A comparative study of methods for estimating hydrocarbon-water interfacial tension

Abstract

Interfacial tension is the force that acts at the boundary between two immiscible phases. This force keeps the two fluids separated, preventing them from mixing together. Hydrocarbon-water interfacial tension is one of the most critical parameters in petroleum engineering calculations, like enhanced oil recovery, phase behavior in reservoirs, separation processes, and so on. Hydrocarbon-water interfacial tension is often measured in laboratories, but it is costly and time-demanding. Consequently, mathematical methods are developed for estimating hydrocarbon-water interfacial tension, which summary in two main groups are empirical correlations and machine learning algorithms. In this paper, the authors carry out a comparative study of mathematical methods for estimating hydrocarbon-water interfacial tension. Empirical correlations are implemented based on the study by Danesh (1998), Sutton (2006), Sutton (2009) and Meybodi et al. (2016). Machine learning algorithms used include Multi-Layer Perceptron (MLP) and Extra Trees (ET). Research data gathered from open literature incorporates features: critical temperature (TC), reservoir temperature (TR), density difference (Δρ), reservoir pressure (P), and interfacial tension (IFT). The estimation outcomes are contrasted using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Coefficient of Determination (R2) to find the optimal model. The results show that the Extra Trees (ET) algorithm is the most accurate model with the lowest MAE and RMSE (0.6391 and 0.9565) and the highest R2 (0.9871) in the group of empirical correlations and machine learning algorithms. Moreover, the solution indicates two machine learning models having better performance than four empirical correlations due to lower MAE and RMSE and the higher R2.