基于最小生成树的粒子群优化算法  

Particle swarm optimization algorithm based on minimum spanning tree

作　　者：李国森 闫李 朱小培 柳琳娜 LI Guo-sen;YAN Li;ZHU Xiao-pei;LIU Lin-na(School of Electronic and Information Engineering,Zhongyuan University of Technology,Zhengzhou 450007,China)

机构地区：[1]中原工学院电子信息学院,河南郑州450007

出　　处：《计算机工程与设计》2022年第7期1972-1980,共9页Computer Engineering and Design

基　　金：国家自然科学基金项目(61673404、61873292);河南省高等学校重点科研基金项目(19A120014);河南省高校创新人才基金项目(16HASTIT033);中国纺织工业联合会科技指导性基金项目(2017054、2018104)。

摘　　要：针对算法多样性丢失,获得的Pareto解集不完整等不足,提出基于最小生成树的粒子群优化算法(particle swarm optimization algorithm based on minimum spanning tree,MSTPSO)。采用最小生成树机制,构建多个不同的邻域,引导粒子在邻域内进化;利用扩散策略,挖掘邻域的有用信息,扩大粒子的搜索范围;引入调和平均距离衡量解在决策空间的拥挤度,维持解的均匀分布。通过12个测试函数和一个实际优化问题,并与8个算法进行比较,其结果表明,MSTPSO在决策空间获得了收敛性好且均匀分布的解集。Aiming at the problem that the algorithm has diversity loss and incomplete Pareto set,a particle swarm optimization algorithm based on minimum spanning tree(MSTPSO)was proposed.The minimum spanning tree mechanism was used to construct multiple neighborhoods and guide particles to evolve in their neighborhood.The diffusion strategy was used to mine the useful information of neighborhood and to expand the search range of particles.The harmonic average distance was introduced to measure the crowding degree of the solution in the decision space and to maintain the uniform distribution of the solution.Through extensive comparison with eight algorithms on the twelve test functions and one practical optimization problem,the results show that MSTPSO obtains a set of well-converged and evenly-distributed optimal solutions in the decision space.

关 键 词：进化算法 粒子群优化 多模态优化 多目标优化 PARETO最优解集 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]