Rapid multiscale modeling of flow in porous media

Increasing our understanding of single and multiphase flow properties requires an accurate description of the pore system. Recent imaging technology provides images at different resolutions by which the pore space can be characterized properly. An accurate prediction of flow properties cannot be achieved computationally if the simulation domain is small. For a large sample, however, one may upscale the fine-scale information through pore-network modeling to expedite the computations. Due to a small field of view of the available images, such modeling can be conducted with ease for simple and heterogeneous rock samples. For complex and multiscale systems, however, one single-resolution image may not be sufficient. Thus, one should integrate the valuable information that lies in both the fine- and coarse scale into the pore network. In this paper, multiscale pore networks are generated which can take the microporosity, either in the solid phase (i.e., grain filling) or pore space (pore filling), into account. The proposed method can use the available high-resolution images for different pores and/or grains and stochastically build several micronetworks, which avoids fewer simplifications as present in the current methods. Next, depending on the location of high-resolution images, the generated micronetworks are implanted in the designated spots. Furthermore, cementation is also quantified and simulated by eroding the pore space. Finally, all the above physics are integrated within several stochastic multiscale networks. The final models are then used to evaluate the flow properties by comparing the statistical characteristics and single or multiphase flow behavior (e.g., capillary pressure, absolute and relative permeability). The results indicate the importance of incorporating the micropores as they can effectively connect the large-scale and isolated macropores. Furthermore, the single- and multiscale pore networks manifested a significant difference between the capillary pressure and residual saturation. Finally, a representative elementary volume study based on the generated multiscale pore networks for different scenarios is performed and the results are compared.