Experimental workflow to estimate model parameters for evaluating long term viscoelastic response of CO2 storage caprocks

Abstract

Understanding the time-dependent behavior of reservoir and sealing formations for geologic carbon storage is critical to assessing geomechanical risks since time-dependent deformation strongly influences the mechanical response of some rock types. Many studies have evaluated the risk of CO2 leakage and induced seismicity by assuming poroelastic rheology in sealing formations. Few have considered viscoelastic or other time-dependent responses, and much of the literature on the long-term mechanical behavior of rocks examines only a 1D uniaxial response. This is primarily because, to date, the general form of a reasonable 3D time-dependent model for rocks remains unclear. In this paper, we propose an approach to address this by using a new workflow to estimate constitutive modeling parameters for the evaluation of a 3D viscoelastic model. The proposed approach uses a 1D power-law response to extrapolate several-hour-long experimental data to the decades-long time frames of interest in geologic carbon storage. Experimental data were obtained by conducting multi-level loading/unloading triaxial relaxation tests with four different rock types. The experimental results showed that the maximum load relaxation observed is approximately 49%, with some rock types considered showing as little as 1.4%. Using the proposed workflow, two linear 3D viscoelastic models, i.e., generalized Maxwell (GM) model and fractional Kelvin-Voigt (FKV) model, were evaluated and their model parameters were chosen with the extrapolated 30-year data such that a maximum deviation from the assumed power-law response for these two 3D models was 2 MPa in axial stress and 7 MPa in radial stress. We provided reasonable ranges for the model parameters to be later used for 3D modeling of rock time-dependent responses. Our results also showed that when GM is selected to analyze the rocks considered here, the relaxation time has a general range of 1–1010 s. This time scale captures the time-dependent behavior as long as centuries, which is much longer than a 10–30 years-long time frame envisioned for CO2 injection projects.