Abstract—This study presents a modified version of the artificial bee colony (ABC) algorithm by including a local search technique for solving the non-convex economic power dispatch problem. The local search step is incorporated at the end of each iteration. Total system losses, valve-point loading effects and prohibited operating zones have been incorporated in the problem formulation. Thus, the problem becomes highly nonlinear and with discontinuous objective function. The proposed technique is validated using an IEEE benchmark system with ten thermal units. Simulation results demonstrate that the proposed optimization algorithm has better convergence characteristics in comparison with the original ABC algorithm. Keywords—Economic power dispatch, artificial bee colony, valve-point loading effects, prohibited operating zones. I. INTRODUCTION UE to the excessive increase of fuel prices, the optimal power generation of thermal units with minimum production cost has become necessary. To satisfy this requirement, several research works have been proposed to solve the economic dispatch (ED). Various studies have considered the traditional ED problem where the production cost function of each thermal unit is approximated by a quadratic function [1], [2]. Unfortunately, modern systems are with units that have prohibited operating zones (POZ) due to physical operation limitations. In addition, practical ED problem includes the valve-point loading effects (VPLE) in the cost function. Thus, traditional optimization techniques proposed in the literature, such as Newton methods [3], lambda iteration [4] and linear programming [5], cannot provide the best solution. In recent years, several intelligent optimization techniques, such as genetic algorithms (GA), particle swarm optimization (PSO), bacterial foraging and simulated annealing have been used to solve non-convex and discontinuous ED problem [6]- [8]. Despite that these metaheuristic methods have shown their ability to converge into reasonable solutions, they do not always guarantee the global optimal solutions. To overcome this drawback, numerous modified algorithms have been appeared. A combination of GA and micro-GA to improve global and local solution of ED problem is proposed in [9]. An elitist real-coded GA based on non-uniform arithmetic crossover and mutation has been used to optimally schedule the generation of all generators of the IEEE 25-bus system [10]. A hybrid differential evolution and sequential quadratic Badr M. Alshammari and T. Guesmi are with the Electrical Engineering Department, College of Engineering, University of Hail, Saudi Arabia (e-mail: badr_ms@hotmail.com, t.guesmi@uoh.edu.sa). programming for solving the power dispatch problem is suggested in [11]. In [12], the variable neighborhood search method is incorporated in the differential evolution algorithm in order to improve the optimal solution. Other modified and hybrid techniques combine different metaheuristic algorithms have been presented in the past two decades to solve various form of the ED problem, such as combined hybrid differential PSO algorithm [13], PSO with Time Varying Operators [14], hybrid PSO and gravitational search algorithm [15] and hybrid ant colony optimization- ABC -harmonic search algorithm [16]. Recently, a swarm based stochastic search algorithm called ABC that imitates the foraging behavior of bee colony [17], [18] is considered as an efficient technique for complex optimization problem. However, the ABC algorithm has been criticized to its poor convergence rate and premature convergence due to the unbalanced exploration-exploitation processes. To overcome this disadvantage, a modified ABC algorithm is proposed in this paper for the ED problem. This algorithm incorporates a local search technique at the end of each iteration to facilitate the convergence into the global optimum. VPLE and POZ constraints have been included in the problem. The validation of the proposed optimization method has been evaluated on the well-known benchmark system with ten units. II. ED PROBLEM FORMULATION The ED problem is considered as an optimization problem that aims to schedule the outputs of the thermal units so as to minimize the total fuel cost subject to the system operating constraints, such as generation capacity, power balance, POZ and VPLE constraints. A. Total Fuel Cost Function Let’s consider a power system with N units. The total fuel cost in $/h including VPLE can be expressed by [19]:       2 min 1 sin N T i i i i i i i i i i C a bP c P d e P P        (1) where, i a , i b , i c , i d and i e are the cost coefficients of the i-th unit. P i is the output power in MW. B. Problem Constraints  Unit capacity constraints min max , 1,..., i i i P P P i N    (2) Badr M. Alshammari, T. Guesmi Improved Artificial Bee Colony Algorithm for Non- Convex Economic Power Dispatch Problem D World Academy of Science, Engineering and Technology International Journal of Electrical and Computer Engineering Vol:12, No:4, 2018 319 International Scholarly and Scientific Research & Innovation 12(4) 2018 scholar.waset.org/1307-6892/10008931 International Science Index, Electrical and Computer Engineering Vol:12, No:4, 2018 waset.org/Publication/10008931