Abstract
Particle swarm optimization (PSO) is an effective method for solving a wide range of problems. However, the most existing PSO algorithms easily trap into local optima when solving complex multimodal function optimization problems. This paper presents a variation, called adaptive PSO based on clustering (APSO-C), by considering the population topology and individual behavior control together to balance local and global search in an optimization process. APSO-C has two steps. First, via a K-means clustering operation, it divides the swarm dynamically in the whole process to construct variable subpopulation clusters and after that adopts a ring neighborhood topology for information sharing among these clusters. Then, an adaption mechanism is proposed to adjust the inertia weight of all individuals based on the evaluation results of the states of clusters and the swarm, thereby giving the individual suitable search power. The experimental results of fourteen benchmark functions show that APSO-C has better performance in the terms of convergence speed, solution accuracy and algorithm reliability than several other PSO algorithms.