Improvements in phase behavior modeling for compositional simulation
Accurate and reliable phase equilibrium calculations are among the most important issues in compositional reservoir simulation of enhanced oil recovery (EOR) processes especially miscible gas floods. The important challenges in equation of state (EOS)-based compositional simulators are the time-consuming nature of the phase equilibrium calculations, e.g. 30%-50% of the total computational time in the UTCOMP simulator (Chang, 1990), and accuracy as well as robustness of these calculations. Thus, increasing the computational speed and robustness of the phase equilibrium calculations is of utmost importance in IMPEC-type and fully implicit reservoir simulators. Furthermore, most current compositional reservoir simulators ignore the effect of capillary pressure in porous media on the fluid’s phase behavior. This assumption may lead to significant errors in performance prediction of tight oil and shale gas reservoirs where the small pore sizes result in very large capillary pressure values. The “tie-simplex-based (TSB) phase behavior modeling” techniques attempt to speed up phase behavior calculations by skipping stability analysis and preconditioning phase-split calculations. We implemented the compositional space adaptive tabulation (CSAT), a TSB phase behavior modeling method, in UTCOMP and compared the computational performance of CSAT when used for skipping stability analysis and generating initial estimates for flash calculations, against the standard phase behavior modeling methods in UTCOMP. The results show that the CSAT method as well as a simple heuristic technique, where stability analysis is skipped for single-phase gridblocks surrounded by single phase neighbors, can improve the total computational time by up to 30% compared to the original UTCOMP. In order to avoid the negative-flash calculations required for adaptive tie-line tabulation during the simulation, a prior set of tie-line tables can be used. We demonstrate that the tie lines from the multiple-mixing-cell (MMC) method are very close to the actual compositional simulation tie lines. Thus, the MMC tie lines were used as prior tieline tables in three tie-line-based K-value simulation methods in order to improve speed and robustness of compositional simulation. Several simulation case studies were performed to compare the computational efficiency of the three MMC-based methods, an extended CSAT method (adaptive K-value simulation) and a method based on pure heuristic techniques against the original UTCOMP formulation. The results show that the MMC-based methods and the extended CSAT method can improve the total computational time by up to 50% with acceptable accuracy for the cases studied. The MMC-based methods, the CSAT method and the heuristic methods were implemented in the natural variable formulation in the fully-implicit General Purpose Adaptive Simulator (GPAS) for speeding up the phase equilibrium calculations. The computational efficiency results for several cases that we studied show that the CSAT method and the MMC-based method improve the computational time of the phase equilibrium calculations by up to 78% in the multi-contact-miscible gas injection cases studied. Finally, we present a Gibbs free energy analysis of capillary equilibrium and demonstrate that there is a limiting maximum capillary pressure (P[subscript cmax]) where gas/oil capillary equilibrium is possible and formulate the P [subscript cmax] limit using the spinodal condition of the phase of smaller pressure in capillary equilibrium. The effect of capillary pressure on phase behavior was implemented in the UTCOMP simulator and several simulation case studies in shale gas and tight oil reservoirs were performed. The simulation results illustrate the effect of capillary pressure on production behavior in shale gas and tight oil reservoirs.