Professional Documents
Culture Documents
1 s2.0 S2215098616313805 Main PDF
1 s2.0 S2215098616313805 Main PDF
a r t i c l e i n f o a b s t r a c t
Article history: In this paper, a new low level with teamwork heterogeneous hybrid particle swarm optimization and
Received 31 December 2016 artificial physics optimization (HPSO-APO) algorithm is proposed to solve the multi-objective security
Revised 9 March 2017 constrained optimal power flow (MO-SCOPF) problem. Being engaged with the environmental and total
Accepted 9 March 2017
production cost concerns, wind energy is highly penetrating to the main grid. The total production cost,
Available online 27 March 2017
active power losses and security index are considered as the objective functions. These are simultane-
ously optimized using the proposed algorithm for base case and contingency cases. Though PSO
Keywords:
algorithm exhibits good convergence characteristic, fails to give near optimal solution. On the other hand,
Multi-objective
Hybrid optimization algorithm
the APO algorithm shows the capability of improving diversity in search space and also to reach a near
Security constrained optimal power flow global optimum point, whereas, APO is prone to premature convergence. The proposed hybrid
Pareto optimal solution HPSO-APO algorithm combines both individual algorithm strengths, to get balance between global and
local search capability. The APO algorithm is improving diversity in the search space of the PSO algorithm.
The hybrid optimization algorithm is employed to alleviate the line overloads by generator rescheduling
during contingencies. The standard IEEE 30-bus and Indian 75-bus practical test systems are considered
to evaluate the robustness of the proposed method. The simulation results reveal that the proposed
HPSO-APO method is more efficient and robust than the standard PSO and APO methods in terms of get-
ting diverse Pareto optimal solutions. Hence, the proposed hybrid method can be used for the large inter-
connected power system to solve MO-SCOPF problem with integration of wind and thermal generators.
Ó 2017 Karabuk University. Publishing services by Elsevier B.V. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction scenario. The wind energy is the most proven available and grow-
ing energy in the entire world. Once wind plant is installed it
Due to rapid increase in electricity demand, the modern power requires zero production cost and less maintenance than conven-
systems run close to their security limits to enhance the system tional plants and also reduces the carbon emission.
security during contingencies. The security assessment and Earlier, a number of conventional (mathematical based)
enhancement are two major concerns in the energy control centers algorithms like gradient method, quadratic programming, linear
[1]. A security assessment is the analysis carried out to determine programming, mixed integer linear programming and nonlinear
what level; a power system is reasonably safe from unforeseen dis- programming (NLP) are proposed to solve the OPF problem
turbances (contingencies). The bottleneck of the security assess- [2–4]. These methods face difficulty in handling the inequality con-
ment is contingency analysis. Security constrained optimal power straints and not guarantee to reach an optimal solution. In addi-
flow (SCOPF) is the highly non-linear OPF problem along with tion, they will struck in the local optimum point in some cases
the contingency analysis. This leads to implementation of preven- and exhibits poor convergence characteristic. Over the last years,
tive control actions in the power system like generation reschedul- many researchers have tended to apply several non-conventional
ing, phase shifter positions, switching of FACTS devices, HVDC line (heuristic) methods for solving the OPF problem like simulated
MW transfer, and load shedding to enhance the system security. annealing (SA) [5], genetic algorithm (GA) [6], enhanced GA
The environmental and cost concerns of the renewable energies (EGA) [7], fuzzy based GA (FGA) [8], differential evolution (DE)
increase the integration to the power sector in the present [9], modified differential evolution (MDE) [9], evolutionary pro-
gramming (EP) [10], improved EP (IEP) [11], particle swarm opti-
⇑ Corresponding author. mization (PSO) [12], ant colony optimization (ACO) [13], honey
E-mail address: Kiran.t39@nitw.ac.in (K. Teeparthi). bee mating optimization (HBMO) [14], modified HBMO (MHBMO)
Peer review under responsibility of Karabuk University. [14], modified shuffle frog leaping algorithm (MSFLA) [15].
