This SPE Live introduces three prominent approaches to generative AI, followed by a presentation of a novel method utilizing two of these approaches to efficiently generate high-quality geomodels essential for understanding the probabilistic distributions of geological properties within a subsurface hydrocarbon play. By combining a hierarchical vector-quantized autoencoder with an autoregressive model, significant acceleration of both unconditional and conditional geomodel generation is achievable through this implementation of generative AI.
Speaker: Dr. Siddharth Misra, Associate Professor at Texas A&M University.
Moderated by Yusuf Ajibola Falola, PhD student, Department of Petroleum Engineering Texas A&M University.