基于高斯核函数的Hammerstein非线性系统辨识  

Identification of Hammerstein Nonlinear System Based on Gaussian Kernel Functions

作　　者：郑天 李峰 罗印升 刘冉冉[2] 顾亚 ZHENG Tian;LI Feng;LUO Yin-sheng;LIU Ran-ran;GU Ya(School of Electrical and Information Engineering,Jiangsu University of Technology,Changzhou 213001,China;School of Automobile and Traffic Engineering,Jiangsu University of Technology,Changzhou 213001,China;School of Electrical Engineering and Automation,Changshu Institute of Technology,Changshu 215500,China)

机构地区：[1]江苏理工学院电气信息工程学院,江苏常州213001 [2]江苏理工学院汽车与交通工程学院,江苏常州213001 [3]常熟理工学院电气与自动化工程学院,江苏常熟215500

出　　处：《控制工程》2022年第11期2034-2041,共8页Control Engineering of China

基　　金：国家自然科学基金资助项目(62003151,62003150,61903050);江苏省自然科学基金资助项目(BK20191035);常州市科技计划项目(CJ20220065);江苏高校“青蓝工程”资助。

摘　　要：提出了一种基于高斯核函数的Hammerstein非线性系统参数辨识方法。Hammerstein非线性系统由一个静态非线性模块和一个动态线性模块串联组成,利用高斯核函数神经网络和传递函数模型分别建立Hammerstein系统的静态非线性模块和动态线性模块。首先,基于可分离信号的输入输出数据,采用相关性分析方法估计动态线性模块的参数,有效抑制噪声的干扰。其次,针对Hammerstein非线性系统的不可测噪声项,利用残差的估计值代替不可测变量,推导了递推增广最小二乘辨识方法,根据随机信号的输入输出数据辨识静态非线性模块和噪声模型的参数。仿真结果表明,针对有色噪声干扰的Hammerstein非线性系统,所提方法具有较好的辨识精度和鲁棒性。A parameter identification method of the Hammerstein nonlinear system based on Gaussian kernel functions is presented in this paper. The Hammerstein nonlinear system consists of a static nonlinear block and a dynamic linear block. The static nonlinear block and dynamic linear block of the Hammerstein system are established by using Gaussian kernel-based neuro network and transfer function model, respectively. To begin with, the parameters of the dynamic linear block are estimated by using the correlation analysis method based on the input-output data of separable signals, which can effectively suppress noise interference. In addition, in view of the unmeasurable noise term of the Hammerstein nonlinear system, the recursive extended least square method is derived by means of residual estimation instead of the unmeasurable variables. Then, the parameters of the static nonlinear block and noise model are identified based on the input-output data of random signals.The simulation results show that the proposed method has good identification accuracy and robustness for the Hammerstein nonlinear system with colored noise interference.

关 键 词：HAMMERSTEIN系统 非线性系统辨识 二进制-随机复合信号 相关性分析法 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]