http://dx.doi.org/10.1016/j.jestch.2017.03.002
2215-0986/Ó 2017 Karabuk University. Publishing services by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
412 K. Teeparthi, D.M. Vinod Kumar / Engineering Science and Technology, an International Journal 20 (2017) 411–426
However, some of these standard algorithms are suffering from The rest of the paper is organized as follows. Section 2 deals
premature convergence due to lack of diversity in search space. with the MO-SCOPF problem with wind energy subject to equality
One way to cope with these drawbacks, hybrid heuristic optimiza- and inequality constraints. The general framework of PSO, APO and
tion methods are implemented in practical and academic prob- proposed hybrid PSO-APO (HPSO-APO) algorithms for solving MO-
lems. The best results are found for OPF problem by hybrid SCOPF is discussed in Sections 3 and 4, whereas Section 5 illus-
methods [16–19]. trates the simulation and results of the proposed method. Finally,
OPF is a nonlinear problem which optimizes the objection func- conclusions are given in Section 6.
tion while satisfying a set of equality and inequality constraints.
The minimization of fuel cost is most generally used as the objec- 2. Problem formulation
tive function. However, other traditional objectives are minimiza-
tion of transmission real power losses, voltage stability index, Multi-objective security constrained OPF is a special type of OPF
voltage deviation, fuel emission, and security index. In traditional problem to minimize the objective functions simultaneously sub-
methods, multi-objective OPF (MO-OPF) problems are solved by jected to equality and inequality constraints by choosing the opti-
evolutionary algorithms (EA) which convert multi-objective prob- mal control variables like phase-shifter angles, active & reactive
lem to a single objective problem by an assigning suitable weight- power generations, transformer tap settings and voltages at PV
ing/scaling factors [20,21]. This approach gives only one solution in buses. In this paper, three objective functions were considered
a single run. It should be separately run for a set of weighting fac- namely, total production cost, security index, and active power loss.
tors to get the Pareto optimal solutions and this increase the com-
putational time. To rectify the above-mentioned problem, EAs have 2.1. Objective functions
been reported to solve MO-OPF problem to attain the Pareto opti-
mal points [22–30]. Kadir Abaci et al. [23] implemented DEA based 2.1.1. Total production cost
OPF for solving single and multi-objective functions and results are The total production cost including thermal and wind generator
compared with other reported methods presented in the literature. units can be expressed as follows [32].
An adaptive group search optimization based MO-OPF problem is
X
Ng X
Nw X
Nw
presented in [24]. A modified decomposition based multi- f TPC ¼ ðai P2gi þ bi P gi þ ci Þ þ C p ðPshed;i Þ þ C r ðPshed;i Þ ð1Þ
objective OPF problem is solved with consideration of different i¼1 i¼1 i¼1
objectives like fuel cost, emission, voltage deviation and power
where P gi is the active power generation at ith bus; N g and N w rep-
losses [25]. The adjustment of control variables like active power
resents the available number of thermal and wind generators; ai , bi
generations and TCSC device are used to alleviate line overloads
and ci are generator cost coefficients. The first part of Eq. (1) implies
during contingencies for enhancing the system security. The
the fuel cost of available thermal generators, and second and third
multi-objective optimization problem is formulated by combing
terms indicate the overestimation (C p ðPshed;i Þ) and underestimation
the installation cost of TCSC device and fuel cost of thermal units
(C r ðPshed;i Þ) costs of wind energy.
[26]. In order to enhance the system security, a modified SFLA is
Mathematically, the stochastic nature of wind speed follows the
proposed to solve OPF problem incorporating UPFC device [27].
Weibull probability density function (PDF) over a time which is
In [28], fuzzy adaptive PSO method is implemented to solve OPF
given by
problem and optimal placement of UPFC device simultaneously.
