Models, Methods, and Algorithms in Operational Research: Applications to Inventory Control, Energy Infrastructure, and Computational Biology
(2025)
Files
Thèse-final.pdf
Open Access - Adobe PDF
- 4.28 MB
Details
- Authors
- Supervisors
- Abstract
- This thesis investigates three distinct applications of optimization in Operational Research (OR), unified by their methodological focus on problem modeling, algorithmic design, and computational analysis. Each chapter addresses a different domain, demonstrating how OR tools can be adapted to diverse data and decision structures. The first chapter applies Deep Reinforcement Learning to dynamic lot-sizing problems with non-stationary demand and continuous action spaces. It identifies critical training instabilities linked to value-function discontinuities and proposes methodological solutions that improve policy performance and training reliability. The second chapter develops a network-flow model for strategic infrastructure planning, focusing on Germany’s deployment of Liquefied Natural Gas regasification capacity under geopolitical and decarbonization scenarios. The third chapter addresses phylogeny inference under the Balanced Minimum Evolution criterion. It introduces a linear programming-based matheuristic as an alternative to Neighbor Joining and improves its performance through integration into a beam search framework, yielding more accurate evolutionary trees on benchmark datasets. Together, these contributions illustrate how optimization can be tailored to address both operational and scientific problems, and provide transferable insights into algorithmic development and modeling practice.
- Affiliations
Citations
Dehaybe, H. (2025). Models, Methods, and Algorithms in Operational Research: Applications to Inventory Control, Energy Infrastructure, and Computational Biology. https://hdl.handle.net/2078.5/249326