Relative permeabilities from simulation in 3D rock models and equivalent pore networks: Critical review and way forward

In the last 15 years, great progress has been achieved in the area of digital rock technology in the domain of 3D rock-model generation using a range of techniques: sedimentation modeling, statistical methods and micro-CT imaging. In the context of sandstones, single-phase prediction from these 3D models or from the equivalent pore networks has become possible for static and, in some cases, for flow petrophysical properties. Nevertheless, according to Sorbie and Skauge (2012), two-phase flow properties such as relative permeabilities cannot be predicted from either the equivalent pore networks or from the 3D models. In the first case, the input choices one has to make throughout the entire prediction workflow would greatly outnumber the actual system parameters; in the second, while the number of input parameters still remains too high, no technique is mature enough to produce calculations "that can be compared meaningfully with experiment". In this work, we extend the observations of Sorbie and Skauge (2012) by actually simulating two-phase flow properties with a number of state-of-the-art pore-network models and one Lattice-Boltzmann simulator. First we show how the network-extraction process and not the distribution of the fluids in the pore network might be at the origin of some recently published counterintuitive results by another group on relative permeabilities obtained by means of pore-network simulation. We then show that the current physics embedded in network models, and the latest refinements to it, are less a concern than other issues, such as, rock representation and wettability characterization. We conclude by comparing results from direct two-phase flow simulation in a micro-CT image and in the network model extracted from that image to focus on the issues that need to be addressed to establish these techniques as industry-ready tools.