ðk1Þ
Cheng- Jin Ye et.al [29] proposed a multi-objective OPF model with k v v k
considering rotor angle stability as the objective function rather
f ðv Þ ¼ ðeÞð c Þ ;0 < v < 1 ð2Þ
c c
than taking as inequality constraint. In [30], fuzzy based grenade
explosion optimization method is successfully applied for solving where v represents the wind speed; k and c are scale and shape fac-
multi-objective OPF problem. tors of the wind speed respectively. The output power of the wind
The main contribution of this paper is to solve multi-objective generator with a given wind speed is expressed as
8
security constrained OPF (MO-SCOPF) using a new hybrid particle >
> 0; 0 < v < v in
swarm optimization and artificial physics optimization (HPSO- >
>
< p v v in ; v in 6 v 6 v r
APO) algorithm. The APO and PSO methods are two meta- p¼ r v r v in ð3Þ
heuristic population-based optimization approaches inspired by >
>
>
>
pr ; vr 6 v 6 vo
:
nature of physics and behavior of bird flocking respectively. The 0; v > vo
proposed hybrid PSO-APO algorithm combines both individual
Here v in , v r and v o are cut-in, rated and cut-out wind speeds
algorithm strengths, to get the balance between global and local
respectively and pr is the rated output power of the wind genera-
search capability. The APO algorithm is improving diversity in
the search space of the PSO algorithm so as to avoid the trapping tor. The probability density function (f P ðpÞ) for linear portion of
of local optima. In this paper, three different objective functions the wind speed is given by
ðk1Þ
have been considered namely, total production cost, active power kðv r v in Þ v in pr þ pðv r v in Þ
f P ðpÞ ¼
losses, and security index. Generally, most of the multi-objective cpr cpr
based optimization methods use non-dominated sorting and !
v in pr þ pðv r v in Þ k
strength Pareto approaches for achieving the optimal trade-off exp ð4Þ
curve. This paper uses non-dominated sorting and crowding dis- cpr
tance approach to maintaining a diverse in Pareto optimal points
For discrete portion, the probabilities of getting no power out-
[31]. Finally, a fuzzy membership approach is used to get compro-
put and rated power output of the wind turbine are as follows
mising solution over the trade-off curve. The proposed multi- k k
objective HPSO-APO (MOHPSO-APO) algorithm is tested on IEEE v in vo
f P ðp ¼ 0Þ ¼ 1 exp þ exp ð5Þ
30-bus and practical Indian 75-bus systems. The simulation results c c
show that the proposed method is more robust and efficient than
k k
the standard multi-objective PSO (MOPSO) and multi-objective vr vo
f P ðp ¼ pr Þ ¼ exp þ exp ð6Þ
APO (MOAPO) methods in terms of obtaining a well distributed c c
Pareto front.
K. Teeparthi, D.M. Vinod Kumar / Engineering Science and Technology, an International Journal 20 (2017) 411–426 413
X
Nb
xi ðt þ 1Þ ¼ xi ðtÞ þ v i ðt þ 1Þ ð20Þ
ðQ gen;i Q load;i Þ V i V j jY ij j sinðdij hij Þ ¼ 0 ð12Þ
j¼1 where v i ðtÞ and xi ðtÞare the velocity and position of ith particle at
iteration t, w is the weighting factor, C 1 and C 2 are acceleration con-
where, Nb represents total number of buses, Pgen;i and Q gen;i are spec-
stants. The first part of (19) indicates the momentum part (history)
ified active and reactive power generations at each bus, P load;i and
of PSO. The second and third terms are indicating cognitive and
Q load;i are active and reactive power loads; V i , V j and Y ij represents
social parts of PSO respectively. The velocities and positions are
the magnitude of bus voltage and bus admittance matrix respec- updated iteratively until limit on iterations reaches its maximum
tively; dij and hij are voltage and admittance phase angles between limit. The general steps involved in finding the optimum values of
buses i-j. generations to solve MO-SCOPF problem using the standard MOPSO
is shown in Fig 1.
2.2.2. Inequality constraints
The inequality constraints for generator and network limits are 3.2. Standard APO algorithm
expressed in terms of lower and upper limits as follows.
Pmin 6 Pgi 6 Pmax i ¼ 1; 2: . . . Ng ð13Þ In APO method [34], each solution in the multidimensional
gi gi
search space is treated as particles which are randomly distributed
414 K. Teeparthi, D.M. Vinod Kumar / Engineering Science and Technology, an International Journal 20 (2017) 411–426
Read system data and Generate the Evaluate the objective Generate new
also initialize the population functions and rank them offspring vector
PSO control randomly within a according to crowding using velocity
parameters given search space distance approach equation
No
Evaluate objective
Sort population based on non-
Update Pbest functions with respective
Meeting end dominated and crowding distance
and Gbest to the new population and
criterion? approach and Select non
positions combines with old
dominated set with best rank
population
Yes
Return the best
values
velocity. Based on the attractive and repulsive forces, each particle F i ðtÞ ¼ F ij ðtÞ ð24Þ
j2best;j–i
tends to larger masses close to global solution. These forces
between the particles can guide the entire population to obtain where G is the gravitational constant and F i ðtÞ is the total force
the near global optimum solution. This algorithm is three-fold exerted on individual i by all other individuals. Finally, using force
namely initialization, calculation of force and motion. The main and mass equations computes the velocities and positions of the
three steps of APO algorithm are mathematically modeled as each individual and updated iteratively using (25) and (26).
