A Grouping Method To Optimize Oil Description for Compositional Simulation of Gas-Injection Processes
- Ali Danesh (Heriot-Watt U.) | Dong-hai Xu (Heriot-Watt U.) | Adrian C. Todd (Heriot-Watt U.)
- Document ID
- Society of Petroleum Engineers
- SPE Reservoir Engineering
- Publication Date
- August 1992
- Document Type
- Journal Paper
- 343 - 348
- 1992. Society of Petroleum Engineers
- 5.3.2 Multiphase Flow, 5.4.2 Gas Injection Methods, 5.5 Reservoir Simulation, 5.2.1 Phase Behavior and PVT Measurements, 5.2.2 Fluid Modeling, Equations of State, 5.2 Reservoir Fluid Dynamics, 5.4.3 Gas Cycling, 5.8.8 Gas-condensate reservoirs, 4.6 Natural Gas, 5.3.1 Flow in Porous Media, 4.3.4 Scale
- 1 in the last 30 days
- 352 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||USD 5.00|
|SPE Non-Member Price:||USD 35.00|
This paper presents a grouping method that optimizes compositionaldescription so that the modeling of the progressive phase behavior of injectiongas and reservoir oil systems requires phase behavior of injection gas andreservoir oil systems requires less computation. The method allows theretrieval of the full composition to select new groups for further calculationswith very little effort and almost no extra computational time. A new mixingrule, based on the parameters of an equation of state (EOS), also has beendeveloped and incorporated into the grouping method. Multistage experimentssimulating the progressive compositional variations of injection gases andreservoir oils were used to evaluate the proposed grouping method.
In gas-injection processes, a significant mass transfer can occur betweenthe injected gas and the reservoir oil. The resulting compositional variationsmay lead to miscibility of the two phases, giving a highly efficientdisplacement. Compositional reservoir models that use cubic EOS to model thephase behavior of fluids are essential for reliable predictions. Naturallyoccurring reservoir hydrocarbons are composed of thousands of constituents, andthe concept of grouping has long been used in fluid description andcompositional calculations. The most conventional method is to describe oilsystems with discrete components up to C6 and to lump the heavy fractions asC7+. However, this is not the most efficient method to the oil in compositionalsimulation studies where it is desirable to minimize the number of componentsdescribing the oil because the computational requirements increase sharply withthe number of components. The required fluid description should vary with thecomplexity of the process to be modeled. For example, the compositionalbehavior of a reservoir oil under pressure depletion only may be modeled byonly two components. Numerous methods have been proposed to group the oilcomponents to reduce the computational time of phase-behavior calculations. Inalmost all the proposed methods, the properties of the groups remain unchangedand equal to the original reservoir oil values during the entire process. Themethods generally are tested for cases where the original oil compositions donot change significantly, such as for predicting the bubblepoint pressure andsingle-flash properties. In gas-injection processes where a significant masstransfer occurs between the injection gas and the reservoir oil, the use of thegroup's properties generated from the original oil composition may beinsufficient for accurate prediction of the phase behavior. prediction of thephase behavior. The main problems with the component groupings are (1) thenumber of groups required and the distribution of components within thesegroups, (2) the mixing rule applied to calculate the group parameters requiredin the EOS modeling, and (3) the retrieval of parameters required in the EOSmodeling, and (3) the retrieval of full description in terms of originalcomponents when needed. The first two problems have been studied extensively bymany investigators, while the third has received little attention. This paperaddresses all these aspects of grouping. The optimum number of groups wereinvestigated for a reliable prediction of various fluid properties. It wasdemonstrated that prediction of various fluid properties. It was demonstratedthat the number of components describing an oil can be reduced significantlywithout impairing the predicted phase-behavior results by an EOS. A newgrouping method, along with new mixing rules to determine the properties of thegroups for EOS calculations, was developed. A proposed method to retrieve fullcompositions from the groups accounts for compositional variations within eachgroup. The proposed methods were compared with other methods and evaluatedagainst experimental data generated at conditions simulating gas-injectionprocesses.
Experimental Apparatus and Procedure
The basic experimental data were obtained by studying the vapor/ liquidequilibria (VLE) in two high-pressure cells housed side by side on a rotatingplate inside an oven. Any fluid phase could be transferred from one cell to theother through a measurement loop for determination of density and composition.The density was measured by a calibrated oscillating densitometer. Thecomposition was measured by a novel high-pressure direct-sampling system with agas chromatograph. The VLE tests were designed to simulate compositionalvariations of a reservoir oil occurring during a gas-injection process.Multiple forward and backward contacts, which simulated process. Multipleforward and backward contacts, which simulated the conditions at the front andthe tail of the injection gas, respectively, as well as conventionalsingle-contact gas-injection and equilibrium flash tests were conducted. Themultiple-contact experiments provided data on the two extreme limits ofcompositional changes occurring in gm-injection processes. All tests wereconducted at 100C. Further information on the equipment and procedures aregiven in Ref. 3. Model fluids were prepared by adding known quantities of purecompounds. The methane purity was better than 99.99%. The purity of the othercompounds was above 99 %, except for nC 18, which was 97% pure. The presence ofminor impurities could markedly affect the predicted results for fluids nearthe critical point. The effect of impurities in the tested fluids on theresults and conclusions of this study, however, are considered to be minimal.Table 1 gives the compositions of the tested fluids. The use of these modelfluids eliminated the ambiguity regarding the properties of compounds ofphase-behavior modeling, often a properties of compounds of phase-behaviormodeling, often a difficulty with real oils.
A number of authors have given recommendations on selecting the number ofpseudocomponents (groups). In general, 4 to 10 pseudocomponents have beenconsidered adequate for simulation purposes. pseudocomponents have beenconsidered adequate for simulation purposes. The methods that involve majordata manipulation, such as matching results predicted by groups and fullcompositions, are not suitable for conditions where frequent regrouping isrequired. The simpler methods mainly group the compounds on the basis of theirvolatilities or concentrations in the original oil. Li et al. proposed groupingcomponents on the basis of their volatility by use of equilibrium ratiosobtained by flashing the fluid at the reservoir temperature and the averageoperating pressure. Pedersen et al. suggested grouping the components onpressure. Pedersen et al. suggested grouping the components on the basis ofweight with each group containing approximately the same weight fraction.Cotterman and Prausnitz used the criterion of approximately equal mole fractionto select groups. A more representative fluid description can be expected whenthe groups are formed by due consideration of the volatility and concentrationof compounds in the oil. The volatility of compounds in a high-pressurereservoir oil at any temperature and pressure depends on their properties aswell as the mixture composition. Fig. 1 shows the equilibrium ratios for aseries of forward contacts.
|File Size||737 KB||Number of Pages||6|