0523
P36 13.7331 12.1332 9.0804 8.8714 V89 0.9908 1.0453 1.0599 1.0600
P40 16.5334 53.9110 49.7472 49.9442 V90 0.9998 1.0197 1.0370 1.0400
P42 23.2765 39.7112 42.4044 42.0478 V91 0.9866 1.0214 1.0389 1.0429
P46 0.6915 18.4851 19.1328 19.1282 V92 0.9508 1.0299 1.0470 1.0523
P49 147.3233 192.6422 193.0716 193.6081 V99 1.0266 1.0091 1.0328 1.0435
P54 45.1897 50.0018 49.6000 50.1543 V100 0.9745 1.0207 1.0347 1.0476
P55 69.0593 29.8026 32.2346 31.6590 V103 0.9566 1.0091 1.0210 1.0371
P56 65.8954 30.3357 34.8159 34.7532 V104 0.9963 0.9933 1.0104 1.0261
P59 152.6823 149.4320 148.8600 149.6018 V105 0.9797 0.9895 1.0065 1.0229
P61 168.1241 147.6822 147.6539 147.8376 V107 0.9421 0.9767 0.9975 1.0166
P62 48.9650 0.0000 0.0208 0.0022 V110 0.9900 1.0073 1.0015 1.0185
P65 367.2027 353.3897 351.7151 353.2842 V111 1.0329 1.0178 1.0074 1.0255
P66 302.8345 349.7794 348.3943 349.1853 V112 1.0494 0.9951 0.9934 1.0101
P69 412.8327 0.0002 453.2653 454.9128 V113 1.0482 1.0315 1.0126 1.0149
P70 53.7545 456.0935 0.3662 0.9058 V116 1.0081 1.0546 1.0580 1.0522
P72 28.3957 0.0000 0.1077 0.0237 T8(8–5) 0.9865 0.9735 1.0030 0.9904
P73 3.6347 0.0504 0.1476 0.2558 T32(26–25) 1.0489 1.0531 1.0969 1.0971
P74 14.3311 20.0946 17.9047 16.2229 T36(30–17) 1.0755 1.0082 1.0182 1.0188
P76 34.9359 21.8083 24.4252 22.2712 T51(38–37) 1.0203 0.9993 1.0165 1.0091
P77 37.9161 0.0606 0.0235 0.0002 T93(63–59) 0.9812 1.0272 1.0240 0.9992
P80 285.7581 436.0302 429.5080 429.6968 T95(64–61) 1.0324 1.0230 1.0357 1.0549
P85 5.1373 0.0078 0.0145 0.0255 T102(65–66) 0.9818 1.0655 0.9058 0.9000
P87 2.2719 3.6376 3.6781 3.6354 T107(68–69) 1.0511 0.9609 0.9661 0.9526
P89 448.5638 503.4463 503.3109 504.4182 T127(81–80) 0.9861 1.0190 1.0001 0.9847
P90 49.7387 0.0069 0.0115 0.0021
Cost ($/h) 132,746.3517 129,756.2275 129,691.6152 129,682.844
P91 26.1496 0.0022 0.0250 0.0000
P92 11.1297 0.0025 0.0127 0.0003
P99 4.4770 0.0240 0.0283 0.0112
P100 193.7171 234.9661 229.8191 231.1064
P103 35.5692 37.8131 37.7738 38.0594 TLBO technique and its effectiveness to solve the OPF problem for
P104 12.2408 0.0012 0.2178 0.0174 large-scale power systems.
