可行性规则动态调整的多目标粒子群算法  

Multi-objective particle swarm algorithm based on dynamic adjustment of feasibility rule

作　　者：赵乃刚[1] 李勇[1] 王振荣 Zhao Naigang;Li Yong;Wang Zhenrong(College of Mathematics&Computer Science,Shanxi Datong University,Datong Shanxi 037009,China;Information Center of Datong People’s Government of Shanxi Province,Datong Shanxi 037009,China)

机构地区：[1]山西大同大学数学与计算机科学学院,山西大同037009 [2]山西省大同市人民政府信息化中心,山西大同037009

出　　处：《计算机应用研究》2018年第5期1304-1306,1336,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(61672331);山西省高等学校教学改革项目(2015090);山西大同大学科学研究项目(2016K1)

摘　　要：针对多目标粒子群算法存在的问题,提出了一种可行性规则动态调整的多目标粒子群算法。在算法中,根据粒子之间的相似度值动态非线性地更新算法的惯性权重,使得算法可以高效地平衡全局和局部搜索之间的矛盾;采用动态加权法解决随机性抽取群体最优粒子的缺陷,保证了种群的多样性;并且动态改变可行性规则的阈值,使得算法可以有效地利用某些不可行解包含的有效信息,提高了算法收敛到Pareto前沿的能力。最后,与其他四种多目标算法的实验比较验证了新算法的性能更好。Aiming at the problems of multi-objective particle swarm algorithm,this paper developed a multi-objective particle swarm algorithm based on dynamic adjustment of feasibility rule.This algorithm dynamically non-linear updated the inertia weight according to the similarity value between particle and best particle.In order to solve the defects of random selection of the best particle,this paper proposed dynamic update method of inertia weight and it could ensure the diversity of the population.It changed the threshold of feasibility rules dynamically so that new algorithm could take advantage of some valid information of some infeasible solutions and it could improve the ability of converging to the Pareto frontier.Experiment comparison with other four algorithms shows better performance of the new algorithm.

关 键 词：粒子群算法 多目标优化 惯性权重 动态加权法 可行性规则 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]