follows.
v i ðt þ 1Þ ¼ w v i ðtÞ þ a F i ðtÞ=mi ð25Þ
1. Population is randomly initialized in the given D – dimensional
search space. Evaluate best (bestðtÞ) and worst (worstðtÞ) fitness
xi ðt þ 1Þ ¼ xi ðtÞ þ v i ðt þ 1Þ ð26Þ
values.
2. Calculate the mass of each individual according to the best and where v i ðtÞ and xi ðtÞ are the velocity and position of ith particle
worst values. respectively at iteration t. a is a random number in the range [0, 1]
and w is the inertia weight. Fig. 2 shows the general steps for find-
bestðtÞ ¼ min fit i ðtÞ ing the optimum values of generations to solve MO-SCOPF problem
i¼1;...pop
ð21Þ using the standard MOAPO method.
worstðtÞ ¼ max fit i ðtÞ
i¼1...pop
bestðtÞ fit i ðtÞ 4. Proposed hybrid PSO-APO algorithm to solve MO-SCOPF
mi ðtÞ ¼ exp ð22Þ problem
worstðtÞ bestðtÞ
where fiti ðtÞ is the fitness of ith population at iteration t; mi ðtÞ is The classifications of hybrid optimization methods are
the mass of each particle at tth iteration. By using mass equation presented in [35]. The hybridization can be done as low level or
compute the forces on each individual (F ij ðtÞ) in each iteration is high level with relay or teamwork method as heterogeneous or
given in (23). homogeneous approaches. In this paper, a low level with team-
work heterogeneous approach is used. The proposed algorithm is
3. Using mass equation compute the forces on each individual low level because of using the combined functionality of both
(F ij ðtÞ) in each iteration is given as follows algorithms. It is teamwork because of both algorithms operated
in parallel and it is a heterogeneous method because of combining
Gmi ðtÞmj ðtÞðxj ðtÞ xi ðtÞÞ ; if fitj ðtÞ < fiti ðtÞ two different algorithms to produce final results. The general pro-
F ij ðtÞ ¼ ð23Þ
Gmi ðtÞmj ðtÞðxj ðtÞ xi ðtÞÞ ; if fitj ðtÞ > fiti ðtÞ cedure to solve security constrained OPF by HPSO-APO method is
given as follows.
No
Yes
Step 1: (a) Read the input data viz. line data, bus data, and gen- distance approach to obtain the Pareto optimal solutions [31].
erator data. Also, initialize the APO and PSO control parameters. Finally, choose one compromising solution out of Pareto optimal
(b) Generate an initial population randomly within the given vector; this can be done using fuzzy model approach. Each objec-
search space. tive function (f i ) of particle p is fuzzified with a membership func-
Step 2: Compute the value of fitness for each particle and tion (lpi ) and it is defined as
update the best and worst values. 8 min
Step 3: Evaluate mass of each particle using (22) and calculate >
> 1 fi 6 fi
<
the total force on each particle using (24). f max f i min max
lpi ¼ i
fi 6 fi 6 fi ð28Þ
Step 4: Update particle’s velocity and positions. The modified >
> f max f min
: i i
max
velocity equation of PSO is as follows. Finally, the positions of all 0 fi P fi
particles are updated by using (26).
max min
where f i and f i are the maximum and minimum objective val-
mi ðt þ 1Þ ¼ w mi ðtÞ þ C 01 randðÞ ðF i ðtÞ=mi ðtÞÞ ues of ith objective function, respectively. The normalized member-
þ C 02 randðÞ ðgbest xi ðtÞÞ ð27Þ ship function of each particle is calculated as
N P
Step 5: Check for maximum iterations count. If yes print the lpi
final results otherwise go to step 2.The general steps involved in
lp ¼ PD i¼1
PN ð29Þ
k¼1 i¼1 lpi
finding the optimum values of generations to solve MO-SCOPF
problem using the proposed method is shown in Fig. 3. The where D represents the number of non-dominated solutions and N
multi-objective problem solves different conflicting objectives is the number of objective functions. The maximum value of lp
simultaneously to find the set of optimal solutions rather than sin- gives the best compromising solution among the total Pareto opti-
gle solution. This paper uses non-dominated sorting and crowding mal points.
