Development of a Permeability Reduction Model Using Deep Learning for CO2 Hydrate Storage

Abstract

Global warming is an important environmental issue, and carbon capture and storage (CCS) is a major technology to reduce the emission of greenhouse gases. Captured carbon dioxide (CO2) can be stored in aquifers onshore or offshore seabed regions. Nevertheless, a small risk exists in which CO2 may leak due to natural phenomena opening cracks in the caprock. The natural formation of CO2 hydrates may create a new impermeable layer managing to block a possible leakage. It is of utmost necessity to understand and evaluate the permeability change due to the hydrate formation. Numerical simulation on different spatial scales has been essential for this purpose. The main objective of this study is to create a new framework for permeability reduction due to CO2 hydrate formation. Using machine learning, a multiscale approach links a large reservoir scale hydrate formation model with a microscale model. Detailed information from the hydrate shape can be obtained from the microscopic range to predict the new permeability reduction coefficient. Initial results have shown that this approach can obtain the permeability change due to CO2 hydrate formation with reasonable accuracy.