A Comparison of Various Formulations for Compositional Reservoir Simulation

Abstract

Abstract Compositional reservoir simulation is a complex task that involves solving a large system of nonlinear equations arising from the discretization of partial differential equations (PDEs) for modeling fluid flow in porous media. In order to solve such a nonlinear system, the unknowns are grouped into primary and secondary variables. Since there are many unknowns in a compositional formulation of fluid flow in a reservoir, there are many approaches for selecting the primary and secondary variables. Such approaches lead to the development of various formulations that have been published in the petroleum literature. This work aims at comparing various formulations to determine their suitability and assess the efficiency of such formulations for solving different scenarios using equation-of-state compositional reservoir simulations (immiscible gas injection, CO2 injection, and condensate gas production). To accomplish these objectives, several formulations for compositional reservoir simulations presented in the literature were implemented into an in-house reservoir simulator called General Purpose Adaptive Reservoir Simulator (GPAS) and their performances were compared. As mentioned above, discretization of the governing PDEs leads to a system of nonlinear equations that needs to be linearized leading to a large system of linear equations that must be solved at every time step. In most cases, a variation of Newton’s method is used for linearizing the resulting nonlinear equations. The results of our comparison studies show that depending on the approach used, the number of Newton’s iterations for the resulting system of linear equations varies. As such, this will result in varied performance for the various formulations and case studies used for the simulation of different processes. There are many papers in the literature that present comparison of different simulators. Although, these comparisons are useful, but in most cases such comparisons do not present various formulations for solving different case studies for compositional reservoir simulations under a common framework. Comparing various formulations under the same framework can lead to a thorough investigation of convergence properties and efficiencies for different formulations. Results of this work suggest that gains in performance, when simulating compositional reservoir simulations, can be obtained by selecting appropriate formulations.