Professional Documents
Culture Documents
Optimal power flow with emission and non-smooth cost functions using
backtracking search optimization algorithm
A.E. Chaib a,⇑, H.R.E.H. Bouchekara a,b, R. Mehasni a, M.A. Abido c
a
Constantine Electrical Engineering Laboratory, LEC, Department of Electrical Engineering, University of Freres Mentouri Constantine, 25000 Constantine, Algeria
b
Laboratory of Electrical Engineering of Constantine, LGEC, Department of Electrical Engineering, University of Freres Mentouri Constantine, 25000 Constantine, Algeria
c
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: Economic operation of electric energy generating systems is one of the prevailing problems in energy
Received 11 December 2014 systems. In this paper, a new method called the Backtracking Search Optimization Algorithm (BSA) is
Received in revised form 14 January 2016 proposed for solving the optimal power flow problem. This method is tested for 16 different cases on
Accepted 4 February 2016
the IEEE 30-bus, IEEE 57-bus, and IEEE 118-bus test systems. In addition to the traditional generating fuel
cost, multi-fuels options, valve-point effect and other complexities have been considered. Furthermore,
different objectives such as voltage profile improvement, voltage stability enhancement and emission
Keywords:
reduction are considered. The obtained results are compared with those obtained using some
Backtracking search optimization
Energy efficiency
well-known optimization algorithms. This comparison highlights the effectiveness of the BSA method
Optimal power flow for solving different OPF problems with complicated and non-smooth objective functions.
Power system optimization Ó 2016 Elsevier Ltd. All rights reserved.
Environmental pollution
Valve-point effect
http://dx.doi.org/10.1016/j.ijepes.2016.02.004
0142-0615/Ó 2016 Elsevier Ltd. All rights reserved.
A.E. Chaib et al. / Electrical Power and Energy Systems 81 (2016) 64–77 65
Nomenclature
based on three basic and well-known operators that are selection, Problem formulation
mutation and crossover. Furthermore, unlike many other meta-
heuristics, the BSA algorithm has only one control parameter and The OPF is a power flow problem which gives the optimal set-
it is not very sensitive to the initial value of this parameter as tings of the control variables for a given settings of load by mini-
reported in [18]. Since it was introduced, the BSA has attracted mizing a predefined objective function such as the cost of active
many researches and it has been applied to various optimization power generation or transmission losses. OPF considers the operat-
problems. The following are some successful examples. In [19] a ing limits of the system and it can be formulated as a nonlinear
comparative analysis of BSA with other evolutionary algorithms constrained optimization problem as follows:
for global continuous optimization is given. In [20] the BSA has
been used for antenna array design. In [21] the BSA was employed Minimize J ðx; uÞ ð1Þ
for the design of robust Power System Stabilizers (PSSs) in
multimachine power systems. In [22] it has been used for the Subject to g ðx; uÞ ¼ 0 ð2Þ
allocation of multi-type distributed generators along distribution
networks. and hðx; uÞ 6 0 ð3Þ
Therefore, this paper seeks to apply to the OPF problem a new
evolutionary algorithm that has not received yet much attention where u is the vector of independent variables or control variables,
in the power systems community that is the BSA. Furthermore, x is the vector of dependent variables or state variables, Jðx; uÞ is the
in this paper, not only the basic OPF problem is investigated, but objective function, g ðx; uÞ is the set of equality constraints and
also some complex formulations with non-smooth cost functions hðx; uÞ is the set of inequality constraints.
and different objective functions are considered along with the The control variables u and the state variables x of the OPF
environmental concern imposing the reduction of emission. problem are defined as follows.
The remainder of the paper is organized as follows. First, the
OPF formulation is presented in brief in section 2. Then, the main
Control variables
features of the BSA algorithm are presented in section 3. Next,
the results after solving different cases of OPF problem using BSA
These are the set of variables which can be adjusted to satisfy
are discussed in section 4. Finally, conclusions are drawn in the last
the load flow equations [9]. The set of control variables in the
section of this paper.
OPF problem formulation are:
66 A.E. Chaib et al. / Electrical Power and Energy Systems 81 (2016) 64–77
V min max
Gi 6 V Gi 6 V Gi ; i ¼ 1; . . . ; NG ð8Þ
PG active power generation at the PV buses except at the
slack bus.
VG voltage magnitude at PV buses. P min max
Gi 6 P Gi 6 P Gi ; i ¼ 1; . . . ; NG ð9Þ
T tap settings of transformer.
