Smart Computational Algorithms for the Prediction of Interfacial Tension between Water and Hydrogen – Insights into Underground Hydrogen Storage

Abstract

Abstract Hydrogen has the potential to play a critical role in the energy transition economy for the next decade, aiding in decarbonization. Hydrogen has a two-pronged utility in the energy mix by acting as a fuel and supporting the distribution of other renewable sources through electrolysis. Nevertheless, a critical hurdle in achieving autonomous hydrogen-based energy transition is the safe, reliable, and economical methods of underground storage mechanisms. Consequently, this requires comprehending interaction processes between hydrogen and subsurface fluids that can affect the storage capacity with a major role of interfacial tension (IFT). Accordingly, this work used smart computational intelligence methods to delineate IFT predictions between H2 and H2O mixture for various pressure/temperature conditions and density variance. A systematic approach was adopted to implement predictive models for IFT prediction by utilizing an experimental data set. A comprehensive statistical analysis is performed to achieve model generalization capabilities and improve control over the most relevant input parameters. Consequently, IFT is demarcated as a function of two readily available inputs of pressure, temperature, and calculated density difference. Various smart approaches in this work are proposed by developing an IFT predictor using Support Vector Regression, XGBoost, and Decision Tree algorithms. Machine learning model training is enhanced using a k-fold cross-validation technique combined with the exhaustive grid search algorithm. Post-training, the developed models are tested for reliability using blind datasets reserved for this purpose. A fair comparison between model efficiency is ensured by using an in-depth error analysis schema that includes various metrics like the correlation of determination, average error analysis, graphical error analysis, and scatter plots. This generates a relative ranking system that weighs various factors to classify one model as the most efficient. For the IFT prediction problem, it was found that the XGBoost was aptly able to yield high efficiency and low errors. This stems from how XGBoost functions map the non-linear relationship between pressure, temperature, density difference, and the IFT. It was also observed that enhanced intelligent model training through multiple techniques resulted in optimized hyperparameters/parameters. Lastly, a trend analysis was conducted to confirm the robustness of the developed XGBoost model.