Vous êtes sur la page 1sur 10

IBP1352_12 THE INFLUENCE OF LUMPING ON THE BEHAVIOR OF RESERVOIR WITH LIGHT OIL AND CO2 Helena F. A.

Scanavini1, Denis J. Schiozer2

Copyright 2012, Brazilian Petroleum, Gas and Biofuels Institute - IBP


This Technical Paper was prepared for presentation at the Rio Oi & Gas Expo and Conference 2012, held between September, 1720, 2012, in Rio de Janeiro. This Technical Paper was selected for presentation by the Technical Committee of the event according to the information contained in the final paper submitted by the author(s). The organizers are not supposed to translate or correct the submitted papers. The material as it is presented, does not necessarily represent Brazilian Petroleum, Gas and Biofuels Institute opinion, or that of its Members or Representatives. Authors consent to the publication of this Technical Paper in the Rio Oil & Gas Expo and Conference 2012 Proceedings.

Abstract
Compositional simulation demands a large number of equations and functions to be solved, once fluid properties depend on reservoir pressure and temperature and also on fluid composition. As a consequence, the number of components used influences considerably in the simulation run time and accuracy: more components yield more equations to be solved with expected higher run time. Giant petroleum fields discovered recently in Brazil (pre-salt reservoirs) demand compositional simulation due to the fluid characteristics (light oil with the presence of CO2). However, the computational time can be a limitation because of the number of grid blocks that are necessary to represent the reservoir. So, reducing the number of components is an important step for the simulation models. Under this context, this paper presents a study on the influence of different lumping clusters, used to reduce the number of components in a volatile oil, on reservoir simulation. Phase diagram, saturation pressure and simulation results were used for comparison purposes. The best results were obtained for the cases with 14, 9 and 7 pseudocomponents, which represented correctly the original fluid, reducing till three times the simulation run time, for the same production volumes of oil and gas.

1. Introduction
In compositional models, the fluid properties do not depend only on the reservoir temperature and pressure (T R e PR, respectively), but also in the fluid variable composition during production. Consequently, the number of unknown functions is higher than in a Black Oil model, once, beyond the composition of each phase, it is necessary to calculate the pressure and saturation at each step (Liu, 2001). Due to these unknown functions, the number of interactions to compute all of them increases. So reducing the number of components becomes a prerequisite for the compositional simulation by decreasing the simulation run time. This time reduction was reported by different authors, such as Liu (2001), Hildebrand et al. (2001) and Li and Johns (2006). Liu (2001) reported that a significant speedup in compositional reservoir simulation can be achieved by using the minimum number of components, while the reservoir production performance remains the same as that calculated using many more components. Using the minimum number of components will also reduce the storage and speed requirements of computers, which enables engineers to perform more detailed and larger scale compositional simulation. This reduction in the number of components is known as lumping or pseudoization and, according to Pedersen and Christensen (2007), it consists on deciding which hydrocarbon fractions will be grouped in the same pseudocomponent, being that the pseudocomponent properties will be calculated based on the component properties of the original fraction. However, pseudoization must be done carefully, because the component lumping can have an impact in the simulation speed and also in the accuracy of the results. It must be done in order to reproduce the results obtained with the complete original characterization that was used to adjust the PVT experimental data (Whitson and Brul, 2000). According to Liu (1999) the best lumping scheme is one of the lumped systems which predicts the closest behavior to the prediction from the original system; the same author implemented the Phase Diagram method for selecting the best-lumped system among all possible lumped schemes. But Liu (2001) defined the Best Lumping

