Well-log attributes assist in the determination of reservoir formation tops in wells with sparse well-log data

The manual picking of reservoir formation boundaries using limited available well-log data in multiple wells across gas and oil reservoirs tends to be subjective unreliable. reasons for this are typically caused by the combined effects spatial boundary complexity availability. Formation characterization classification can improved when treated as a binary task based on two or three recorded well logs assisted their calculated derivative volatility attributes assessed machine learning. Two example wellbores penetrating complex boundary, one with gamma-ray, compressional-sonic, bulk-density recorded, other just gamma-ray used illustrate more rigorous proposed methodology. By combining attribute calculation, optimized feature selection, multi-k-fold cross validation, confusion matrices, feature-influence analysis, learning models it is possible improve boundary. With plus selected attributes. K-nearest neighbour, support vector classification, extreme gradient boosting able achieve high accuracy: greater than 0.97 training/validation well; 0.94 testing another well. analysis reveals that most important predictions but these likely vary from reservoir. results study suggest has potential provide systematic definition relying only few curves. Cited as: Wood., D. A. Well-log assist determination tops sparse data. Advances Geo-Energy Research, 2023, 8(1): 45-60. https://doi.org/10.46690/ager.2023.04.05