A Hybrid Particle Swarm Optimization to solve Economic Dispatch with Valve-Point Effect

Abstract Scarcity of energy resources, increasing power generation cost and evergrowing demand for electric energy necessitates optimal economic dispatch in todays power systems. The main objective of economic dispatch is to reduce the total power generation cost, while satisfying various equality and inequality constraints. Traditionally in economic dispatch problems, the cost function for generating units has been approximated as a quadratic function which doesnt provide accurate results. Moreover, to obtain accurate fuel cost, valve-point effect in thermal power plant has to be taken into account. The inclusion of valve-point effect makes the modeling of the fuel cost functions of generating units more practical. In this paper a new hybrid evolutionary algorithm called Hybrid PSO, has been employed to solve economic dispatch problem with the valve-point effect. The hybrid technique combines Genetic Algorithms (GA) and Particle Swarm Optimization (PSO). The proposed algorithm is modeled on the concepts of Darwins theory based on natural selection and evolution, and on cultural and social rules derived from the swarm intelligence. Using Hybrid PSO technique the non-linear cost function is solved for three unit system and the results are compared with the traditional PSO and GA method. These results prove that Hybrid PSO method is capable of getting higher quality solution including mathematical simplicity, fast convergence, and robustness to solve hard optimization problems. Keywords Economic Dispatch, Genetic Algorithm, Particle Swarm Optimization, Valve point effect, Hybrid PSO. I. INTRODUCTION Economic dispatch is one of the most important problems to be solved in the operation
1

and planning of a power system. The primary objective of the electric power generation is to schedule the committed generating unit outputs so as to meet the required load demand at minimum operating cost while satisfying all unit and system equality and inequality constraints. In the traditional ED problem, the cost function for each generator has been approximately represented by a single quadratic function and is solved using mathematical programming based on the optimization techniques such as lambda-iteration method, gradient-based method, etc. These mathematical methods require incremental or marginal fuel cost curves which should be monotonically increasing to find global optimal solution. The fuel cost functions of generating units can be modeled in a more practical fashion by including the valve-point effects. The valvepoint effects result in the ripples in the fuel cost function; thereby the number of local optima is increased. Thus, the practical ED problem is represented as a non smooth optimization problem with equality and inequality constraints, which cannot be solved by the traditional mathematical methods. Dynamic programming method can solve such types of problems, but it suffers from so-called the curse of dimensionality. Over the past few years, in order to solve this problem, many salient methods have been developed such as genetic algorithm, evolutionary programming, tabu search, neural network approaches, and particle swarm optimization. Considering Genetic Algorithms and Particle Swarm Optimization algorithms, most of the times, PSO have faster convergence rate than GA early in the run, but they are often outperformed by GA for long simulation runs, or when the number of unknowns increase. This is due to the different types of search, adopted by the two algorithms. The proposed Hybrid PSO method consists in a strong co-operation of GA and PSO, since it

maintains the integration of the two techniques for the entire run of simulation. In each iteration, in fact, some of the individuals are substituted by new generated ones by means of GA, while the remaining part is the same of the previous generation but moved on the solution space by PSO. Doing so, the problem of premature convergence of the best individuals of the population to a local optimum, one of the most known drawbacks found in tests of hybrid global-local strategies, has been cancelled. Hence Hybrid PSO technique is used to solve power system economic load dispatch problem (ELD) with valve-point effect. The convergence characteristics and robustness of this method proves efficient when compared to other methods. II. PROBLEM FORMULATION A. BASIC OBJECTIVE FUNCTION The objective of the economic dispatch problem is to minimize the total fuel cost of thermal power plants subjected to the operating constraints of a power system. In general, it can be formulated mathematically with an objective function and two constraints as follows:
N

should be the same as total load demand plus the total line loss
N

Pi=PD+PLoss
i=1

(3)

where PD is the total system demand and PLoss is the total line loss. However, the transmission loss is not considered in this paper for simplicity (i.e., PLoss=0). 2) Minimum and maximum generation constraints Generation output of each generator should lie between maximum and minimum limits. The corresponding inequality constraints for each generator are Pi,min Pi Pi,max (4)

where Pi,min and Pi,max are the minimum and maximum output of generator i, respectively. C. VALVE-POINT EFFECTS The generating units with multi-valve steam turbines exhibit a greater variation in the fuel-cost functions. Since the valve point results in the ripples as shown in Fig. 1, a cost function contains higher order nonlinearity [9]. Therefore, the equation (2) should be replaced as the equation (5) to consider the valve-point effects. Here, the sinusoidal functions are thus added to the quadratic cost functions as follows. Fi(Pi) = ai + biPi +ciPi +|ei sin(fi (Pi,min Pi))| (5) Where ei and fi are the coefficients of generator i reflecting valve-point effects.

