一种基于搜索历史信息的粒子群算法

A particle swarm algorithm based on search history information

针对粒子群算法易过早收敛、陷入局部最优,从而导致收敛精度不足等问题,提出一种基于搜索历史信息的粒子群算法。该算法利用粒子群算法速度迭代公式产生的新速度与上一代飞行速度协同学习,以此作为新的粒子速度更新粒子个体;对历史飞行速度进行学习可以扩大粒子搜索区域,增强算法寻优能力,有效改善早熟收敛问题;构建多种策略对学习因子进行差异化选取,达到多样化搜索路径的目的。采用CEC 2014不同类型基准测试函数进行仿真试验,与其他经典粒子群算法进行对比表明,所提算法具有更稳定、更优异的综合性能。

Due to the problems that the particle swarm algorithm is easy to converge prematurely and fall into the local optimum, a new particle swarm variant based on search history information is proposed. The algorithm constructs the new speed generation formula aiming to learn the previous flying beneficial information to update the individual particle.Learning the historical flying information can expand the particle search area and enhance the algorithm′s optimization ability.And it finally improves the phenomenon of premature convergence.The proposed algorithm designs a variety of strategies to select different learning factors so as to achieve the purpose of diversified search paths.Different types of benchmark functions of CEC2014 are used for simulation experiments, and the comparison with other classical particle swarm algorithms shows that the proposed algorithm has more stable and better comprehensive performance.