Coal Wettability Prediction for Carbon Geo-Storage Through Data-Driven Machine Learning Approaches

Abstract One promising strategy for reducing carbon output involves the sequestration of CO2 in subsurface geological formations, particularly within coal seams. This approach not only stores CO2 but also facilitate enhanced coal bed methane recovery (ECBM). The performance of CO2 sequestration in coal seams is critically linked to the wettability of coal when it interacts with both CO2 and water. Although traditional methods—such as measuring an equivalent contact angle—are commonly employed to estimate wettability, they often suffer from significant uncertainties, require specialized equipment, and can be time-intensive. This study explores the wettability of coal surfaces in the context of carbon dioxide (CO2) injection for storage purposes. It examines how factors such as temperature, vitrinite reflectance, and salinity influence coal's wetting behavior, addressing previous discrepancies in the literature and shedding light on the complex interactions among CO2, water, and coal. To overcome the limitations of conventional experimental methods, this study employed advanced machine learning and deep learning techniques—including tree-based algorithms and neural networks—to evaluate wettability in coal-water-CO2 systems. Using a dataset comprising around 300 contact angle measurements taken under temperature conditions ranging from 25 to 75°C (reflective of subsurface coal seam environments), 80% of the data was allocated for training the models while the remaining 20% was reserved for testing. Validation with unseen data revealed that wettability tends to decrease with increasing temperature but shows a positive correlation with both salinity and vitrinite reflectance—although the behavior changes notably at higher reflectance values and lower salinities. These insights demonstrate the value of innovative computational methodologies in enhancing our understanding of coal wettability, which is crucial for optimizing both CO2 storage and methane recovery processes.