基于Maximin的动态种群多目标粒子群算法

针对粒子群优化算法在处理多目标函数优化问题的过程中,往往会出现局部收敛现象,在MOPSO算法基础上提出了一种新的多目标粒子群优化算法.该算法在运行过程中采用动态调整粒子群种群数目的方式使粒子摆脱局部最优解对其的吸引;同时为了克服粒子种群多样性降低带来的影响,将粒子的相对适应度方差引入到Maximin计算公式中.然后基于Pareto最优的概念,利用方差Maximin策略来评价最优解,并保存在可变的外部精英集中,以保证结果的分布性良好.最后,该方法在仿真中取得了良好效果,可以更广泛地应用到复杂工业多目标优化领域中.

Dynamic Swarm Multiobjective Optimization PSO-Based Maximin Function

Since the local convergence was often found in the process that the PSO algorithm is used to deal with multiobjective optimization problems, an new algorithm based on Maximin PSO algorithm is presented for multiobjective PSO, where the swarm size is adjusted dynamically to avoid converging at a false Pareto optimal solutions. Meanwhile, the relative fitness variance of particle is introduced into the formula of Maximin to get rid of the effect of reducing the diversity of particle swarm. Then, based on the idea about Pareto optimum, the variance Maximin strategy is used to evaluate the optimal solutions and the non-dominated solutions are stored in a size-variable elitism repository so as to provide a good diversity in its solutions. Experimental results showed the feasibility of the algorithm in solving complicated multi-objective optimization in industry.