Lithology Identification Method Based on Machine Learning and Geophysical Well Logging

Abstract

Lithology identification assumes an absolutely crucial role in the realm of reservoir delineation, functioning as a fundamental prerequisite for accurately determining porosity, oil saturation, and sundry other parameters. The precise identification of lithology effectively forms a foundation for ensuring the effective exploration and development of oil and gas fields in the subsequent stages. Traditional logging lithology identification methods have many limitations, and therefore many challenges. Therefore rapid and accurate lithology identification is a matter of significant concern and importance. This study identified the initial lithology using a cross-plot approach, capitalizing on the distinctive characteristics of the lithology logging response. Subsequently, a rapid lithology identification method was developed for marine carbonate rocks by integrating artificial neural networks and the logging curves. To evaluate its accuracy and precision, the outcomes were compared with the core data and the lithology scanning logging techniques. The neural network–based method proposed in this paper can enable swift and accurate identification of lithologies, encompassing even transitional lithologies such as silty limestone and limy dolomite. Therefore, it provides novel theoretical and technical support that is of tremendous significance for subsequent research and applications.