A method of piecewise affine model identification and predictive control based on data-driven

A multi-input multi-output piecewise affine model identification method based on data-driven and a real-time predictive control method based on multi-parametric programming are proposed for a class of nonlinear and multivariable processes represented by quadruple tank system. A clustering-based algorithm is designed to solve a MIMO piecewise affine autoregressive exogenous (PWARX) model identification problem. With the use of HYSDEL modeling language and its compiler tools, MIMO-PWARX model is transformed into piecewise affine (PWA) model. A multi-parametric programming method is adopted to design a predictive controller based on PWA model. The experiment with quadruple tank shows the capability of obtaining PWARX model information directly through analysis of process data based on data-driven and the effect of real-time nonlinear predictive control based on the model in this scheme.