A Chance–constrained Robust Mpc and Application to Hot Rolling Mills

Most approaches of chance–constrained robust MPC utilise an open– loop model for uncertainty prediction; this results in conservative control. To avoid this some approaches propose robust MPC (RMPC) based on closed–loop prediction, and these often give improved dynamic performance at a small expense to constraint handling performance. The aim of this paper is to extend the methodology of (Warren, 2005), which is a chance–constrained RMPC based on closed–loop prediction, through the use of state space models. Moreover, we adopt dual mode MPC in the constrained optimisation, which allows us to guarantee stability for the infinite horizon. The use of the closed–loop paradigm for dual mode MPC has advantages in numerical conditioning and robustness analysis. A case study, looper–tension control, is used to evaluate the robustness of the proposed RMPC strategy on a realistic process as well as extending previous studies of MPC for rolling mills (Choi et al., 2004) to the robust case.