Simulation Of Enhanced Recovery Projects-The Problems And The Pitfalls Of The Current Solutions

Abstract

Abstract Enhanced recovery schemes such as gas, solvent and CO2 injection are difficult to simulate. The paper discusses the problems of realistically modelling such processes with respect to:controlling the numerical dispersion and its influence on the displacement pattern;grid orientation effect and the means for its reduction; andsimulation of recovery in unstable displacements where viscous fingering, channeling and physical dispersion are important. Numerical methods for control of dispersion, modelling of fingering and elimination of the grid orientation effects are discussed in the light of their practical application. The important results which demonstrate the difficulties of obtaining correct solutions and need for further model improvements are:In unstable displacements it is important to control both phase and component dispersion. An example of multiple-contact-miscible (MCM) flood in a Canadian reef shows that small changes in the numerical treatment can produce drastic changes in the displacement pattern.The success of nine-point approximations is problem dependent. The commonly used Yanosik and McCracken method overcorrects the solution in miscible displacement. A new alternative method derived from variational considerations gives better solution in this case.The popular method of matching unfavourable mobility displacements by increased dispersion via the Todd-Longstaff or Koval method may in many cases be the wrong solution to the problems. The correct solution may require low-dispersion, high-accuracy modelling~ or significant improvements of the mixing-type models. Introduction Much research has been done in the past in the area of simulation of enhanced recovery processes. The applications have included immiscible and miscible gas injection, solvent and CO2 flooding, and other less common techniques (the "chemical" group of processes are not discussed in this paper). As the demands of the applications increased, modelling techniques have also evolved. Although some applications can be treated with extended black-oil models, the largest development occurred in the compositional simulation. Most of this work has been concerned with the formulation of the equations and the manner in which the phase equilibrium was represented in the model. The early simulators were based on an IMPES formulation and used K-value correlations for phase equilibrium (Roebuck et al.(1), Nolen(2), Kazemi et al.(3)). Much of the later work was directed toward improving the PVT treatment. Several authors(4, 5, 6) have developed models which use the thermodynamic equilibrium conditions instead of K-values in conjunction with equation of state, while others have pursued simplified PVT treatment(7, 8) for special classes of problems. New methods of organizing the computations and new formulations have been proposed(8–11) and stability of the models has been improved by the use of implicit techniques(4) and by explicit Runge-Kutta approximations(12). In contrast to this, relatively little attention has been paid to the accuracy of the computations. Coats(4) has reported serious numerical dispersion in a linear MCM flood. The problems of dispersion and grid orientation effect for multiphase flow as well as in miscible displacement have been long recognized. However, the known numerical techniques which can alleviate considerably the problems, have not been incorporated in existing simulators in a coherent fashion.