Capillary Pressure Estimation in Carbonate Reservoirs via Machine Learning: An Experimental Data-Driven Approach

Accurate prediction of capillary pressure (Pc) is crucial for optimizing hydrocarbon recovery in carbonate reservoirs. This study presents a machine learning (ML) framework based on experimental data, where Categorical Boosting (CatBoost) outperforms Support Vector Regression (SVR), Random Forest (RF), K-Nearest Neighbors (KNN), and Decision Tree (DT). Engineered features effectively capture the non-linear relationship between Pc and wetting phase saturation (Sw) in heterogeneous carbonate rocks. The model's scalability and reduced reliance on extensive laboratory tests enhance reservoir characterization. Despite limited data diversity, incorporating varied lithologies and deep learning techniques could further improve performance and provide valuable insights for sustainable hydrocarbon recovery.

Keywords:

Capillary Pressure , Carbonate Reservoirs , Machine Learning , Predictive Modeling , Feature Engineering