Temperature Dependence of L1/L2/V Behavior in CO2/Hydrocarbon Systems

Abstract

Summary The temperature dependence of liquid/liquid (L1/L2) and liquid/liquid/vapor (L1/L2/V) phase behavior in CO2 /hydrocarbon systems is investigated with a combination of measurement of L1/L2 and L1/L2/V phase compositions and calculations of two- and three-phase behavior with the Peng-Robinson equation of state (PREOS). New phase-composition data are reported for mixtures of CO2 with normal hexadecane (C16) and squalane (C30) and for mixtures of CO2 with methane (C1) and hexadecane. Measurements of three-phase pressures for mixtures of CO2 with C14, C16, and C30 are also reported. Calculations of L1/L2 and L1/L2/V phase behavior with the PREOS are compared with all available experimental phase-composition data for CO2/C1/C16 and CO2/propane (C3)/C16 systems. The PREOS predicts such behavior with reasonable quantitative accuracy if binary interaction parameters are selected to match L1/L2 data. Details of the disappearance of L1/L2/V behavior with increasing temperature are investigated with the validated PREOS. For the CO2/C3/C16 system, three-phase behavior disappears at a tricritical point. For the CO2/C1/C16 system, the disappearance occurs at the upper critical endpoint (UCEP) for the CO2 /C16 binary system. Finally, qualitative similarities between CO2/crude-oil systems and the ternary systems described here are discussed. Introduction Phase behavior plays a fundamental role in determining the efficiency of EOR processes in which high-pressure CO2 displaces crude oil from reservoir rocks.1–3 The displacement process depends on the transfer of hydrocarbons from the crude oil to a CO2-rich phase. When the CO2-rich phase is dense enough that light and intermediate hydrocarbons partition into it relatively efficiently, high displacement efficiency results. Analysis of the mechanisms of CO2 flooding is complicated by the fact that mixtures of CO2 and most crude oils show complex phase behavior at temperatures near the critical temperature of CO2 (88°F [31°C]). At such temperatures, liquid/vapor (L1 /V), L1/L2/V, and L1/L2 phase behavior can occur.1–7 In addition, L1/L2/V regions on pressure-composition (p-x) phase diagrams for CO2 /crude-oil mixtures show a variety of patterns.1 Many of the crude oils from reservoirs in the Permian Basin, for example, exhibit such phase behavior. Thus, understanding L1/L2/V phase behavior and its impact on displacement efficiency is important in field application of CO2 flooding. In this paper, we use a combination of experimental observations and computations with an equation of state (EOS) for model CO2 /hydrocarbon systems to analyze phase-behavior patterns observed for more complex CO2/crude-oil mixtures. EOS's have come into general use in recent years for calculations of L/V equilibrium phase behavior. Such calculations are useful because they can be performed far more rapidly than phase-behavior experiments and because they permit, in compositional simulators, relatively straightforward calculations of the interactions of phase behavior and flow. Applications to L1/L2/V and L1/L2 systems, however, have been more limited.8–13 Here we use the PREOS14 to examine details of phase diagrams for a model system, CO2/C3/C16, that shows L/V, L1/L2/V, and L1/L2 behavior similar to that of CO2/crude-oil systems when the crude oil does not contain dissolved gas,7,15 and a second model system, CO2/C1 /C16, that shows behavior similar to CO2/crude-oil systems with dissolved gas present. Phase-Composition Measurements CO2/C16 and CO2/C30. To check the accuracy of the PREOS predictions in the L/L region, limited PVT experiments were performed for compositions and temperatures in Table 1. Saturation pressures were determined visually,16 and phase compositions and densities were estimated from material-balance calculations. Phase densities can be estimated from the total mass present in the cell and the phase volumes at a given pressure, provided that data for two different CO2 concentrations are available. Phase densities are obtained from the material-balance equationsEquations 1a and 1b Similarly, the weight fractions of CO2 in each phase can be obtained by solvingEquations 2a and 2b Results of the binary PVT experiments are shown in Figs. 1 and 2 and Tables 2 and 3. For most of the material-balance calculations, several pairings of volumetric data were possible. Phase-density and -composition values for all available pairs were calculated and averaged to give the data points shown. Horizontal error bars shown for the material-balance values represent plus or minus one standard deviation. No error bars are shown for the dewpoint compositions because the standard deviations of the L2 mole fractions were <0.005 (see Tables 2 and 3) and, hence, were too small to show on the scales of Figs. 1 and 2. Error bars shown for the visual observations of saturation pressures represent the maximum uncertainty in the saturation