 Open Access
 Total Downloads : 118
 Authors : S. Brahma Reddy, A. Anjaneyulu
 Paper ID : IJERTV4IS110473
 Volume & Issue : Volume 04, Issue 11 (November 2015)
 DOI : http://dx.doi.org/10.17577/IJERTV4IS110473
 Published (First Online): 27112015
 ISSN (Online) : 22780181
 Publisher Name : IJERT
 License: This work is licensed under a Creative Commons Attribution 4.0 International License
Multiobjective Optimization using NonDominated Sorting Improved Particle Swarm Optimization
S.Brahma Reddy
EEE Department
NOVA College of Engg and Technology, Jangareddygudem A.P, India
Abstract: This paper mainly concentrates in finding enhanced optimal solution for MultiObjective Problem (MOP) formulated using generation fuel cost, emission, and loss objectives. Improved Particle swarm optimization (IPSO) is proposed to select best value as compared with existing evaluation algorithms. Optimizing multiple objectives simultaneously and selecting a best compromised solution as per the requirements of decision maker needs an application of MOP along with fuzzy decision making tool. The proposed Non dominated Sorting Improved Particle swarm optimization (NSIPSO) is tested on IEEE 30 bus test system and corresponding results are analyzed.
Key words: Multi object, fuel cost, emission, total power loss, nondominated sorting, IPSO.

INTRODUCTION
The aim of optimal power flow is to determine the optimal combination of real power generation, voltage magnitudes, compensator capacitors and transformer tap position to minimize the specific objective function like total generation cost in power systems. The mentioned conditions make the OPF problem a large scale nonlinear constrained optimization problem [1].
A.Anjaneyulu
Assistant Prof`essor, EEE Department NOVA College of Engg and Technology,
Jangareddygudem A.P, India
studies on evolutionary algorithms have revealed that these methods can be efficiently used for solving the multi objective optimization problem, some of these algorithms are multiobjective evolutionary algorithm [7], strength Pareto evolutionary algorithm (SPEA) [8], nondominating sorting genetic algorithm (NSGA) [9] and multiobjective PSO algorithm [10]. Since these algorithms are population based techniques, multiple Paretooptimal solutions can be found in one program run.
In this paper the proposed NSIPSO algorithm is used for solving multi objective optimization problems and tested with standard IEEE test systems compared with existing methods. The result shows proposed method gives the best compared with existing literature.

GENERAL OPF PROBLEM FORMULATION
The standard OPF problem can be written in the following form
Single objective optimization
(1)
The literature on OPF is vast and [5] presented the major contributions in this area, where a review of literature is done on Optimal Power Flow up to 1993. Dommel and Tinney [6] has given a practical method for solving the power flow program with control variables such as real and reactive power and transformer ratios automatically adjusted to minimize instantaneous costs or losses.
There are several techniques that have been considered in the literature to solve multiobjective problems. One of these methods is reducing the multiobjective problem into a single objective problem by considering one objective as a target and others as a constraint. Another strategy is combining all objective functions into one objective function. The above strategies have some weak points such as the limitation of the available choices and their priori selection need of weights for each objective function. Besides the above drawbacks, finding just one solution for the multiobjective problem is known as the most important weak point of these strategies. Over the past few years, the
Subject to: = 0 and 0
Where,
is vector of state variables, is vector of control variables,
= Reactive power supplied by all shunt reactors, = Transformer load tap changer magnitudes, Voltage magnitude at PV buses, = Active power generated at the PV buses, Voltage magnitude at PQ buses, = Voltage angles of all buses, except the slack bus, =
Active generating power of the slack bus, = Reactive
power of all generator units, and is the vector of control variables, the control variable can be generated active and reactive power, generation bus voltage magnitudes, transformer taps etc.
Multi objective Problem formulation
Let be objective
functions defined over dimensional search space. A multi
objective optimization problem can then be formulated as [18]:
(2)
Subjected to the constraints, this will give a set of Pareto optimal solutions. A decision vector, (a set of control parameters) is said to be Pareto optimal, if there is no other decision vector, dominating with respect to the set of objective functions. The decision vector is said to strictly
dominate the another vector y (denoted by ) if;
for at least one i.

