Prediction of shale wettability using different machine learning techniques for the application of CO2 sequestration

Abstract

Understanding the behavior of carbon dioxide (CO2) wetting in shale formations is crucial for various applications such as CO2-enhanced oil recovery, CO2 foam hydraulic fracturing, and CO2 storage in saline aquifers and shale formations. However, traditional laboratory methods to determine shale wettability, including contact angle measurements and nuclear magnetic resonance spectroscopy, can be time-consuming and expensive. This study aims to overcome these limitations by utilizing machine learning (ML) techniques to estimate shale wettability based on shale characteristics and experimental conditions.A dataset was collected, encompassing different shale samples under various conditions, including pressure, temperature, and brine salinity, to predict shale wettability. Pearson's correlation coefficient and linear regression analysis were employed to assess the relationship between the contact angle value and other input parameters. ML models, including decision tree, random forests, function networks, and gradient boosting regressor, were utilized for shale wettability prediction.Operating pressure, temperature, rock mineralogy, and total organic content were identified as influential factors on shale wettability. While the linear regression model showed limited accuracy, the ML models, especially random forests, decision tree, gradient boosting regressor, and function networks, effectively predicted the contact angle value with high R2 values. Gradient boosting regressor demonstrated the best performance, achieving R2 values of 0.99 and 0.98 for the training and testing datasets, respectively, with a root mean square error below 5 degrees. Sensitivity analysis revealed the significant impact of pressure on shale wettability, with low pressures resulting in water-wet shale regardless of other shale properties.The study highlights the efficacy of the developed ML models in predicting shale wettability in the context of CO2-water-shale systems, providing a faster and more affordable alternative to complex experimental analyses.