论文部分内容阅读
This paper discusses data-driven subspace predictive control and control performance monitoring based on the historical objective function benchmark. A data-driven subspace model predictive control method is used to design the controller. No prior knowledge of model structure and system rank but the I/O data of an open-loop test are required. Then we propose a new criterion for the selection of the historical data, which is used to monitoring the controller’s performance instead of the traditional method based on prior knowledge. The proposed algorithms are illustrated through a distillation column simulation example.