多目标优化的一种改进微粒群算法  

Improved Particle Swarm Optimization Algorithm for Multi-Objective Optimization

作　　者：袁代林[1,2] 陈虬[1] 

机构地区：[1]西南交通大学力学与工程学院,四川成都610031 [2]西南交通大学数学学院,四川成都610031

出　　处：《计算机仿真》2010年第6期234-238,共5页Computer Simulation

基　　金：西南交通大学校基金(2008B07)

摘　　要：微粒群算法是解决多目标优化问题的一个重要方法。为了多目标目标优化求解问题,常用的微粒群算法在处理多目标优化问题时,存在所得Pareto最优解集的分散性和实用性较差的缺点。针对上述问题,提出了微粒群算法的一种改进形式。改进算法引入了个体精英解集,从中选择更合适的个体最优位置。同时,在评价个体适应度时,考虑了目标函数值差异这一信息。个体对应的目标函数值差异大,则其适应度就小。这样能避免各目标函数值差异过大的最优解存在。三个典型的多目标测试函数表明,改进方法得到最优解集具有更好的分散性和实用性。测得结果证明,改进方法是有效的。Particle swarm optimization （PSO） algorithm is an important measure for solving multi -objective （MO） problems. The diversity and practicability of Pareto optimal solution set are not good when using common PSO to MO problems. In view of this problem, an improved PSO algorithm is advanced. In the improved PSO, individual elitism set is advanced which is good for the choice of more individual optimal position. At the same time, the variancebetween objective function values is considered when the individual fitness is evaluated. If the variance is bigger, the individual fitness is also bigger. So the optimal solutions with big variance from objective functionvalues are avoi- ded. Simulation results from three classical test functions show that the diversity and practicability of the optimal solution set are better by using the improved PSO than using common PSO. So, the improved measure is valid.

关 键 词：多目标优化 微粒群算法 改进算法 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]