QC shunt VAR compensation. Q min max
Gi 6 Q Gi 6 Q Gi ; i ¼ 1; . . . ; NG ð10Þ
where NG, NT and NC are the number of generators, the number of T min
i 6 T i 6 T max
i ; i ¼ 1; . . . ; NT ð11Þ
regulating transformers and the number of VAR compensators,
(c) Shunt VAR compensator constraints
respectively.
Shunt VAR compensators must be bounded by their lower
and upper limits as follows:
State variables
Q min max
Ci 6 Q Ci 6 Q Ci ; i ¼ 1; . . . ; NC ð12Þ
These are the set of variables which describe any unique state of
the system [9]. The set of state variables for the OPF problem for- (d) Security constraints
mulation are: These constraints can be mathematically formulated as
follows:
P G1 active power at slack bus. V min 6 V Li 6 V max
Li Li ; i ¼ 1; . . . ; NL ð13Þ
VL voltage magnitude at PQ buses (load buses).
QG reactive power generation of all generator units.
Sli 6 Smax
li ; i ¼ 1; . . . ; nl ð14Þ
Sl transmission line loading (or line flow).
X
NB
where kP ; kV ; kQ and kS are penalty factors and xlim is the violated
PGi PDi V i V j Gij cos hij þ Bij sin hij ¼ 0 ð6Þ limit value of the dependent variable x and can be defined as
j¼i
follows:
(b) Reactive power constraints:
xmax ; x > xmax
xlim ¼ ð16Þ
X
NB
xmin ; x < xmin
Q Gi Q Di V i V j Gij sin hij Bij cos hij ¼ 0 ð7Þ
j¼i
Objective function
where hij ¼ hi hj , NB is the number of buses, P D is the active load
demand, Q D is the reactive load demand, Gij and Bij are the ele- Several objective functions are considered in this paper which
ments of the admittance matrix Y ij ¼ Gij þ jBij representing the are: cost reduction (with and without multi-fuels options or
conductance and susceptance between bus i and bus j, valve-point effect), voltage profile improvement, voltage stability
respectively. enhancement and emission reduction. Moreover, some combina-
tions between these objective functions are considered also.
BSA is achieved following five steps namely: initialization, Algorithm 2. Pseudocode of BSA [18].
selection-I, mutation, crossover, and selection-II. The following
subsections describe in details these steps. 1 function bsa (ObjFun, N, D, maxcycle, low, up)
Input: ObjFun, N, D, maxcycle, mixrate, low1:D, up1: D
Algorithm 1. General structure of BSA [18]. Output: globalminimum, globalminimizer
// INITIALIZATION
1. Initialization 2 globalminimum = inf
repeat 3 for i from 1 to N do
2. Selection-I 4 for j from 1 to D do
Generation of Trial-Population 5 Pi;j ¼ rnd:ðupj — lowj Þ þ lowj // Initialization of
3. Mutation population, P.
4. Crossover 6 oldPi;j ¼ rnd:ðupj — lowj Þ þ lowj // Initialization of oldP.
end 7 end
5. Selection-II 8 fitness Pi = ObjFun ðPi Þ // Initial-fitness values of P.
until stopping conditions are met 9 end
10 for iteration from 1 to maxcycle do
// SELECTION-I
11 if (a < bja, b U(0,1)) then oldP:= P end
Initialization
12 oldP: = permuting (oldP)
13 Generation of Trial-Population
As many other metaheuristics, in the initialization step the BSA
// MUTATION
generates a random initial population P of N individuals in the
14 mutant = P + F. (oldP — P)
search space that has D dimensions as shown in lines (3–9) of
// CROSSOVER
Algorithm 2.
15 Map1:N;l:D ¼ 1 // Initial-map is an N-by-D matrix of
ones.
