基于周期性演化策略的粒子群优化算法

Particle swarm optimization based on periodic evolution strategy

针对标准粒子群优化算法(PSO)在寻优过程中容易出现早熟的问题,提出一种基于周期性演化策略的粒子群优化算法,该策略通过在速度更新方程中构建基于粒子群能量的粒子群最优值扰动项,使得粒子群能量在演化过程中可以周期性变化,相比标准PSO算法,当粒子群能量较大时,能够增强局部搜索能力;当粒子群能量较小时,能够增强全局搜索能力,典型优化问题的仿真结果表明,所提出的算法与线性下降惯性权重粒子群优化(LWPSO)和PSO算法相比,优化性能得到了显著提高。

To overcome the problem of premature convergence frequently appeared in standard particle swarm optimization（PSO） algorithm, an improved PSO algorithm based on periodic evolution strategy（PSO-PES） is proposed. Thus, the problem of premature convergence of the PSO algorithm is prevented. The local search ability and global search ability are enhanced by an adaptive function based on the information of velocities of the particles. Simulations results for several typical test functions show that PSO-PES possesses more powerful global search capabilities, better convergence rate and better performance of optimization.