Compositional simulation of reservoir performance by a reduced thermodynamic model

Peng Wang, E. H. Stenby

Research output: Contribution to journalJournal articleResearchpeer-review

Abstract

In compositional reservoir simulations, it is usually assumed that a local thermodynamic equilibrium between all phases exists throughout the reservoir within each time step. Phase behavior calculations are conducted based on an equation of state (EOS) in each block. This computation normally takes a very significant percentage of the total CPU time. In this work, an extension of the reduced thermodynamic model (Wang and Stenby, Computers chem. Engng 16, 5449-5456, 1992) to the processes in which both pressure and feed composition vary is demonstrated, This modified model has been tested against the experimental data of gas condensates and the PVT data computed from the EOS. The results indicate that the new model can well represent the influences of changes in the feed composition and the pressure on the K-values.The new model has been implemented into a compositional reservoir simulator, UTCOMP (Chang Ph.D. Thesis, Univ. of Texas, Austin, 1990). Simulations of natural depletion for three well-defined mixtures and of a dry gas recycle process for a gas condensate in the North Sea are performed. The results simulated with the new method are comparable to those computed by the PR EOS. But the CPU time is reduced by approximately 50% by means of the new method compared to the one required by the PR EOS. In addition, this work shows that the CPU time can be further decreased by using the newly suggested procedure (Leibovici and Neoschil, Fluid Phase Equil. 74, 303-308, 1992) for solving the Rachford-Rice equation for phase split.
Original languageEnglish
JournalComputers and Chemical Engineering
Volume18
Issue number2
Pages (from-to)75-81
Number of pages7
ISSN0098-1354
DOIs
Publication statusPublished - 1994

Fingerprint Dive into the research topics of 'Compositional simulation of reservoir performance by a reduced thermodynamic model'. Together they form a unique fingerprint.

Cite this