Selection-I
16 if (c < djc, d U(0,1)) then
17 for i from 1 to N do
In this second step, a historical population called oldP is deter-
18
mined for calculation of the search direction. oldP is initialized like
mapi;uð1:dmixrate:rnd:De ¼ 0ju ¼ permutingðh1; 2; 3; . . . . . . ; DiÞ
the population P as shown in line (6) of Algorithm 2. However, oldP
can be redefined in every iteration using an ‘if–then’ rule as shown 19 end
in line (11) of Algorithm 2. After oldP is determined, the order of 20 else
individuals inside it is randomly permutated (line (12) of 21 for i from 1 to N do, mapi, rndi(D) = 0, end
Algorithm 2). 22 end
// Generation of Trial Population, T
Mutation 23 T: = mutant
24 for i from 1 to N do
In the mutation step, a mutant is generated as the initial form of 25 for j from i to D do
trial population (T) using the old population (oldP) and the popula- 26 if mapi, j = 1 then Ti,j:= Pij,
tion (P) itself as follows [18]: 27 end
28 end
Mutant ¼ P þ F ðoldP P Þ ð17Þ // Boundary Control Mechanism
where F is the scale factor used to control the search direction 29 for i from 1 to N do
matrix (oldP-P). It is worth to mention here that using the old pop- 30 for j from 1to D do
ulation in mutation allows BSA to benefit from the experience of 31 if (Ti, j <lowj) or (Ti, j > upj) then
previous generations [18]. The mutation step is shown in line 14 32 Ti, j= rnd. (upj —lowj) + lowj
in Algorithm 2. 33 end
34 end
Crossover 35 end
36 end
The crossover operator used in BSA is different from the one // SELECTION-II
used in DE or its variants. The crossover process represents the last 37 fitness T = ObjFn (T)
step while generating the trial population. This process starts by 38 for i from 1 to N do
generating a binary integer-valued matrix of size (N:D) denoted 39 if fitness Ti < fitness Pi then
mapn;m using two strategies as shown in Algorithm 2 (lines from 40 fitness Pi:= fitness Ti
(15) through (22)). This matrix indicates the individuals of T to Pi = Ti
41 end
be manipulated using the relevant individuals of P as shown in line
(26) of Algorithm 2 [18]. 42 end
43 fitness Pbest = min (fitness P) n best 2 {1, 2, 3. . . N}
Selection II 44 if fitness Pbest < globalminimum then
45 globalminimum: =fitness Pbest
In this step, the trial population T is used to update the popula- globalminimizer: = Pbest
// Export globalminimum and globalminimizer
tion P in a greedy manner as shown in Algorithm 2 (lines from (37)
through (42)). If two individuals have the same order, one from the 46 end
47 end
trial population T i and one from the population Pi , and T i has a bet-
ter fitness than P i , therefore, P i is replaced by T i [18].
68 A.E. Chaib et al. / Electrical Power and Energy Systems 81 (2016) 64–77
It is worth mentioning that, the developed software program is System characteristics Value Details
written in the commercial MATLAB computing environment using Buses 30 –
some MATPOWER programs. It is applied on a 2.20 GHz i7 personal Branches 41 –
computer with 8.00 GB-RAM and using parallel processing to run Generators 6 Buses: 1, 2, 5, 8, 11 and 13
Shunts 9 Buses: 10, 12, 15, 17, 20, 21, 23, 24 and 29
the different runs.
Transformers 4 Branches: 11, 12, 15 and 36
In this paper, the proposed BSA algorithm has been applied on Control variables 24 –
three test systems that are the IEEE 30-bus, the IEEE 57-bus, and
IEEE-118 bus test systems. Moreover, 16 different case studies
are investigated. These cases start from basic cases then different in Table 5 where the optimal generation fuel cost obtained is
complexities are included and other objectives are considered. (799.0760 $/h). Furthermore, the evolution of the objective func-
These cases are summarized in Table 1. In our implementation, tion over iterations is sketched in Fig. 1. This figure shows the evo-
the transformer tap settings have been considered as discrete vari- lution of cost (without penalty) and the penalty term over
able with a step of 0.125 p.u. iterations. It can be seen that the penalty term tends to zero after
34 iterations only. In other words, the BSA finds a feasible solution
IEEE 30-bus test system for this case after 34 iterations.
The IEEE 30-bus test system has a total generation capacity of CASE 2: OPF by considering cost reduction and voltage profile
435.0 MW and its main characteristics are given in Table 2. improvement
Detailed data can be derived from [24]. Cost and emission coeffi- Bus voltage is one of the most important and significant indica-
cients used in this paper are given in Table 3. Table 4 gives cost tion of safety and service quality of power systems. The solution
coefficients when considering multi-fuels options for generators found in CASE 1 is a purely cost based objective function. Such
1 and 2. solution may have an undesirable voltage profile [25]. Thus, this
example aims at minimizing fuel cost with a flatter voltage profile
CASE 1: OPF by considering cost reduction by considering a twofold objective function [26]. The voltage pro-
The first case investigated in this paper is the base case, which file is optimized by minimizing the load bus voltage deviation (VD)
consists of minimizing the generation fuel cost expressed by a from 1.0 p.u, which is given by:
quadratic function. Hence, the objective function (or more pre-
cisely the augmented objective function) for this case is: X
NL
VD ¼ V L 1 ð19Þ
! i
X
NG j¼1
Jðx; uÞ ¼ ai þ bi P Gi þ ci P2Gi þ Penalty ð18Þ
i¼1 Therefore, the objective function can be expressed as:
!
where ai ; bi and ci are the cost coefficients of the ith generator. X
NG
The proposed BSA technique has been run for this case and the
Jðx; uÞ ¼ ai þ bi PGi þ ci P2Gi þ kVD ðVDÞ þ Penalty ð20Þ
i¼1
optimal control variables obtained after optimization are tabulated
Table 1
Different case studies investigated in this paper.
