(685e) Probabilistic Analysis of CO2 Storage Mechanisms in a CO2-EOR Field Using Polynomial Chaos Expansion | AIChE

(685e) Probabilistic Analysis of CO2 Storage Mechanisms in a CO2-EOR Field Using Polynomial Chaos Expansion


Jia, W. - Presenter, University of Utah
McPherson, B., University of Utah
Xiao, T., University of Utah
Bromhal, G., National Energy Technology Laboratory
Depleted oil fields are already used for storing CO2 via injection of CO2 for enhanced oil recovery (CO2-EOR). Such storage is an outcome of CO2-EOR, albeit not necessarily by design. A next step would be intentional storage via post-EOR CO2 injection. Following injection for EOR or post-EOR, CO2 may be trapped by dissolution in the oil or aqueous phases, or remain in the supercritical CO2 phase. Forecasting the nature of trapping and the ultimate CO2 distribution in a reservoir is hindered by uncertainty in reservoir properties. The purpose of this study is to develop and apply reduced order models (ROMs) integrated with Monte Carlo simulations to quantify the three primary types of storage mechanisms in an active CO2-EOR field, including oil solubility trapping (oil phase), aqueous solubility trapping (aqueous phase), and hydrodynamic trapping (CO2 in supercritical phase). The case study site for this analysis is the SACROC unit in western Texas. A Polynomial Chaos Expansion (PCE) technique was used to develop the ROMs. The sources of uncertainty considered are porosity and permeability. Model outputs of interest include dissolved mass of CO2 in oil phase, mass of CO2 in supercritical phase, dissolved mass of CO2 in aqueous phase, and oil saturation in the rock formation. Reduced order models were developed for all cells in five selected layers of the model, which are adjacent to injection wells, at three specific time points of interest, including the end of a simulated CO2-EOR period, the end of a post-EOR continuous CO2 injection period, and the end of a post-injection monitoring period. Results of regression fit and validation analysis yield high R2 values and low NRMSE values, indicating that the ROMs derived from PCE are able to make meaningful predictions (compared to conventional reservoir models) and effectively represent relationships between model inputs and model outputs of interest. Results of 1000 Monte Carlo simulations indicate a dominant upward transport of CO2 during injection, driven by buoyancy, and a dominant downward transport of dissolved CO2 after injection stops, caused by the increase in both oil and brine density. At the end of the 100-year simulation of the SACROC case study, the selected five layers store between 54% and 61% of total trapped CO2 in the entire domain in the oil phase, between 21% and 24% in the supercritical phase, and between 11% and 15% in the aqueous phase. The selected five layers store between 89% and 98% of total trapped CO2 in the entire domain by the end of simulation.