Full text loading...
-
Revealing Patterns in the Digital Age: Advancing Seismic Reservoir Characterization
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, Fifth EAGE Digitalization Conference & Exhibition, Mar 2025, Volume 2025, p.1 - 5
Abstract
Summary
This paper explores advancements in seismic reservoir characterization using machine learning. While traditional seismic inversion methods have improved reservoir models, they struggle in complex geological conditions. Convolutional Neural Networks (CNNs) provide a promising alternative by capturing complex, nonlinear patterns that traditional methods may miss. The study demonstrates how synthetic well data, generated using rock-physics models, enhances machine learning models by supplementing real well data, improving generalization and resolution. Case studies from the Gulf of Mexico show how this approach predicts reservoir properties like porosity, water saturation, and lithology, offering faster and more accurate predictions compared to traditional inversion methods. By combining rock-physics models with machine learning, this work reveals complex geological patterns and advances reservoir characterization, improving the accuracy of predictions.
© EAGE Publications BV