微粒群算法的参数选择及收敛性分析

Convergence analysis and parameter selection in particle swarm optimization

微粒群算法是相对较新颖的优化算法,已经成功应用于许多优化问题。然而算法的参数选择及收敛性分析研究不足,为此首先认真研究了现有微粒群算法粒子轨迹及其收敛性的文献,在此基础上,根据递减惯性权重和递增惯性权重微粒群算法各自的特点,结合算法的收敛区间,提出了一种具有先增后减惯性权重的新的微粒群算法,既保留了具有递增和递减惯性权重的优点,也克服了它们的缺点,取得了比较好的效果。

Particle Swarm Optimization（PSO） is a novel optimization technology and has been applied successfully to various optimization problems.But the convergence of the algorithm has been studied insufficiently.The author studies the particle＇s trajectory and convergence of the existing PSO algorithm,on the basis of their characteristics,proposes a new PSO algorithm with a new inertia weight along broken line.This algorithm preserves the advantages of the incremental and reduced inertia weight ,and overcomes their shortcomings.Simulated experiments achieve good results.