计算机工程与应用 ›› 2016, Vol. 52 ›› Issue (24): 171-177.

• 模式识别与人工智能 • 上一篇    下一篇

多策略融合粒子群算法及其收敛性分析

叶洪涛1,2,皮倩瑛1,2   

  1. 1.广西科技大学 电气与信息工程学院,广西 柳州 545006
    2.广西汽车零部件与整车技术重点实验室,广西 柳州 545006
  • 出版日期:2016-12-15 发布日期:2016-12-20

Multi-strategy particle swarm optimization algorithm and its convergence analysis

YE Hongtao1,2, PI Qianying1,2   

  1. 1.School of Electrical and Information Engineering, Guangxi University of Science and Technology, Liuzhou, Guangxi 545006,China
    2.Guangxi Key Laboratory of Automobile Components and Vehicle Technology, Liuzhou, Guangxi 545006, China
  • Online:2016-12-15 Published:2016-12-20

摘要: 针对使用经典线性递减策略来确定惯性权重的粒子群优化算法在实际运算过程中与粒子寻优的非线性变化特点不匹配的问题,提出一种改进的粒子群算法。该算法采用多次随机初始化的策略初始种群位置,再对惯性权重引入随机因子,使其基于粒子适应度大小来动态调节惯性权重,更好地引导粒子进行搜索,提高算法的收敛精度,并证明其能以概率1全局收敛。为了验证该算法的寻优性能,通过8个经典测试函数将标准粒子群算法、惯性权重递减的粒子群算法及提出的改进算法在不同维度下进行测试比较。结果表明,该算法的寻优精度更高。

关键词: 粒子群算法, 初始化策略, 动态调节, 惯性权重

Abstract: For the use of classical linear decreasing strategy to determine the inertia weight in the actual operation process optimization and particle nonlinear changes of mismatch, this paper proposes an improved particle swarm optimization. The algorithm uses the initial population policy position random initialization times, and then introduces random factor to inertia weight, based on particle fitness size dynamically adjusts inertia weight, better guides the search for particles, improves the convergence precision and proves it is capable of global convergence with probability 1. In order to verify the performance of the optimization algorithm, by 8 classic test functions, PSO, decreasing inertia weight particle swarm optimization and the proposed algorithm are tested and compared with different dimensions. The results show higher optimization accuracy of the algorithm.

Key words: particle swarm algorithm, initialization strategy, dynamic adjustment, inertia weight