start
Read System data like line parameters, bus data Generate new population using modified PSO
and generator data velocity equation
Initialize PSO and APO control parameters Evaluate objective functions with respective to
the new population and combines with old
population
Set iter=1 No
Iter=iter_max
End
Iter=iter+1
5. Simulation results and discussions 40 MW is connected at bus no 22, which is arbitrarily considered.
In this paper, wind generator is modeled as PV bus with unity
To validate the robustness and effectiveness of the proposed power factor for load flow studies. Throughout the simulation,
hybrid PSO-APO algorithm for solving MO-SCOPF problem, stan- total load demand of the system is assumed to be constant. The
dard IEEE 30-bus and practical Indian 75-bus test systems are to total electrical energy demand of the system is 283.4 MW. For a
be considered. The test results are compared with individual stan- particular time horizon, the Weibull distribution factors are taken
dard PSO and APO methods. The proposed algorithm is coded in as scale factor (c) = 5.92 and shape factor (k) = 1.12. According to
MATLAB environment and all simulations were carried out on these factors the scheduled wind power generation is taken as
Intel, 2.40 GHz and core i5-4210U with 4 GB personal computer. 12 MW. The following parameters to be consider for a wind gener-
For each test case, 20 independent trail runs are performed to val- ator. They are, cut-in speed (v in ) = 3 m/s, cut-out speed (v o ) = 25 m/
idate the effectiveness of the proposed algorithm. Before HPSO- s and rated output speed (v r ) = 10.28 m/s.
APO method is applied, following control parameters need to be In order to test the robustness of the proposed algorithm, we
chosen effectively. For PSO, acceleration constants, C1 = C2 = 2 made a comparison of the proposed optimization method with
and w = 1 to be considered. For APO, gravitational constant (G) other existing optimization methods presented in the literature.
adopted is 0.01. In order to get fair comparison between proposed Here, we considered fuel cost of thermal generators as an objective
and standard PSO and APO algorithms in terms of reaching best function. The comparison of minimum fuel cost obtained by the
value, the population size (NP = 50) and number of iterations (iter proposed hybrid algorithm along with existing algorithms has
Max. = 100) are taken same for three optimization methods. The been presented in Table 1. From the results, the minimum fuel cost
simulation results of MO-SCOPF with base and contingency cases obtained by the proposed method is 801.708 $/h, which is better
are discussed in the subsequent sections. than standard individual PSO and APO algorithms as well as previ-
ously reported methods. The convergence characteristics of the
proposed hybrid method, standard PSO and APO methods are
5.1. IEEE 30-bus system
shown in Fig. 4. The results reveal that PSO algorithm stuck at local
optimum value but exhibits faster convergence characteristic,
The generator cost coefficients with their limits, line data and
whereas the APO method reaches the near optimal solution
bus data are taken from [32,36]. The total wind farm of capacity
Table 1
Comparison of fuel cost obtained by proposed hybrid algorithm and other reported algorithms.
807
PSO
806
APO
HPSO-APO
805
Fuel Cost ($/h)
804
803
802
801
0 10 20 30 40 50 60 70 80 90 100
Number of Iterations
Fig. 4. IEEE 30-bus system: Convergence plot for proposed HPSO-APO, PSO and APO methods.
K. Teeparthi, D.M. Vinod Kumar / Engineering Science and Technology, an International Journal 20 (2017) 411–426 417
7.5 HPSO-APO
APO
7 PSO
Los s (MW )
6.5
5.5
4.5
670 680 690 700 710 720 730 740 750
Total production cost ($/h)
Fig. 5. IEEE 30-bus system: Pareto optimal fronts for case 1 (base case).
