Abstract: 

This presentation deals with a real-world case study of Multi-Objective Optimization from Energy management. Firstly, industrial context and constraints of electricity production are presented, which is also nowadays an issue for citizens. Secondly, the optimization problem of maintenance planning and refueling of nuclear power plants is presented with a stochastic mathematical programming formulation. Thirdly, methodological insights are shown to solve this complex mono-objective optimization on large scale real-world instances. Lastly, multi-objective optimization extensions are discussed. A first extension dealing with stability in monthly reoptimization is straightforward and efficient. A robust optimization extension, taking into account uncertainty in maintenance durations, can be modeled using bi-level programming; numerical issues induce the use a multi-objective optimization approach as 100% robust solutions can be unfeasible, diverse solutions with different risk/cost are provided along a Pareto Front instead.

About the speaker: 

Nicolas Dupin is currently Assistant Professor in Université Paris-Saclay and will be Associate Professor in Université d'Angers from September. His research focuses on solving discrete and/or combinatorial, mono and multi-objective optimization problems, using hybrid algorithms, and especially matheuristics to combine in several manners strengths of meta-heuristics and mathematical programming. His industrial collaborations include maintenance planning problems (for power plants and also for military aircrafts), electricity production, train time-tabling and transportation applications.