Name Cost Voltage profile Voltage stability Multi-fuels Valve-point effect Emission System
CASE 1 – – – – – IEEE 30-bus test system
CASE 2 – – – –
CASE 3 – – – –
CASE 4 – – – –
CASE 5 – – –
CASE 6 – – –
CASE 7 – – – –
CASE 8 – – –
CASE 9 – – –
CASE 10 – – – –
CASE 12 – – – –
CASE 13 – – – –
CASE 14 – – – –
CASE 15 – – – –
Table 3
Cost and emission coefficients for the IEEE 30-bus test system.
Generator Bus a b c d e a b c x l
G1 1 0 2 0.00375 18 0.037 4.091 5.554 6.49 2.00E04 2.857
G2 2 0 1.75 0.0175 16 0.038 2.543 6.047 5.638 5.00E04 3.333
G3 5 0 1 0.0625 14 0.04 4.258 5.094 4.586 1.00E06 8
G4 8 0 3.25 0.00834 12 0.045 5.326 3.55 3.38 2.00E03 2
G5 11 0 3 0.025 13 0.042 4.258 5.094 4.586 1.00E06 8
G6 13 0 3 0.025 13.5 0.041 6.131 5.555 5.151 1.00E05 6.667
Table 4
Cost coefficients for generator 1 and 2 of the IEEE 30-bus test system for multi-fuels sources.
Table 5
Optimal settings of control variables found for CASE 1 through CASE 10.
CASE 1 CASE 2 CASE 3 CASE 4 CASE 5 CASE 6 CASE 7 CASE 8 CASE 9 CASE 10
PG1 177.3838 173.1714 172.0277 139.9204 139.5920 139.7202 198.7223 197.0323 197.6731 112.9189
PG2 48.8335 48.3093 48.8610 54.9886 54.2062 54.7933 44.3031 41.8254 46.0134 59.3719
PG5 21.2907 23.4476 22.3206 23.2095 24.4755 27.7068 18.5637 22.0463 15.6953 27.6576
PG8 21.0186 22.1097 20.9493 35.0000 28.6711 31.0154 10.0000 10.7320 12.4700 34.9989
PG11 11.4675 13.3916 13.5766 18.5930 22.2083 16.7496 10.1017 10.8690 10.0000 27.0652
PG13 12.0602 12.3455 14.1551 18.3118 21.5854 19.8743 12.0000 12.0000 12.0000 26.4502
VG1 1.1000 1.0441 1.0880 1.0863 1.0356 1.0902 1.1000 1.0514 1.0954 1.1000
VG2 1.0806 1.0245 1.0678 1.0699 1.0177 1.0720 1.0778 1.0278 1.0754 1.0855
VG5 1.0545 1.0037 1.0488 1.0403 1.0037 1.0463 1.0520 1.0086 1.0455 1.0606
VG8 1.0633 1.0005 1.0510 1.0532 1.0012 1.0568 1.0574 1.0018 1.0382 1.0757
VG11 1.0946 1.0316 1.0979 1.0679 1.0240 1.1000 1.0802 1.0263 1.1000 1.1000
VG13 1.1000 1.0049 1.0972 1.0541 1.0111 1.0903 1.0803 1.0061 1.0985 1.1000
T11(69) 1.0250 1.0500 0.9750 1.0625 1.0375 0.9625 1.0000 1.0250 0.9750 1.0000
T12(610) 0.9000 0.9000 0.9000 0.9125 0.9000 0.9250 1.0125 0.9125 0.9000 0.9500
T15(412) 0.9625 0.9625 0.9625 1.0000 0.9875 0.9625 1.0250 0.9625 0.9750 1.0000
T36(2827) 0.9625 0.9625 0.9500 0.9875 0.9625 0.9500 1.0000 0.9750 0.9375 0.9625
QC10 4.2998 5.0000 4.5315 5.0000 4.3687 4.4714 4.3411 3.8622 4.5090 3.4844
QC12 4.6378 0.7241 4.5338 5.0000 5.0000 4.5103 4.9527 1.9742 4.9177 4.5129
QC15 4.9106 3.7630 4.6750 5.0000 3.3418 5.0000 4.2358 2.4068 4.1622 4.7990
QC17 5.0000 2.3539 4.2197 5.0000 0.0000 5.0000 4.7605 0.0000 5.0000 4.9965
QC20 4.0889 4.9912 5.0000 3.1123 5.0000 4.9360 4.0597 5.0000 5.0000 3.9809
QC21 5.0000 3.6589 5.0000 5.0000 3.9460 3.8837 4.5901 5.0000 4.6871 4.7684
QC23 3.1843 4.9775 4.7058 3.9314 4.9261 4.8542 4.1971 4.5359 4.5496 3.8535
QC24 4.8423 4.8500 3.6102 5.0000 5.0000 3.4071 5.0000 4.9781 2.9377 4.2332
QC29 2.5810 2.2713 4.8119 1.4393 3.7995 4.1926 4.1450 4.2463 3.9664 1.6339
Cost ($/h) 799.0760 803.4294 800.3340 646.1504 653.1019 648.1424 830.7779 836.8811 832.7029 835.0199
VD 1.9129 0.1147 1.9855 1.0273 0.1161 1.9825 1.2050 0.1194 1.8985 1.9214
