基于改进PSO算法的Elman网络系统辨识  

Elman network system identification based on improved PSO algorithm

作　　者：岳颀[1] 

机构地区：[1]西安邮电大学,陕西西安710121

出　　处：《自动化与仪器仪表》2015年第8期201-203,共3页Automation & Instrumentation

摘　　要：针对动态BP算法训练Elman网络时需要通过经验确定网络结构的缺点,提出一种可同时训练网络结构和网络权值的改进种群小生境PSO算法.并实现了基于该算法的Elman神经网络对非线性模型的辨识.仿真结果表明,改进算法扩展了PSO算法的全局优化能力和局部搜索能力,提高了收敛率和解的质量,并有效提高了收敛速度。A new improved species niching particle swarm optimization algorithm is presented, which can optimize the structure and parameters of Elman simultaneously, aiming the flaws that usual algorithm of deciding the initial structure of Elman depends on expert knowledge. The simulation results show that the improved method have strong global search ability while improving the con-vergence speed and accuracy, and achieve a better effect.

关 键 词：ELMAN 种群小生境PSO 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]