Simulation of Linear Displacement Experiments on Massively Parallel Computers

Abstract

Abstract We have developed a throat-bulb network model to simulate multiphase flow experiments. The model has been efficiently implemented on massively parallel computers (MasPar and CM5). This model has allowed us to study the dynamics of displacement processes. Using grids of size 64 × 64 × 128 (half a million cells), we are able to incorporate random variations in the throat and pore size distributions and observe flow patterns such as viscous fingering, stable front displacement, and invasion percolation. Furthermore, we are able to discern certain scaling rules in terms of macroscale variables. Displacement processes in a network model with random distribution of throat and pore size can be characterized by three parameters: M (viscosity ratio), Ca (capillary number), and Cg (the ratio of capillary number and Bond number). Macroscale variables, Kr and Sor, are strongly dependent on flow patterns, whereas the dynamic capillary pressure, Pc, is independent of flow patterns. Analysis of laboratory measurements of these macroscopic variables is generally based on steady state approximations and moreover, it offers no clue as to how one might extrapolate the results to the reservoir scale. Using the pore network simulator, we can model dynamic flow behavior and understand the scaling rules for experimental measurements. The simulator will eventually become a practical tool to investigate multi-phase flow in various flow conditions. Furthermore, a better understanding of experimental measurements will help us construct effective grid-block properties for reservoir simulation.