Identification Of Hammerstein Model Based On The Principle Of Compressed Sensing

The problem how to identify the Hammerstein CAR model orders and parameters at the same time is mainly explored in this paper.The domestic and foreign reach status about the identification methods of the Hammerstein model is summarized.And two new identification methods about Hammerstein CAR model are proposed on basis of the compressive sensing(CS)principle: one is based on the key term separation principle and the orthogonal matching pursuit(OMP)algorithm of the CS principle,the other is the instrumental variable base on OMP(IV-OMP)identification method.Simulation results illustrate that the two methods,both based on the CS,are effective.The content,explored in this paper,is as follows:First,introducing and overviewing the structure and identification methods of the modular system respectively.The basic least squares(LS)?the recursive least squares(RLS)?auxiliary model identification idea?and the key term separation principle are also introduced in Introduction.Second,in terms of the Hammerstein CAR model,OMP identification method,based on the key term separation principle is proposed by combining the key term separation principle and OMP.And using the key term separation principle to get the linear parametric model,which can avoid overparametrization problem caused by coupling between nonlinear part and linear part.To reduce the influence of noise,threshold orthogonal matching pursuit(TH-OMP)algorithm is adopted to identified system parameters and orders at the same time.Third,in terms of the Hammerstein CARMA model,another OMP identification method based on instrumental variable is also proposed through combining auxiliary model identification idea and OMP.Using the filtering technique and instrumental variable idea to convert the form of system equation and decouple parameters,so system output can be expressed as a linear regression equation with all parameters to be identified.Replacing respectively unknown parameters with their estimates,to solve the problem caused by those unknown parameters.A method,about identifying system parameters and orders at the same time and based on filtering technique and instrumental variable idea,is proposed.At last,Both two proposed methods are simulated in MATLAB.Simulation results illustrate that two methods has higher identification efficiency?small amount of arithmetic labor and can be used in online identification.