This paper presents a new approach for non-linear predictive control based on the local model ideas. The algorithm uses a non-linear (NL) model of the plant for internal simulation. When a change in the operating point is required this NL model is used to identify a linear ARX model of the process. The optimization procedure is then executed using this linear model and a control weighting factor adapted to the nonlinear model static gain obtained from the inverse static characteristic. The proposed controller is applied to some simulated NL process and the obtained results compared to some MPC algorithms recently presented in literature to show the main advantages of this new approach: simplicity, low processing time and uniform dynamic behaviour in the operating range.
Predictive Control non-linear; local models; identification; penalization control
