Quantum Particle Swarm Optimization for

Multiobjective Combined Economic Emission

Dispatch Problem Using Cubic Criterion Function
Fahad Parvez Mahdi, Pandian Vasant M. Abdullah-Al-Wadud
Department of Fundamental and Applied Sciences Department of Software Engineering
Universiti Teknologi PETRONAS King Saud University
Bandar Seri Iskandar, Perak, Malaysia Riyadh, Saudi Arabia
[email protected], [email protected] [email protected]

Md. Mushfiqur Rahman Junzo Watada, Vish Kallimani

Department of Electrical and Electronic Engineering Department of Computer and Information Sciences
University of Dhaka Universiti Teknologi PETRONAS
Dhaka, Bangladesh Bandar Seri Iskandar, Perak, Malaysia
[email protected] [email protected], [email protected]

Abstract—In this research, quantum particle swarm to solve this multiobjective problem were goal programming
optimization (QPSO) is utilized to solve multiobjective [1], Newton-Raphson [2] and Lagrangian relaxation [3].
combined economic emission dispatch (CEED) problem Nanda et al. [1] explored coordination equation based
formulated using cubic criterion function considering a uni classical method to solve multiobjective CEED problem
wise max/max price penalty factor. QPSO is implemented on a considering line flow constraints. However, nonlinearities in
6-unit power generation system and compared with the actual power generating systems and many real-world
Lagrangian relaxation, particle swarm optimization (PSO) and complicating factors in multiobjective CEED problem make
simulated annealing (SA). The obtained results verified the it hard for the classical models to solve this problem
effectiveness and demonstrate the robustness of QPSO method.
efficiently.
This research suggests that QPSO can be used as an effective
and robust tool in other power dispatch problems. Intelligent techniques have emerged as a replacement of
the classical methods to solve the multiobjective CEED
Keywords-quantum particle swarm optimization; combined problem. Various stand-alone and hybrid intelligent
economic emission dispatch; cubic function; penalty factor; algorithms have been developed and exploited to get quality
multiobjective optimization solution with high computational efficiency. Genetic
algorithm (GA) [4], PSO [5], SA [6], differential evolution
I. INTRODUCTION (DE) [7] and their variants are some of the mostly used
In electrical power generation systems, economic load methods in this power dispatch problem. Hybrid techniques
dispatch (ELD) is one of the most crucial problems. The goal have been developed in order to combine the benefits of two
of ELD is to find an optimal combination of power or more algorithms and diminish each other’s limitations.
generation in order to minimize the generation cost by Many well-known methods have been hybridized to solve
satisfying the load demand and all other practical equal and CEED problem in a powerful way. Some of them are
unequal constraints. On the other hand, due to environmental differential evolution with biogeography-based optimization
pollution from the fossil fuels fired thermal plant, ELD as a (DE-BBO) [8], PSO with GA (PSO-GA) [9] and artificial
single objective is not sufficient to be considered alone. bee colony with simulated annealing (ABC-SA) [10].
Thus, reduction of emission from thermal plants is also a Recently, the idea of quantum computing (QC) is being used
crucial objective in power dispatch. Combined economic with existed intelligent techniques to develop
emission dispatch (CEED) is a multiobjective optimization computationally efficient and powerful algorithm. One of the
problem where the goal is to minimize the generation cost advantages of QC is that it can show massive quantum
and at the same time emission of pollutants from thermal parallelism, which makes it computationally very powerful
power generation system. and efficient. Some previous works regarding solving CEED
problem using quantum computational intelligence technique
Many techniques have been proposed to obtain an can be found in [11]. In this paper, QC integrated PSO
optimal solution with sound computational efficiency in technique (QPSO) is used to solve multiobjective CEED
multiobjective CEED problem. Some of the early approaches problem using cubic criterion function.

