Particle Swarm Optimization Algorithm based on Differential Mutation and Thermodynamic Selection

Abstract

In order to improve the convergence speed and the drawback of easily converging to the local optimum of the standard particle swarm optimization (PSO), two improved PSO algorithms are presented based on the simple particle swarm optimization algorithm without speed attribute. One is introducing differential mutation technology of differential evolution algorithm into the simple PSO algorithm for the disturbance of the particle position update, so the particles have the more opportunity to escape from local extreme points and reach the global optimum. The other is based on the law of free energy minimization in the statistical physics and thermodynamics. The particles with the smaller free energy component are chosen to retain in the new population. This selection strategy effectively maintains the diversity of the population and improves the search performance of the PSO algorithm. The experimental results on the twelve classical test functions show that the two improved simple PSO algorithms have the better convergence rate, convergence precision and stability.