Bayesian network model for assessing hydrogen ignition probability

The global energy transition is accelerating the adoption of hydrogen as a clean, versatile fuel. Hydrogen offers environmental benefits but introduces safety challenges due to its high flammability. Accurate estimation of hydrogen ignition probability is crucial for effective safety management, especially in environments where ignition sources cannot be fully controlled. Existing approaches often rely on surrogate models, such as methane-based models, which may not adequately capture hydrogen behavior. Computational fluid dynamics simulations can provide detailed ignition assessments. They are, however, computationally expensive and impractical for multi-scenario analyses. This paper presents a novel quantitative Bayesian network-based ignition probability model aimed at delivering hydrogen-specific ignition probability estimates with improved accuracy and efficiency. Case studies involving hydrogen electrolyzers demonstrate the model’s practicality and effectiveness for rapid, scenario-based risk assessments. For a 1 MW PEM electrolyzer operating at 30 bar, the probability of immediate ignition was 0.068 for a leak (1% of pipe cross-section) and 0.091 for a rupture (100% of pipe cross-section). The probability of delayed ignition was 0.115 for a leak and 0.167 for a rupture. This model provides a valuable tool to enhance hydrogen safety evaluations and supports the safer deployment of hydrogen technologies.