978-1-5090-6004-7/17/$31.00 ©2017 IEEE

 The following section of this paper includes there are many equal and unequal constraints that need to be
mathematical formulation of multiobjective CEED problem considered in order to optimize the actual scenario of the
with constraints. The next section presents a brief discussion system. Power balance and generator limit constraints are
of the proposed method. Result and analysis section shows two most important constraints, which are considered in this
the obtained results with comparative analysis of the work. The constraints are discussed in the following
proposed method as compared to other approaches for subsections.
solving CEED problem using cubic criterion function.
Finally, a concluding section highlights the achievements A. Power Balance Constraint
and drawbacks of the proposed method along with future
Total output power generation (in MW) must satisfy
research directions.
total load demand (in MW). Thus, total output power must
be equal to the summation of total load demand and total
II. PROBLEM FORMULATION power loss (in MW). It can be define as
Combined economic emission dispatch problem is a n
combination of both economic load dispatch and emission P Pi PD PL (5)
dispatch problem. In this paper, cubic criterion function is i 1
used instead of quadratic function to represent the CEED where P, PD and PL are total generated power, total load
problem. Cubic criterion function has been found to be more demand and total loss, respectively.
effective against the nonlinearities of actual power
generating system. Economic dispatch problem can be B. Generator Limit Constraint
defined as Each generating unit in a power generation system has
n
its upper and lower limits. The output of the generating unit
F ( P) ai Pi 3 bi Pi 2 ci Pi di (1) must lie within this limit in order to work properly. This
i 1
constraint can be defined as
Pi ,min Pi Pi ,max (6)
where F(Pi) is the generation cost (in $/hr) of generating unit
i when the output power is Pi; ai, bi, ci and di are the cost where Pi,min and Pi,max denote the minimum and maximum
coefficients of the generating unit i. limits, respectively, of generating unit i.
Emission dispatch problem can also be defined as a cubic III. METHODOLOGY
criterion function with four emission coefficients as
Before getting into the discussion of QPSO, it is
n necessary to understand the concept of PSO. The following
E ( P) ei Pi 3 fi Pi 2 gi Pi hi (2) subsections will introduce the concept of PSO and then
i 1 QPSO to understand how it operates to solve multiobjective
where E(Pi) is the emission (in Kg/hr) and Pi is the power CEED problem.
generated by unit i while ei, fi, gi and hi are the emission A. Particle Swarm Optimization (PSO)
coefficients. The objectives of minimizing generation cost
Kennedy and Eberhart [12] pioneered PSO technique in
and emission of pollutants can be converted into a single
objective using a price penalty factor. The max/max penalty 1995, which is based on the movement and intelligence of
factor is considered in this research to solve CEED problem. swarms. A group of particles makes up a swarm or
CEED problem with max/max penalty factor can be population. Particles fly over a search space of possible
described as solutions and record their best individual solutions. In each
step, a particle changes its position based on its velocity,
n n
and updates the velocity in accordance with its own
OF FT F ( Pi ) hi ,max/max E ( Pi ) (3) experience (the individual best (pbest) solution achieved by
i 1 i 1
it) and the experiences of the other particles (the global best
where OF stands for objective function (CEED), FT refers to (gbest) solution) achieved so far. In this way, the particles
total cost and hi,max/max is max/max penalty factor of continue to move and converge until they reach one of the
generating unit i. The max/max penalty factor can be defined stopping criteria of the given problem.
as
B. Quantum Particle Swarm Optimization (QPSO)
n n
QPSO, proposed and developed by Sun et al. [13], is an
hi ,max/max F ( Pi ,max ) / E ( Pi ,max ) (4) extension of PSO in the domain of quantum computing. The
i 1 i 1
concept of quantum bit (qubit) and rotation gate are
where Pi,max refers to the maximum amount of power (in introduced here to improve the characteristics of population
MW) that can be generated by the generating unit i. diversity. Furthermore, to avoid being trapped into any local
optima, the self-adaptive probability selection and chaotic
The goal of this paper is to minimize both the generation cost
sequences mutation are exploited. Quantum bit and angle
and emission of pollutant gases, i.e., the total cost, while
satisfying all other constraints. In power generation system, represent the state of a particle instead of the position and
the velocity of a particle as done in the basic PSO. Thus, reaches the stopping criteria, e.g., maximum number of
QPSO shows stronger search capability with powerful and iterations. Otherwise, the algorithm iterates again from the
fast convergence characteristics. second step.
The basic difference between qubit and classical bit is
that the later one can simultaneously stay in the
superposition of two different quantum states, Start

