ESN-Based Multi-Condition Model Predictive Control Method for Chemical Processes

In both academia and industry, it is shown that model predictive control is very effective for handling complex chemical processes with nonlinearities and large time lags. However, the establishment of predictive models usually requires online test experiments and affects normal production. In addition, chemical processes can be affected by feedstock conditions and scheduling strategies to generate multiple operating conditions, which can cause large deviations in a single predictive model. In the face of these problems, this paper proposes a data-driven deep learning-based predictive control method for chemical process models with multiple operating conditions. Multi-parallel ESN are trained with historical data of different operating conditions to integrate a data-driven prediction model. Based on the LM algorithm, the objective function is solved by rolling optimization of the working conditions for the future response of the controlled object, and the optimal control strategy is obtained. A distillation tower simulation model is used as the object. Simulation experiments are conducted for the proposed method, and satisfactory control effects are obtained to verify the effectiveness of the method.