混合粒子群优化算法及其收敛性分析

Composite Particle Swarm Optimization and Analysis of Convergence

摘要:

为解决标准粒子群优化算法不能保证全局收敛,寻优精度低,尤其在高维函数优化方面易陷入局部极小值等问题,提出一种融合Kent混沌映射、云模型理论和布谷鸟搜索的混合粒子群优化算法(CPSO);CPSO算法采用混沌初始化种群位置、全局开发及局部开采的均衡搜索、多子种群协同进化等改进策略,同时从随机优化算法的全局收敛准则角度对CPSO算法的全局收敛性进行证明,并给出了CPSO算法的时间复杂度分析;经典的benchmark测试函数的实验统计结果表明,CPSO算法在收敛性、寻优精度、稳定性等方面均优于经典算法。

Abstract:

In order to cope with low accuracy and disability in convergence of PSO,a composite PSO algorithm is proposed,combined with Kent mapping,cloud model and cuckoo search. Furthermore, chaotic initialization, exploration and exploitation as well as multi-swarm strategies are adopted. In addition,convergence analysis and time complexity of CPSO are conducted. Finally,experiment results of standard benchmark function show that compared with traditional methods,our proposed algorithm is excellent in both convergence, accuracy and robustness.