Capturing Coupled Effects of CO2 and Brine Leakage in Cemented Wellbores at CO2 Storage Sites using Decoupled Reduced Order Models
Abstract
Understanding the potential significance and risk of CO2 and brine leakage from wells at proposed geologic storage sites is a key goal of the National Risk Assessment Partnership (NRAP). In this study, we developed reduced order models (ROMs) of cemented-wellbore leakage that are being incorporated into systems models in order to determine CO2 storage risk profiles. ROMs are used to capture general trends in numerically simulated leakage in a computationally efficient manner, allowing for large numbers of evaluations required by stochastic evaluations of risk. In this study, we perform the wellbore leakage numerical simulations using the multi-phase Finite Element Heat and Mass (FEHM) code. The constituents of interest are water, supercritical CO2, dissolved CO2, liquid CO2, and gaseous CO2. The computational domain is a full 3D mesh with a 0.1 m wellbore at the center. A storage reservoir is located below the impermeable rock and the wellbore is connected directly to the atmosphere or to a shallow aquifer above. Injector wells introduce CO2 into the reservoir for 20 years followed by a 30 year relaxation period. Reservoir depth, cement permeability, relative permeability model parameters, and injection rate are sampled by Latin Hypercube sampling. For each sample, simulations are performed with and without a wellbore in the model. Transient pressures and CO2 saturations are collected at the top of the reservoir at the location where the wellbore would be located from the model without a wellbore. CO2 and brine flow rates (leakage) are collected near the top of the wellbore from the models with a wellbore. ROM model inputs are the sampled parameters, transient pressures and CO2 saturations and their first and second derivatives calculated by backward finite differences. Using these inputs, ROMs are generated for CO2 and brine leakage using the Multivariate Adaptive Regression Splines (MARS) regression technique. ROM development, evaluation, and use will be discussed. Key results show that ROMs can be developed that capture coupled effects, such as the effects of the reduction in CO2 saturation following leakage on the leakage rate, within a decoupled ROM.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2013
- Bibcode:
- 2013AGUFM.H23B1271H
- Keywords:
-
- 1847 HYDROLOGY Modeling;
- 1849 HYDROLOGY Numerical approximations and analysis;
- 1942 INFORMATICS Machine learning