Table 2
IEEE 30-bus system: Contingency Ranking. HPSO-APO and standard PSO, APO algorithms. The optimal values
of active power generations with best optimal objectives and com-
Line outage (between buses) Performance Index Rank
promise solution (comp. sol) for the base case are tabulated in
6–8 497.30 1 Table 3. This table gives the comparison of optimal values for total
6–28 230.20 2
28–27 131.21 3
production cost and active power losses, and their compromising
10–21 82.58 4 solutions by the proposed and standard methods. Fig. 5 shows
4–12 62.73 5 the Pareto optimal fronts obtained by the implementation of
HPSO-APO, PSO and APO methods to the base case of MO-SCOPF
problem. It is clear that the proposed HPSO-APO approach gives
(minimum fitness value) but prone to premature convergence. The the well distributed Pareto front than other two reported standard
proposed hybrid optimization method superior and robust com- methods.
pared to APO and PSO algorithms in terms of getting near global The contingency ranking has been done based on the perfor-
optimum point and faster convergence. mance index, which is determined by Newton- Raphson load flow
The proposed HPAO-APO algorithm is tested for base case as (NRLF) method. Only the line outages are considered in this work.
well as (N-1) contingency cases with simultaneously considering The thermal and voltage limits are considered as 120% and ±5% of
different objective functions and its performance is compared with the base case values respectively. The top five severe contingencies
the standard PSO and APO algorithms. To verify the effectiveness of are tabulated in Table 2 in descending order. In this paper, only a
the proposed algorithm of solving MO- SCOPF, three different case top severe contingency case is presented in detail. After security
studies are considered as follows. assessment, it is noticed that the outage of line connected between
the buses 6–8 causes severe contingency condition in the network
case 1. Case 1: Minimization of total production cost (TPC) and [32]. The proposed HPSO-APO algorithm is implemented to solve
active power losses simultaneously MO-SCOPF problem for taking preventive control actions like gen-
case 2. Case 2: Minimization of total production cost (TPC) and erator rescheduling to regain the system’s normal state with satis-
security index (SI) simultaneously fying both equality and inequality constraints. The optimal active
case 3. Case 3: Minimization of total production cost (TPC), power generations along with the best objective and compromis-
security index (SI) and active power losses ing solution vectors by the proposed and other reported methods
simultaneously are shown in Table 4. Fig. 6 shows the Pareto optimal front
obtained by the proposed HPSO-APO, PSO and APO methods. It is
5.1.1. Case 1: Minimization of total production cost and active power revealed that the proposed method gives better diverse Pareto
losses simultaneously optimal solutions than the other reported methods. The CPU time
In this case, the conflicting objectives like total production cost taken for completion of 100 iterations of each algorithm is
and active power losses were considered for the MO-SCOPF prob- listed in Table 5. Hence, the proposed HPSO-APO method can solve
lem. This objective function was solved by the proposed MO- SCOPF problem effectively with consideration of both wind
Table 3
Base case: Minimization of total production cost and active power losses (Case 1).
Table 4
Contingency case: Minimization of total production cost and losses (case 1).
8.5
8 Hybrid PSO-APO
APO
7.5 PSO
7
Loss (MW)
6.5
5.5
4.5
4
680 700 720 740 760 780 800
Total production cost ($/h)
Fig. 6. IEEE 30-bus system: Pareto optimal fronts for case 1 (contingency case).
Table 5 (comp. sol) vectors for the base case is shown in Table 6. After
Comparison of CPU times for case 1. security enhancement, the compromising solution vectors
Algorithm MOPSO MOAPO Proposed MOHPSO-APO obtained by using the proposed and other reported methods are
CPU time (s) 33.24 30.62 30.26
listed in Table 7. The optimal trade-off curves obtained for base
case as well as contingency case by using the proposed HPSO-
APO, PSO and APO methods are shown in Figs. 7and 8. These
graphs show that the proposed method gives well diverse Pareto
and thermal generators and also minimize the fuel consumption
optimal solutions than PSO and APO methods. The CPU time taken
and shorten carbon emission impact on environment.
for completion of 100 iterations for each algorithm is listed in
Table 8.
5.1.2. Case 2: Minimization of total production cost and security index
(SI) simultaneously
In this case, total production cost and security index were con- 5.1.3. Case 3: Minimization of total production cost, security index (SI)
sidered as the multi objective function for SCOPF problem. The pro- and active power losses simultaneously
posed HPSO-APO, standard PSO and APO algorithms have been In the present case, all three conflicting objectives were opti-
implemented for the MO-SCOPF problem. The active power gener- mized simultaneously by using the proposed HPSO-APO and other
ations along with optimal objective and compromise solution reported methods. The best compromising solution vectors for
Table 6
Base case: Minimization of total production cost and security index (Case 2).
