Compositional Flow Simulation Using Reduced-Variables and Stability-Analysis Bypassing

Abstract

Abstract The Phase Equilibrium Computations (PEC) associated with compositional flow simulation include: (1) Phase-Stability Analysis (PSA) of single-phase cells, and (2) Two-Phase Flash (TPF) for cells that change from one to two phases over a Newton iteration. For simulation of multi-contact miscible displacements, the PEC cost increases dramatically with the number of components, and for highly detailed models, PEC can become the most time consuming part of a simulation. In this work, we integrated the Reduced-Variables (RV) method (Pan and Firoozabadi 2002, 2003) for PEC with a compositional simulator that uses the natural-variables formulation (Coats, 1980). We also implemented an RV-based form of Rasmussen et al.'s (2006) method for bypassing PSA computations during a simulation run. We present and analyze simulation results for a wide range of problems. Specifically, we used three reservoir models: the top layer of the SPE10 reservoir model, a discrete fracture model, and a 3D model with 400,000 cells. Three different fluids were investigated, namely, the SPE3 fluid (9 components in the critical region), a 15-component gas-condensate system at high pressure, and a 26-component oil. We focused on multi-contact miscible displacement involving complex phase behavior, both in space and time. Compositional flow simulation with the RV-based method for bypassing PSA computations proved to be robust and efficient. Compared with the standard approach, over 80% of the PSA computations are bypassed, and where PSA is needed, the RV-based Newton method converges quadratically. In a typical simulation, the PEC computations take up only about 10% of the total CPU time. The exceptional performance of the RV-based method with PSA bypassing allows for modeling complex EOR processes of practical interest, where large numbers of components and highly detailed reservoir descriptions are needed. Implementation of the RV-based approach with PSA bypassing in existing reservoir simulators, especially those based on the natural-variables set, is straight forward.