0 1 (8)
Initialize the Particles
In the above equation, α and β are complex numbers that
satisfy the equation
2 2 Compute Fitness of each particle
1 (9)
Spin up state is represented by |0> and spin down state is
Update pbest
represented by |1>. From (1), it can be seen that one qubit is
representing two state of information (|0> and |1>)
Calculate mbest
simultaneously. This superposition state can also be Gen = Gen + 1
expressed as
Update gbest
sin 0 cos 1 (10)
where the phase of the qubit is represented by θ. The
relation among α, β and θ can be defined as Update Position of the Particle
No
arctan( / ) (11)
The position of the particle in QPSO can be described as
Stopping
L 1 Criteria Met?
xid pid ln( ) (12)
2 u Yes
where xid is the position of the ith particle, pid is the local
Stop
attractor of particle i that lies in between the pbest and gbest
and u is uniformly distributed random number in the range Figure 1. Flowchart of standard Quantum PSO
of [0, 1]. The value of L can be determined using the
following equation IV. RESULT AND ANALYSIS
L 2 | xid pid | (13)
where α is the only parameter of QPSO, which can be QPSO has been exploited in this research to solve CEED
problem for a 6-unit power generation system using cubic
calculated using the following equation
criterion function, where the total load demand is 150 MW.
tmax t Authors of this paper have implemented QPSO in
(1 0.5). 0.5 (14)
tmax MATLAB R2015a and executed with core i5-3470 CPU @
3.20 GHz (4 CPUs), ~3.2GHz and 4GB RAM personal
where t and tmax refer to number of iteration and maximum computer. Table I shows the parameter settings of QPSO
number of iterations, respectively. Equation (12) can be and PSO. Total 100 runs have been considered as a fair test
rewritten as of robustness, and the average of the outcomes of these runs
are reported in this section.
1
xid pid | xid pid | ln( ) (15)
u TABLE I PARAMETER SETTINGS FOR QPSO
and the local attractor p can be represented as below
p . pbest (1 ).gbest (16) Parameters Values
where φ refers to a uniformly distributed random number. Population Size 1000
The range of φ is [0, 1]. Maximum iterations 100
Fig. 1 depicts the flowchart of QPSO. In the first step,
Number of runs 100
algorithm parameters such as population size, particle
dimension and maximum number of iterations are initialized. Dimension 6
The second step evaluates fitness value for each particle and
records pbest and gbest. In the next step, particles are
updated using QPSO algorithm formula. The algorithm Tables II and III show the generator limit, fuel
coefficients and emission coefficients of 6-unit power
terminates and gives the optimal output if the algorithm
generation system respectively. All the data are taken from
[14].
TABLE II MINIMUM AND MAXIMUM LIMITS OF THE OUTPUT POWERS
GENERATED BY THE 6 UNITS OF THE POWER GENERATION SYSTEM

Generating Pmin Pmax

Unit (MW) (MW)
1 50 200
2 20 80
3 15 50
4 10 50
5 10 50
6 12 40

Convergence graph of PSO and QPSO is presented in

fig. 2 to compare the convergence characteristics of the
above mentioned techniques using cubic criterion function. Figure 3. Total cost ($/hr) at each run of PSO as well as QPSO algorithm
It can be seen from the fig. 2 that QPSO shows better to solve the 6-unit CEED problem
convergence characteristics by showing very fast and
smooth convergence to the optimal point than PSO. . Fig. 3 shows the fuel costs for the 100 different runs in
the experiment. It can be seen that the fluctuations are very
less for QPSO (the standard deviation is ~0.164) than that in
the case of PSO (the standard deviation is ~2.579). The
obtained results also confirm that the solutions obtained by
QPSO are stable, and hence, reliable.
TABLE IV COMPARISON OF CEED SOLUTIONS (PD=150 MW)
CONSIDERING MAX-MAX PENALTY FACTOR