P105 11.4288 1.7058 7.0826 4.5931
P107 16.2340 32.6088 28.8557 30.4984
P110 43.4340 36.4924 8.3921 0.4828
5. Conclusion
P111 17.5178 35.6210 35.0042 35.4211
P112 48.5944 0.0003 35.9265 39.8141
P113 19.4258 0.0082 0.1107 0.0152 In this paper, we have introduced and presented a novel tech-
P116 22.6387 0.0844 0.0147 0.0036 nique to solve the OPF problem in power systems that is TLBO. It is
V1 0.9707 1.0212 0.9998 1.0106
based on passing knowledge from teacher to learners and between
V4 1.0214 1.0365 1.0165 1.0269
V6 0.9757 1.0291 1.0077 1.0180
learners themselves. The outcome of this research helps not only to
V8 1.0269 1.0365 1.0375 1.0340 determine the optimal generation cost in power systems, but also
V10 1.0138 1.0590 1.0570 1.0600 helps consider some practical constraints like piecewise quadratic
V12 0.9706 1.0265 1.0051 1.0133 fuel cost curves and valve point effect. TLBO has been success-
V15 1.0120 1.0224 1.0025 1.0063
fully and effectively implemented and applied to the IEEE 30-bus
V18 1.0085 1.0230 1.0044 1.0033
V19 1.0090 1.0214 1.0011 1.0039 and the IEEE 118-bus test systems where objective functions have
V24 0.9842 1.0319 1.0308 1.0221 been considered for the minimization of generation fuel cost, volt-
V25 0.9709 1.0560 1.0589 1.0385 age profile improvement, voltage stability enhancement, voltage
V26 1.0276 1.0600 1.0600 1.0600 stability enhancement during contingency condition, piecewise
 H.R.E.H. Bouchekara et al. / Electric Power Systems Research 114 (2014) 49–59
quadratic fuel cost curve and fuel cost minimization of generators
with valve-point loading.
This paper is significant for the following reasons among oth-
ers: 1) the high ranking of the proposed technique among other
techniques which has been proven when compared with other
techniques, 2) the efficiency of the TLBO which has been proven by
carrying out a statistical analysis that has revealed its strength by
converging every time the optimum value or very near to it 3) unlike
other optimization techniques TLBO has no internal parameter to
tune.
Acknowledgement
Dr. M.A. Abido would like to acknowledge the support of Dean-
ship of Scientific Research, King Fahd University of Petroleum
and Minerals, through the Electrical Power and Energy Systems
Research Group.
References
[1] R.B. Squires, Economic dispatch of generation directly from power sys-
tem voltages and admittances, IEEE Trans. Power App. Syst. 79 (3) (1961)
1235–1245.
[2] J. Carpentier, Contribution à l’étude du Dispatching Economique, Bull. Soc. Fran-
caise Electriciens (1962) 431–447.
[3] T. Niknam, M.R. Narimani, M. Jabbari, A.R. Malekpour, A modified shuffle frog
leaping algorithm for multi-objective optimal power flow, Energy 36 (11)
(2011) 6420–6432.
[4] M. Huneault, F.D. Galiana, A survey of the optimal power flow literature, IEEE
Trans. Power Syst. 6 (2) (1991) 762–770.
[5] S. Frank, I. Steponavice, S. Rebennack, Optimal power flow: a bibliographic
survey I, formulations and deterministic methods, Energy Syst. 3 (3) (2012)
221–258.
[6] M.R. AlRashidi, M.E. El-Hawary, Applications of computational intelligence
techniques for solving the revived optimal power flow problem, Electr. Power
Syst. Res. 79 (4) (2009) 694–702.
[7] S. Frank, I. Steponavice, S. Rebennack, Optimal power flow: a bibliographic
survey II, non-deterministic and hybrid methods, Energy Syst. 3 (3) (2012)
259–289.
[8] R.V. Rao, V.J. Savsani, D.P. Vakharia, Teaching-learning-based optimization: a
novel method for constrained mechanical design optimization problems, Com-
put. Aided Des. 43 (3) (2011) 303–315.
[9] M. Črepinšek, S.H. Liu, L. Mernik, A note on teaching-learning-based optimiza-
tion algorithm, Inform. Sci. 212 (2012) 79–93.
[10] M.A. Abido, Optimal power flow using particle swarm optimization, Int. J. Electr.
Power Energy Syst. 24 (7) (2002) 563–571.
[11] K. Lee, Y. Park, J. Ortiz, A united approach to optimal real and reactive power
dispatch, IEEE Trans. Power App. Syst. 104 (5) (1985) 1147–1153.
[12] A.A. Abou El Ela, M.A. Abido, Optimal power flow using differential evolution
algorithm, Electr. Power Syst. Res. 80 (7) (2010) 878–885.
59
[13] O. Alsac, B. Stot, Optimal load flow with steady-state security, IEEE Trans. Power
App. Syst. PAS-93 (3) (1974) 745–751.
[14] C.A. Belhadj, M.A. Abido, An optimized fast voltage stability indicator, in: Elec-
tric Power International Conference on Engineering, PowerTech, Budapest,
1999, pp. 79–83.