Lmax 0.1273 0.1484 0.1259 0.1378 0.1478 0.1258 0.1363 0.1491 0.1262 0.1266
Ploss (MW) 8.6543 9.3751 8.4904 6.6233 7.3386 6.4595 10.2908 11.1050 10.4519 5.0626
Qloss (MVar) 36.5617 41.0206 34.0897 30.6217 32.6230 26.2444 40.4990 45.2534 41.0987 22.1368
Emission (ton/h) 0.3671 0.3546 0.3514 0.2833 0.2805 0.2821 0.4377 0.4300 0.4343 0.2425
Penalty
1
1.9129 p.u to 0.1147 p.u. This reduction can be visually verified
in Fig. 2 where the optimized voltage profile of load buses is greatly
improved compared to that of CASE 1. Furthermore, the evolution
of both objective functions over iterations is sketched in Fig. 3.
1.1
1.05
Voltage Profile (p.u.)
0.95
0.9
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
(a)
1.1
1.05
Voltage Profile (p.u.)
0.95
0.9
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
(b)
Fig. 2. Comparison of voltage profiles between (a) CASE 1 and (b) CASE 2.
860 1.2 where H matrix is generated by the partial inversion of Y bus . More
Cost details can be found in [27].
850 VD 1 The indicator Lmax varies between 0 and 1 where the lower the
indicator, the more the system stable. Therefore, enhancing voltage
840 0.8 stability can be achieved through the minimization of Lmax of the
VD (p.u.)
Cost ($/h)
whole system [28]. Thus, the objective function in this case is:
830 0.6
!
X
NG
Jðx; uÞ ¼ ai þ bi PGi þ ci P2Gi þ kLmax ðLmax Þ þ Penalty ð23Þ
820 0.4 i¼1
!
900 0.14 NG
X
2 min
Jðx; uÞ ¼ ai þ bi P Gi þ ci PGi þ di sin ei PGi PGi
Cost i¼1
L max
þ Penalty
ð28Þ
Cost ($/h)
Lmax
where di and ei are the coefficients that represent the valve-point
850 0.13
effect.
The OPF problem of CASE 7 has been solved using the BSA algo-
rithm and the optimal control variables obtained are displayed in
Table 5. The optimal fuel cost obtained in this case has increased
from 799.0760 $/h to 830.7779 $/h compared to the basic case
i.e., CASE 1 due to the consideration of the valve-point effect.
800 0.12
1 125 250 375 500
CASE 8: OPF by considering valve-point effect and voltage profile
Iterations
improvement
Fig. 4. Objective function variation for CASE 3. This case is similar to CASE 2 where the aim is to optimize both
the cost and to improve the voltage profile. However, in this case
the valve point effect is taken into consideration. The optimal
Jðx; uÞ ¼ J CostG1 and G2
þ J CostRemaining Generators
þ Penalty ð25Þ results obtained using BSA for this case are given in Table 5.
Emission (ton/h)
obtained are tabulated in Table 5. We can see from this table that
Cost ($/h)
Table 7
Cost and emission coefficients for the IEEE 57-bus test system.
Generator Bus a b c d e a b c x l
G1 1 0 2 0.00375 18 0.037 4.091 5.554 6.49 2.00E04 2.86E01
G2 2 0 1.75 0.0175 16 0.038 2.543 6.047 5.638 5.00E04 3.33E01
G3 3 0 3 0.025 13.5 0.041 6.131 5.555 5.151 1.00E05 6.67E01
G4 6 0 2 0.00375 18 0.037 3.491 5.754 6.39 3.00E04 2.66E01
G5 8 0 1 0.0625 14 0.04 4.258 5.094 4.586 1.00E06 8.00E01
G6 9 0 1.75 0.0195 15 0.039 2.754 5.847 5.238 4.00E04 2.88E01
G7 12 0 3.25 0.00834 12 0.045 5.326 3.555 3.38 2.00E03 2.00E01
Table 8
Optimal settings of control variables found for CASE 11 through CASE 15.