Table 7
Contingency case: Minimization of total production cost and security index (Case 2).
1.8
1.6
APO
PSO
1.4 Hybrid PSO-APO
Security Index
1.2
0.8
0.6
0.4
670 680 690 700 710 720 730 740 750 760 770 780
Total production cost ($/h)
Fig. 7. IEEE 30-bus system: Pareto optimal fronts for case 2 (base case).
1.4
1.3 PSO
APO
1.2 dHybridPSO-APO
1.1
Security Index
0.9
0.8
0.7
0.6
0.5
680 690 700 710 720 730 740 750 760 770 780
Total production cost ($/h)
Fig. 8. IEEE 30-bus system: Pareto optimal fronts for case 2 (contingency case).
Table 8 ods. The CPU time taken for completion of 100 iterations for each
Comparison of CPU times for case 2. algorithm is listed in Table 11.
Algorithm MOPSO MOAPO Proposed MOHPSO-APO
CPU time (s) 33.34 31.11 30.22
5.2. Indian 75-bus system
Table 9
Base case: Minimization of total production cost, losses and security index (Case 3).
Table 10
Contingency case: Minimization of total production cost, losses and security index (Case 3).
9 Hybrid PSO-APO
PSO
8 APO
Loss (MW)
5
2
4
1.5
3 1
660 680 700 Security Index
720 740 760 780 800 820 0.5
840 860
Total fuel cost ($/h)
Fig. 9. IEEE 30-bus system: Pareto optimal fronts for case 3 (base case).
demand of the system is 5568.12 MW. The multi-objective security superior and robust compared to APO and PSO algorithms in terms
constrained based OPF is solved by the proposed and other of getting near global optimum point and faster convergence.
reported methods as outlined in the following subsections.
To validate the robustness and effectiveness of the proposed 5.2.1. Case 1: Minimization of total production cost and active power
hybrid PSO-APO algorithm, first, we considered single objective losses simultaneously
(fuel cost of thermal generators) as an objective function. The com- The proposed HPSO-APO method is employed to solve MO-
parison of minimum fuel cost obtained by the proposed hybrid SCOPF problem for base case as well as (N-1) contingency cases.
algorithm along with existing PSO and APO algorithms has been The contingency ranking has been done based on the performance
reported in Table 12. From the results, the minimum fuel cost index, which is determined by NRLF method. The top five severe
obtained by the proposed method is 8826.04 $/h, which is better contingencies are listed in Table 13. After the security assessment,
than standard individual PSO and APO algorithms. The conver- it is noticed that outage of the line connected to the buses 16–46
gence characteristics are shown in Fig. 11. The results reveal that causes severe contingency in the network [32]. The optimal real
PSO algorithm stuck at local optimum value but exhibits faster power generator settings obtained by the proposed HPSO-APO
convergence characteristic, whereas the APO method reaches the for considering two different objective functions i.e., total produc-
near optimal solution (minimum fitness value) but prone to pre- tion cost and active power losses of both base and contingency
mature convergence. The proposed hybrid optimization method cases are listed in Tables 14 and 15. The results are compared with
K. Teeparthi, D.M. Vinod Kumar / Engineering Science and Technology, an International Journal 20 (2017) 411–426 421
10
700
3
750 2.5
2
800 1.5
1
0.5
Total production cost ($/h) 850 0
Security Index
Fig. 10. IEEE 30-bus system: Pareto optimal fronts for case 3 (contingency case).
Table 11 Table 13
Comparison of CPU times for case 3. Indian 75-bus system: Contingency Ranking.
Algorithm MOPSO MOAPO Proposed MOHPSO-APO Line outage (between buses) Performance Index Rank
CPU time (s) 34.34 33.02 32.38 16–46 17.36 1
22–25 10.80 2
16–50 5.39 3
Table 12
Comparison of fuel cost obtained by proposed hybrid algorithm and other reported 29–30 5.16 4
algorithms. 35–36 4.97 5
11000
PSO
APO
10500
HPSO-APO
Fuel Cost ($/h)
10000
9500
9000
8500
0 10 20 30 40 50 60 70 80 90 100
Number of Iterations
Fig. 11. IEEE 30-bus system: Convergence plot for proposed HPSO-APO, PSO and APO methods.