OBJECTIVES FORMULATION
The three considered objective functions are described as follows
Objective1: Generation cost
The generation cost function can be mathematically stated as follows [3].
(3)
where is the total fuel cost , , , are fuel cost coefficients of the unit, is the real power
generation of the unit, is the voltage magnitude of the generator, is the tap of transformer,
is the reactive power of the compensator capacitor,
is the total number of generation units, is the number of tap transformer and is the number of the compensation capacitor.
Objective 2: Emission
The emission function can be presented as the sum of all types of emissions considered, such as , ,
thermal emission, etc. In the present study, two important types of emission gases are taken into account. The amount
of and emission is given as a function of generator output that is the sum of a quadratic and exponential function as follows.
(4)
Where is the total emission ( ), are the emission coefficients of the unit.
Objective 3: Transmission loss
The power flow solution gives all bus voltage magnitudes and angles. Then, the active power loss in transmission line can be computed as follows.
(5)
Where is the total transmission loss (MW), is the number of transmission lines, and are the bus voltage angles at the two ends of the line, and
are bus voltage amplitudes at the two ends of the line and is the conductance of the line.

CONSTRAINTS

Equality constraints:
The OPF equality constraints reflect the physics of the power systems. Equality constraints are expressed in the following equations
Where

Inequality constraints:
The inequality constraints of the OPF reflect the limits on physical devices in the power system as well as the limits created to ensure system security. They are presented in the following.
Where is the number of load bus and is the power that flows between bus i and bus j.
are the maximum and minimum valid voltages for bus. is the maximum power flow
through the branch. and are the maximum and minimum active power values of the bus, respectively. and are the maximum and minimum reactive power values of the bus.


SINGLE OBJECTIVE IPSO BASED OPF An algorithm for single objective IPSO based OPF is
given bellow.
Algorithm
Step 1: Initialize the population and PSO parameters.
Step 2: Read the input system data and select the PSO control variables.
Step 3: Randomly generate the velocities and populations. Step4: Update the bus and line datas of given system according to the population generation and run the NR load flow.
Step 5: After load flow calculation check the equality and inequality constraints; if any violets add the penalty terms to the objective function.
Step 6: Compute the objectivefunction and fitness values. Step 7: Do the same process of step4 and step5 for all populations and select the best fitness value as globalfit value and corresponding particles are Gbest values
Step 8: Initialize the iteration counter Iter, and start the iteration process
Step 9: Update the velocities and position values, check the updated velocities and positions within limit or not. Fix those values min or max according to their violation.
Step 10: Repeat the step4 to step6 for all populations. Step 11: Update the localbest and globalbest values. Step 12: Repeat the step9 to step11 until Iter < IterMax. Step 13: Stop the process and print the Gbest values.

MULTI OBJECTIVE NSIPSO BASED OPF
An algorithm and flow chart for multi objective NSIPSO based OPF is given bellow.
Algorithm
Step 1: Initialize the random population and velocities.
Step 2: Update the system data according to the population generation.
Step 3: Run the load flow solution for updated system.
Step 4: Check the equality and inequality constraints and calculate penalty terms.
Step 5: Select the optimized objectives and calculate their objective function values and fitness value. Save these values in a repository.
Step 6: Initialize the Pbest values and found the Gbest value.
Step 7: Start the iteration process and update the velocities and positions, check their limits and fix the values, Repeat the process step3 to step5 generate new population.
Step 8: The new populations add with old repository and there by apply nondominate sorting technique.
Step 9: After sorting, save the non dominated pareto fronts and apply the crowding distance and crowding sort techniques.
Step 10: After non dominated set re arrangement select top 10% values are Gbest values and update Pbest values.
Step 11: Repeat the process step7 to step10 until maximum iteration value. And stop the iteration process
Step 12: Finally we get the non dominated repository; it is apply to fuzzy decision maker tool and calculated the best feasible compromised multi objective solution according to their weighting factors.

