Thesis start: 2023
Thesis end: 2026
Expected defense date: –
Abstract
The current international context and environmental challenges make energy issues crucial, whether in terms of sobriety, autonomy, security of supply, or the gradual transition to renewable energy sources. Responsible for over 50% of global energy consumption in 2019, industry is the most affected sector. This explains the growing attention paid by the scientific community to the energy efficiency of manufacturing systems: different aspects (such as energy consumption and associated cost, electricity consumption shifting, consideration of peak power) are investigated, from design to planning and scheduling of operations, in the form of constraints and/or criteria to optimize. New industrial paradigms related to Industry 4.0/5.0 further broaden the number of potential levers for action. However, the resulting optimization problems seem difficult to address exactly using traditional techniques. This thesis aims to develop advanced tools for the exact resolution of decision problems in manufacturing systems, taking energy into account, in order to tackle large-scale instances and promote their use in real-world cases. These tools would be based on Combinatorial Optimization techniques, such as Column Generation or Cutting Plane Algorithms, which have already proven effective in other fields but are still rarely used for problems like those under study, and whose potential therefore remains to be explored. The development of such tools would also help to better understand the structural properties of these problems and open up further avenues of investigation, such as the design of metaheuristic algorithms or the integration of Machine Learning techniques.
Keywords
Combinatorial optimization, Exact resolution, Column generation, Valid inequalities.
Sustainable Development Goals concerned
Publications
- the job-shop scheduling problem under energy considerations, specifically focusing on minimizing total energy costs within a Time-of-Use pricing framework, denoted as Jm||TEC. We propose a period-indexed Mixed-Integer Linear Programming formulation, which proves advantageous due to its smaller model size compared to traditional time-indexed approaches. Initial studies highlight that while our model can rapidly find feasible […]
- The Job-Shop Scheduling Problem (JSSP) is a challenging problem in combinatorial optimization, ranking among the hardest to solve in practice within its complexity class. In this work, we address the JSSP under energy considerations, specifically focusing on minimizing total energy costs under a Time-of-Use (ToU) pricing scheme and a peak power limit. ToU pricing schemes […]