Machine-learning based technique for lithology and fluid content prediction: Case study from offshore West Africa

Abstract

Machine learning and artificial intelligence are new trends in earth science applications, opening immense opportunities to explore, analyze and leverage various types of data to identify important insights. However, one main concern underlying these advances, in the geoscience community, is how to reliably interpret machine learning (purely data-driven algorithm) outputs on geophysical datasets (physics-based input). The algorithm itself cannot be an alternative for physical modeling, but when integrated with and trained by a detailed prior-knowledge of the physical problem and geological constraints, it is likely to lead to successful solutions. To develop improved predictive workflows for complex geological problems, we integrate seismic attributes, well log data, texture attributes and machine learning to predict lithology and fluid content. Blind tests from various field datasets verify the reliability of predicted results, and the potential and applicability of the proposed workflow for exploration and reservoir management. Presentation Date: Wednesday, October 17, 2018 Start Time: 1:50:00 PM Location: Poster Station 2 Presentation Type: Poster