A stochastic optimization algorithm for the supply vessel planning problem under uncertain demand and uncertain weather conditions

Abstract

The Supply Vessel Planning Problem (SVPP) with stochastic demands and uncertain weather conditions is a transportation problem occurring in offshore oil and gas logistics. A fleet of supply vessels based at an onshore depot delivers supplies to a set of offshore oil platforms in a weekly sailing schedule. However, the schedules are frequently disrupted due to adverse weather conditions and uncertain demand for cargo from the oil platforms. The two sources of uncertainty are generally addressed in separate, most often through the use of two-phased methods, where simulation is combined with an optimization algorithm. The most common approach to incorporate robustness in the constructed schedules is to use a subjective penalized cost for non-robust voyages, with explicit modelling of recourse actions. In contrast, this paper proposes a two-stage stochastic programming algorithm accounting for both uncertain demand and uncertain weather conditions, allowing for the incorporation of the cost of recourse in the objective function. The cost of each solution is approximated through the use of discrete event simulation within a genetic algorithm. For the tested problem instances, the potential benefit from solving the stochastic program over solving the corresponding deterministic version leads to average relative annual cost savings of approximately 12%.