.png){kind=icon}
Return
DK-MAOS: A Data-Knowledge-Driven Multiobjective Adaptive Optimization Method for Process Industries Under Multiple Operating Conditions
.png){kind=icon}
PRE
AI
DOI：10.1109/tii.2026.3722022.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Process industries operate under varying operating conditions, which lead to different operational states. Therefore, it is crucial to design a state identification-based optimization strategy to ensure reliable operation. This article proposes a data-knowledge-driven multiobjective adaptive optimization strategy (DK-MAOS) for process industries operational state prediction and optimal control under varying operating conditions. First, a prototype temporal network is constructed to predict future operational states. The designed network can achieve accurate prediction in few-shot scenarios through soft condition embedding and a prototype vector mechanism. Second, an adaptive objective model is designed for satisfying the operational requirements in each operating condition by adaptively adjusting the weights of operating indicators. It integrates operational requirements and tracking errors to ensure controller feasibility during setpoint optimization. Finally, a dynamic step-size adaptive sequential quadratic programming method is developed for the synergistic multiobjective solution of optimization and control. The effectiveness of the proposed DK-MAOS method is validated on the alcohol distillation process and the diesel hydrofining process. Results demonstrate that it effectively maintains product quality within prescribed limits while reducing energy consumption.
Keywords:
Adaptive objective function (AOF)
multiobjective optimization
multiple operating conditions
operation status prediction
process industries
Journal
IF:
9.9
Papers:
8.6K
Citations:
6.0W
Organization
Cited Papers
No cited papers available