粒子群算法稳定性的参数选择策略分析

Analysis on Parameter Selection for Stability of Particle Swarm Optimization Algorithm

在对标准粒子群算法位置期望及位置方差进行稳定性分析的基础上,研究加速因子的取值对位置期望及位置方差的影响。从提高系统稳定性的角度考虑,建议个体认知加速因子为1.85,社会认知加速因子为2,通过五个经典测试函数与Kennedy提出的设定两个加速因子均为2的建议做比较。大量实验统计结果表明对于同一种算法本文建议的参数设置在寻优速度和方差上均优于Kennedy提出的参数设置,更有利于工程应用。

The effect of acceleration factors on expectation and standard deviation values of the position was studied based on the stability analysis of Particle Swarm Optimization Algorithm. From the viewpoint of improving system stability, setting the cognitive acceleration factor as 1.85 and the social acceleration factor as 2 is a better choice than that Kennedy proposed both as 2. Moreover, five benchmark functions were used to test the influence of these parameters on system stability in the same strategy with numerous experiments. The statistical results from the new strategy of parameters selection also show higher efficiency than that Kennedy suggested.