|
|||||
|
|
A machine learning framework for improving refinery production planning |
|
| Authors: | Omar Santander Vidyashankar Kuppuraj Christopher A. Harrison Michael Baldea |
| Affiliation: | 1. McKetta Department of Chemical Engineering, The University of Texas at Austin, Austin, Texas, USA Contribution: Conceptualization (lead), Formal analysis (lead), Investigation (lead), Methodology (equal), Software (lead), Visualization (lead), Writing - original draft (lead);2. Marathon Petroleum Corporation, Garyville, Louisiana, USA Contribution: Data curation (supporting), Formal analysis (supporting), Methodology (supporting), Validation (supporting), Writing - review & editing (supporting);3. Marathon Petroleum Corporation, Garyville, Louisiana, USA Contribution: Conceptualization (supporting), Funding acquisition (equal), Methodology (supporting), Project administration (equal), Supervision (equal), Writing - review & editing (supporting);4. McKetta Department of Chemical Engineering, The University of Texas at Austin, Austin, Texas, USA |
| Abstract: | We propose a framework that relies on machine learning techniques and statistical modeling to enhance industrial production planning. Supervised learning is employed to improve the production planning model, whereas unsupervised learning is used to achieve economic synchronization between the process control and production planning layers. Finally, an upgraded production planning decision-making structure is formulated where model uncertainty, the effect of process control/disturbances, and time correlation are considered. The proposed framework is implemented on an industry-relevant refinery model demonstrating that the performance of the framework is substantially better than established industrial production planning techniques. |
| Keywords: | fluid catalytic cracker machine learning production planning uncertainty |