基于BP神经网络对时滞系统的模型参数辨识仿真  

Simulation of Model Parameter Identification for Time-delayed Systems Based on BP Neural Network

作　　者：武俊丽[1] 李建辉[1] 史庆武[1] 

机构地区：[1]佳木斯大学信息电子技术学院

出　　处：《佳木斯大学学报（自然科学版）》2010年第4期496-498,501,共4页Journal of Jiamusi University：Natural Science Edition

基　　金：佳木斯大学科研项目(L2009-149,L07,L2007-43)

摘　　要：主要研究了基于BP神经网络对时滞系统的参数辨识,分析了两种辨识结构和两种建模方法,对系统被控对象的建模采用了神经网络正模型,辨识结构为串-并联型.考虑加强BP网络的泛化能力,用随机数据去训练网络,然后得到训练后的权值,给一个阶跃信号,利用交叉两点法,从而得到时滞系统的特征参数.通过仿真,基于BP网络对时滞系统的参数辨识是有效的.In this paper,the simulation of model parameter identification for time-delayed systems based on BP neural network was studied. The positive model of NN and the serial-parallel type were employed. In order to enhance the generalization,the random data were used to train the BP NN and the weights of NN were obtained. Then the characteristic parameters of the time-delayed systems were obtained on the cross two-point algorithm when the input was step signal. The simulation results indicate that the proposed method is effective.

关 键 词：神经网络 BP算法 系统辨识 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]