多目标PSO算法综述.docVIP

多目标微粒群优化算法研究 王艳1,2，曾建潮2( （1 兰州理工大学电信工程学院 甘肃 兰州730050） （2 太原科技大学复杂系统和智能计算实验室 山西 太原 030024） 摘要：作为一种有效的多目标优化工具，微粒群优化(PSO)算法已经得到广泛研究与认可。本文首先对多目标优化问题进行了形式化描述，简要介绍了微粒群优化算法与遗传算法的区别，并对多目标微粒群优化算法(MOPSO)进行了分类，着重分析了各类算法的主要思想、特点及其代表性算法。其次，针对非支配解的选择、外部档案集的修剪、解集多样性的保持以及微粒个体历史最优解和群体最优解的选取等热点问题进行了详细论述，并在此基础上对各类典型算法进行了比较。最后，根据当前MOPSO算法的研究状况，提出了该领域的发展方向。 关键词： 多目标优化； 微粒群优化算法；非支配解；外部档案；多样性 中图分类号： TP301 文献标识码：A A Survey of Multi-Objective Particle Swarm Optimization Algorithm WANG Yan1,2, ZENG Jian-chao2 (1 College of Electrical and Information Engineering, Lanzhou University of Technology, Lanzhou 730050, China.) (2 Complex System and Computational Intelligence Laboratory, Taiyuan University of Science and Technology, Taiyuan, 030024, China. Correspondent: ZENG Jian-chao, E-mail: zengjianchao@263.net) Abstract As an effective multi-objective optimizer, Particle Swarm Optimization (PSO) algorithms have been widely studied and approbated. This paper firstly described multi-objective problems formally and introduced the difference between PSO and Genetic Algorithm (GA). Then a classification of multi-objective PSO (MOPSO) algorithms was presented and the main idea, features and representative algorithms of each approach were analyzed. Secondly, such hot topics in MOPSO algorithms as selecting nondominated solutions, pruning archive set, holding diversity of the solutions set and selecting both personal best solution and global best solution were discussed in details, on base of which all typical algorithms were compared. Finally, several viewpoints for the future research of MOPSO were proposed according to the present studies. Key words multi-objective optimization; particle swarm optimization; nondominated solutions; archive; diversity 1引言 在科学实践、工程系统设计及社会生产活动中，许多问题都是多目标优化问题。通常多目标优化问题中的各个目标函数之间可能会存在冲突，这就意味着多目标优化问题不存在唯一的全局最优解，使得所有目标函数同时达到最优。为了达到总目标的最优化，需要对相互冲突的目标进行综合考虑，对各子目标进行折衷。最初，多目标优化问题往往通过加权等方式转化为单目标优化问题，但这样需要事先知道每个目标函数所占的权重，并且对目标给定的次序也比较敏感。 微粒群优化(PSO)算法是1995年由Kennedy和Eberhart提出的