From empirical to intelligent: ML-enhanced parachor modeling of gas (H2, CO2, CH4, and N2)–water interfacial tension for applications in energy storage and carbon sequestration

In subsurface energy applications such as gas injection and storage, Interfacial Tension (IFT) between gas and water critically influences fluid distribution and trapping in porous media. Thus, accurate IFT prediction is essential for reliable simulation of multiphase flow in formations such as saline aquifers and depleted reservoirs. Traditional IFT prediction methods often rely on experimental measurements or solubility-based correlations, which can be impractical under high-pressure reservoir conditions. The classical parachor model offers a simple and computationally efficient approach for estimating interfacial tension. However, since it was originally developed for hydrocarbon systems, its accuracy declines in gas–aqueous environments—especially for polar or low-density gases. To expand its practical utility in large-scale gas injection and storage projects, it is crucial to enhance the model’s applicability to such systems, enabling more reliable integration into reservoir simulation tools. In this study, we propose a machine learning-enhanced reformulation of the parachor model to predict IFT in gas–water systems involving H₂, CH₄, N₂, and CO₂, without requiring solubility or compositional inputs. Using an extensive experimental database, parachor values were back-calculated and modeled using key thermophysical inputs, then used to estimate IFT. The models achieved high accuracy, with R² values up to 0.9897 and AARE as low as 0.8619 % across different systems. Comparative evaluations against existing empirical models revealed significant improvements. The proposed framework offers a robust and solubility-independent tool for IFT prediction, suitable for a wide range of energy applications including hydrogen storage, CO₂ sequestration, and gas-based reservoir processes.