Seismic Signal Interpretation for Reservoir Facies Classification

Abstract

Understanding facies distribution in a hydrocarbon reservoir is an important aspect to characterise a hydrocarbon reservoir. Facies classes are basically based on rock characteristics that can indicate the locations of good quality of sand and presence of hydrocarbon, thereby helping the geologists to decide the location to drill a production well in a hydrocarbon field. However, due to the heterogeneous and nonlinear nature of earth subsurface, gathering facies information becomes a critical task. Researchers from different domains such as Machine learning, Geology, and Geophysics work towards the understanding of facies distribution. However, with the increased complexity of the reservoir, its interpretation becomes difficult. This work describes a case study that involves a framework to classify the facies categories in a reservoir using seismic data by employing different machine learning models. The framework is also capable to handle the data imbalance problem that occurs quite often while studying these kinds of datasets. Moreover, for gaining more confidence in the developed model, we used Local Interpretable Model-Agnostic Explanations to provide the interpretation of the model. The interpretations generated can be helpful for geologists to rate the applicability of our developed model in their domain.