Characterization of gas flow through low-permeability claystone: Laboratory experiments and two-phase flow analyses

For the characterization of gas migration through a low-permeability clay host rock for deep underground repositories, a comprehensive understanding of the relevant phenomena of gas and fluid flow through low-permeability clay is required. The National Cooperative for the Disposal of Radioactive Waste (Nagra) in Switzerland has developed a comprehensive programme to characterize gas flow in low-permeability Opalinus Clay through laboratory tests and detailed numerical analyses for developing appropriate constitutive models. Laboratory tests were performed on cores by two different laboratories, the Laboratory for Soil Mechanics at EPFL and the Department of Geotechnical Engineering and Geosciences at UPC. Loading tests were performed by both laboratories to study rock compressibility at different stress levels and water permeability dependence on void ratio. The water retention behaviour demonstrated by EPFL and UPC produced comparable results. Water permeability tests and fast controlled-volume air injection experiments were performed in a triaxial cell under isotropic stress conditions on two samples with flow parallel and normal to the bedding planes. A confining stress of 15 MPa was applied during gas testing, corresponding to a lithostatic pressure at a depth of c. 600 m below ground. For detailed analyses, the two-phase flow code TOUGH2 (Pruess et al. 1999) was used. This considers fluid flow in both liquid and gas phases under the influence of pressure, viscous and gravity forces, according to Darcy's law. The standard analyses could not reproduce the measured pressure responses well, and the calibrated hydraulic and two-phase parameters were not consistent with the preceding water test and laboratory analyses. Implementing the non-linear behaviour in terms of the observed relationship between changes in void ratio and associated changes in permeability under different stress conditions significantly improved the simulated results, resulting in a conceptual model that well reproduced the observed injection pressure and outflow responses for both tests, parallel and normal to bedding, using a consistent parameter set.