Capacitated facility location problem under uncertainty with service level constraints

Classic facility location models often assume customer demands to be deterministic, although real-world demand is usually uncertain, especially in long-term strategic planning. While stochastic programming models are widely used to address uncertainty, the default approach of ensuring that the facil...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:European journal of operational research
Hauptverfasser: Zhang, Haoyue, Kalcsics, Jörg
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier B.V 01.09.2025
Schlagworte:
Location
Sample average approximation
Service level
Stochastic programming
Uncertainty modelling
Service level
Uncertainty modelling
Location
Stochastic programming
Sample average approximation
ISSN:0377-2217
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Classic facility location models often assume customer demands to be deterministic, although real-world demand is usually uncertain, especially in long-term strategic planning. While stochastic programming models are widely used to address uncertainty, the default approach of ensuring that the facility capacities are met at all times, i.e., for every scenario, can sometimes produce overly conservative solutions. This paper presents a novel stochastic programming model that incorporates a range of service level restrictions that allow demand to be unsatisfied with a certain probability and up to a certain amount. Concerning the former, we use two α-service level constraints, a well-known local and a new global constraint, while the latter is controlled through two β-service level constraints that take the expected value and the maximum value of the excess demand into account. The service levels are incorporated in the stochastic programming model using chance constraints. To solve the model’s deterministic equivalent, we implement a Benders’ decomposition and a modified sample average approximation algorithm with concentration sets. We carry out experiments on randomly generated data sets and a real-world inspired case study in Scotland to compare the performance of models with different service level combinations, as well as with the classical penalty model. •A novel CFLP model with 4 different service levels to tackle demand uncertainty.•Service levels offer intuitive and independent control of different aspects of risk.•Develops a modified SAA, enhancing efficiency and solution quality and robustness.•Incorporates BD with valid inequalities for efficient problem solving.•Conducts comprehensive experiments on random and real-world inspired datasets.
ISSN:0377-2217
DOI:10.1016/j.ejor.2025.08.056