AI-driven optimization approaches of metal–organic frameworks for enhanced methane delivery

Methane, the primary component of natural gas, emits less carbon dioxide than other petroleum-based fuels but faces challenges in efficient storage and transportation. Advanced adsorption materials provide a safe and cost-effective solution, with metal–organic frameworks (MOFs) emerging as promising candidates for natural gas storage and delivery in vehicles. This research employed AI-Driven Optimization (AiDO) to identify optimal parameters for enhancing methane uptake while simultaneously improving both gravimetric and volumetric delivery. We developed and validated three machine learning models: eXtreme Gradient Boosting (XGBoost), Kolmogorov–Arnold Network (KAN), and Convolutional Neural Network (CNN), using experimental data. All models demonstrated strong predictive performance, with XGBoost achieving outstanding results, including a Root Mean Squared Error (RMSE) of 0.0103 and a coefficient of determination ( R 2 ) of 0.9722. When integrated into an optimization framework, the XGBoost model identified optimal conditions for methane delivery, predicting a room temperature gravimetric delivery of 724.14 cm 3 /g, and a volumetric delivery of 602.21 cm 3 /cm 3 from 65 to 5 bar. Sensitivity analysis validated the robustness of the AiDO methodology, highlighting its potential to effectively reduce costs and enhance the performance of porous MOFs. • Three AI models developed and validated for predicting methane uptake and delivery in porous MOFs. • XGBoost demonstrated the highest predictive performance (RMSE = 0.0103, R 2 = 0.9722). • AI-Driven Optimization (AiDO) framework identified optimal gravimetric and volumetric methane delivery conditions. • Methodology is robust, scalable, and extendable to other gas storage system.