Machine learning approaches for the prediction of hydrogen uptake in metal-organic-frameworks: A comprehensive review

Hydrogen (H2) is a potential energy source for achieving net-zero carbon emissions. Metal Organic Frameworks (MOFs) are promising for hydrogen storage due to their high specific surface area, large pore volume, and adaptable structure. Machine Learning (ML) offers adaptability, scalability, and automation, making it effective for analyzing MOF performance in H2 storage prediction. Databases like HyMARC and CoRE MOF 2019 provide suitable data for this purpose. ML-based screening techniques outperform HTCS and GCMC calculations. Key descriptors influencing hydrogen uptake include pressure, temperature, pore volume, and BET surface area. ML algorithms such as LASSO, MLR, RF, ERT, GB, MLP, DNN, LS-SVM, and CMIS have shown superior prediction performance with R2 values over 0.95. This review highlights the impact of MOF datasets, screening methodologies, and ML algorithms on predicting MOF hydrogen absorption capabilities, offering insights and guidance for future ML-based MOF research.