Abstract:

Like many other intelligent optimization algorithms,Particle Swarm Optimization (PSO) algorithm often suffers from premature convergence,especially in multipeak problems.In addition,the convergence accuracy of its local search is unsatisfactory.To overcome these problems,a novel improved PSO,named Multistrategy Particle Swarm Optimization (MPSO) algorithm,is proposed.In the evolution of particle swarms,each particle chooses its own contemporary optimal search strategy from multiple alternative strategies according to the changes of optimal position it finds.Among these strategies,the steepest descent strategy,the corrective decline strategy and the random mobile strategy are able to be chosen by the optimal particle,while the aggregation strategy and the diffusion strategy are available for nonoptimal particles.In the end,the performance of MPSO is tested with four typical test functions and numerical results indicate that the proposed MPSO algorithm has a stronger and more stable global search ability than the standard PSO algorithm.

Key words: particle swarm optimization algorithm;the steepest descent method using difference quotient;diffusion, decision making