Inverse analysis for detection of high permeability zones in porous media

The objective of this dissertation is to estimate possible leakage pathways such as abandoned wells and fault zones in the deep subsurface for CO2 storage using inverse analysis. Leakage pathways through a cap rock may cause CO2 to migrate into the layers above cap rock. An inverse analysis using iTOUGH2 was applied to estimate possible leakage pathways using pressure anomalies in the overlying formation induced by brine and/or CO2 leaks. Prior to applying inverse analysis, sensitivity analysis and forward modeling were conducted. In addition, an inverse model was developed for single-phase flow and it was applied to the leakage pathway estimation in a brine/CO2 system. Migration of brine/CO2 through the leakage pathway was simulated in the generic homogeneous and heterogeneous domains. The increased pressure gradient due to CO2 injection continuously induced brine leaks through the leakage pathway. Capillary pressure was induced by the migration of CO2 along the leakage pathway saturated by brine. Pressure anomalies due to capillary pressures were propagated to the entire overlying formation. The sensitivity analysis was focused on how the hydrogeological properties affect the pressure signals at monitoring wells. Parameter estimation using the iTOUGH2 model was applied to detect locations of leakage pathways in homogeneous and heterogeneous model domains. For homogeneous models, the parameterization of uncertain permeability in an overlying formation could improve location estimation accuracy. Residual analysis illustrated that pressure anomalies in the overlying formation induced by leaks are critical information for the leakage pathway estimation. For heterogeneous models, the calibration of renormalized permeability values could reduce systematic modeling errors and should improve the leakage pathway location estimation accuracy. The weighting factors significantly influenced the accuracy of the leakage pathway estimation. The developed inverse model was applied to estimate the leakage pathway in a brine/CO2 system using pressure anomalies induced by only brine leaks. To estimate a possible leakage pathway, the developed inverse model calibrated each integrated parameter (of both cross-sectional area and vertical hydraulic conductivity) of initial guesses of the leakage pathway. This application can provide warning before the CO2 leaks, and will be useful in mitigating the risk of CO2 leaks.