X
NG The results obtained after optimization using the BSA algorithm
Emission ¼ 102 ai þ bi PGi þ ci P 2Gi þ xi eðli PGi Þ ð29Þ are displayed in Table 5 and the trend of optimization is repre-
i¼1
sented in Fig. 5. The results show that the emission has been
reduced from (0.3671 ton/h) to (0.2425 ton/h) i.e. by 33.94%, how-
where ai ; bi ; ci ; xi and li are coefficients of the ith generator emis-
ever, the total fuel cost has increased from (799.0760 $/h) to
sion characteristics.
(835.0199 $/h) i.e. by 4.50% compared with CASE 1.
Therefore, the objective function for this case can be expressed
by:
IEEE 57-bus test system
!
X
NG
Jðx; uÞ ¼ ai þ bi P Gi þ ci P2Gi þ kEmission ðEmissionÞ þ Penalty In order to test the scalability of the proposed BSA algorithm, a
i¼1 larger test system is considered in this paper, which is the IEEE 57-
ð30Þ bus test system. This system has a total generation capacity of
1975.9 MW and its main characteristics are given in Table 6. Cost
where kEmission is a weighting factor and it is equal to 1000 in this and emission coefficients for this system are given in Table 7.
case. Detailed data can be derived from [33].
A.E. Chaib et al. / Electrical Power and Energy Systems 81 (2016) 64–77 73
1.1
1.05
Voltage Profile (p.u.)
0.95
0.9
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 42 43 44 45 46 47 48 49 50
PQ Bus Number
(a)
1.1
1.05
Voltage Profile (p.u.)
0.95
0.9
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 42 43 44 45 46 47 48 49 50
PQ Bus Number
(b)
Fig. 6. Comparison of voltage profile between (a) CASE 11 and (b) CASE 12.
Table 9
The main characteristics of the IEEE 118-bus test system. is given by (20) where kVD ¼ 10; 000 in this case to balance
between the objectives. The results of such optimization using
System Value Details
characteristics
the proposed BSA algorithm are tabulated in Table 8. It appears
from this table that the VD has been reduced from (1.1009 p.u.)
Buses 118 [33]
Branches 186 [33]
to (0.6888 p.u.) compared with CASE 11. However, the cost has
Generators 54 Buses: 1, 4, 6, 8, 10, 12, 15, 18, 19, 24, 25, 26, 27, slightly increased from (6411.0043 $/h) to (6436.7551 $/h) com-
31, 32, 34, 36, 40, 42, 46, 49, 54, 55, 56, 59, 61, pared with CASE 11. The improvement of the voltage profile is
62, 65, 66, 69, 70, 72, 73, 74, 76, 77, 80, 85, 87, illustrated in Fig. 6.
89, 90, 91, 92, 99, 100, 103, 104, 105, 107, 110,
111, 112, 113 and 116
Shunts 14 Buses: 5, 34, 37, 44, 45, 46, 48, 74, 79, 82, 83,
105, 107 and 110
CASE 13: OPF by considering cost reduction and voltage stability
Transformers 9 Branches: 8, 32, 36, 51, 93, 95, 102, 107 and 127 enhancement
Control variables 130 – The voltage stability enhancement has been also tested for the
IEEE 57-test system in this case where the aim is to minimize
the cost and at the same time to improve the vulnerability of the
system by enhancing its voltage stability. Thus, the objective func-
CASE 11: OPF by considering cost reduction tion of CASE 13 is given by (23) with kLmax ¼ 10; 000. The BSA is run
The aim here is to minimize the total generating fuel cost. for this case and the obtained control variables are tabulated in
Therefore, the objective function of this case is given by (18). The Table 8. These results show clearly that the voltage stability of
BSA is run in order to find the optimal settings for CASE 11 and the system is enhanced with a slight augmentation in cost.
the obtained results are given in Table 8. The cost obtained for this
case is (6411.0043 $/h) while VD and Lmax are (1.1009 p.u.) and
(0.2819), respectively. CASE 14: OPF by considering valve-point effect
The effect of valve-point loading is tested in this case. Therefore,
CASE 12: OPF by considering cost reduction and voltage profile the objective function of this case is given by (28). The obtained
improvement results using BSA are given in Table 8. In this case the cost has mar-
As in CASE 2, the goal here is to optimize both the total gener- ginally augmented from (6411.0043 $/h) to (6462.4093 $/h) com-
ating fuel cost and the voltage profile. Hence the objective function pared with CASE 11.
