Spatiotemporal uncertainty-aware predictive control for industrial distributed parameters systems

Distributed Parameter Systems (DPS) play a crucial role in modern chemical industry processes, requiring efficient control strategies to ensure performance. In practical industrial applications, the unknown intricate dynamics of DPS often coexist with complex spatiotemporal uncertainties, which hinder precise control. To address the challenges, a novel learning-based spatiotemporal uncertainty-aware Model Predictive Control (MPC) method is proposed for DPS in this paper, which formulates a scenario-based MPC with spatiotemporal uncertainty modeling. Specifically, a spatiotemporal prediction uncertainty quantification approach is designed to mitigate the intricate uncertainties when predicting the dynamics of DPS by applying the time-space separation technique. Subsequently, a predictive control strategy based on spatiotemporal scenarios is developed to take probable DPS responses into account, thereby optimizing the control inputs to enhance the control performance. Theoretical analysis is given to validate the stability of the proposed control paradigm. Extensive experiments on two chemical process industrial cases are conducted to verify the effectiveness of the proposed method. The experimental results show that the proposed control paradigm can benefit from the spatiotemporal uncertainty analysis and thus significantly improve the control precision of industrial DPS. © 2025 Elsevier Ltd.