pressure. The values reported in Fig. 1 were obtained by extrapolating phase volumes plotted against pressure to zero volume of one of the phases. The error bars indicate pressures at which actual observations of one or two phases were made. Fig. 1 shows that phase compositions determined by material balance agree well with the phase-composition data of Stewart and Nielsen.17 Fig. 2 indicates that phase compositions for the CO2/C30 system were qualitatively consistent with values interpolated from the data of Liphard and Schneider.18 Quantitative agreement, however, was not particularly good. Because the observations reported here were obtained directly rather than by interpolation of data at other temperatures, they are more likely to be accurate. Fig. 3 reports three-phase pressures for CO2/C14,19 CO2/C16, and CO2/C30 mixtures. For each system, the three-phase pressure was determined for a series of temperatures by visual observation of the existence of three phases. Also shown is the vapor-pressure correlation reported by Newitt et al.20 As the size of the hydrocarbon molecule increases, the three-phase pressures lie closer to the vapor pressure of CO2, and the UCEP, the point at which the L2 and V phases become identical, falls closer to the CO2 critical point. The UCEP for C14 was at 99°F [37.2°C], while that of C30 was at ∼88.5°F [∼31.4°C]. The C16 value was not measured, but PREOS calculations, dashed line in Fig. 3, place it at ∼95.6°F [∼35.3°C]. For C30, the three-phase pressures actually were as much as 6 pzsia [40 kPa] higher than the vapor-pressure correlation. From the shapes of the phase diagrams, the three-phase pressure must be below the vapor pressure. Apparently, errors in the pressure measurements were large enough to give slightly higher readings. CO2/C16 and CO2/C30. To check the accuracy of the PREOS predictions in the L/L region, limited PVT experiments were performed for compositions and temperatures in Table 1. Saturation pressures were determined visually,16 and phase compositions and densities were estimated from material-balance calculations. Phase densities can be estimated from the total mass present in the cell and the phase volumes at a given pressure, provided that data for two different CO2 concentrations are available. Phase densities are obtained from the material-balance equationsEquations 1a and 1b Similarly, the weight fractions of CO2 in each phase can be obtained by solvingEquations 2a and 2b Results of the binary PVT experiments are shown in Figs. 1 and 2 and Tables 2 and 3. For most of the material-balance calculations, several pairings of volumetric data were possible. Phase-density and -composition values for all available pairs were calculated and averaged to give the data points shown. Horizontal error bars shown for the material-balance values represent plus or minus one standard deviation. No error bars are shown for the dewpoint compositions because the standard deviations of the L2 mole fractions were <0.005 (see Tables 2 and 3) and, hence, were too small to show on the scales of Figs. 1 and 2. Error bars shown for the visual observations of saturation pressures represent the maximum uncertainty in the saturation pressure. The values reported in Fig. 1 were obtained by extrapolating phase volumes plotted against pressure to zero volume of one of the phases. The error bars indicate pressures at which actual observations of one or two phases were made. Fig. 1 shows that phase compositions determined by material balance agree well with the phase-composition data of Stewart and Nielsen.17 Fig. 2 indicates that phase compositions for the CO2/C30 system were qualitatively consistent with values interpolated from the data of Liphard and Schneider.18 Quantitative agreement, however, was not particularly good. Because the observations reported here were obtained directly rather than by interpolation of data at other temperatures, they are more likely to be accurate. Fig. 3 reports three-phase pressures for CO2/C14,19 CO2/C16, and CO2/C30 mixtures. For each system, the three-phase pressure was determined for a series of temperatures by visual observation of the existence of three phases. Also shown is the vapor-pressure correlation reported by Newitt et al.20 As the size of the hydrocarbon molecule increases, the three-phase pressures lie closer to the vapor pressure of CO2, and the UCEP, the point at which the L2 and V phases become identical, falls closer to the CO2 critical point. The UCEP for C14 was at 99°F [37.2°C], while that of C30 was at ∼88.5°F [∼31.4°C]. The C16 value was not measured, but PREOS calculations, dashed line in Fig. 3, place it at ∼95.6°F [∼35.3°C]. For C30, the three-phase pressures actually were as much as 6 pzsia [40 kPa] higher than the vapor-pressure correlation. From the shapes of the phase diagrams, the three-phase pressure must be below the vapor pressure. Apparently, errors in the pressure measurements were large enough to give slightly higher readings.