New strategies and procedures for efficient compositional reservoir simulation

New strategies and procedures have been proposed upon which a new three-dimensional, three-phase, isothermal, compositional simulator (AGGIECOMP) has been developed for compositional reservoir simulation.;The overall computational procedure of the simulator is noniterative-sequential in nature over a time step. The numerical solution scheme is implicit in pressure and explicit on saturation and composition (IMPESC). Instead of solving the reservoir mass balance equation directly, the reservoir volume balance equation is used to solve the reference phase pressure.;A minimization criterion is proved to satisfy the necessary and sufficient conditions for phases to reach equilibrium, which enables us to split a complicated phase equilibrium problem into simpler optimization problems. Rule of maximum number of phases, rule of phase disappearance, and rule of phase continuity are proposed to perform the strategy analysis involved in compositional reservoir simulation and to fast-find the correct number and types of phases. The combination of the theories above and K-value method yields an efficient fluid phase equilibrium hybrid algorithm with which the false solution problem involved in the traditional K-value method is avoided and the slow computational efficiency problem associated with the Gibbs energy minimization method is solved. The solution satisfies both necessary and sufficient fluid phase equilibrium conditions.;Three methods are proposed to solve the computational problems efficiently: an SS-NR hybrid method (a combination of the successive substitution and Newton-Raphson methods) for phase stability analysis, a modified method (MM method) for trivial solution problem, and an SS-MM hybrid method (a combination of the successive substitution and modified methods) for phase equilibrium calculations. The applications of the methods proposed can significantly improve the computational efficiency for near critical point systems.;With the models, strategies, and algorithms proposed in this work, the numerical results from AGGIECOMP closely match those from VIP (a commercial simulator developed by Western Atlas International Inc.) and COMPDU2.0 (an in-house simulator available at Texas A&M University), and AGGIECOMP is much faster than COMPDU2.0 for the simulation runs performed in this study.