Electrofacies and Lithofacies Fusion for Supervised Rock Typing in Highly Heterogeneous Reservoirs

Accurate rock typing is critical for assigning reservoir properties and improving volumetric estimates and reservoir models. Traditional methods—based on petrophysical cutoffs or indices like RQI and FZI—often struggle in geologically complex settings such as fluvio-marine reservoirs, where petrophysical properties and lithofacies show weak correlation. This study addresses that challenge by integrating lithofacies and electrofacies data using a supervised machine learning approach.

Focusing on the Group XY Sands of the Alpha Field, a highly heterogeneous fluvio-marine reservoir, lithofacies descriptions from five wells were combined with well log data to train a Random Forest classification model. The model effectively harmonized geological and petrophysical inputs, producing consistent and geologically reasonable rock type predictions.

Results showed strong agreement with core descriptions in cored intervals and logical facies assignments in uncored zones. At Alpha-3, the model identified subtle rock quality variations missed by traditional methods. Across all wells, the model demonstrated robust generalization and improved resolution in medium to poor quality intervals.

This integrated machine learning workflow enhances rock typing accuracy in complex reservoirs and supports more confident reservoir characterization and development planning.