改进的多目标粒子群优化算法

摘要

为提高解决多目标优化问题的能力,提出一种改进的多目标粒子群优化算法.该算法采用均匀随机初始化方法初始种群,采用快速支配策略选取非支配解,生成外部档案;通过比较粒子连续几代的更新情况来判断是否陷入局部最优并相应地采取不同的更新策略,同时引入变异因子对粒子进行扰动.实验结果表明,在世代距离GD(Generational Distance)和空间评价方法SP(Spacing)性能指标上,改进之后的算法与另外几种对等算法相比,具有显著的整体优势.%In order to improve the ability to solve the problem of multi-objective optimization (MOPSO),an improved multi-objective particle swarm optimization algorithm (IMOPSO) is proposed.Using IMSPSO,initial population was produced by a uniformly random initialization approach,and non-dominated solutions were selected by fast control strategy to generate the external archive.By comparing the successive generations of particles,we could judge whether they felled into local optima and adopted different updating strategies.At the same time,a disturbance item was added to the particle's updating.The experimental results show that the proposed algorithm significantly surpasses other algorithms in terms of GD(Generational Distance),SP(Spacing).