FT = Fi(Pi)
i=1

(1) (2)

Fi (Pi) = ai + biPi + ciPi2 where,

FT total generation cost, Fi cost function of generator i, ai ,bi , ci cost coefficients of generator i, Pi power of generator i, N number of generators. B. THE CONSTRAINTS or Active power balance equation For power balance, an equality constraint should be satisfied. The total generated power
2 1) Equality

Vik+1=Vik + c1r1(Pbestik Xik) + c2r2(GbestkXik) (6) where, Vi c1, c2 r1,r 2 Xik Pbestik Gbestk velocity of individual i at iteration k, inertia weight parameter, acceleration coefficients, random numbers between 0 and 1, position of individual i at iteration k, best position of individual i until iteration k, best position of the group until iteration k.

Fig. 1 Cost Curve for large steam power plant with valve-point effect III. OVERVIEW OF THE PSO AND GA A. PARTICLE SWARM OPTIMIZATION Kennedy and Eberhart developed a PSO algorithm based on the behavior of individuals (i.e., particles or agents) of a swarm [11]. Its roots are in zoologists modeling of the movement of individuals (i.e., fish, birds, and insects) within a group. It has been noticed that members of the group seem to share information among them, a fact that leads to increased efficiency of the group. The PSO algorithm searches in parallel using a group of individuals similar to other AI-based heuristic optimization techniques [12]. Each individual corresponds to a candidate solution to the problem. Individuals in a swarm approach to the optimum through its present velocity, previous experience, and the experience of its neighbours. In a physical n-dimensional search space, the position and velocity of individual i are represented as the vectors Xi=( xi1,, xin) and Vi= (vi1,,vin) in the PSO algorithm. Let Pbesti=( xi1Pbest,, xinPbest) and Gbest=( x1Gbest,, xnGbest) be the best position of individual i and its neighbors best position so far, respectively. Using the information, the updated velocity of individual i is modified under the following equation in the PSO algorithm:

In this velocity updating process, the values of parameters such as , c1, and c2 should be determined in advance. In general, the weight is set according to the following equation [1], [9]: =max- ((max-min)/Itermax)*Iter where, max ,min Itermax Iter (7)

initial and final weights, maximum iteration number, current iteration number.

Each individual moves from the current position to the next one by the modified velocity in (6) using the following equation: Xik+1=Xik+Vik+1 B.GENETIC ALGORITHM Genetic algorithm (GAs) were invented by John Holland in the 1960s and were developed with his students and colleagues at the University of Michigan in the 70s. Hollands original goal was to investigate the mechanisms of adaptation in nature to develop methods in which these mechanisms could be imported into computer systems. GA is a method for deriving from one population of chromosomes (e.g., strings of ones and zeroes, or bits) a new population. This is achieved by employing natural selection together with the genetics inspired operators of recombination (crossover), mutation, and inversion. Each chromosome consists of genes (8)

(e.g. bits), and each gene is an instance of a particular allele(e.g.,0 or 1). The selection operator chooses those chromosomes in the population that will be allowed to reproduce, and on average those chromosomes that have a higher fitness factor (defined bellow),produce more offspring than the less fit ones. Crossover swaps subparts of two chromosomes, roughly imitating biological recombination between two single chromosome (haploid) organisms; mutation randomly changes the allele values of some locations (locus) in the chromosome; and inversion reverses the order of a contiguous section of chromosome. Operators of Genetic Algorithm A basic genetic algorithm comprises three genetic operators. Selection Crossover Mutation Starting from an initial population of strings (representing possible solutions), the GA uses these operators to calculate successive generations. First, pairs of individuals of the current population are selected to mate with each other to form the offspring, which then form the next generation. 1) Selection This operator selects the chromosome in the population for reproduction. The more fit the chromosome, the higher its probability of being selected for reproduction. The various methods of selecting chromosomes for parents to crossover are Roulette-wheel selection Boltzmann selection Tournament selection Rank selection Steady-state selection Roulette-wheel selection The commonly used reproduction operator is the proportionate reproductive operator where a string is selected from the mating pool with a probability proportional to Fi where Fi is the fitness value for that string. Since the population size is usually kept fixed in a
4

simple GA, The sum of the probabilities of each string being selected for the mating pool must be one. The probability of the ith selected string is Fi pi = (6.1)
n

