标准粒子群优化算法收敛性能分析与参数选择

Based on the discrete time linear dynamic system theory, the formal sufficient and necessary condition for the deterministic standard PSO algorithm to converge to equilibrium point, diverge to infinity or oscillate within a range is derived. Several simulation results of particle trajectories are given to illustrate and verify the theory analysis. Based on the theory analysis, a general guideline for parameters selection is provided; the effect of randomness is discussed and the exploration and exploitation tradeoff of PSO algorithm at different stages is analyzed. It is also pointed out that, strictly speaking, the currently popular view that small inertia weight will facilitate a local search is not accurate enough. It should be used carefully in practice.