Operation of Distillation Columns Using Model Predictive Control Based on Dynamic Mode Decomposition Method
Xing Qian, Qingmei Dang, Shengkun Jia, Yang Yuan, Kejin Huang, Haisheng Chen, Liang Zhang
Abstract
As a typical nonlinear system, the prediction of state variables and system control in distillation column systems face numerous challenges. The dynamic mode decomposition (DMD) method can construct an approximate linearized model of the system based on the state and input variables of the nonlinear system. In this paper, the DMD method is used to obtain an approximate linearized model of the distillation column system, which can predict the system’s state variables. This model is then utilized as the prediction model for designing a model predictive controller in model predictive control (MPC). The aim is to achieve the prediction of the state variables of simple and complex distillation systems and control the system effectively. By comparing the approximate linearized model obtained using the DMD method with the local linearized model obtained by linear expansion at the steady-state point, it is found that the approximate linearized model is more accurate for system reconstruction and prediction of the future outputs of the system. The feasibility and effectiveness of the DMD-MPC algorithm are investigated using a binary column system and four product Kaibel dividing-wall distillation column systems as case studies. It is found that the MPC designed using this approximate linearized model as the prediction model has good dynamic performance with small maximum and steady-state deviations and can suppress disturbances well.