基于改进帝王蝶算法的群优化算法加速框架  

Swarm Optimization Algorithm Acceleration Framework Based on Improved Monarch Butterfly Algorithm

作　　者：巨重阳 刘立群[1] JU Chong-yang;LIU Li-qun(School of Information Science and Technology,Gansu Agricultural University,Lanzhou 730070,China)

机构地区：[1]甘肃农业大学信息科学技术学院,甘肃兰州730070

出　　处：《计算机技术与发展》2024年第10期16-23,共8页Computer Technology and Development

基　　金：甘肃省高校教师创新基金项目(2023A-051);甘肃农业大学青年导师基金资助项目(GAU-QDFC-2020-08);甘肃省科技计划资助(20JR5RA032)。

摘　　要：该文提出了一个通过改进的帝王蝶算法的群优化算法加速框架,该架构旨在改善群智能算法的优化特性和收敛效率。传统的帝王蝶算法具有收敛速度慢和容易进入局部最优的问题。为解决上述问题,该文在帝王蝶算法中引入了一系列改进措施,改进帝王蝶算法并使其作为加速框架与其它群智能算法组合使用。首先,通过应用混沌映射来更新群体起始状态以增强其成员的多元性,这能有效扩大搜寻范围并采用反向学习和随机干扰取代传统的移动操作,从而提升整体的稳定性,防止算法被困于局部最优。此外,采用非线性的自适应运算因子,初期强化了变异力以避开局部最优,后期减弱它以便深入寻找更好的结果,进而提高了精度。通过组合不同的群智能优化算法30维下寻优在10个测试函数的综合评估,验证了该算法框架可以有效提升其它群智能优化算法的收敛速度和精度。We propose a swarm optimization algorithm acceleration framework through the improved monarch butterfly algorithm,which aims to improve the optimization characteristics and convergence efficiency of swarm intelligence algorithms.The traditional monarch butterfly algorithm has the problems of slow convergence and easy entry into local optimality.In order to solve the above problems,we introduce a series of improvement measures into the monarch butterfly algorithm,which improve the monarch butterfly algorithm and use it as an acceleration framework to integrate with other swarm intelligence algorithm.First,chaos mapping is applied to update the starting state of the group to enhance the diversity of its members,which can effectively expand the search range and use reverse learning and random interference to replace traditional movement operations,thereby improving the overall stability and preventing the algorithm from being trapped in the local optimum.In addition,nonlinear adaptive operation factors are used to strengthen the variability in the early stage to avoid local optimality,and weaken it in the later stage to search for better results in depth,thereby improving accuracy.By combining different swarm intelligence optimization algorithms for comprehensive evaluation of 10 test functions in 30 dimensions,it is verified that the proposed algorithm framework can effectively improve the convergence speed and accuracy of other swarm intelligence optimization algorithms.

关 键 词：帝王蝶算法 反向学习 混沌映射 群优化算法 自适应算子 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]