Modified Particle Swarm Optimization Algorithm Based on Particle Classification

Abstract

In order to overcome the shortcoming of local convergence of simple particle swarm optimization algorithm (SPSO) in optimization of high-dimensional and complex function, a modified particle swarm optimization algorithm based on particle classification (PCPSO) is proposed. The particle classification is based on the particle fitness and the particle distance which is firstly proposed and defined in this paper. The coordinate system is established based on the mean value of particle distance and the mean value of particle fitness. Particle swarm is divided into four subclasses in four quadrants. The properties of the particles in each quadrant and their influence on the optimization process are analyzed, and different formulas of particle velocity updating are established to improve the particle update efficiency. In addition, the algorithm has the function of dynamically adjusting the origin of the classification coordinates as the iteration number increases to balance the algorithm’s global search performance and local search ability and improve the algorithm’s optimization efficiency. The optimization results of the typical test functions show that the accuracy and success rate of the modified algorithm in high-dimensional complex function optimization are greatly improved, and the premature convergence phenomenon can be overcome effectively.