Fj
j=1

Where n is the population size. Crossover The cross over operator involves the swapping of genetic material (bit-values) between the two parent strings. This operator randomly chooses a locus (a bit position along the two chromosomes) and exchanges the subsequences before and after that locus between two chromosomes to create two offspring. For example, the strings 1110 0001 0011 and 1000 0110 0111. The crossover operator roughly imitates biological recombination between two haploid (single chromosome) organisms. The crossover may be a single bit cross over or two bit cross over. In case of two bit crossover two points are chosen where the binary digits are swapped. 3) Mutation The two individuals (children) resulting from each crossover operation will now be subjected to the mutation operator in the final step to forming the new generation. This operator randomly flips or alters one or more bit values at randomly selected locations in a chromosome. For example, the string 1000 0001 0011 might be mutated in its second position to yield 1100 0001 0011. Mutation can occur at each bit position in a string with some probability and in accordance with its biological equivalent; usually this is very small, for example, 0.001. If 100% mutation occurs, then all of the bits in the chromosome have been inverted. The mutation operator enhances the ability of the GA to find a near optimal solution to a given problem by maintaining a sufficient level of genetic variety in the population, which is needed to make sure that the entire solution space is used in the search for the best solution. In a sense, it serves as an insurance policy; it helps prevent the loss of genetic material. 2)

IV. OVERVIEW OF HYBRID PSO In recent years several evolutionary algorithms have been developed for optimization of ED problems. The general goal of the optimization is to find a solution that represents the global maximum or minimum of a fitness function. A natural evolution of the particle swarm algorithm can be achieved by incorporating methods that have already been tested in other evolutionary computation techniques. Many authors have considered incorporating selection, mutation and crossover, as well as the differential evolution (DE), into the PSO algorithm. The main goal is to increase the diversity of the population by: 1) either preventing the particles to move too close to each other and collide or 2) to self-adapt parameters such as the constriction factor, acceleration constants, or inertia weight. As a result, hybrid versions of PSO have been created and tested in different applications. The most common ones include hybrid of genetic algorithm and PSO (GA-PSO), evolutionary PSO (EPSO) and differential evolution PSO (DEPSO and C-PSO). Considering Genetic Algorithms and Particle Swarm Optimization algorithms, most of the times, PSO have faster convergence rate than GA early in the run, but they are often outperformed by GA for long simulation runs, or when the number of unknowns increases. This is due to the different types of search, adopted by the two algorithms. The proposed method consists in a strong co-operation of GA and PSO, since it maintains the integration of the two techniques for the entire run of simulation. In each iteration, in fact, the population is divided into two parts and it is evolved with the two techniques respectively. The two parts are then recombined in the updated population, that is again divided randomly into another two parts in the next iteration for another run of genetic or particle swarm operators. Doing so, the problem of premature convergence of the best individuals of the population to a local optimum, one of the most known drawbacks found in tests of hybrid
5

global-local strategies, has been cancelled. Fig.2 shows the flow chart of the developed algorithm. GA-PSO combines the advantages of swarm intelligence and a natural selection mechanism, such as GA, in order to increase the number of highly evaluated agents, while decreasing the number of lowly evaluated agents at each iteration step. Therefore, not only is it possible to successively change the current searching area by considering pbest and gbest values, but also to jump from one area to another by the selection mechanism, which results in accelerating the convergence speed of the whole algorithm. The driving parameter of the GA-PSO algorithm is the Hybridization Coefficient (HC); it expresses the percentage of population that in each iteration is evolved with GA: so HC=0 means the procedure is a pure PSO (the whole population is updated according to PSO operators), HC=1 means pure GA, while 0<HC<1 means that the corresponding percentage of the population is developed by GA, the rest with PSO.

Fig. 2 Flowchart for Hybrid PSO (GA-PSO) V. IMPLEMENTATION OF HYBRID PSO ALGORITHM The procedure for Hybrid PSO is as follows, Step 1: Read the given data.

