Facies prediction with Bayesian inference using supervised and semisupervised deep learning
Sagar Singh, Ilya Tsvankin, Ehsan Zabihi Naeini · 2021
Delineation of geologic facies from seismic reflection data plays an important role in reservoir characterization during hydrocarbon exploration and development. Facies classification is often done manually by an experienced interpreter, which makes this process subjective and inefficient. Several machine-learning (ML) models have been proposed to automate this interpretation but there are still significant practical challenges in multiclass facies segmentation. We present supervised and semisupervised Bayesian deep-learning methodologies designed to improve analysis of seismic facies depending on the scope of the labeled data. The developed networks reliably predict facies distribution using seismic reflection data and estimate the corresponding uncertainty. Two deep neural networks are successfully tested on seismic data from the Dutch sector of the North Sea. As expected, in the case of sufficient availability of manually interpreted labels (or facies), the supervised learning model accurately recovers the facies distribution. When the interpreted labels are limited, the semisupervised algorithm can be efficiently applied to avoid overfitting.