多策略融合的粒子群优化算法  

Multi-strategy fusion particle swarm optimization

作　　者：王惠敏 孙滢 高岳林 WANG Hui-min;SUN Ying;GAO Yue-lin(Collage of Computer Science and Engineering,North Minzu University,Yinchuan 750021,Ningxia,China;Collage of Mathematics and Information Science,North Minzu University,Yinchuan 750021,Ningxia,China;Ningxia Key Laboratory of Intelligence Information and Big Data Processing,Yinchuan 750021,Ningxia,China)

机构地区：[1]北方民族大学计算机科学与工程学院,宁夏银川750021 [2]北方民族大学数学与信息科学学院,宁夏银川750021 [3]宁夏智能信息与大数据处理重点实验室,宁夏银川750021

出　　处：《宝鸡文理学院学报（自然科学版）》2023年第4期1-9,14,共10页Journal of Baoji University of Arts and Sciences(Natural Science Edition)

基　　金：宁夏自然科学基金项目(2021AAC03185);宁夏高等教育一流学科教育基金项目(NXYLXK2017B09);北方民族大学重大科研专项项目(ZDZX201901);南京证券支持基础学科研究项目(NJZQJCXK202201);北方民族大学研究生创新项目(YCX22192)。

摘　　要：目的针对传统粒子群优化算法存在全局搜索时解的质量较低和局部搜索时易陷入局部最优的缺陷,提出一种多策略融合的粒子群优化算法。方法首先,均值自适应判断机制将进化过程有效地划分为全局勘探和局部开发阶段,在不同阶段使用逆向思维的惯性权重。其次,在勘探阶段采用线性组合速度更新公式和自适应位置更新公式。最后,在开发阶段引入爬山算法。结果将多策略融合的粒子群优化算法与标准粒子群优化算法、2种粒子群优化算法变体、3种群智能优化算法在12组测试函数中进行对比实验。所提算法在11组测试函数中完全优于6种对比算法且可以收敛到全局最优。结论多策略融合的粒子群优化算法有效提升了传统粒子群优化算法的收敛精度和速度。Purposes—To propose a multi-strategy fusion particle swarm optimization to address the shortcomings of traditional particle swarm optimization in that the solution quality is low in global search and prone to fall into local optimum in local search.Methods—Firstly,the mean-adaptive judgement mechanism effectively divides the evolutionary process into global exploration and local exploitation phases,while inertia weights of inverse thinking are used in the different phases.Secondly,a linear combination of velocity update formulas and adaptive position update formulas are used in the exploration phase.Finally,the hill-climbing algorithm is introduced in the exploration phase.Results—The multi-strategy fusion particle swarm optimization is compared with the standard particle swarm optimization,2 variants of the particle swarm optimization and 3 kinds of state-of-the-art intellgence algorithms in 12 sets of benchmark functions for experiments.The experimental results show that the proposed algorithm completely outperforms the 6 compared algorithms in 11 sets of benchmark functions and converges to the theoretical global optimum.Conclusions—The multi-strategy fusion particle swarm optimization effectively improves the convergence accuracy and speed of the traditional particle swarm optimization.

关 键 词：粒子群优化算法 均值自适应判断机制 逆向思维的惯性权重 爬山算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]