Building informative priors for the subsurface with generative adversarial networks and graphs
Abstract/Contents
- Abstract
- Uncertainty quantification of flow in porous media in the subsurface is important for applications such as aquifer management, CO2 storage, hydrocarbon production and contaminant transport modeling. The main goal of this dissertation is to develop new approaches to build fast and geologically consistent priors for the subsurface. We first review the physics of computational sediment transport and the consequences for simulation in different depositional environments. A comparison of a data set of 110 modern analogs is made with a large number of Delft3D realizations of river deltas using the channel networks as a framework. Because of the prohibitive computational cost associated with using computational hydrodynamics on geological time scales, it is necessary to speed up such experiments significantly or to somehow extract the essence of the information. One possibility is to use the final output as training images. Thus, a Generative Adversarial Network, which is a modern family of generative machine learning models, is here trained using a data set of satellite images of river delta networks. These types of methods hold much promise for the future but are shown to have limitations when realizations need to be generated in data-dense environments. As an alternative to artifical neural networks, which require large amounts of training data, we propose a computationally cheaper and grid-free alternative, namely the framework of spectral graph theory. The mapped graph Laplacian spectrum is shown to hold strong discriminative power both for real and synthetic river deltas, digital rock samples and heterogeneous porous media. We argue that there should be done more research on generating such geological graphs directly based on prior distributions of more detailed geological patterns.
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource. |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2019; ©2019 |
| Publication date | 2019; 2019 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Nesvold, Erik |
|---|---|
| Degree supervisor | Mukerji, Tapan, 1965- |
| Thesis advisor | Mukerji, Tapan, 1965- |
| Thesis advisor | Caers, Jef |
| Thesis advisor | Kitanidis, P. K. (Peter K.) |
| Degree committee member | Caers, Jef |
| Degree committee member | Kitanidis, P. K. (Peter K.) |
| Associated with | Stanford University, Department of Energy Resources Engineering. |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Erik Nesvold. |
|---|---|
| Note | Submitted to the Department of Energy Resources Engineering. |
| Thesis | Thesis Ph.D. Stanford University 2019. |
| Location | electronic resource |
Access conditions
- Copyright
- © 2019 by Erik Nesvold
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...