基于差分演化的改进多目标粒子群优化算法  

Improved Multi-objective Particle Swarm Optimization Based on Differential Evolution

作　　者：吴亚丽[1] 徐丽青[2] 

机构地区：[1]西安理工大学自动化与信息工程学院,西安710048 [2]中国电南京自动化股份有限公司,南京211100

出　　处：《系统仿真学报》2011年第10期2211-2215,共5页Journal of System Simulation

基　　金：陕西省自然科学基金(2010JQ8006);陕西省教育厅科学研究专项(2010JK711)

摘　　要：提出一种基于差分演化的改进多目标粒子群优化算法来求解多目标优化问题。算法通过对Pareto最优解集的差分演化来增加Pareto解集的多样性;通过循环拥挤距离来控制归档集中非劣解的分布,提高对种群空间的均匀采样;采用一种新的多目标适应值轮盘赌法选择粒子的全局最优位置,使其更逼近Pareto最优前沿;自适应惯性权重和加速度因子的设计增强了算法的全局搜索能力。多个多目标测试函数的仿真结果表明,改进的多目标粒子群算法能够在保持Pareto最优解多样性的同时具有较好的收敛性能。An improved multi-objective particle swarm optimization algorithm based on differential evolution （DE-IMOPSO） was proposed to solve multi-objective optimization problem. Differential evolution was used for Pareto set to increase its diversity. And a circular crowded sorting approach was adopted to improve the uniformity of the population distribution. A new multi-objective fitness roulette algorithm was applied to select the global best location of each particle to make it approach to Pareto frontier more closely. Adaptive inertia weight and acceleration coefficients enhanced the global exploratory capability. The simulation results of benchmark test functions show that DE-IMOPSO not only obtains the more diversity of the Pareto solutions but also possesses the better global convergence.

关 键 词：多目标优化 差分演化 粒子群优化算法 循环拥挤排序 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]