Interpretation of Laboratory Gasfloods With Multidimensional Compositional Modeling

Publisher Website
Google Scholar
Abstract
Summary Corefloods were conducted and simulated to evaluate oil recovery from field cores by injection of hydrocarbon solvents. Heterogeneity and fingering were evaluated separately by conducting computed-tomography (CT) -aided single-phase stable and unstable floods. In the absence of bypassing, oil displacement by an equimolar mixture of C2, propane, and butane (C234) is first-contact miscible and C2 (C2) flood is multicontact miscible. In corefloods, heterogeneity and fingering cause oil bypassing. Bypassed oil remains miscible with injectant C234 but becomes immiscible with injectant C2. Two heaviest pseudocomponents drop out in the bypassed region to form a residual oil saturation (ROS) of ˜25%. Introduction Selection of a hydrocarbon solvent for oil recovery often involves corefloods at the laboratory scale. In our laboratory, we are evaluating several solvents for recovery of a moderately viscous oil. Several hydrocarbon corefloods have been conducted in field cores. They are conducted vertically with the solvent injected from the top to avoid gravity override, but the displacement is far from 1D. These corefloods incorporate a complex interaction of phase behavior, heterogeneity, fingering, and multiphase flow. The objective of this work is to simulate these corefloods by a compositional simulator and to study the interaction between bypassing and phase behavior. Bypassing caused by fingering and heterogeneity in miscible floods has been modeled empirically, statistically, or numerically. Koval,1 Todd and Longstaff,2 and Fayers and Newley3 modeled unstable fluid propagation empirically with one or several mixing parameters. This approach is simple to use in noncompositional models but is neither mechanistic nor easy to use in fully compositional models. Young4 suggested modifying dispersion coefficients to account for bypassing. Kremsec and Sebastian5 used numerical dispersion in 1D simulations to account for fingering and dispersion. However, the assumption of complete mixing can be erroneous and its effect on phase behavior is significant.6,7 A dual-zone mixing procedure proposed by Nghiem et al.8 and Fayers et al.7 accounts for bypassing more mechanistically than complete mixing models. This kind of modeling is needed at field scales because fine-grid compositional simulation with details of every finger may not be feasible with existing computers. Araktingi and Orr9 and King10 proposed probabilistic ways to model fingering. These methods use statistical approximation of the Laplace operator and introduce perturbations into the system at every timestep. They have not been generalized to fully compositional models. Peaceman and Rachford,11 Hatziavramidis,12 Giordano et al.,13 and Christie and Bond14 used direct numerical methods for single-phase, two-component flow. If grids are fine enough and numerical dispersion is small enough, then these simulations can match laboratory fingering experiments. Gardner and Ypma15 used such an approach to understand the mechanisms of a CO2 flood. The objective of this work is not just to match coreflood data but to understand the compositional effects. Thus, we have chosen a multidimensional compositional simulator to model explicitly the bypassing of oil and its interaction with phase behavior. Fine-grid simulation has been used in the past9 to match 2D laboratory miscible fingering tests successfully.13 But corefloods (including the experiments described here) are carried out in cylindrical cores that are not 2D. Many cores are layered parallel to the main flow direction. If the permeability variation is significant, it can affect coreflood behavior. Heterogeneity and fingering can be accounted for explicitly in a 3D grid, but doing so is computationally expensive. It is desirable to have a 2D method that models the essential mechanisms. Gardner and Ypma15 used a special perturbation scheme to match fingering in a cylindrical core by 2D modeling. Both the 3D and 2D approximations are evaluated later. For multidimensional compositional simulation, cores must be characterized multidimensionally and fluids must be characterized in terms of an equation of state (EOS). Before the advent of CT scanning, there was no way to characterize the permeability or porosity variation inside a core nondestructively. This study uses CT scanning19 to characterize heterogeneity and fingering behavior of cores. The following procedure is developed to estimate all the input parameters for the hydrocarbon flood simulations. Single-contact, multicontact, and slim-tube experiments reported by DeRuiter et al.20 were modeled to develop an EOS characterization of fluids used in this study. Two single-phase miscible floods, one at unit mobility ratio and the second at an unfavorable mobility ratio, are conducted in a core, and in-situ concentration of the injectant is monitored by a CT scanner. The heterogeneity and dispersivity of the cores are evaluated from the CT scan concentration maps of the unit mobility ratio floods. The transverse dispersion, boundary conditions, and gridding are adjusted so that the fingering behavior of the model matches that of the unfavorable-mobility-ratio experiment. The relative permeability between the dense hydrocarbon phases can be determined directly from flow of equilibrated fluids. The phase behavior, core description, and relative permeability then can be incorporated to simulate multicontact gasflood experiments. In this paper, we describe core characterization, hydrocarbon gasflood experiments, and simulations of characterization floods and gasfloods. We do not discuss the development of fluid...