RESULT AND ANALYSIS
In this paper, the multi objective OPF solution using Non Dominated Sorting IPSO (NSIPSO) is given. The proposed method is tested on IEEE 30 bus test systems with bus voltages, real and reactive power and line flow constraints. On all optimization runs, the IPSO population size and the maximum number of iterations which are set on 100 each considered.
The IEEE30 bus system is used throughout this work to test the proposed algorithm (IPSO based OPF for single objective and NSIPSO based OPF for multi objective). This system consists of 6 generator units as well as 41 transmission lines. The detailed bus and line parameters are presented in [22].
In this section we describe the single and multi objective analysis.

Single objective minimization Case Study 1: Fuel cost minimization
In this case, developed algorithm is applied to minimize the fuel cost objective. The obtained results are compared with Ant Colony Algorithm (ACA) are tabulated in table 1. The convergence pattern is shown in Fig 1.
Table 1: Fuel cost minimization
Control variable
Existing ACA method [3]
Proposed method
PG1, MW
181.945
176.6491
PG2, MW
47.001
48.83991
PG3, MW
20.553
21.52671
PG4, MW
21.146
21.73635
PG5, MW
10.433
12.16658
PG6, MW
12.173
12
Fuel Cost ($/h)
802.578
802.4029
Generation Cost ($/h)
806
805
804
803
GENERATION COST MINIMIZATION (LIMITED CONTROL VARIABLES)
proposed method is 0.204838 ton/h. It is clear that the proposed method can achieve better result when compared to GA method. In Fig 2, (a) and (b) shows the emission variation and the fitness variation with respect to number of iterations respectively. The final solution of the proposed method is converged with in 15 iterations.
EMISSION MINIMIZATION (LIMITED CONTROL VARIABLES)
Emission (ton/h)
0.212
802
0 10 20 30 40 50 60 70 80 90 100
0.21
1.246
Fitness
1.244
3
x 10
Iterations
Fig 1 (a)
0.208
0.206
0.204
0 10 20 30 40 50 60 70 80 90 100
1.242
1.24
0 10 20 30 40 50 60 70 80 90 100
0.83
Iterations
Fig 2 (a)
Iterations
Fig 1 (b)
Fitness
0.828
Fig 1 Fuel cost convergence pattern
It can be easily seen from the Table 1, the fuel cost with existing ACA [3] method is 802.578 $/h and with the
0.826
0 10 20 30 40 50 60 70 80 90 100
proposed method is 802.4029 $/h. It is clear that the proposed method can achieve better result when compared to ACA
Iterations
Fig 2 (b)
method. In Fig 1, (a) and (b) shows the fuel cost variation and the fitness variation with respect to number of iterations respectively. The final solution of the proposed method is converged with in 15 iterations.
Case Study 2: Emission minimization
In this case, developed algorithm is applied to minimize the emission objective. The obtained results are compared with Genetic Algorithm (GA) are tabulated in Table 2. The convergence pattern is shown in Fig 2.
Table 2: Emission minimization
Control variable
Existing GA
method [3]
Proposed
method
PG1, MW
69.73
64.32621
PG2, MW
67.84
67.76814
PG3, MW
49.73
50
PG4, MW
34.42
35
PG5, MW
29.15
30
PG6, MW
39.29
40
Emission (ton/h)
0.2072
0.204838
It can be easily seen from the Table 2, the emission with existing GA [3] method is 0.2072 ton/h and with the
Fig 2 Emission convergence pattern
Case Study 3: Transmission loss minimization
In this case, developed algorithm is applied to minimize the transmission loss objective. The obtained results are compared with Genetic Algorithm (GA) are tabulated in Table 3. The convergence pattern is shown in Fig 3.
Table 3: Transmission loss minimization
Control variable
Existing GA
method [3]
Proposed method
VG1, p.u.
1.03
1.1
VG2, p.u.
1.00
1.07135
VG3, p.u.
1.00
1.06827
VG4, p.u.
1.02
1.0735
VG5, p.u.
1.04
0.95708
VG6, p.u.
1.00
1.03229
T69, p.u.
1.00
1
T610, p.u.
1.01
1.08182
T412, p.u.
1.00
1.1
T2728, p.u.
1.04
1.03477
Transmission Loss, MW
5.3513
4.97153
It can be easily seen from the Table 3, the transmission loss with existing GA [3] method is 5.3513 MW and with the proposed method is 4.97153 MW. It is clear that the proposed method can achieve better result when compared to GA method. In Fig 3, (a) and (b) shows the transmission loss variation and the fitness variation with respect to number of iterations respectively. The final solution of the proposed method is converged with in 15 iterations.
TRANSMISSION LOSS (MW
TRANSMISSION LOSS MINIMIZATION (LIMITED CONTROL VARIABLES)
5.2
5.1
5
4.9
From table 4 it is observed that cost is 800.17746
$/h, emission is 0.0204683 ton/h and total power loss is 2.99099MW.

