Comparison of Simulation and Experiments for Compositionally Well-Defined Corefloods

Abstract

Abstract The mechanisms which affect the efficiency of chemical flooding processes have been the focus of considerable research and speculation in recent years. One method to identify these mechanisms is the comparison of carefully conducted experiments with results generated using a mathematical model (simulator). In this paper we report both experimental and simulation results for II-, II+, and three-phase systems run so that the entire flood remains on a single phase diagram. In order to do this rigorously, we have used simple systems in which the complexities of non-pseudoternary behavior are absent. We simulated experimental core floods using a modified version of INTERCOMP's Chemical Flooding Simulator CFTE and tested the sensitivity of the results to a variety of process mechanisms. We also employed a method-of-characteristics solution to independently check the simulator. In doing so we extended the method-of-characteristics to three-phase, multicomponent flow in porous media. Our major objective is to show the interactions among the phase behavior, fluid flow, mixing, and mobility mechanisms which are present in chemical flooding processes. Introduction Over the past ten years INTERCOMP has been developing and improving their chemical flood simulator(s) by incorporating concepts and mechanisms as they are suggested by the companies funding the development program. The current version of the simulator includes a wide variety of mechanisms, both proven and speculative. The simulator has a tremendous amount of very diverse input data since each added mechanism requires a new block of input data. In general, a simulator with a wide variety of mechanisms can match a single experimental variable (say oil production) to within experimental accuracy with a variety of mechanistically different inputs. This -non-uniqueness- increases the difficulty of interpreting floods. Thus, for the user to be confident that he has the correct mechanistic interpretation of the process, he needs to compare a variety of experimental parameters from a suite of carefully analyzed flooding parameters from a suite of carefully analyzed flooding experiments. We are reporting the initial results of such a combined simulytion-experimental program here. Camilleri et al. 1 used a similar approach in benchmarking their micellar-polymer simulator, although they did not make as many experimental comparisons as presented here. The work presented here divides conveniently into two parts. First, results from the simulator are compared to problems in which the equations admit simple solutions - for example, single-phase "tracer" floods which have a well-known analytic solution as well as simplified chemical floods which can be modelled using the method-of-characteristics (MOC). Such comparisons test whether the simulator is performing its numerical computations correctly as well as indicating the number of grid blocks required to accurately model a physical problem. Second, we compare simulation physical problem. Second, we compare simulation results with a series of compositionally well-defined core floods. Many micellar fluid systems must be modelled using four or more pseudocomponents. We wanted fluid systems where we could model two-phase plait-point-right (type II-), two-phase plait-point-left plait-point-right (type II-), two-phase plait-point-left (type II+), and three-phase (type 111) phase behavior- where each of the floods would phase behavior- where each of the floods would remain on a single, ternary phase diagram. Furthermore, we looked for systems in which the complexities of adsorption, inaccessible pore volume, and low tension relative permeability were absent. Thus, we developed a series of analog alcohol/brine/hydrocarbon systems which gave the desired three types of phase behavior and thereby allowed us to evaluate the fluid flow and phase behavior mechanisms of the simulator unambiguously. These systems met our requirement of true ternary phase behavior and provided the advantage of very rapid equilibration and phase separation. Selection of these systems also involved handling certain questions associated with viscous fingering, i.e. they are unstable. This selection was undertaken knowingly, because another goal of our research effort is the modelling of such processes. The sequel to this paper deals with our processes. The sequel to this paper deals with our attempts to accurately model the fingering process using this same simulator. P. 233