基于区域动态概率变异及二分法的粒子群算法

Particle swarm optimization algorithm based on regional dynamic probability variation and dichotomy

针对粒子群算法容易陷入局部最优的问题,加入变异机制,结合新的变异条件,将搜索域划分为有限子区域,依据子区域的历史访问量,确定变异粒子变异到该区域的概率,使得算法能更加有效地跳出局部最优,提出区域二分法提高搜索精度。实验结果表明,基于区域动态概率的变异机制以及区域二分法的粒子群算法能更加有效地跳出局部最优,得到更加精确的目标解。

According to the problem of easily falling into local optimum of particle swarm optimization algorithm,a variation mechanism with new mutation conditions was added.The search area was divided into limited sub-regions and in accordance with the historical region-visited times of each sub-region,the probabilities of the sub-regions that particle varied to were determined.Combining with the regional dichotomy,it becomes a method which jumps out of local optimal more effectively,and obtains more accurate targets.