Abstract

Development of oil & gas fields requires reservoir characterization followed by geo-cellular property modeling (GCM). Facies interpretation using suites of recorded logs is an integral part of GCM which consumes a significant amount of project time. Facies interpretation using advanced artificial intelligence/machine learning methods provides good accuracy and saves a lot of time. But these methods of facies classification fail if the facies class distribution is imbalanced, i.e.count of hydrocarbon facies are, in general, very less compared to that of shale or sand.