Pore-scale simulation of flow of CO2 and brine in reconstructed and actual 3D rock cores

Sequestration of CO2 in deep underground saline formations is currently under study as a practical approach for reducing emissions of CO2 from power plants into the atmosphere, thereby helping to mitigate problems of climate change and global warming. Such formations are required to be overlain by an impermeable formation in order to prevent CO2 from leakage and migration, while also having sufficient permeability to allow injection without excessive pressure build up. Thus, any potential sequestration site should be extensively characterized to assess its properties for tight and permanent isolation of CO2. Achieving this goal is difficult without considering the heterogeneity and variability of the formations. Thus, a considerable number of samples covering the range of variability must be studied; pore morphology and connectivity in these samples are assessed using three-dimensional (3D) imaging techniques, which can be very costly and time consuming. As an alternative, one can use 2D images, which are widely available, as the input data to stochastically reconstruct realizations of the porous formation. To this end, we present a method by which one generates an ensemble of 3D realizations using one or very few 2D images by an iterative multiscale approach. The method first generates an initial 3D model and then progressively improves it by removing artifacts and refining the spatial connectivities. Thus, various possible realizations of the connectivity of the pore space can be modeled without much experimentation with actual 3D samples. These realizations can then be used in pore network or lattice Boltzmann simulation of CO2-brine flow processes. The method is utilized to reconstruct realizations of the Mt. Simon sandstone, the target reservoir for a pilot CO2 sequestration project conducted by the Illinois State Geological Survey, for which there is a large data base of rock core and 3D structure. Then, the flow behavior of a CO2-brine system is simulated in the reconstructed realizations. Quantitative comparisons for the multiple point connectivity function, single-phase permeability, capillary-pressure-saturation functions, two-phase relative permeabilities, and the statistical properties of the pore network extracted from the data are presented. Good agreement is found in all the cases between the computed results for the reconstructed models and those of the actual sample.