A robustly model predictive control strategy applied in the control of a simulated industrial polyethylene polymerization process

Abstract

This work investigates the quality control problem of polyethylene grades produced in an industrial polymerization plant. The plant is represented by a phenomenological model previously validated using industrial data. Three advanced control strategies were developed in order to propose strategies capable of dealing with different grades of products. A stability guaranteed [IHMPC (Infinite Horizon Model Predictive Control)] technique and a robust stabilized [RIHMPC (Robust IHMPC)] technique were designed and compared with a conventional MPC. The results show that RIHMPC outperformed the other techniques, providing smoother and faster grade transitions and setpoint tracking. This result provides a significative contribution, since no report of application of RIHMPC on polymerization plants was found. The RHIMPC controller seems to be a promising option for other polymerization reactor control problems in which the diversity of resins produced, coupled with the nonlinear characteristic of the system, lend themselves to the use of different linear models.