______________________________ 1 Ph.D, UNISIM / CEPETRO / UNICAMP 2 Ph.D, Professor, DEP/ FEM / UNICAMP

Rio Oil & Gas Expo and Conference 2012 Scheme also as one of the lumped systems that can predict the available PVT lab tests with the most accuracy. Since an equation of state (EOS) of the original fluid system should be tuned to match relevant lab data available, the above new definition does not conflicts with the original definition given before. In the literature it can be found some lumping suggestions. Some authors (Whitson and Brul, 2000; Pedersen and Christensen, 2007) suggest N2 to be grouped with methane, CO2 with ethane, iso-butane with n-butane, and isopentane with n-pentane. But it should be emphasized that non-hydrocarbon compounds should not be lumped with other components if they will be used as injection gas, such as CO2 (Li et al., 1985). For heavier compounds (from C7 on), the grouping suggestion gave by Whitson and Brul (2000) is: C7 C10, C11 C15, C16 C20 and C21+. According to these authors, lumping is better than splitting only the first hydrocarbons of the heavy fraction and grouping the others, as for example, C7, C8, C9 e C10+. Shealy (1989) recommends the minimum lumping for the more concentrated components and the maximum for the more diluted. The calculations of the saturation pressures obtained before and after lumping and also the PT envelopes consist of a quick way to analyze the sensitivity of the phase behavior for different lumping schemes. One can investigate the possible lumping effect on the phase equilibrium and fluid behavior in a wide range of temperature, pressure and composition, making sure the grouping is adequate or not (Li et al., 1985; Leibovici, 1993, Whitson and Brul, 2000). This paper presents an investigation on how the lumping influences the fluid properties and also the simulation results when trying to match it to PVT experimental data. Different lumping clusters on a volatile oil were tested to tune the equation of state to simulate the reservoir production. Phase diagram, saturation pressure and simulation results were used for comparison purposes.

2. Methodology
In this study, 9 different forms of lumping were compared for a volatile fluid found in the literature. After the lumping and EOS tuning, each case was submitted through compositional simulation and the results were compared. For this study, it was considered a fluid from Danesh (1998) which is a volatile oil, with 36API and 1.49% (molar) of CO2. The composition is shown in Table 1 and the PVT data can be found at the literature. The reservoir is under an initial pressure of 5712 psig and 195F. Table 1 Oil Composition Component CO2 N2 C1 C2 Molar % Component 1.49 C3 0.90 iC4 51.54 nC4 6.57 iC5 Molar % Component 4.83 nC5 0.68 C6 2.39 C7 0.91 C8 Molar % Component 1.47 C9 2.17 C10 4.30 C11 3.96 C12+ Molar % 1.93 1.66 1.38 13.82

Different lumping schemes were evaluated, as shown in Table 2. The EOS tuning was performed for the Peng Robinson equation, using a commercial software for fluid adjustment. All the component properties were obtained from the software database. After the tuning, deviations of PVT properties and saturation pressure were calculated. It was also considered the P-T diagrams for comparison purposes. The simulations were performed for all lumping cases, using a commercial compositional simulator. The reservoir model is based on a GEM template that was modified by Ligero (2011), to be homogeneous in porosity (30%) and permeability (Kx = Ky = 500 and Kz = 5mD), with connate water saturation constant at 30%. The block size is 110x110x8m. There are two wells: a producer (operating with a target oil production rate of 400 bbl / day and minimum BHP of 5000 psi) and a water injector (operating with a target water rate of 400 bbl/day and maximum BHP of 6000 psi). The model was run for 18 years. In order to confirm the lumping results obtained for the homogenous reservoir with water injection, tests were run under other conditions, such as: a) Water injection in a heterogeneous reservoir; b) CO2 injection in homogeneous and heterogeneous reservoirs. The heterogeneous model is based on the SPE 10 model (Christie and Blunt, 2001). The block size was maintained equal to the homogeneous reservoir: 110 x 110 x 8. There are also two wells, a producer (same operating conditions as in the homogeneous reservoir) and an injector (water or CO 2, depending on the case). In the case of the CO2 injector, the operating conditions were maximum BHP of 8500 psi and target gas rate of 1x10 6 bbl/day. The CO2 was obtained from an external source. In the simulations with water injection, the model was run till the condition of maximum water cut (0.9) was reached, while with CO 2 injection, till GOR reached 10000 ft3/bbl. 2

