Data-Driven Bayesian Network for Risk Analysis of Urban Hydrogen Refueling Station Accident

Hydrogen refueling station (HRS) safety is receiving increasing attention with the growth of hydrogen energy application. Existing risk assessment methods of HRS are primarily based on expert knowledge to develop failure processes. It may lead to insufficient accuracy due to potential subjectivity. This paper aims to conduct a new hybrid risk assessment method by incorporating the latest HRS accident data and physical knowledge into a Bayesian network (BN) model to analyze the key risk influencing factors (RIFs). In this paper, the latest HRS accident data in HIAD 2.1 from 1980 to 2023 is collected. 30 RIFs are identified based on the accident report and physical knowledge. Use Bayesian Search (BS) for structure learning. The expectation maximation algorithm is designed in the parameter learning stage to obtain the data-driven BN model. Additionally, K-fold cross validation is dedicated to test the performance of different BN models. With these developments, new findings and implications are revealed beyond the state-of-the art of HRS risk analysis.

Keywords: Hydrogen refueling stations, Risk analysis, Data-driven Bayesian network, CTGAN, K-fold cross validation, Hydrogen energy.