422 K. Teeparthi, D.M. Vinod Kumar / Engineering Science and Technology, an International Journal 20 (2017) 411–426
Table 14
Base case: Minimization of total production cost and active power losses (Case 1).
Table 15
Contingency case: Minimization of total production cost and active power losses (Case 1).
190
185 HPSO-APO
APO
PSO
180
Loss (MW)
175
170
165
7800 8000 8200 8400 8600 8800 9000 9200 9400
Fig. 12. Indian 75-bus system: Pareto optimal fronts for case 1 (base case).
K. Teeparthi, D.M. Vinod Kumar / Engineering Science and Technology, an International Journal 20 (2017) 411–426 423
195
HPSO-APO
PSO
190 APO
185
Loss (MW)
180
175
170
7600 7800 8000 8200 8400 8600 8800 9000
Total production cost ($/h)
Fig. 13. Indian 75-bus system: Pareto optimal fronts for case 1 (contingency case).
Table 17
Base case: Minimization of total production cost and security index (Case 2).
Table 18
Contingency case: Minimization of total production cost and security index (Case 2).
14
13 HPSO-APO
APO
PSO
12
Security Index
11
10
8
8000 8100 8200 8300 8400 8500 8600 8700 8800 8900 9000
Total production cost ($/h)
Fig. 14. Indian 75-bus system: Pareto optimal fronts for case 2 (base case).
15
PSO
14 HPSO-APO
APO
13
Security Index
12
11
10
8
7500 7700 7900 8100 8300 8500 8700 8900 9100 9300 9500
Total production cost ($/h)
Fig. 15. Indian 75-bus system: Pareto optimal fronts for case 2 (contingency case).
6. Conclusions
APO are shown in Figs. 14 and 15. These graphs show that the best
Pareto optimal front is obtained through the proposed approach.
In this paper, three different conflicting objective functions viz.
The CPU time taken for completion of 100 iterations for each algo-
total production cost, active power losses and security index were
rithm is listed in Table 19.
considered for MO-SCOPF problem. The three multi-objective func-
tions have been solved simultaneously using the proposed HPSO-
5.2.3. Case 3: Minimization of total production cost, security index (SI) APO, PSO and APO methods. Non-dominated sorting and crowding
and active power losses simultaneously distance approach is used to maintain a diverse in Pareto optimal
The best compromising solution vectors obtained by using the points. Finally, a fuzzy membership approach is used to get com-
proposed algorithm along with PSO and APO methods for both promising solution over the trade-off curve. The proposed opti-
base and contingency cases are tabulated in Tables 20 and 21. mization algorithm integrates strength of the both the algorithms
Three conflicting objectives were optimized simultaneously by to get a balance between exploration and exploitation. The validity
using the proposed HPSO-APO method, PSO and APO methods. of the proposed algorithm is demonstrated on IEEE 30-bus and
For both the cases, the Pareto optimal solutions obtained by the practical Indian 75-bus systems. The test results show the pro-
proposed hybrid, PSO and APO methods are shown in Figs. 16 posed HPSO-APO method gives well distributed Pareto optimal
and 17. The CPU time taken for completion of 100 iterations for solutions than standard PSO and APO methods for MO-SCOPF
each algorithm is listed in Table 22. It is clear that the proposed problem of base and contingency cases with different objectives.
K. Teeparthi, D.M. Vinod Kumar / Engineering Science and Technology, an International Journal 20 (2017) 411–426 425
Table 20
Base case: Minimization of total production cost, losses and security index (Case 3).
Table 21
Contingency case: Minimization of total production cost, losses and security index (Case 3).
240 HPSO-APO
APO
220 PSO
Los s (MW )
200
180
160
8000
8500
9000
9500 22
20
10000 16 18
12 14
Total production cost ($/h) 10500 10
6 8
Security Index
Fig. 16. Indian 75-bus system: Pareto optimal fronts for case 3 (base case).
426 K. Teeparthi, D.M. Vinod Kumar / Engineering Science and Technology, an International Journal 20 (2017) 411–426
APO
HPSO-APO
PSO
18
200
Loss (MW)
16
14
180
12
160 10
7500 8000 8500 8 Security Index
9000 9500 10000
10500 11000 6
11500 12000
Total production cost ($/h)
Fig. 17. Indian 75-bus system: Pareto optimal fronts for case 3 (contingency case).