The Research On Identification And Control For Nonlinear Hammerstein-Wiener System

Ideal linear control system does not exist in practical industrial system, and industrial objects often have obviously nonlinear characteristics. Therefore, the identification, modeling and adaptive control of nonlinear systems has become the research focus for the control theory of non-linear model.It has been found that Hammerstein-Wiener type usually may account for experimental model of nonlinear systems. It is composed of a nonlinear static element, a linear dynamic element and another nonlinear static element connected in series, characterized by nonlinear static character and linear dynamic character considered respectively, thus varieties of nonlinear systems can be described simply and effectively so that the majority of nonlinear characteristics of the process of industrial production can be reflected precisely. For example, the system mode would be more similar to the actual situation with the higher accuracy by representing the electrode regulator system of the EAF steelmaking process as Hammerstein-Wiener model.The paper first introduces the research background, describes the nonlinear system identification and control method of the research status, and briefly explains the content and the structure of the paper. The next, it research based hybrid neural network and predictive function control theory to Hammerstein-Wiener of nonlinear system identification and control method.By the hybrid neural network structure on the Hammerstein-Wiener model is identified, the hybrid neural network using two multilayer network respectively approach two nonlinear gain, and with linear network estimate linear dynamic subsystems, and deduced the unified back propagation algorithm, the feedforward networks and linear network weight coefficient and threshold training, synchronization identification effect; At the same time, to the actual industrial production has Hammerstein-Wiener model structure of the electrode regulation system is identified, with MATLAB simulation demonstrate the efficiency of this method.Predictive function control is controlled object basis functions output responses can stack to be premise, and therefore only apply to linear dynamic system, it has the small amount of calculation, high real-time control performance and characteristic of good. Using the nonlinear static gain the inverse of the object and the nonlinear series, offset the nonlinear static nonlinear object gain part, nonlinear control problem into the object of linear control problem of the object, so as to realize the object of nonlinear predictive function control. At the same time, to the actual industrial production has Hammerstein-Wiener model structure of the electrode adjustment system controller design, with MATLAB simulation demonstrate the efficiency of this method.