基于多准则决策的塔状粒子群优化算法

Pyramid Particle Swarm Optimization Based on Multi Criterions

针对经典粒子群优化算法存在早熟、收敛精度低和收敛速度慢的问题,提出了一种新的改进算法.该算法采用了塔状优化互联机制,底层粒子群负责寻找局部最优解,顶层粒子负责收集、反馈全局最优解,为底层种群提供全局最优信息,建立共享学习机制.顶层粒子一旦发现停滞现象,将通知底层粒子群采用细菌觅食优化、随机初始化等停滞优化策略,以改善粒子群的收敛速度.实验结果表明,与同类算法相比,改进算法具有更好的寻优能力,改善了粒子群的收敛精度和收敛速度.

Aiming at the problems of Particle Swarm Optimization （PSO）, such as premature, low convergence precision and slow convergence rate, a newly improved algorithm is proposed in which the whole particles are organized in a pyramid structure. In the optimizing process, the swarms in the bottom layer share information among groups under the coordination of the global optimal particle in the top layer. Once the swarms stop moving, the groups in the bottom layer will take multi criterions to optimize the swarms＇ convergence rate. Experimental results suggest that the proposed algorithm bears better efficient and stronger optimizing ability, and improves optimizing precision and convergence rate more than some other existing optimization algorithms.