74 A.E. Chaib et al. / Electrical Power and Energy Systems 81 (2016) 64–77
Table 12
comparison of the BSA with DE, PSO, GA ABC and BBO for solving different OPF problems.
Table 12 (continued)
In order to evaluate the convergence speed of the BSA, the fol- more than, 95%, 97%, and 99% of the final value objective function,
lowing approach is proposed in this paper. First, the values of the respectively.
objective function (average values over all runs) are recorded at
four cut-points of the search corresponding to 20%, 40%, 60% and Conclusion
80% of the maximum number of iterations. Then, the percentage
that represents the value of this objective function compared with In this paper, a newly developed algorithm called the BSA is
the final value (average of final values over all runs) reached by the implemented to solve several OPF problems. Three test systems
algorithm is calculated. These percentages are reported in Table 13. and sixteen cases have been investigated in order to evaluate the
The final row is the average, calculated on all cases for each cut performance of the proposed algorithm. Furthermore, the obtained
point. From this table we can say that the BSA is very efficient in results have been compared to those obtained using other well-
solving several OPF problems with different complexities. For the known optimization algorithms such as DE, PSO, ABC, GA and
first cut-point, the BSA has reached already more than 85% of the BBO. The main conclusion that can be drawn from this paper is that
final value of the objective function for all cases except for CASE the BSA is a very effective and robust algorithm for solving OPF
16. For cut-point 2, cut-point 3 and cut-point 4 the BSA has reached problems. It has good convergence characteristics and can achieve
A.E. Chaib et al. / Electrical Power and Energy Systems 81 (2016) 64–77 77
Table 13 [8] Bouchekara HREH, Abido Ma, Boucherma M. Optimal power flow using
Convergence speed of the BSA algorithm. teaching-learning-based optimization technique. Electr Power Syst Res
2014;114:49–59.
Case Cut-point 1 Cut-point 2 Cut-point 3 Cut-point 4 [9] Bouchekara HREH, Abido Ma, Chaib aE, Mehasni R. Optimal power flow using
the league championship algorithm: a case study of the Algerian power
CASE 1 99.12 99.66 99.86 99.95
system. Energy Convers Manag 2014;87:58–70.
CASE 2 91.81 95.89 98.25 99.38
[10] Bouchekara HRE-H, Abido MA. Optimal power flow using differential search
CASE 3 97.35 99.16 99.76 99.92 algorithm. Electr Power Components Syst 2014;42:1683–99.
CASE 4 98.19 99.37 99.72 99.89 [11] Mukherjee A, Mukherjee V. Solution of optimal power flow using chaotic krill
CASE 5 89.11 95.34 97.56 99.00 herd algorithm. Chaos, Solitons Fract 2015;78:10–21.
CASE 6 96.18 98.79 99.63 99.87 [12] Duman S, Güvenç U. Optimal power flow using gravitational search algorithm.
CASE 7 98.07 99.28 99.76 99.91 Energy Convers Manag 2012;59:86–95.
CASE 8 91.46 95.55 97.75 99.09 [13] Bhowmik AR, Chakraborty AK. Solution of optimal power flow using non
CASE 9 96.85 98.72 99.45 99.81 dominated sorting multi objective opposition based gravitational search
CASE 10 99.46 99.82 99.90 99.96 algorithm. Int J Electr Power Energy Syst 2015;64:1237–50.
CASE 11 93.65 99.70 99.93 99.98 [14] Ghasemi M, Ghavidel S, Ghanbarian MM, Massrur HR, Gharibzadeh M.
CASE 12 86.04 96.56 98.81 99.64 Application of imperialist competitive algorithm with its modified
CASE 13 94.98 99.14 99.68 99.89 techniques for multi-objective optimal power flow problem: a comparative
study. Inf Sci (Ny) 2014;281:225–47.
CASE 14 92.89 99.72 99.92 99.98
[15] Daryani N, Hagh MT, Teimourzadeh S. Adaptive group search optimization
CASE 15 93.14 99.78 99.93 99.98
algorithm for multi-objective optimal power flow problem. Appl Soft Comput
CASE 16 0.96 96.28 98.92 99.74
2016;38:1012–24.
Average 88.70 98.30 99.30 99.75 [16] AlRashidi M, El-Hawary M. Applications of computational intelligence
techniques for solving the revived optimal power flow problem. Electr
Power Syst Res 2009;79:694–702.