LAGRANGE
SA [16] PSO [16] QPSO
[14]
P1 50.65 50 50 50.00

P2 21.20 20.0009 20 20.00

P3 15.46 15.0001 15 15.00

P4 22.6846 20.6141 22.11 22.9

Figure 2. Convergence characteristics of PSO and QPSO for 6-unit CEED P5 21.3002 22.4910 20.6 20.04
dispatch problem
P6 21.1181 21.8940 22.31 22.03
TABLE III FUEL COEFFICIENTS AND EMISSION COEFFICIENTS FOR 6-UNIT Fuel Cost
2734.21 2702.7841 2701.796 2701.476
POWER GENERATION SYSTEM ($/h)
ai bi ci di ei fi gi hi Emission
2642.702 2607.4606 2593.1844 2583.6485
0.0010 0.0920 14.50 -136.00 0.0015 0.0920 14.0 -16.0 ($/h)
0.0004 0.0250 22.00 -3.50 0.0014 0.0250 12.5 -93.5 Total Cost
5007.941 4949.7958 4946.8 4944.5
($/h)
0.0006 0.0750 23.00 -81.00 0.0016 0.0550 13.5 -85.0
0.0002 0.1000 13.50 -14.50 0.0012 0.0100 13.5 -24.5
0.0013 0.1200 11.50 -9.75 0.0023 0.0400 21.0 -59.0
V. CONCLUSION
0.0004 0.0840 12.50 75.60 0.0014 0.0800 22.0 -70.0 In this paper, QPSO technique is successfully
implemented to solve multiobjective CEED problem. Here,
A comparison with three other available methods, CEED is represented using cubic criterion function. Unit
namely Lagrangian relaxation [14], PSO [15] and simulated wise max/max price penalty factor is considered to convert
annealing (SA) [16], is done to evaluate and verify QPSO both objectives into a single objective. Simulation results
method. Table V shows the costs incurred by all these verify the effectiveness of QPSO in solving this
approaches. It demonstrates that QPSO successfully multiobjective problem by achieving reliable, robust and
outperforms the other three methods in terms of minimizing suitable solutions with fast convergence characteristics. The
the total cost, which shows the better performance of QPSO obtained results are compared with other well-known
in solving the multiobjective CEED problem than the methods like Lagrangian relaxation, PSO and SA which
others. One of the problems with PSO as well as other demonstrate QPSO’s superiority over these methods to
existing algorithms is that it sometimes trapped into the solve the CEED problem. Small number of generating units
local optima [17]. QPSO can avoid such local minima in have been considered in this research due to the
many cases, which helps to improve the performance. unavailability of data for large power generation systems.
Future work will be directed to use QPSO to solve Point Effect," in Swarm, Evolutionary, and
manyobjective CEED problem. In the same time, other Memetic Computing, Pt I. vol. 8297, B. K.
well-known recent techniques like bat algorithm (BA) and Panigrahi, P. N. Suganthan, S. Das, and S. S. Dash,
cuckoo search (CS) algorithm shall be investigated along Eds., ed, 2013, pp. 354-365.
with quantum computing phenomena. [11] F. P. Mahdi, P. Vasant, V. Kallimani, and M.
Abdullah-Al-Wadud, "A review on economic
REFERENCES emission dispatch problems using quantum
[1] J. Nanda, D. Kothari, and K. Lingamurthy, computational intelligence," AIP Conference
"Economic-emission load dispatch through goal Proceedings, vol. 1787, p. 020002, 2016.
programming techniques," 1988. [12] J. Kennedy and R. Eberhart, "Particle swarm
[2] J.-F. Chen and S.-D. Chen, "Multiobjective power optimization," in Neural Networks, 1995.
