Anti-disturbance Identification Of Hammerstein Type Model For Nonlinear Systems With Time Delay

Time delay and load disturbances are ubiquitous in industrial production processes.In this thesis,the nonlinear Hammerstein output error model(HOE model)with time delay is used to study the model parameter identification under different load disturbances.For the time-delay Hammerstein nonlinear system under time-varying load disturbance.A forgetting factor least squares algorithm is proposed.The idea of the algorithm views the load disturbance a time-varying parameter,by augmenting the information vector and the parameter vector,obtaining the linear regression form of the output response.The error criterion function of an adaptive forgetting factor is defined.The forgetting factor least squares recursive algorithm is obtained.But the parameter vector contains two types of model parameters and load disturbance.Therefore,the scalar adaptive forgetting factor is transformed into a adaptive forgetting factors matrix.The forgetting factor matrix least squares recursive algorithm is obtained.In the adaptive forgetting factor matrix,two adaptive forgetting factors are included,which are used to speed up the model parameter convergence rate and track the time-varying load disturbance.For the estimation of the time-delay parameter,the minimum value of the error criterion function in the time-delay interval is found by giving the time-delay search interval,and the time-delay parameter is obtained.Based on the separation strategy and the interaction estimation criterion,the Hammerstein nonlinear system is decomposed into two subsystems.These two subsystems contain model parameters and time-varying load disturbances,respectively.The load is treated as a parameter in one of the subsystems.Then the adaptive forgetting factor error criterion function are constructed for the two subsystems respectively.Two least squares algorithms is presented for estimating the model parameters and load disturbance.The two algorithms perform interaction estimation iteration,and the relevant parameters can be obtained.Based on the persistent excitation condition,the convergence of the proposed separation strategy algorithm is analyzed.Finally,the Hammerstein system is decomposed into two subsystems based on the separation strategy.In order to estimate the system model parameters,a multi-innovation error criterion function with adaptive forgetting factor is defined by using multiple innovation methods,and a multi-innovation recursive least squares algorithm is obtained.The idea of this method is to expand the scalar innovation into a matrix innovation,thus improving the accuracy of parameter estimation.Due to the uncertainty of the disturbance,multi-innovation will cause error accumulation.Therefore,the standard least squares algorithm is used for estimation.Through the interaction estimation of these two algorithms,the convergence speed of model parameter estimation and real-time tracking of load disturbance are accelerated.For the unknown intermediate variables contained in the information vector in the above method,the output of the model is reconstructed by the auxiliary model.