多粒子群协同优化算法

Particle Swarms Cooperative Optimizer

提出一种多粒子群协同优化(PSCO)方法.PSCO是2层结构:底层用多个粒子群相互独立地搜索解空间以扩大搜索范围;上层用1个粒子群追逐当前全局最优解以加快算法收敛.这些粒子群含的粒子数以及粒子状态更新策略不要求相同.为改善粒子群容易陷入局部极小的弱点,提出扰动策略,当1个粒子群的当前全局最优解未更新时间大于扰动因子时,重置粒子的速度,迫使粒子群摆脱局部极小.用Rosenbrock函数等3种基准函数做优化实验表明,PSCO性能优于经典PSO,FPSO和HPSO等算法.

A particle swarms cooperative optimizer (PSCO) algorithm with two layers framework is proposed. Particle swarms are employed to search best solution in the solution space independently in the bottom layer, and a single swarm is employed in top layer. Sates of the particles of the top swarm are updated based on global best solution has been searched by all the particle swarms both in bottom and top layer. Both the particle numbers of the swarms and updating schemes of particle states are independence. A disturbance factor is added to a particle swarm optimizer (PSO) for improving PSO algorithms' performance. When the time of the current global best solution having not been updated is longer than the disturbance factor, the particles' velocities will be reset in order to force swarms getting out of locally minimizers. Three benchmark functions are used in experiments, and the experimental results show that the performances of PSCO are superior to that of classical PSO and fuzzy PSO and hybrid PSO.