Modeling capillary bound water in limestone reservoirs

Precise evaluation of capillary bound water is critical for accurate hydrocarbon reserve estimation and reliable forecasting of reservoir behavior. Traditional laboratory techniques and empirical models often struggle with limitations in both accuracy and efficiency. This study introduces a data-driven approach to precisely assess capillary bound water saturation (CABW) in limestone reservoirs, a critical yet challenging aspect of hydrocarbon reserve estimation. This research leverages comprehensive core-analysis data, such as He-porosity, Gas permeability, and Nuclear Magnetic Resonance (NMR) T2lm​ data from diverse limestone core samples to train and validate various machine learning models. We evaluated the performance of several machine learning models. Rigorous data integrity checks were performed using the Leverage technique for outlier detection, while sensitivity analysis quantified the impact of features on CABW​. Model robustness was confirmed through K-fold cross-validation. The CNN model demonstrated superior performance and gas permeability as the strongest interpreter of CABW​, a finding that underscores the physical validity of our model at the core scale. This study highlights the strong capability of soft computing approaches to enhance petrophysical modeling in complex reservoirs, providing a scalable and economical alternative to traditional techniques. By integrating core analysis with computational modeling, the methodology advances reservoir characterization and reserve estimation with greater accuracy and reliability. • Accurate prediction of Capillary Bound Water in limestone reservoirs • Sensitivity analysis of effective parameters using SHAP and relevancy index • Outlier detection using a renowned approach