Leveraging Machine Learning Capabilities in Seismic Reservoir Characterization

Abstract Objectives & Scope The primary objective of this study is to explore the application and advantages of machine learning (ML) in seismic petrophysics and reservoir characterization, with a focus on quantitative interpretation, facies prediction, and petro-elastic modeling. The scope includes evaluating ML techniques for facies classification and the prediction of essential reservoir properties, such as porosity, lithology, and fluid saturation, from seismic inversion attributes. Challenges & Procedures Traditional approaches to seismic reservoir characterization involve complex, time-intensive workflows that often struggle to capture the heterogeneity of subsurface formations. By leveraging ML, this work aims to improve prediction accuracy, streamline workflows, and enhance the reliability of interpretations for efficient reservoir management and development. This study will also show a comparative analysis between machine learning algorithms and multilinear regressions for petro-elastic modeling (PEM) and seismic reservoir properties prediction and will compare the machine learning algorithms and Bayesian inversion technique for seismic facies classification model.