SPE Journal
Volume 15, Number 2, June 2010, pp. 313-325

SPE-117752-PA

A New Algorithm for Rachford-Rice for Multiphase Compositional Simulation

View full textPDF ( 986 KB )

DOI  More information 10.2118/117752-PA http://dx.doi.org/10.2118/117752-PA

Citation

  • Okuno, R., Johns, R.T., and Sepehrnoori, K. 2010. A New Algorithm for Rachford-Rice for Multiphase Compositional Simulation. SPE J.  15 (2): 313-325. SPE-117752-PA. doi: 10.2118/117752-PA.

Discipline Categories

  • 6.2.1 Phase Behavior and PVT Measurements
  • 6.2.2 Fluid Modeling, Equations of State
  • 6.5.1 Simulator Development
  • 6.4.2 Gas-Injection Methods
  • 6.3.1 Flow in Porous Media

Keywords

  • multiphase flash calculations, Rachford-Rice equations, multiphase behavior, compositional simulation, gasflooding

Summary

Flash calculations for use in compositional simulation are more difficult and time-consuming as the number of equilibrium phases increases beyond two. Because of its complexity many simulators do not even attempt to incorporate three or more hydrocarbon phases even though such cases are important in many low-temperature gas floods, or for high temperatures where hydrocarbons can partition into water. Multiphase flash algorithms typically use successive substitution (SS) followed by Newton's method. For NP-phase flash calculations, (NP-1) Rachford-Rice (RR) equations are solved in every iteration step in SS, and depending on the choice of independent variables, in Newton's method. Solution of RR equations determines both compositions and amounts of phases for a fixed overall composition and set of K -values. A robust algorithm for RR is critical to obtain convergence in multiphase compositional simulation, and has not been satisfactorily developed unlike the traditional two-phase flash. In this paper, we develop an algorithm for RR equations for multiphase compositional simulation that is guaranteed to converge to the correct solution independent of the number of phases for both positive and negative flash calculations.

We derive a function, whose gradient vector consists of RR equations. This correct solution to the RR equations is formulated as a minimization of the non-monotonic convex function using the independent variables of (NP-1) phase mole fractions. The key to obtaining a robust algorithm is that we specify non-negative constraints for the resulting equilibrium phase compositions, which are described by a very small region with no poles. The minimization uses Newton's direction with a line search technique to exhibit super-linear convergence. We show a case in which a previously developed method cannot converge while our algorithm rapidly converges in a few iterations. We implement the algorithm both in a stand-alone flash code and in UTCOMP, a multiphase compositional simulator, and show that the algorithm is guaranteed to converge when a multiphase region exists as indicated by stability analysis.

View full textPDF ( 986 KB )

History

  • Original manuscript received: 16 June 2009
  • Revised manuscript received: 11 August 2009
  • Manuscript approved: 19 August 2009
  • Published online: 16 February 2010
  • Version of record: 17 June 2010