Seismic interpretation of geological units is a tedious task difficult to solve with deep learning. Current methods often require considerable labelled data to provide satisfactory results. We introduce an innovative approach based on Physics-Informed Neural Network (PINN) to constrain the temporal stacking of predicted geological units. Our hybrid learning strategy trains the network in a supervised manner on a few labelled lines and in an unsupervised manner on the entire dataset. We showed, on a real seismic dataset, that our approach significantly improves the prediction’s accuracy and geological consistency while requiring minimal amounts of labelled data.

Keywords:
deep learning, stratigraphy, geology, prediction, geologist, neural network, reservoir characterization, upstream oil & gas, artificial intelligence, machine learning
Subjects:
Reservoir Characterization, Information Management and Systems, Seismic processing and interpretation, Neural networks
This content is only available via PDF.
Copyright 2023, Society of Exploration Geophysicists and the American Association of Petroleum Geologists
You can access this article if you purchase or spend a download.