Disturbance Modeling and State Estimation for Predictive Control with Different State-Space Process Models

Abstract The paper is concerned with disturbance modeling and observer design for Model Predictive Control (MPC) with different formulations of state-space process models. Systematic discussion is given, presenting the ways the deterministic disturbances most important in process control applications should be treated in the MPC algorithm, to obtain disturbance attenuation and offset-free control. The closely related problem of observer design and understanding is explained. It is shown that a simple approach with a process state observer only, in the presence of deterministic disturbances, can work better than the conventional approach of extended process-disturbance state estimation. The observer design is also considered for the case of extended velocity form state-space modeling including integrators. A simplified approach with reduced order observer is given. The results are illustrated on a 2×2 example process problem.