[17] Frank S, Steponavice I. Optimal power flow?: a bibliographic survey II non-
better performances than some well-known optimization algo- deterministic and hybrid methods. Energy Syst 2012:259–89.
rithms. This is more pronounced when dealing with large scale [18] Civicioglu P. Backtracking search optimization rithm for numerical
power systems. Finally, it is recommended that a multiobjective optimization problems. Appl Math Comput 2013;219:8121–44.
[19] Mandal S, Sinha RK, Mittal K. Comparative analysis of backtrack search
BSA algorithm based on Pareto optimal solutions has to be devel- optimization algorithm (BSA) with other evolutionary algorithms for global
oped and applied to solve OPF problems in further studies. continuous optimization. Int J Comput Sci Inf Technol 2015;6:3237–41.
[20] Civicioglu P. Circular antenna array design by using evolutionary search
algorithms. Prog Electromagn Res B 2013;54:265–84.
Acknowledgments [21] Shafiullah M, Abido MA, Coelho LS. Design of robust PSS in multimachine
power systems using backtracking search algorithm; 2015.
[22] El-Fergany A. Multi-objective allocation of multi-type distributed generators
Dr. M. A. Abido would like to acknowledge the support provided along distribution networks using backtracking search algorithm and fuzzy
by King Abdulaziz City for Science and Technology (KACST) expert rules. Electr Power Components Syst 2015;5008:1–16.
through the Science and Technology Unit at King Fahd University [23] Hendrix EMT, G.-Tóth B. Introduction to nonlinear and global optimization,
vol. 37. Springer; 2010.
of Petroleum and Minerals (KFUPM) for funding this work through [24] Ray D. Zimmerman CEM-S& D (David) G. MATPOWER. <http://www.
project# 14-ENE265-04 as a part of the National Science, Technol- pserc.cornell.edu/matpower/#docsn.d>.
ogy and Innovation Plan (NSTIP). [25] Yuryevich J. Evolutionary programming based optimal power flow algorithm.
IEEE Trans Power Syst 1999;14:1245–50.
[26] He S, Went JY, Prempaint E, Wut QH, Fitchs J, Mand S. An improved particle
swarm optimization for optimal power flow; 2004. p. 21–4.
References
[27] Kessel P, Glavitsch H. Estimating the voltage stability of a power system. IEEE
Trans Power Deliv 1986;1:346–54.
[1] Cain M, O’neill R, Castillo A. History of optimal power flow and formulations. [28] Abou El Ela AA, Abido MA, Spea SR. Optimal power flow using differential
FERC Staff Tech Pap; 2012. p. 1–36. evolution algorithm. Electr Power Syst Res 2010;80:878–85.
[2] Surender Reddy S, Bijwe PR, Abhyankar aR. Faster evolutionary algorithm [29] Hardiansyah H. A modified particle swarm optimization technique for
based optimal power flow using incremental variables. Int J Electr Power economic load dispatch with valve-point effect. Int J Intell Syst Appl
Energy Syst 2014;54:198–210. 2013;5:32–41.
[3] Pandya KS, Joshi SK. Survey Opt Power Flow Methods 2008;0:450–8. [30] Niknam T, Narimani MR, Azizipanah-abarghooee R. A new hybrid algorithm
[4] Frank S, Steponavice I. Optimal power flow?: a bibliographic survey I. for optimal power flow considering prohibited zones and valve point effect.
Formulations and deterministic methods; 2012. p. 221–58. Energy Convers Manag 2012;58:197–206.
[5] Khorsandi a, Hosseinian SH, Ghazanfari a. Modified artificial bee colony [31] Alsumait JS, Sykulski JK, Al-Othman aK. A hybrid GA–PS–SQP method to solve
algorithm based on fuzzy multi-objective technique for optimal power flow power system valve-point economic dispatch problems. Appl Energy
problem. Electr Power Syst Res 2013;95:206–13. 2010;87:1773–81.
[6] Ayan K, Kılıç U, Baraklı B. Chaotic artificial bee colony algorithm based solution [32] Abido Ma. Environmental/economic power dispatch using multiobjective
of security and transient stability constrained optimal power flow. Int J Electr evolutionary algorithms. IEEE Trans Power Syst 2003;18:1529–37.
Power Energy Syst 2015;64:136–47. [33] 2014 OPF PROBLEMS. <https://www.uni-due.de/ieee-wgmho/competition
[7] Bouchekara HREH. Optimal power flow using black-hole-based optimization 2014n.d>.
approach. Appl Soft Comput J 2014;24:879–88.