A Machine Learning Based Framework for Brine-Gas Interfacial Tension Prediction: Implications for H2, CH4 and CO2 Geo-Storage

Abstract

Brine-gas interfacial tension (γ) is an important parameter to determine fluid dynamics, trapping and distributions at pore-scale, thus influencing gas (H2, CH4 and CO2) geo-storage (GGS) capacity and security at reservoir-scale. However, γ is a complex function of pressure, temperature, ionic strength, gas type and mole fraction, thus time-consuming to measure experimentally and challenging to predict theoretically. Therefore herein, a genetic algorithm-based automatic machine learning and symbolic regression (GA-AutoML-SR) framework was developed to predict γ systematically under GGS conditions. In addition, the sensitivity of γ to all influencing factors was analyzed. The prediction results have shown that: the GA-AutoML-SR model prediction accuracy was high with the coefficient of determination (R2) of 0.994 and 0.978 for the training and testing sets, respectively;a quantitative mathematical correlation was derived as a function of pressure, temperature, ionic strength, gas type and mole fraction, withR2= 0.72;the most dominant influencing factor for γ was identified as pressure. These insights will promote the energy transition, balance energy supply-demand and reduce carbon emissions.