A Deep Learning-Based Framework for High Certainty CO2 Storage Properties Estimation

This research addresses the challenges of accurately estimating carbon capture and storage (CCS) potential in subsurface geological formations by proposing a novel deep learning framework for predicting CO2 storage properties, specifically porosity and permeability, using well log and seismic acoustic impedance data. The approach overcomes the limitations of traditional variogram-based methods by employing specialized deep neural networks for properties estimation and a self-organizing map for rock type classification. The framework’s effectiveness is demonstrated through a successful application of characterizing a real depleted gas reservoir, highlighting its potential to enhance and derisk CCS strategies.