Multi-Source Data-Based Hydrogen Refuelling Station Location Optimization - a Case Study of Guangdong, China

Abstract

Hydrogen is a promising alternative to fossil fuels in transportation; however, a significant gap remains between the current number of hydrogen refuelling stations and development targets, primarily due to challenges in optimizing station locations, demand patterns, and costs. This study develops an economy-society-environment assessment framework to optimize station siting. A simulated annealing-based Maximal Covering Location Problem (MCLP) model identifies optimal strategies for new, oil-hydrogen combined, and hybrid stations, integrating demand intensity and economic considerations for reliable supply allocation. Additionally, a machine learning approach combining random forest and SVM forecasts hydrogen demand and station locations for 2030 and 2050. Findings indicate that, as of 2023, stations are concentrated in central Guangdong, but by 2030, demand extends to northern, western, and eastern clusters due to shifts in passenger vehicle use. Oil-hydrogen combined stations are the most cost-efficient, reducing costs by 17.2% to 18.5% compared to other strategies. However, average transport costs per station increase from $0.33 million in 2023 to $0.80 million in 2050, highlighting the need for early expansion of hydrogen production facilities to control future costs.

Keywords: Hydrogen refuelling station, location optimisation, MCLP, demand prediction, multi-sources data