LNG inventory routing problem under uncertain weather
Master thesis
Permanent lenke
https://hdl.handle.net/11250/2736669Utgivelsesdato
2020Metadata
Vis full innførselSamlinger
- Master Thesis [4490]
Sammendrag
The inventory routing problem (IRP) in Liquefied Natural Gas (LNG) is one of the representative maritime IRP. In this problem, how to hedge the risk of uncertain voyage time brought by uncertain weather has long been a challenging issue for LNG suppliers. Given high costs in LNG delivery operations and storage, efficient inventory management and scheduling can yield substantial savings. This paper first introduces the LNG supply chain, describing how the uncertain weather conditions influence the shipment and then establishes two stochastic models to find the optimal solution.
The first one focuses on reducing the uncertain influence through speed and ship schedule adjustments when the weather condition is realized in the second stage. In contrast, the second one extends the first model, adding a path choice so that ships can bypass the area with rough weather rather than go through it. A deterministic model is generated at the same time and works as a reference to compare with the two stochastic models. Finally, a real-world case computation is conducted to evaluate the models.
From the computational results, we see that both stochastic models result in higher costs than the deterministic one. However, when adapting the deterministic solution to stochastic settings, the result is quite different. Although the original schedule cost of the deterministic solution is still the lowest, the value of the stochastic part (expected value of extra costs of changing the original schedule under different possible scenarios) is significantly higher than the two stochastic models' solutions, leading to higher total costs. It means that the schedule chosen by the deterministic model is not the most economical one facing the uncertain weather influence.
When comparing the results of the two stochastic models, we find that when the rough weather days are scattered and short, there's nearly no difference between the two models. And if the stormy days are concentrated and continuous, the model with path choice has better performance because it can avoid all the negative influences brought by the bad weather. However, when the long detour which can bypass the rough area is too long, this choice becomes meaningless. At this time, the solutions of these two stochastic models become almost consistent.