Low-carbon advancement through cleaner production: A machine learning approach for enhanced hydrogen storage predictions in coal seams

Hydrogen is recognized as a crucial participant in the ever-changing field of global energy sustainability due to its exceptional energy density and environmentally friendly combustion. This study aimed to improve the prediction of hydrogen adsorption in coal using advanced machine learning (ML) approaches. The focus of this research is on the General Regression Neural Network (GRNN) and XGBoost, commonly used ML tools, highlighting the improvements made to these methods to overcome problems unsolved by traditional approaches. The GRNN, known for its outstanding proficiency during the training phase, demonstrated strong predictive ability with an R2 score of 0.971 and low error margins in the testing phase, as evidenced by RMSE and MAE values of 0.004 and 0.001, respectively. Remarkably, the GRNN achieved unmatched prediction accuracy with an R2 of 0.999 during the test phase. Similarly, the XGBoost technique showed its ability to reduce errors, achieving RMSE and MAE values of 0.151 and 0.073, respectively. These results underscore the reliability of GRNN and XGBoost as tools for measuring hydrogen storage capabilities in coal beds. The study demonstrates the GRNN model's accuracy in representing the complexities of hydrogen uptake in coal, emphasizing the feasibility of ML models in improving hydrogen storage methods. This research contributes significantly to sustainable energy applications, showing how machine learning techniques can support sustainable energy strategy.