The Nonlinear Predictive Control Based On Hammerstein Model

Model predictive control is a linear control algorithm produced in the industrial process. After decades of development, the model predictive control no matter in theory or in industrial applications have been developed very mature. However, with the development of the industrial process, more and more high requirement in control performance and widespread nonlinear characteristics, model predictive control can’t satisfy people’s needs, then, nonlinear predictive control is produced. Nonlinear predictive control mainly replaces the linear prediction model with nonlinear prediction model. The commonly used models usually are modeled by mechanism modeling or test modeling. Lots of complex nonlinear systems can’t get the relatively accurate model, or even can not get the model of the controlled system using the mechanism modeling. While test modeling apply input and output data to model. In common, there are input and output constraints in the controlled system such as to ensure the normal running process or achieve the control effect we want. T-S fuzzy neural model is a very useful identification algorithm. It only need the system input and output data to identify a set of fuzzy rules, conversing nonlinear problem into a linear rules. The output of the T-S fuzzy neural model is a weighted sum of a series of linear rules.This article introduced the T-S fuzzy neural algorithm process and the identification method of the model. The nonlinear inverse part use the T-S fuzzy neural model to describe. First of all, use generalized predictive control to design linear controller of the linear part of Hammerstein model and to deal with the constraints. Get the intermediate variable. Then apply T-S fuzzy neural model to get the actual input of the system. And the nonlinear predictive control is applied to the continuous stirred tank reactor to research. The simulation results show that the method has good feasibility and control effect.