基于融合指标分区的多策略竞争群优化算法  

Multi-strategy Competitive Swarm Optimization Algorithm Based on Fusion Index Partition

作　　者：张伟[1] 郑洪轩 ZHANG Wei;ZHENG Hongxuan(School of Electrical Engineering and Automation,Henan Polytechnic University,Jiaozuo 454003,China)

机构地区：[1]河南理工大学电气工程与自动化学院,河南焦作454003

出　　处：《信息与控制》2025年第1期97-109,共13页Information and Control

基　　金：国家自然科学基金项目(61703145);河南省科技攻关项目(222102210213)。

摘　　要：在粒子群优化算法中,多样性丢失会导致算法过早收敛,造成种群多样性和收敛性不足。为更好平衡种群多样性与收敛性,提出一种基于融合指标分区的多策略竞争群优化算法(multi-strategy competitive swarm optimization algorithm based on fusion index partition,FIMSCSO)。首先,为同时评估不同子种群的多样性和收敛性,提出一种利用融合指标的分区方法,更有效区分子种群性能,提升子种群粒子的搜索效率。其次,对通过分区获得的子种群设计多重学习机制。子种群内部学习时,为使子种群朝向更利于平衡收敛性与多样性的方向学习,将粒子性能与寻优过程结合,提出新的获胜粒子学习方法,提高种群粒子参与度和算法寻优效率;子种群间学习时,为减少整个种群陷入局部最优的可能性,引入改进三重竞争机制促进子种群间信息交流,帮助提高收敛精度。最后,设计实时停滞检测和变异策略帮助种群跳出局部最优。理论证明了所提算法的收敛性。实验结果表明,相比其他改进算法,所提FIMSCSO算法具有良好的收敛精度和寻优效率。Loss of diversity will lead to premature convergence in particle swarm optimization,which can lead to insufficient diversity and convergence of the swarm.To balance iversity and convergence,a multi-strategy competitive swarm optimization algorithm based on fusion index partition(FIMSCO)is presented.Firstly,a swarm partition technique using the fusion index is suggested,and the diversity and convergence of the sub-swarms are evaluated simultaneously,which enhances the search efficiency of the particles in the sub-swarms.While internal learning in sub-swarms,particle performance is merged with algorithm and a new winner particle learning method is proposed,which can guide the sub-swarms toward the more favorable balancing between the convergence and the diversity.And then the participation of the particles and the optimization efficiency of the algorithm are improved.While learning among the sub-swarms,to prevent the whole swarm from falling into the local optima,an improved triple-competition mechanism is introduced,which can promote the information exchange between the sub-swarms and can assist the improvement of the convergence accuracy.Finally,the real-time stagnation detection and the mutation strategy are designed to prevent the swarm from falling into the local optima.The convergence of the proposed algorithm is proved theoretically.Experimental results reveal that the proposed FIMSCSO exhibits excellent convergence accuracy and optimization efficiency compared with other algorithms.

关 键 词：融合指标分区 多策略学习 竞争群优化算法 粒子群优化算法 实时停滞检测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]