Parameter Identification Mehtods Rea-Search For A Class Of Hammerstein System

System identification refers to the experimental approach that con-sists of determining system models by fitting experimental data to a suit-able model structure. As the basis of automatic control, system identifica-tion, control theory and state estimation are three major components of modern control theory. Parameter identification is an important part of system identification. In industry process, most systems have nonlinear characteristic, so the research for nonlinear system parameter identifica-tion has important practical significance. Hammerstein system, which consists the interaction of linear time-invariant dynamic subsystems and static nonlinear elements, is mainly discussed in the study. Some parame-ter identification methods are researched for Hammerstein system in complex nonlinear characteristic conditions. The research for a complex linear system parameter identification problem is also discussed, and can lay the foundation for which as the linear part of Hammerstein system. This dissertation introduces the corresponding algorithm theory deriva-tion and simulation experiment under the situations above, and gets the corresponding results. The main work of this dissertation is as following. 1. The development process and research current situation of system identification particularly the Hammerstein nonlinear parameter identifi-cation is introduced. The basic knowledge of identification is also de-scribed in the study such as convectional signal, common models and methods.2. A memory-less saturation nonlinear element with hysteresis cha-racteristic is researched. And let this nonlinear element be as the nonli-near part of Hammerstein equation error model system or Hammerstein controlled autoregressive model (Hammerstein CAR model). The Ham-merstein model parameters are identified by the stochastic gradient me-thod and its improved algorithm with forgetting factor. The mul-ti-innovation method is also used while the multi-innovation gradient al-gorithm with forgetting factor is researched to estimate the model para-meters. The estimation result can be achieved by simulation.3. A uniform description which can describe four typical disconti-nuous nonlinear characteristics such as dead-zone, saturation, hysteresis and preload is introduced. Then this nonlinear description combined with the Hammerstein output error model (Hammerstein OE model) is re-searched. The recursive least squares method is used for this system and the simulation shows the identification results.4. Considering a complex linear output error model. The parameters of this system is estimated by the auxiliary model least squares algorithm and stochastic gradient algorithm. Then combining with the iterative me-thod, the iterative least squares algorithm and the iterative stochastic gra- dient algorithm are researched for the parameter identification problem of this model. We can regard this model as the linear part of Hammerstein system and lay the foundation of the parameter identification method re-search in the future.