Predictive modeling for hydrogen storage in functionalized carbonaceous nanomaterials using machine learning

Hydrogen, valued for its high energy density and environmental friendliness, stands out as a promising energy carrier. Traditional storage methods, such as high-pressure gas cylinders and cryogenic liquid tanks, face challenges related to transportation risks, low energy efficiency, and economic feasibility. In this context, solid-state materials, particularly carbonaceous nanomaterials, have gained prominence as alternatives for hydrogen storage due to their notable capacity, cost-effectiveness, and reliability. However, the identification of efficient carbonaceous adsorbents presents challenges, requiring expensive and time-consuming experimental efforts. This study introduces a machine learning (ML) model designed to predict hydrogen storage capacity in functionalized carbonaceous nanomaterials. Leveraging the ability of ML technology to unravel complex relationships within extensive datasets, the collected comprehensive dataset undergoes rigorous statistical analysis and mechanistic discussions. Feature importance analysis is used to provide insights into the influence of input features on hydrogen storage capacity. The Gaussian process regression model emerges as a standout performer, achieving impressive metrics with an R² > 0.955, RMSE < 0.121, and MAE < 0.064 for hydrogen adsorption/desorption during the testing phase. The impact of pressure and temperature on hydrogen adsorption, along with pressure and sorbent average crystal size for hydrogen desorption, is elucidated. This ML model has the potential to revolutionize the identification of optimal material properties and processing conditions for hydrogen adsorption in carbonaceous nanomaterials, offering significant time and cost savings.