Deep Learning Model Predictive Control Frameworks: Application to a Fluid Catalytic Cracker–Fractionator Process

The present study proposes two model predictive control (MPC) frameworks that incorporate deep learning models: (i) deep learning MPC (implements only deep learning models) and (ii) hybrid MPC (implements deep learning and linear models, which provide improved model-based interpretability). The prop...

Full description

Saved in:
Bibliographic Details
Published inIndustrial & engineering chemistry research Vol. 62; no. 27; pp. 10587 - 10600
Main Authors Santander, Omar, Kuppuraj, Vidyashankar, Harrison, Christopher A., Baldea, Michael
Format Journal Article
LanguageEnglish
Published American Chemical Society 12.07.2023
Subjects
Process Systems Engineering
Online AccessGet full text

Cover

More Information
Summary:The present study proposes two model predictive control (MPC) frameworks that incorporate deep learning models: (i) deep learning MPC (implements only deep learning models) and (ii) hybrid MPC (implements deep learning and linear models, which provide improved model-based interpretability). The proposed frameworks are successfully applied to a large-scale fluid catalytic cracker–fractionator process. The results demonstrate economic improvement (with respect to the current industrial MPC practice) under various disturbance scenarios. The solution time of the proposed frameworks is promising, particularly for hybrid MPC, which is expected to promote industrial implementation.
ISSN:0888-5885
1520-5045
DOI:10.1021/acs.iecr.3c01110