可行域连通性对智能算法的寻优性能影响研究  

Study on the Influence of Feasible Domain Connectivity on the Optimization Performance of Intelligent Algorithms

作　　者：王亚丹 赵启林[1] 马森 WANG Yadan;ZHAO Qilin;MA Sen(College of Mechanical and Power Engineering,Nanjing Tech University,Nanjing 211816,Jiangsu,China;Training Base,Officers College of PAP,Guangzhou 510440,Guangdong,China)

机构地区：[1]南京工业大学机械与动力工程学院,江苏南京211816 [2]武警警官学院训练基地,广东广州510440

出　　处：《山西师范大学学报（自然科学版）》2024年第2期9-21,共13页Journal of Shanxi Normal University(Natural Science Edition)

基　　金：军委科技委技术加强基金(2020-JCJQ-JJ-518);军队后勤开放研究科研项目“高墩大跨桥梁的快速化抢修技术研究”资助(ALJ18C003).

摘　　要：目前对于可行域连通性等搜索空间特性变化对算法寻优效率与全局寻优能力的影响规律研究较少.基于以上背景,对比分析了GA(Genetic Algorithm)、DE(Differential Evolution)、PSO(Particle Swarm Optimization)三种典型智能算法相互之间的差异性,指出DE算法具有更强全局寻优能力的基本机理;而后采用三种算法求解了利用四组测试函数设计出的连通可行域、不连通可行子域大小一致以及不连通可行子域大小不一致三种优化模型,并将其结果进行了对比分析.分析表明:DE算法的寻优性能要远远优于GA和PSO算法,这与前人研究结论基本一致;同时发现对于可行域不连通情况下全局最优解位于较小可行域的高维优化问题,DE算法也很难找到全局最优解,现有智能算法仍旧不能完全满足实际需求,由此提出了一种新的基于全局寻优的自适应差分进化算法.At present,there are few studies on the influence of changes in search space characteristics such as feasible domain connectivity on the optimization efficiency and global optimization ability of algorithms.Based on the above background,the differences between the three typical intelligent algorithms of GA(Genetic Algo-rithm),DE(Differential Evolution)and PSO(Particle Swarm Optimization)are compared and analyzed,and the basic mechanism of DE algorithms with stronger global optimization ability is pointed out.Then,three optimi-zation models of connected feasible domain,incommensurable subdomain size consistency and incommensible feasible subdomain size inconsistency designed by using four sets of test functions are solved using three algo-rithms,and the results are compared and analyzed.The analysis shows that the optimization performance of the DE algorithm is far better than that of the GA and PSO algorithms,which is basically consistent with the conclu-sions of previous research.At the same time,it is found that for the high-dimensional optimization problem of the global optimal solution in the smaller feasible domain when the feasible domain is not connected,the DE algo-rithm is also difficult to find the global optimal solution,and the existing intelligent algorithm still cannot fully meet the actual needs,so a new adaptive differential evolution algorithm based on global optimization is pro-posed.

关 键 词：可行域连通性 差分进化算法 寻优性能 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]