Optimization of a hydrogen supply chain network design under demand uncertainty by multi-objective genetic algorithms
Résumé
Hydrogen is currently considered one of the most promising sustainable energy carriers for mobility ap- plications. A model of the hydrogen supply chain (HSC) based on MILP formulation (mixed integer linear programming) in a multi-objective, multi-period formulation, implemented via the ε-constraint method to generate the Pareto front, was conducted in a previous work and applied to the Occitania region of France. Three objective functions have been considered, i.e., the levelized hydrogen cost, the global warm- ing potential, and a safety risk index. However, the size of the problem mainly induced by the number of binary variables often leads to difficulties in problem solution. The first innovative part of this work explores the potential of genetic algorithms (GAs) via a variant of the non-dominated sorting genetic al- gorithm (NSGA-II) to manage multi-objective formulation to produce compromise solutions automatically. The values of the objective functions obtained by the GAs in the mono-objective formulation exhibit the same order of magnitude as those obtained with MILP, and the multi-objective GA yields a Pareto front of better quality with well-distributed compromise solutions. The differences observed between the GA and the MILP approaches can be explained by way of managing the constraints and their different logics. The second innovative contribution is the modelling of demand uncertainty using fuzzy concepts for HSC design. The solutions are compared with the original crisp models based on either MILP or GA, giving more robustness to the proposed approach.
Origine : Fichiers produits par l'(les) auteur(s)