种群熵启动反向学习的动态多种群粒子群算法  

Dynamic multi-swarm particle swarm optimization using population entropy to start reverse learning

作　　者：梁晓磊 张孟镝 周文峰 武建国 LIANG Xiaolei;ZHANG Mengdi;ZHOU Wenfeng;WU Jianguo(School of Automobile and Traffic Engineering,Wuhan University of Science and Technology,Wuhan 430065,China;School of Transportation and Logistics Engineering,Wuhan University of Technology,Wuhan 430065,China)

机构地区：[1]武汉科技大学汽车与交通工程学院,武汉430065 [2]武汉理工大学交通与物流工程学院,武汉430065

出　　处：《智能计算机与应用》2024年第2期9-17,共9页Intelligent Computer and Applications

摘　　要：针对传统粒子群优化算法在求解复杂优化问题时容易陷入局部最优和停滞的问题,提出采用种群熵启动反向学习的动态多种群粒子群算法。借鉴狮群算法划分狮群的思想,采用动态多种群划分策略,将粒子划分成3个不同行为子群,对其实施不同的位置更新公式,保持粒子在搜索过程中的多样性;在迭代阶段,为避免算法早熟,构建了各维重心反向变异策略丰富变异备选个体,并结合种群熵指标进行种群状态评价适时启动变异策略,帮助粒子跳出局部最优。最后,通过8个基准测试函数与同种类6种经典和新型改进算法,在不同维度下进行测试对比。数值实验结果表明,改进策略显著提升了粒子群算法搜索能力,在搜索精度和搜索速度方面均优于其他对比算法。Due to particle swarm optimization(PSO)easily being trapped in local optimum,a dynamic multi-type particle swarm optimization based on population entropy to start each dimension mutation adaptively is proposed.Inspired from the lion group searching behaviors,a dynamic multi-type particle division strategy is provided to divide the particles into three different classes,and implements different information updating models on them,which can maintain the diversity of particles in the search process.In the iterative stage,the algorithm introduces a population entropy value as the judgment condition to start the reverse learning strategy for the global optimal particles,thereby helping the particles to jump out of local optimal.Numerical experiments show that the proposed algorithm has stronger search ability then other selected algorithms with excellent performance in terms of accuracy and search speed.

关 键 词：粒子群算法 狮群算法 种群熵 反向学习 动态多种群划分 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]