Offset-Free Nonlinear Model Predictive Control

Abstract

Offset-free model predictive control (MPC) algorithms for nonlinear state-space process models, with modeling errors and under asymptotically constant external disturbances, is the subject of the paper. A brief formulation of the MPC formulation used is first given, followed by a brief remainder of the case with measured state vector. The case with process outputs measured only and thus the necessity of state estimation is further considered.The main result of the paper is the presentation of a novel technique with process state estimation only, despite the presence of deterministic disturbances. The core of the technique is the state disturbance model used for the state prediction. It was introduced originally for linear state-space models and is generalized to the nonlinear case in the paper. This leads to a simpler design without the need for decisions of disturbance structure and placement in the model and to simpler (lower dimensional) control structure with process state observer only. A theoretical analysis of the proposed algorithm is provided, under applicability conditions which are weaker than in the conventional approach of extended process-and-disturbance state estimation. The presented theory is illustrated by simulation results of a nonlinear process.