张伟, 张润雨,等.基于精英知识引导的多种群协作粒子群优化算法[J].河南理工大学学报（自然科学版）,2024,43(6):116-128.

ZHANG W, ZHANG R Y,et al.Multi-group cooperation particle swarm optimization algorithm based on elite knowledge guidance[J].Journal of Henan Polytechnic University(Natural Science) ,2024,43(6):116-128.

基于精英知识引导的多种群协作粒子群优化算法

张伟, 张润雨

河南理工大学 电气工程与自动化学院，河南 焦作 454000

摘要: 目的 为了解决粒子群优化（particle swarm optimization，PSO）算法易早熟收敛、后期收敛速度慢、精度低等问题，  方法 提出一种基于精英知识引导的多种群协作粒子群优化算法 （multi-group cooperation particle swarm optimization algorithm，MGCPSO）。首先，采用基于幂函数约束的logistic映射得到分布均匀的初始种群，加快寻优速度并提高找到最优解的概率；其次，在算法执行阶段动态划分多种群，并利用精英知识引导劣势粒子飞行，实现粒子间的信息共享和协同进化，降低粒子在解空间探索的盲目性；最后，综合融入精英知识的反向学习和极值扰动策略对粒子施加变异，帮助粒子扩大搜索区域并加强对最优邻域的精细探索。  结果 为验证MGCPSO的性能，在30维和100维的基准测试函数上进行了仿真实验研究，结果表明，相比于其他几种改进算法，提出的算法在收敛速度和收敛精度上均有良好表现。  结论  多种群协作粒子群优化可以有效避免算法早熟收敛和陷入局部最优，同时可以提高算法的全局搜索能力和局部开发能力。

关键词:粒子群优化算法;Logistic映射;多种群;精英知识;反向学习;极值扰动

doi:10.16186/j.cnki.1673-9787.2022070038

基金项目:国家自然科学基金资助项目（61703145）；河南省科技攻关项目（222102210213）

收稿日期:2022/07/17

修回日期:2022/12/29

出版日期:2024-09-24

Multi-group cooperation particle swarm optimization algorithm based on elite knowledge guidance

ZHANG Wei, ZHANG Runyu

School of Electrical Engineering and Automation， Henan Polytechnic University，Jiaozuo 454000，Henan，China

Abstract: Objectives To overcome premature convergence and improve the speed and accuracy of PSO， Methods a multi-group cooperation particle swarm optimization algorithm（MGCPSO） based on elite knowledge guidance was proposed.Firstly，the logistic mapping based on the power function constraint was used for obtaining a uniform initial distribution， which could speed up and improve the probability of finding the optimal solution.Secondly，multi-populations were dynamically divided during the algorithm execution phase and the elite knowledge was utilized for guiding the inferior particles，that could effectively realize information sharing and reduce the exploration blindness of particles.Finally，the mutation operation was carried out by combining the opposition-based learning and the extreme perturbation strategy with elite knowledge， which could help the particles expand their search area and strengthen the fine searching within the optimal neighborhood. Results  In order to verify the performance of MGCPSO，simulation experiments were conducted on 30-dimension and 100-dimension test functions.Simulation results showed that MGCPSO performed well in both convergence speed and convergence accuracy compared with other improved algorithms. Conclusions  The multi-group cooperative optimization could effectively avoid the problem of premature convergence and falling into local optimum，and could also improve the global search ability and local exploitation ability.

Key words:particle swarm optimization algorithm;logistic mapping;multi-group;elite knowledge;opposition-based learning;extreme disturbance