Step 2: Generate the initial random population. Step 3: Evaluate the fuel cost using the fitness function for all individuals. Step 4: Check whether a stopping criterion occurred or not. If occurred compare fuel cost of each individual and find optimal solution. Then go to step 8. Step 5: The population was splitted according to HC value if HC is within 0 and 1. Apply the splitted population for PSO and GA algorithm. Step 6: If HC=0, apply PSO algorithm for pure PSO and if HC=1, apply GA algorithm for pure GA. Step 7: Evaluate the new resulting population and go to step 3. Step 8: Stop. VI. SIMULATED RESULT ANALYSIS: To verify the feasibility of the proposed Hybrid PSO method, three unit power systems were tested. The results obtained from the Hybrid PSO are compared with Particle Swarm Optimization (PSO) and genetic algorithm (GA). This system comprises of 3 generating units and the input data of 3-generator system are given in Table I. Here, the total demand for the system is set to 850MW. The obtained results for the 3-generator system using the Hybrid PSO are given in Table II and the results are compared with PSO and GA. TABLE II
COMPARISON OF SIMULATION RESULTS OF GA, PSO AND HYBRID PSO METHOD CONSIDERING VALVE-POINT EFFECT (3-UNIT SYSTEM)

VII. CONCLUSION This paper presents a new approach for solving the ED problems with valve-point effect based on the Hybrid PSO algorithm. The Hybrid Particle Swarm Optimization algorithm emerges as a fast method for optimization of complex nonlinear objective functions and its generality makes it suitable for wide use in ED applications. The hybrid pso method is capable of dealing directly with load demand at various intervals of time in the scheduled horizon with no restrictions on the shape of the input-output cost function of the generating unit. It is very fast compared to other evolutionary techniques in exploring the search space.

TABLE I
DATA FOR TEST CASE (3-UNIT SYSTEM)

The Hybrid PSO has provided the global solution in the 3-unit test system and the better solution than GA and PSO. The result shows that the convergence property was not affected based on the shape of the incremental fuel cost function. The advantage of this method is its ability in finding high quality solutions reliably with fast converging characteristics. The transmission loss is not considered in this work for simplicity. To improve accuracy, programming can be done by including the transmission losses.

[7] Boggs, P.T. and Tolle, J.W. (1995) "Sequential Quadratic Programming", Acta Numerica, Cambridge University Press, Cambridge, vol.no 4, pp. 1-52. [8] Chang C.S., Liew A.C., Srinivasan D., and Wen F. (1996) "A survey of evolutionary computing in power systems", IEEE Proceedings, pp. 35-41. [9] Chia-Feng Juang, A Hybrid of Genetic Algorithm and Particle Swarm Optimization for Recurrent Network Design, IEEE Transactions On Systems, Man And Cybernetics-Part B: Cybernetics, vol. 34, n. 2, pp. 997-1006, April 2004. [10] Chowdhury, B.H. and Rahman, S. (2001) "A review of recent advances in economic dispatch", IEEE Transaction on Power System, vol.5, no.4, pp.1248-1259. [11] Fonseca N. and Miranda V. New evolutionary particle swarm algorithm (EPSO) applied to voltage/VAR control, in Proc. 14th Power Syst. Comput. Conf., Jun. 2002. [12] Fukuyama Y., Genji T., Naka S., and Yura T., A hybrid particle swarm optimization for distribution state estimation, IEEE Trans. Power Syst., pp. 6068, Feb. 2003. [13] El-Dib A., El-Metwally M., Osman Z., and Youssef H., Load flow solution using hybrid particle swarm optimization, in Proc. Int. Conf. Elect., Electron., Comput. Eng., Sep. 2004, pp. 742746.

REFERENCES [1] Aruldoss Albert victorie, T. Jeyakumar, A. (2004) "Hybrid economic dispatch with valve Electric power system research, 51-59. and Ebenezer PSP-SQP for point effect", vol.no.71, pp.

[2] Attaviriyanupap P., Hasegawa J., Kita H., and Tanaka E. (2002)"A hybrid EP and SQP for dynamic economic dispatch with non-smooth incremental fuel cost function", IEEE Transactions on Power System, vol. 17, no. 2, pp. 411-416. [3] Aug and Zwe- Lee Gaing (2003) Particles swarm optimization to solving the Economic Dispatch Considering the Generator constraints, IEEE Transactions on power systems, vol. 18, No. 3, pp.1157-1195. [4] Bentley, P. and Blackwell, T. Dont push me! Collision-avoiding swarms, in Proc. IEEE Congr. Evol. Comput., May 2002, vol. 2, pp. 16911696. [5] Bhagwan Das, D. and Patvardhan, C. (1999) "Solution of Economic Load Dispatch using real coded Hybrid Stochastic Search", Electrc. Power Energy System, vol.no 21, pp. 165-170. [6] Boeringer, D.W. and Werner, D.H. Particle Swarm Optimization Versus Genetic Algorithms for Phased Array Synthesis, IEEE Transactions on antennas and propagation, vol. 52, n. 3, pp. 771-779, March 2004.
7