面向单目标优化的集成粒子群算法

串行粒子群算法广泛应用于多个领域,出现了多个变种,但解决不同种类的优化问题时性能有差异.为提高串行粒子群算法对各种优化问题的适应能力,提出一种集成粒子群优化算法.新算法使用Matlab的单程序多数据并行结构发挥单节点多核计算能力,通过设置外部档案分享不同粒子群的全局最佳位置,促进不同串行粒子群算法之间的信息交流,综合利用不同串行粒子群算法在解决不同类型优化问题的优势.在广泛使用的测试函数集上开展仿真实验,结果验证了新算法的有效性,与多个知名的串行粒子群算法相比,新算法在寻优性能上优势明显.新算法不仅能够提高粒子群算法的适应能力,而且,所采用的算法框架也适应于其他群智能算法,改善了算法的性能.

Ensemble particle swarm optimizer for single objective optimization

Serial particle swarm optimizer (SPSO) is popularly applied in many areas.SPSO variants are proposed to solve different kinds of optimization problems,but there are differences between the performances.Therefore, an ensemble particle swarm optimizer (EPSO) was proposed to improve SPSO's ability to adapt to problems.The parallel structure single program multiple data (SPMD) in Matlab was utilized to play single node multicore computing power.An outside archive was set to share the global best positions of different particle swarms and further facilitate information exchange of different SPSOs.SPSOs' advantages to solve different kinds of optimization problems were synthesized by the new algorithm.Simulation experiments conducted on a set of widely used benchmark functions verify the effectiveness of the new algorithm.Compared with several well-known SPSO variants, the new algorithm has a significant advantage in optimizing performance.The new algorithm can not only improve the adaptability of PSO, but also adapt to the other swarm intelligent algorithms to improve the algorithm performance.