Novel integrated EMPC framework for fluid catalytic cracking unit under feedstock uncertainty

The conventional economic model predictive control (EMPC) framework faces challenges in achieving optimal operation of fluid catalytic cracking units (FCCUs) under feedstock uncertainty. To overcome this limitation, a novel hierarchical EMPC framework integrating a soft sensor for real-time estimation of feedstock properties is proposed in this work. Specifically, an innovative soft sensor is developed, combining oil classification strategy with ensemble learning to accurately predict feedstock properties. This soft sensor is coupled with pseudo-components method to mitigate the effects of feedstock variability, thereby enabling robust dynamic optimization within the economic optimization layer. Simulation results across various feedstock disturbance scenarios demonstrate that the proposed framework outperforms the conventional EMPC approach by rapidly identifying the plant optimum in response to feedstock variations. These validate the proposed framework’s ability to significantly enhance the economic performance of FCCUs under conditions of feedstock uncertainty.