混合粒子群算法在高维复杂函数寻优中的应用

Hybrid particle swarm optimization algorithm for high-dimension complex functions

针对粒子群算法应用于复杂函数优化时可能出现过早收敛于局部最优解的情况,提出了一种改进的算法结构。通过构造单个粒子的最优序列代替单一的进化方向和类似于蚁群算法信息素表的选择机制,保留了粒子的多种进化可能方向,提高了粒子间的多样性差异,从而改善算法能力。算法同时设计了最优序列的加入规则和基于粒子群聚度的最优序列动态长度控制方法。改进后的混合粒子群算法保证了算法拥有更强的搜索能力,也保留了粒子群算法高效优化的特点。仿真实验证明,混合粒子群方法相对传统方法而言具有明显的精度优势。

To improve the PSO algorithm which is a new population based optimization algorithm against trapping into local minima, a hybrid method combining ant colony method with PSO-named hybrid PSO （HPSO） is presented. In HPSO the local best query is created to store lbest information for each particle. A strategy is also designed to choose which one in the lbest query may be the local best for PSO evolution process just like pheromone table in ACS. Lbest query keeps some potential good evolving directions in memory and therefore improves the diversity of particles. Furthermore congregation of all particles is calculated to decide the length of lbest query dynamically and acts as the criterion on whether a new solution should be added to the lbest query or not. With these ways PSO can break away from the local minima greatly, Experimental results show that the hybrid PSO is an appropriate method for highdimension complex functions with higher accuracy and speed.