摘要
针对高维复杂函数的标准粒子群算法常存在早熟收敛问题,提出一种让初始化粒子群的位置"相对均匀"并且随着搜索阶段不同而改变认知学习因子和社会学习因子的算法。该算法可以在搜索前期增强全局搜索,使之不陷入局部最优,而到搜索后期增强局部搜索能力,使之得到更精确全局最优解。通过五个典型测试函数的实验结果对比,可以清楚地表明改进后的算法得到的最优解更加接近真实的最优解。
Partical swarm optimization usually lead to premature convergence, especially in optimizing high-dimensional functions. In this paper, particle relatively uniform distribute in search space when initializing them, two acceleration coefficients vary with iterations and searching stage.The algorithm can make the search behave well in globle searching avoiding local optimum at preactive stage.meanwhile, it strengthens local searching to get more precise globle optimum.Five typical benchmark functions＇ experiment simulation show that proposed algorithm can more better globle optimum.
关键词
粒子群算法
相对均匀
学习因子
全局搜索与局部搜索的平衡
particle swarm optimization algorithm
relatively uniform
acceleration coefficient
balance in local and globle searching
