Stochastic single-source capacitated facility location model with service level requirements

Abstract

A stochastic version of single-source capacitated facility location problem is considered. A set of capacitated facilities is to be selected to provide service to demand points with stochastic demand at the minimal total cost. The facilities have service level requirements modeled by chance constraints. For Poisson demand, the problem is proved equivalent to a known solvable deterministic problem. For Normally distributed demand, it is equivalent to a deterministic mixed integer non-linear programming problem. A hybrid heuristic of Lagrangean relaxation with a single-customer-multi-exchange heuristic is embedded within a branch-and-bound framework to find upper and lower bounds for this problem. From test instances created from benchmark problems (10–20 facilities and 50 demand nodes) and real-life data on the deterministic problem, the gap between the bounds is within 6.5% with an average of 2.5%.