Rio Oil & Gas Expo and Conference 2012

3. Results and Discussion


3.1. Lumping Tests First, in order to control the lumping and tuning processes quality, EOS was fitted through the previous methodologies. Each property from the differential liberation and separator data available was regressed to the smallest deviation possible from least squares. Pedersen and Christensen (2007) recommends a quality control checking, which determines that if the deviation between the experimental and the adjusted data is below 10%, then the adjustment can proceed. Analyzing Table 3, it is possible to observe that all the absolute deviations from regression are below 10%, so the properties were adjusted for the EOS fitting. Another important property calculated is the saturation pressure. The values calculated show low deviations from the experimental value (4580.7 psia), as shown in Table 3. Once the properties are adjusted, the phase envelope, showed in Figure 1, can be elaborated. It can be seen that the envelopes overlap at the reservoir temperature (T res=195F), what proves that all of the lumped systems predicted the same fluid properties which are the same as those from the original system (25) in the common single phase region. The same Psat value for all cases proves this envelope analysis that, at the reservoir conditions of temperature and pressure (5712 psig), the fluid in all cases will be liquid (oil). Table 2 Lumping Scheme
25 CO2 N2 C1 C2 C3 IC4 NC4 IC5 NC5 C6 C7 C8 C9 C10 C11 C12 C13 C14 C15 C16 C17 C18 C19 C20 C21+ 14 CO2 N2 C1 C2 C3 IC4 NC4 IC5 NC5 C6 C7 to C10 C11 to C15 C16 to C20 C21+ 9 CO2 N2 to C1 C2 C3 to NC4 IC5 to C6 C7 to C10 C11 to C15 C16 to C20 C21+ 7A CO2 N2 to C2 C3 to C6 C7 to C10 C11 to C15 C16 to C20 C21+ 7B CO2 N2 to C1 C2 to NC4 IC5 to C6 C7 to C12 C13 to C19 C20+ 7C CO2 N2 to C1 C2 C3 IC4 to NC4 IC5 to NC5 C6+ 6 CO2 N2 to C2 C3 to C6 C7 to C10 C11 to C20 C21+ 5 CO2 N2 to C2 C3 to C6 C7 to C15 C16+ 4 CO2 N2 to NC5 C6 to C15 C16+

However, it is important to notice that there is a difference between the case 7C, 4 and all the other lumping choices. For 7C, Psat shows a little higher deviation (0.30%) and the envelope shrinks as the temperature increases. It can be a problem if the reservoir temperature varies during production; then the well would start to produce gas sooner than in the other cases. For case 4, the P sat deviation was very low (0.11%), but the envelope is wider than the other. In this case, the differences could be appearing due to the reduced number of components. But as in this study the reservoir operates at constant temperature, these cases were still considered for simulation, for comparison purposes. The simulations results are shown in Figures 2 to 4. In general a good agreement can be observed among the lumped cases. But it is possible to notice that some curves are closer to case 25 than others (14, 9 and 7 B are closer than 7C, 6 and 5), as in Figure 2 that shows the oil production. Table 4 shows the deviations of oil and gas production (maximum of 2.04% for case 5 and 5.72% for case 7C, respectively) and proves the small differences observed in the graphics. The drop in the oil rate observed in the beginning of the production in Figure2b is caused by the decrease in the reservoir pressure (Figure 3a) the same sequence in the pressure decrease is observed in the rate. Figure 3a shows the pressure variation along the years; it was always kept over the saturation pressure (4580.7 psia), maintaining the GOR constant (at around 1526, in Table 4) guaranteeing the gas production only at surface conditions.