dispatch with line flow constraints using the fast Proceedings., IEEE International Conference on,
Newton-Raphson method," IEEE Transactions on 1995, pp. 1942-1948 vol.4.
Energy conversion, vol. 12, pp. 86-93, 1997. [13] J. Sun, W. Fang, V. Palade, X. Wu, and W. Xu,
[3] A. El-Keib, H. Ma, and J. Hart, "Environmentally "Quantum-behaved particle swarm optimization
constrained economic dispatch using the with Gaussian distributed local attractor point,"
Lagrangian relaxation method," IEEE transactions Applied Mathematics and Computation, vol. 218,
on Power Systems, vol. 9, pp. 1723-1729, 1994. pp. 3763-3775, 2011.
[4] Y. Song, G. Wang, P. Wang, and A. Johns, [14] S. Krishnamurthy, R. Tzoneva, and Ieee, Impact of
"Environmental/economic dispatch using fuzzy Price Penalty Factors on the Solution of the
logic controlled genetic algorithms," IEE Combined Economic Emission Dispatch Problem
Proceedings-Generation, Transmission and using Cubic Criterion Functions, 2012.
Distribution, vol. 144, pp. 377-382, 1997. [15] J. Pitono, A. Soepriyanto, and M. H. Purnomo,
[5] A. I. S. Kumar, K. Dhanushkodi, J. J. Kumar, and "Advance optimization of economic emission
C. K. C. Paul, "Particle swarm optimization dispatch by particle swarm optimization (PSO)
solution to emission and economic dispatch using cubic criterion functions and various price
problem," in TENCON 2003. Conference on penalty factors," Kursor, vol. 7, 2014.
Convergent Technologies for the Asia-Pacific [16] I. Ziane, F. Benhamida, and A. Graa, "Simulated
Region, 2003, pp. 435-439. annealing algorithm for combined economic and
[6] M. Basu, "A simulated annealing-based goal- emission power dispatch using max/max price
attainment method for economic emission load penalty factor," Neural Computing and
dispatch of fixed head hydrothermal power Applications, pp. 1-9, 2016.
systems," International Journal of Electrical [17] F. Mahdi, P. Vasant, V. Kallimani, P. Fai, and M.
Power & Energy Systems, vol. 27, pp. 147-153, Wadud, "Emission Dispatch Problem with Cubic
Feb 2005. Function Considering Transmission Loss using
[7] A. A. Abou El Ela, M. A. Abido, and S. R. Spea, Particle Swarm Optimization," Journal of
"Differential evolution algorithm for emission Telecommunication, Electronic and Computer
constrained economic power dispatch problem," Engineering (JTEC), vol. 8, pp. 17-21, 2016.
Electric Power Systems Research, vol. 80, pp.
1286-1292, Oct 2010.
[8] A. Bhattacharya and P. K. Chattopadhyay, "Hybrid
differential evolution with biogeography-based
optimization algorithm for solution of economic
emission load dispatch problems," Expert Systems
with Applications, vol. 38, pp. 14001-14010, Oct
2011.
[9] J. P. Roselyn, D. Devaraj, and S. S. Dash,
"Economic Emission OPF Using Hybrid GA-
Particle Swarm Optimization," in Swarm,
Evolutionary, and Memetic Computing, Pt I. vol.
7076, B. K. Panigrahi, P. N. Suganthan, S. Das,
and S. C. Satapathy, Eds., ed Berlin: Springer-
Verlag Berlin, 2011, pp. 167-175.
[10] S. Arunachalam, R. Saranya, and N. Sangeetha,
"Hybrid Artificial Bee Colony Algorithm and
Simulated Annealing Algorithm for Combined
Economic and Emission Dispatch Including Valve