Data Mining for Enhanced PEM Electrolysis

Abstract

Despite numerous publications on PEM electrolyser production, a lack of standardized methodologies and reporting guidelines complicates direct comparisons. We conducted a systematic exploratory data analysis, creating a database comprising more than 1,000 samples from 127 publications, which considered over 85 parameters encompassing material selection, MEA fabrication, cell assembly, and characterization. Through statistical analysis, we identified trends and hidden influencing factors. Furthermore, we utilized an Extreme Gradient Boosting model quantifying feature importance, revealing critical factors often underreported relative to their impact. This work provides a foundation for standardizing research data, showing that systematic data management is key in overcoming the comparability challenges and accelerating development. Database of >1000 water electrolysis experiments links materials to performance. Critical fabrication parameters are found to be underreported in literature. XGBoost model identifies most critical influencing parameters. Statistical analysis of metadata enables data-driven standardization.