Generalized predictive control of a heat exchanger using fuzzy model

This paper proposes a new approach to predictive control of highly nonlinear processes based on the Takagi-Sugeno fuzzy model. It is shown how the Takagi-Sugeno fuzzy models can be linked to a special type of model based predictive control algorithm, the generalized predictive control (GPC). In GPC design, a purely linear transfer function model is used for long-range prediction. The advantage of GPC and other linear MBPC methods is the guaranteed convergence within each time sample, but they are not able to deal with strong process nonlinearities. In our approach, approximate linear models are extracted at each time sample by instantaneous linearization of the nonlinear fuzzy model, and adaptive GPC is used. Applicability of this approach to control a real world process (nonlinear laboratory-scale thermal plant) with operating point dependent gain and time constants is demonstrated in the paper.