Enhancing local search of differential evolution algorithm for high dimensional optimization problem

Differential evolution (DE) algorithm is very simpe, robust but efficient. However, the convergence speed and solution accuracy of DE algorithm significantly lower when solving high-dimension(more than 100) optimization problems. for this problem, A novel local search operation was proposed. This local operation combines both advantage of orthogonal crossover and opposition-based learning strategy. In the new algorithm, only one random individual was chose to undergo the local search operation. The purpose of this operation is to improve the local search ability, at the same time without adding too much computing resources. The simulation experiments on 9 benchmark functions show that the new algorithm improved optimization ability for high-dimensional problem. Compared with DE and OXDE, the result show that the proposed algorithm is an efficient method for the high-dimensional optimization problem.