Scheduling tank trucks at a fuel distribution terminal using max-plus model-based predictive control

• Published in: Journal of process control Vol. 103; pp. 8 - 18
• Main Authors: Gonçalves, Marcos Vinícios, da Cunha, Antonio Eduardo Carrilho
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01-07-2021
• Subjects: Discrete event systems; Machine learning; Max-plus algebra; Model predictive control; Scheduling problems; Model predictive control; Discrete event systems; Max-plus algebra; Scheduling problems; Machine learning
• ISSN: 0959-1524; 1873-2771
• DOI: 10.1016/j.jprocont.2021.05.005

This study addresses the problem of scheduling tank trucks at a fuel distribution terminal. The plant was modeled in the max-plus algebra, applying machine learning to determine process times. With this model, and based on a just-in-time approach, we have developed a predictive controller with two operation modes. This control system aims to prevent the excess of tank trucks inside the loading yard, thus achieving a better flow, efficiency, and safety in the process. Next, we have investigated the case study of a realistic and representative fuel distribution terminal, developing a simulator to enable a performance comparison between the proposed algorithms and the current heuristic. There was a 42.7% reduction in the work-in-progress (WIP) and 41.4% in the lead time, while productivity suffered a 2.8% loss. Bearing in mind, however, that there is flexibility in parametrization to mitigate this loss of productivity. In doing so, the reductions in WIP and lead time are slightly lower, at 34.7% for both metrics. The results show that the proposed control system can contribute significantly to improving the company’s performance indicators.
•Model predictive control and max-plus algebra applied to a scheduling problem.•A pre-trained neural network provides the process times for the max-plus model.•Optimizing the sequencing and timing of entry through switching max-plus-linear systems.•Notes on the implementation and performance analysis of the control strategies.