Improved pore-scale models for multi-phase flow in natural rocks with applications to geological sequestration of carbon dioxide
Understanding physics of two-phase flows in porous media is essential in many applications such as carbon dioxide (CO2) sequestration in deep saline reservoirs, recovery of oil from hydrocarbon reservoirs, etc. Reservoir-scale simulation models used in engineering practice require the continuum-scale constitutive relations. These relations depend upon the pore-scale rock geometry, interfacial tension, wetting properties, and physics of two-phase immiscible flow. Since laboratory measurement of these properties is very expensive and time-consuming, pore-scale numerical simulation has recently emerged as an alternative approach. X-ray micro-CT scanning technology can provide the necessary input of pore- scale models in terms of pore space of rock.
Although direct numerical simulation (DNS) can in principle be used, there are severe computational challenges for rock samples large enough to constitute a representative elementary volume (REV). Because of the computational burden of DNS, pore-network modeling (PNM) is one of the most common tools in pore-scale modeling. PNM achieves its computational advantage by simplifying the pore structure into idealized familiar shapes that mimic the pore space allowing use of predefined flow equations. Since PNM has proven effective under many circumstances in petroleum engineering, it has been applied to the new problem of geological sequestration of carbon dioxide, in which the greenhouse gas CO2 is captured from a conventional power plant and injected as a supercritical liquid into a deep saline reservoir. However, it has been found that conventional PNM fails in predicting trapped CO2 during the full drainage-imbibition cycle. This error becomes severe when the contact angle is small and the rock is more heterogeneous e.g. CO2-brine flow in Mount Simon sandstone.
We hypothesize that the source of this error is related to defined imbibition events in PNM and their corresponding idealized flow, capillarity, and trapping rules. To address this issue, we propose testing and modifying PNM imbibition rules with the aid of DNS using the lattice Boltzmann method (LBM), which can predict the exact interface location for several pore-network configurations and scenarios. LBM can provide velocity, pressure distribution, and phase field at each point that are required for quantified modification of PNM rules.
However, the computational cost of DNS for tracking interfaces in a full drainage-imbibition cycle is very high and can only be performed on high-end parallel computers such as Blue Waters. This research allocation request will enable us to understand the physics behind the deficiencies of PNM in CO2-brine systems and improve its predictive capabilities, especially the predicted amount of trapped CO2 in a natural reservoir rock. As PNM is widely used in a variety of applications, our research has the potential for significant impact by allowing for physically based modifications of existing PNM codes.