Multi Objective minimization
The results are obtained from the developed algorithm for multiobjective OPF based on NSIPSO method which has been discussed in the above sections. The multi objective OPF problem has been formulated with different combinations of objectives namely fuel cost emission, fuel costlosses and emissionloss combinations are considered.
In this, the proposed methodology handles two different objectives together as multiobjective optimization problem. There are ten possible combinations with the five objectives. The obtained results are compared with existing weighted sum method which is given in Table 5.
0.175
Fitness
0.17
0.165
0.16
0 10 20 30 40 50 60 70 80 90 100
Iterations
0 10 20 30 40 50 60 70 80 90 100
Iterations
Fig 3 Transmission loss convergence pattern
The obtained results for single objective OPF
Table 5: Multiobjective obtained best compromised results for two different objectives
problem based on PSO algorithm is given in Table 4.
Control variables
Fuel cost
minimization
Emission minimization
Loss
minimizatio n
PG1, MW
177.22929
64.00868
51.39099
PG2, MW
48.550303
67.59438
80
PG3, MW
21.462934
50
50
PG4, MW
21.211045
35
35
PG5, MW
11.881975
30
30
PG6, MW
12.000032
40
40
VG1, p.u.
1.1
1.092719
1.1
VG2, p.u.
1.0370108
1.082577
1.041686
VG3, p.u.
1.0646606
1.057189
1.083148
VG4, p.u.
1.0544999
1.068489
1.087906
VG5, p.u.
0.9634969
0.944209
1.099556
VG6, p.u.
1.1
1.093477
1.1
T69, p.u.
0.9514214
1.015055
1.017291
T610, p.u.
0.9910521
0.9562
0.968865
T412,pu
0.9919611
0.994948
0.983142
T2728, p.u.
0.9679805
0.966505
0.970435
Qc10,MVA
15.974439
17.78494
21.07306
Qc24,MVA
10.460198
17.53809
11.67689
Fuel cost ($/h)
800.17747
944.3457
967.4024
Emission (ton/h)
0.3664768
0.204683
0.207122
Power Loss (MW)
8.9355744
3.203066
2.99099
Table 4: PSO based OPF Solutions for five different objectives individually
Optimized Two Objectives
Weighting factors
Proposed
method
W1
W2
Cost
Emission
($/h)
(ton/h)
Cost
0.8
0.2
805.9989
0.311993
Emission
0.5
0.5
830.0619
0.251936
0.2
0.8
880.9416
0.217372
W1
W2
Cost ($/h)
Loss (MW)
CostLoss
0.8
0.2
809.8782
6.951288
0.5
0.5
824.0478
5.695692
0.2
0.8
860.88
4.573103
W1
W2
Emission (ton/h)
Loss (MW)
Emission
Loss
0.8
0.2
0.204742
3.120038
0.5
0.5
0.205361
3.073873
0.2
0.8
0.206237
3.039198
From the table 5 it is observed that more waited objective will minimize more. It is also observed that cost is 805.9989 $/h and emission is 0.217372 ton/h for the weight is 0.8 in costemission combination, cost is 809.8782 $/h and loss is 4.573103 MW for the weight is 0.8 in costloss combination and emission is 0.204742 ton/h and loss is 3.039198MW for the weight is 0.8 in emissionloss combination,
The alignments of the generated two dimensional Pareto solutions are shown Fig 4.
Fig 4: Twodimensional Paretooptimal fronts for two different objectives


