Reliability prediction for proton exchange membrane water electrolyzer using heterogeneous data sources

Water electrolysis is increasingly recognized as a viable method for producing green hydrogen to support the decarbonization of global energy utilization. In reliability assessment of electrolyzers, the central component of the production process, a significant challenge lies in the lack of historical failure data. This study introduces a novel methodology for predicting the failure probability of proton exchange membrane (PEM) electrolyzers using the limited data from different sources. Given that PEM electrolyzers and PEM fuel cells share the same core membrane technology, they exhibit similar degradation mechanisms and failure modes. This work leverages existing degradation data from PEM fuel cells to model the degradation trajectory of PEM electrolyzers and introduces adjustment factors to adapt it to electrolyzers. By establishing appropriate failure thresholds, it becomes possible to estimate both the failure rate and the probability of failure within a defined operational timeframe. These individual failure probabilities are then integrated into a Bayesian network to calculate the overall failure probability of a PEM electrolyzer. This work can be helpful for reliability assessment of PEM electrolyzers in their design and earlier phases of lifecycle, and informative for the relevant works on other novel technologies.