Estimating Shear Wave Velocity Using Deep Learning and Mineralogical Proportion Logs in the Búzios Field

Shear wave velocity (VS) is crucial in the oil and gas industry for reservoir characterization and seismic interpretation but is often missing from basic well logs. This study focuses on using deep neural networks (DNNs), alongside XGBoost and CatBoost algorithms, to estimate VS in 12 wells from the Búzios Field, the world’s largest deepwater oil producer. The approach stands out by incorporating mineralogical profiles (calcite, dolomite, quartz) as input features, enhancing model adaptability to complex geological scenarios. DNNs showed superior performance, achieving an R² of ~94%, significantly outperforming classical methods. By leveraging mineralogical logs, this methodology can better handle data gaps and geological variability. The combination of deep learning and detailed mineralogical profiles marks a significant advance for predicting VS and improving reservoir characterization in challenging environments.