Rio Oil & Gas Expo and Conference 2012

Figure 1 P-T Phase Diagrams. Table 3 Properties and Deviations Case 25 14 9 7A 7B 7C 6 5 4 SG 2.29 2.36 2.38 2.80 1.86 1.19 2.82 2.14 0.07 Absolute Deviation (%) Bo cg 4.28 1.91 6.24 4.36 1.78 6.12 4.28 1.95 6.62 4.78 3.07 6.70 3.36 2.37 6.00 2.14 0.98 6.31 4.85 2.98 6.75 4.22 2.48 6.47 2.61 8.44 4.60 GORseparator 2.19 2.02 2.37 1.71 4.13 1.91 1.75 1.65 6.90 Value (psia) 4570.1 4570.1 4570.0 4570.0 4572.7 4566.8 4568.5 4567.5 4575.5 Psat Deviation (%) 0.23 0.23 0.23 0.23 0.17 0.30 0.27 0.29 0.11

Observing Table 4, the difference in the results mostly for the case 7C, but also for 7A, 6 and 4 is very clear. Cases 7A and 6 gave the higher GOR values (1579.4 and 1580, respectively), while for the others cases cited (7C and 4), the lowest values were obtained: 1469.6 and 1481.5. High values of GOR normally indicate that the reservoir is depleted (what is not the case) or a higher concentration of volatile compounds in the liquid mixture in the reservoir (Thomas et al., 2001). This last justification can be true. The analysis of Figure 3b and Table 4 prove it: cases 7A and 6 produce more gas than all while case 7C produces less. It could be related to the envelope, as the cases 7A and 6 envelopes are the larger ones, indicating that, at a higher temperature, it would start to produce gas sooner than others.

Figure 2 In the producer: (a) Cumulative Oil; (b) Oil Rate in the producer.

Rio Oil & Gas Expo and Conference 2012 Table 4 Oil and Gas Cumulative Production and Deviations Case 25 14 9 7A 7B 7C 6 5 4 GOR 1528.3 1514.0 1519.3 1579.4 1503.6 1469.6 1580.0 1561.4 1481.5 Oil Production (106 bbl) 1.178 1.185 1.183 1.156 1.189 1.155 1.156 1.154 1.19 Deviation 0.00% -0.59% -0.42% 1.87% -0.93% 1.95% 1.87% 2.04% -1.02% Gas Production (109 ft3) 1.801 1.794 1.797 1.826 1.787 1.698 1.826 1.803 1.764 Deviation 0.00% 0.39% 0.22% -1.39% 0.78% 5.72% -1.39% -0.11% 2.05%

Figure 3 (a) Average Reservoir Pressure; (b) Cumulative Gas Produced. For case 4, as the reduced number of components considered might have interfered in the simulation not representing the fluid correctly. While a plausible explanation for these differences in case 7C could be the heavy components lumping from C6 to C21+. This grouping caused the fluid to become heavier, affecting the envelope, P sat and the critical point, and the fluids production as well. Figure 4 shows that the oil produced in case 7C, for example, is heavy, containing a higher amount of C6+ than other components, resulting in a lower GOR and less gas produced.

Figure 4 Oil Production by pseudocomponent, in mass. The water cut in the production well is shown in Figure 5. It can be observed that the water starts to be produced after around 12 years of oil extraction and its increase coincides with the decrease in the oil rate (Figure 2b). 5

Rio Oil & Gas Expo and Conference 2012 As the production of oil reduces, the water starts to flow faster. In terms of pseudocomponents, it does not seem to be affect by the lumping clusters considered in the simulations. Another parameter analyzed was the simulation time (CPU time), shown in Figure 6. A clear tendency of decrease in the simulation run time can be observed with the reduction of the number of components: while for case 25, it took 181.9 s, a simulation with 7 pseudocomponents (case 7B, for example) took only a third of the time (64.9 s). With four pseudocomponents, a decrease of 72% in the simulation time was achieved. This reduction was expected once with fewer components, the number of equations and functions systems to be solved is smaller, speeding up the simulation.

Figure 5 Water Cut during production.

