Adaptive Lyapunov-based MPC With Constraint Updating For FCC Unit

This paper focuses on the Learning MPC design of the FCCU regenerator for automatically searching for the optimal stabilizing control law, which contains strong nonlinear characteristics and slow time-varying dynamics due to the catalyst aging or component changes. In order to guarantee the decreasing of the variance of the prediction of the identified model, which could relax the stability constraint and reduce conservatism, the FCCU is transformed into a parametric linear form by means of the polynomial fitting, then Set-Membership identification is used to update the system model, the variance bound of the updated model is deduced. According to the decreasing variance bound, the Lyapunov constraint is designed and is used in the optimization problem of MPC. The proposed method could still guarantee feasibility and stability during updating. Besides, the decreasing of the unstable region, which represents the steady-state error, could also be achieved based on the decreasing identification variance. The simulation results demonstrate that the proposed ALMPC with constraint updating performs better than the MPC without constraint updating.