CONCLUSION
In this paper proposes an optimal power flow technique with three competitive objectives, cost of generation, emission and loss of thermal plants. A multi objective Non Dominating sorting improved particle swarm optimization technique has been proposed to solve this optimization problem. To maintain diversity among Pareto optimal solutions a Non Dominating sorting technique has been proposed. The goal of the proposed multiobjective OPF problem is to compute advised set points for power system controls that satisfy the security, the environment and the economical conditions simultaneously. The most important privilege of the proposed approach for the multiobjective formulation is obtaining several nondominated solutions allowing the system operator to use his personal preference in
selecting one solution for implementation. Furthermore the proposed fuzzy decision method helps the power system operator to apply his decisions very easily. Also in single objective cases, the proposed approach can obtain better results with respect to other algorithm in the literature. In multiobjective cases, the proposed method proves its ability to obtain welldistributed Pareto fronts. The IEEE 30 bus systems are considered for experimentation.
REFERENCES

Niknam, T., Narimani, M.R., Jabbari, M., Malekpour, A.R.: A modified shuffle frog leaping algorithm for multiobjective optimal power flow, Energy, 2011, 36, pp. 64206432.

Momoh J. A., M. E. ElHawary and Adapa R., A review of selected optimal power flow literature to 1993, IEEE Transactions on Power System, Part I and II, Vol. 14, pp. 96111, 1999.

Dommel H. W. and Tinney W. F., Optimal power flow solutions, IEEE ransactions on Power Apparatus and Systems, Vol. PAS87, pp. 18661876,1968.

Andrew, K., Haiyang, Z.: Optimization of wind turbine energy and power factor with an evolutionary computation algorithm, Energy, 2010, 35, (3), pp. 13241332.

Abido, M.A.: A novel multi objective evolutionary algorithm for environmental /economic power dispatch, Electr. Power Syst. Res., 2003, 65, (1), pp. 7181.

Elbeltagi, E., Hegazy, T., Grierson, D.: Comparison among five evolutionarybased optimization algorithms, Adv. Eng. Inf., 2005, 19, (1), pp. 4353.

Niknam, T., ZeinoddiniMeymand, H., DoagouMojarrad, H.: A practical multi objective PSO algorithm for optimal operation management of distribution network with regard to fuel cell power plants, Renew.Energy, 2011, 36, (5), pp. 15291544.

J. Hazra, A. K. Sinha, A multiobjective optimal power flow using particle swarm optimization, Euro. Trans. Electr. Power(2010).

T. Niknam, M.R. Narimani, J. Aghaei R. AzizipanahAbarghooee, Improved particle swarm optimisation for multiobjective optimal power flow considering the cost, loss, emission and voltage stability index, IET Gener. Transm. Distrib., 2012, Vol. 6, Iss. 6, pp. 515 527.

Alsac, O., Stott, B.: Optimal load flow with steadystate security,IEEE Trans. Power Appar. Syst., 1974, 93, (3), pp. 745751.