Deep-learning-based coupled flow-geomechanics surrogate model for CO2 sequestration

Abstract

A deep-learning-based surrogate model capable of predicting flow and geomechanical responses in CO2 storage operations is presented and applied. The 3D recurrent R-U-Net model combines deep convolutional and recurrent neural networks to capture the spatial distribution and temporal evolution of 3D saturation and pressure fields and 2D surface displacement maps. The method is trained using high-fidelity simulation results for 2000 storage-aquifer realizations characterized by multi-Gaussian porosity and log-permeability fields. Detailed comparisons between surrogate model and full-order simulation results for new storage-aquifer realizations are presented. The saturation, pressure and surface displacement fields provided by the surrogate model display a high degree of accuracy, for both individual realizations and ensemble statistics. The recurrent R-U-Net surrogate model is applied with a rejection sampling procedure for data assimilation. Although the (synthetic) observations consist of only a small number of surface displacement measurements, significant uncertainty reduction in pressure buildup at the caprock is achieved.