[15] P. Kessel, H. Glavitsch, Estimating the voltage stability of a power system, IEEE
Trans. Power Deliv. 1 (3) (1986) 346–354.
[16] A. Bhattacharya, P.K. Chattopadhyay, Application of biogeography-based opti-
misation to solve different optimal power flow problems, IET Gener. Transm.
Distrib. 5 (1) (2011) 70–80.
[17] J. Yuryevich, K.P. Wong, Evolutionary programming based optimal power flow
algorithm, IEEE Trans. Power Syst. 14 (4) (1999) 1245–1250.
[18] S. Duman, U. Güvenç, Y. Sönmez, N. Yörükeren, Optimal power flow using
gravitational search algorithm, Energy Convers. Manag. 59 (2012) 86–95.
[19] H.R.E.H. Bouchekara, M.A. Abido, Optimal power flow using differential search
algorithm, Electric Power Compon. Syst. (2013) (under review).
[20] C.A. Roa-Sepulveda, B.J. Pavez-Lazo, A solution to the optimal power flow using
simulated annealing, Int. J. Electr. Power Energy Syst. 25 (1) (2003) 47–57.
[21] A.F. Attia, Y.A. Al-Turki, A.M. Abusorrah, Optimal power flow using adapted
genetic algorithm with adjusting population size, Electric Power Compon. Syst.
40 (11) (2012).
[22] H.R.E.H. Bouchekara, Optimal power flow using black-hole-based optimization
approach, Appl. Soft Comput. (2013) (under review).
[23] H.R.E.H. Bouchekara, M.A. Abido, Optimal power flow using electromagnetism-
like mechanism, Electr. Energy (2013) (under review).
[24] K. Vaisakh, L.R. Srinivas, Genetic evolving ant direction HDE for OPF with non-
smooth cost functions and statistical analysis, Expert Syst. Appl. 38 (2011)
2046–2062.
[25] K. Vaisakh, L.R. Srinivas, Evolving ant direction differential evolution for OPF
with non-smooth cost functions, Eng. Appl. Artif. Intell. 24 (2011) 426–436.
[26] R.H. Liang, S.R. Tsai, Y.T. Chen, T. Wan-Tsun, Optimal power flow by a fuzzy
based hybrid particle swarm optimization approach, Electr. Power Syst. Res.
81 (7) (2011) 1466–1474.
[27] L.L. Lai, J.T. Ma, R. Yokoyama, M. Zhao, Improved genetic algorithms for optimal
power flow under both normal and contingent operation states, Int. J. Electr.
Power Energy Syst. 19 (5) (1997) 287–292.
[28] A.J. Ghanizadeh, G. Mokhtari1, M. Abedi, G.B. Gharehpetian, Optimal power
flow based on imperialist competitive algorithm, Int. Rev. Electr. Eng. 6 (June
(4)) (2011), Papers Part B, 2011.
[29] A.G. Bakirtzis, P.N. Biskas, C.E. Zoumas, V. Petridis, Optimal power flow by
enhanced genetic algorithm, IEEE Trans. Power Syst. 17 (2) (2002) 229–236.
[30] M.A. Abido, Optimal power flow using Tabu search algorithm, Electric Power
Compon. Syst. 30 (2002) 469–483.
[31] S. Sayah, K. Zehar, Modified differential evolution algorithm for optimal power
flow with non-smooth cost functions, Energy Convers. Manag. 49 (11) (2008)
3036–3042.
[32] W. Ongsakul, T. Tantimaporn, Optimal power flow by improved evolutionary
programming, Electric Power Compon. Syst. 34 (1) (2006) 79–95.
[33] S.R. Paranjothi, K. Anburaja, Optimal power flow using refined genetic algo-
rithm, Electric Power Compon. Syst. 30 (2002) 1055–1063.
[34] P.N. Biskas, N.P. Ziogos, A. Tellidou, C.E. Zoumas, A.G. Bakirtzis, V. Petridis,
Comparison of two metaheuristics with mathematical programming methods
for the solution of OPF, Proc. Gener. Transm. Distrib. 153 (January (1)) (2006)
16–24.
[35] Matpower Matlab toolbox http://www.pserc.cornell.edu/matpower/
matpower.html