Nonlinear predictive control based on a global model identified off-line

A model predictive control (MPC) strategy for the non-stationary nonlinear systems with operating point-dependent dynamics is presented. The MPC proposed does not require on-line parameters estimation, because its internal model is an off-line identified global (RBF-ARX.) model, which is a generalized ARX model with Gaussian radial basis function networks-based functional coefficients. The RBF-ARX model parameters are estimated using a quickly-convergent structured nonlinear parameter optimization method (SNPOM). The quadratic programming routines may be used to solve the MPC problem with constraints. Simulation study on a chemical process shows satisfactory modeling and control performance.