一种面向大规模复杂全局优化的流场吸引动态涡流搜索算法  

A Dynamic Vortex Search Algorithm of Flow Field Attraction for Large-Scale Complex Global Optimization

作　　者：刘景森[1,2] 李浩然 李煜[3,4] 周欢 LIU Jing-sen;LI Hao-ran;LI Yu;ZHOU Huan(Henan International Joint Laboratory of Intelligent Network Theory and Key Technology,Henan University,Kaifeng,Henan 475004,China;College of Software,Henan University,Kaifeng,Henan 475004,China;Institute of Management Science and Engineering,Henan University,Kaifeng,Henan 475004,China;Business School,Henan University,Kaifeng,Henan 475004,China)

机构地区：[1]河南大学河南省智能网络理论与关键技术国际联合实验室,河南开封475004 [2]河南大学软件学院,河南开封475004 [3]河南大学管理科学与工程研究所,河南开封475004 [4]河南大学商学院,河南开封475004

出　　处：《电子学报》2023年第7期1949-1955,共7页Acta Electronica Sinica

基　　金：国家自然科学基金(No.72104069);河南省重点研发与推广专项(No.222102210065);河南省重大科技专项(No.201300210400);河南大学研究生培养创新与质量提升行动计划项目(No.SYLYC2022150)。

摘　　要：为了拓展涡流搜索算法的应用能力,提升其求解复杂优化尤其是大规模复杂优化问题的性能,本文提出了一种基于流场吸引流动、逐维半径试探更新和领导层决策机制的动态涡流搜索算法.首先,本文在算法中引入压强差的概念,使候选解依据压强差进一步向着较优解移动,提高算法整体的搜索质量;然后,算法通过逐维半径更新策略,有效避免了在某一维陷入局部极值的情况;最后,本文在中心点的更新中引入领导层决策机制,提高算法快速确定最佳区域的能力.在计算机仿真部分,本文将该改进算法与多组具有不同代表性的对比算法分别在CEC2017套件的100维和CEC2010套件的1000维上进行了极值优化分析,结果表明改进后的算法无论是在高维问题还是大规模复杂问题上的寻优结果都能领先其他代表性对比算法多个数量级,具有很好的收敛性能.In order to expand the application capabilities of the vortex search algorithm and improve its performance in solving complex optimization problems,especially large-scale complex optimization problems,a vortex search algorithm is proposed based on attractive flow field operation,dimension-by-dimension dynamic radius,and leadership decision-mak⁃ing mechanism.Firstly,this paper introduces the concept of pressure difference in the algorithm.Candidate solutions fur⁃ther move towards the optimal solution according to pressure difference,which improves the overall search quality of the al⁃gorithm.Then,a dimension-by-dimension radius updating strategy is used to avoid trapping into the local minima in a cer⁃tain dimension.Finally,the leadership decision-making mechanism is introduced into updating the circle center,which im⁃proves the algorithm's ability and quickly determines the optimal region.In the simulation section,the improved algorithm and multiple sets of representative comparison algorithms are analyzed for extreme value optimization on the 100 dimen⁃sions of the CEC2017 suite and 1000 dimensions of the CEC2010 suite,respectively.The results show that the improved algorithm can outperform other representative algorithms by multiple orders of magnitude in both high-dimensional and large-scale complex problems,and has good convergence performance.

关 键 词：涡流搜索算法 流场吸引 逐维更新 领导层决策机制 CEC2017 极值优化 大规模全局优化 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]