摘要
针对传统神经网络非线性系统辨识算法存在收敛速度慢、易早熟、需人工设置网络结构及初始参数等问题,提出自适应小生境PSO非线性系统辨识方法。改进算法融合分层递阶算法和小生境PSO算法思想,联合优化网络结构及初始化参数,引入自适应灾变因子提高寻优精度。仿真实验表明,改进算法可提高辨识精度和收敛速度,能有效避免早熟现象,并可显著提高大空间、多峰值函数寻优效率。
An adaptive population niche optimization algorithm( PNOA) is for nonlinear identification problem,aiming at the problems of slow convergence speed,easy getting into local minimum and manually setting the initialization parameters and the Elman neural network structure. The proposed algorithm based on hierarchical algorithm and niche PSO algorithm can optimize the Elman neural network structure and initialization parameters together. By introducing the adaptive catastrophe,it improves the optimization precision. Simulation results show that using proposed algorithm can improve the identification accuracy and convergence speed,and can effectively avoid premature phenomenon,and significantly improve the large space and the optimization efficency of multimodal function.
出处
《西北大学学报(自然科学版)》
CAS
CSCD
北大核心
2015年第5期749-751,共3页
Journal of Northwest University(Natural Science Edition)
基金
国家"863"高技术研究发展计划资助项目(2010AA7080302)
西安市科技计划基金资助项目(XY1436)
关键词
微粒群算法
参数估计
非线性模型
PSO algorithm
parameter estimation
nonlinear model