一种多目标粒子群改进算法的研究  

Research on modified multi-objective particle swarm optimization

作　　者：徐鸣[1] 沈希[2] 马龙华[1] 黄跃进[2] 胡娟[3] 顾江萍[2] 金华强[2] 钱积新[1] 

机构地区：[1]浙江大学信息科学与工程学院,杭州310027 [2]浙江工业大学机械工程学院,杭州310014 [3]加西贝拉压缩机有限公司,嘉兴314000

出　　处：《控制与决策》2009年第11期1713-1718,1728,共7页Control and Decision

基　　金：国家自然科学基金项目(60474064);浙江省杰出青年基金项目(R106768);浙江省重大科技专项基金项目(2006C11002)

摘　　要：针对多目标粒子群优化过程中的粒子飞行偏向性和多样性损失问题,提出一种基于最大最小适应函数的改进算法.该算法在最大最小适应函数的计算中引入了函数相对值算法和ε-支配的概念,并提出了变ε-支配的策略,改进了最大最小适应函数的计算方法,解决了粒子飞行过程中的偏向性和多样性损失问题,加快了算法的收敛速度.将该改进算法应用于直流变频压缩机启动时峰值电流和启动转速的优化问题,应用结果表明该算法收敛速度快且效果良好.According to multi-objective optimization problem, a modified multi-objective particle swarm optimization algorithm is proposed based on maximin fitness function. Maximum-minimum relative algorithm is introduced to the computation of maximin fitness function value, which solves the skewed popularity of the particles in the flight process. Besides, the conception of ε-dominance is imported to the computation of maximin fitness function value. A modified computation method and an alterable ε-dominance strategy are proposed, which effectively improved the convergence speed of the algorithm and the diversity of the particles. The modified multi objective particle swarm optimization algorithm is applied to starting performance optimization of direct current （DC） inverter compressor. The results show that the convergent speed of the algorithm is fast and the result is valid.

关 键 词：多目标优化 粒子群算法 最大最小适应函数 变ε-支配 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]