多目标非支配邻近免疫粒子群算法  

Multi-objective Particle Swarm Optimization with Non-dominated Neighbor Immune Strategy

作　　者：刘俊华[1] 高岳林[1] 

机构地区：[1]北方民族大学信息与系统科学研究所,宁夏银川750021

出　　处：《太原理工大学学报》2014年第6期769-775,共7页Journal of Taiyuan University of Technology

基　　金：国家自然科学基金资助项目:融合粒子群优化和差分进化的混合智能算法研究(60962006);国家民委科研资助项目:两类非线性规划问题的PSO-DE的算法研究(12DFZ005)

摘　　要：针对PSO算法求解多目标优化问题时易陷入局部最优解的问题,采用非支配邻近免疫算子来对粒子群的外部档案进行维护和变异操作,有效提高了Pareto解的多样性,从而提出一种多目标非支配邻近免疫粒子群算法(NICPSO)。采用动态加权法选择全局最优粒子,提高全局搜索能力;当粒子群趋于早熟时,采用优势邻域认知的个体极值更新策略;同时将学习因子表示为惯性权重的非线性函数,有效提高算法的统一性。通过ZDT1—ZDT4及ZDT6基准测试函数验证,该算法有效的提高了优化解的收敛性和多样性,与其他多目标进化算法和多目标粒子群优化算法相比,具有较好的性能。To solve the problem that PSO algorithm is easy to entrap pre-mature convergencewhen solving multi-objective problem, non-dominated neighbor immune strategy is adopted tomaintain and mutate the external elitist archive of population, which improves the diversity of Pa-reto solutions effectively, so a PSO algorithm with non-dominated neighbor immune strategy（NICPSO） is proposed for multi-objective problem. Pynamic weighting method is used to chooseglobal best particle and improve overall searching ability. Superior neighborhood consciousness isadopted to update personal best particle when the population tends to pre-mature. Learning fac-tors are expressed as functions of inertia weight, effectively improving the uniformity of the algo-rithm. Four well-known benchmark test functions ZDTI-ZDT4 and ZDT6 are used to test theperformance of the proposed algorithm. Simulation results show that the NICPSO algorithm canreceive better Pareto solutions than some other multi-objective evolutionary algorithm.

关 键 词：多目标优化 粒子群算法 非支配邻近免疫 学习因子 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]