Figure 6 CPU time. 3.2. Different reservoir conditions In these next simulations results, only the cases 25, 7A, 7C, 6, 5 and 4 were considered for comparison purposes. a) Water injection in a heterogeneous reservoir Observing the results presented in Figure 7, the curves from the heterogeneous reservoir are more close to each other than the ones from the homogenous case (showed previously), in almost every property analyzed. Although in the homogenous cases, the distance among the curves is not very big. But in both cases, all the lumping cases described well the fluid behavior. b) CO2 injection in homogeneous and heterogeneous reservoirs Unlikely in the water injection case, under CO2 injection, a difference can be noticed for the results obtained in cases 7C and 4, compared to the others simulations, especially more close to the end of the simulations. In Figures 8a and 10a, which show the cumulative oil and gas productions, the differences are not as evident as they are in the oil rate (Figure 8b) and gas rate (Figure 10b) graphics, for example. 6

Rio Oil & Gas Expo and Conference 2012 These two cases were the ones with the most poor EOS adjustments and distant envelope in the modeling step. So the lumping choice might be the cause for this behavior under CO 2 injection. In the case of 7C, the lumping of the heavy fraction till the C6+ was not a good choice; the proportion of light and heavy components is not well balanced, as all the heavy ones are together in only one group, while the lighters correspond to the other 5. And it is not what happened in the other cases with 7 pseudocomponents (7A and 7B), where the pseudoizations are more equalized, given results closer to the more complete case (25). And in the case of 4, the small number of components used should be what interfered in the fluid behavior representation, and, therefore, in the simulation results. Since the EOS adjustments and envelope analysis, it could be seen that something might go wrong with these two cases, but we decided to study them anyway, in order to evaluate their behavior. And by these results, it can be seen that they are not the best lumping choices. The simulation program first maintains the oil rate equals to 400 bbl/day for a while, in the beginning, as shown in Figure 8b, while the BHP (Figure 9b) decreases. When the simulation cannot maintain the oil rate at 400 any longer, this rate started to fall too, and at this moment the BHP stabilized, as the average pressure (Figure 9a), staying always above 5000 psi (superior to the Psat of the fluid) till the end of the process in this case, the oil rate varies with time. Figure 10b shows a large variation in the gas rate during the simulation. And an increase in the rate can be noticed around the year 2010, in the heterogeneous case, for example. But in Figure 11, an increase in the GOR can also be observed in the same period. A larger value for GOR indicates a higher production of gas and a smaller production of oil. This can be confirmed through the graphics, once there is a reduction in the growth of oil production (Figure 8 the curve of heterogeneous presents a less accentuated growth from 2010); while the increase in the GOR, associated with a smaller increase in the reservoir pressure, indicate the presence of free gas in the reservoir, i.e, that the injection gas CO2 has reached the production well, what can be seen in Figure 12.

Figure 7 - (a) Cumulative Oil; (b) Oil rate in the producer.

Figure 8 In the producer: (a) Cumulative Oil; (b) Oil Rate in the Homogeneous and Heterogeneous Reservoirs with CO2 injection. 7

Rio Oil & Gas Expo and Conference 2012

Figure 9 Pressures: (a) Average Pressure in the reservoir; (b) Bottom Hole Pressure in the Production Well.

Figure 10 (a) Cumulative Gas; (b) Gas Rate in the producer in the Homogeneous and Heterogeneous Reservoirs with CO2 injection.

Figure 11 Gas Oil Ratio in the Homogeneous and Heterogeneous Reservoirs with CO2 injection.

Rio Oil & Gas Expo and Conference 2012

Figure 12 Global Mole Fraction Map of Case 7B. CPU Time Another parameter compared for all cases was the CPU time, presented in Figure 13. As it was observed for the homogeneous reservoir with water injection (Figure 6), in these last cases simulated, the use of a reduced number of components does guarantee a quicker simulation. The reduction can be of 5 times (around 81%) in the case of a CO2 injection in a heterogeneous reservoir, when using 5 pseudocomponents (45.8 s), instead of 25 (237.9 s), without losing simulation performance. These were the expected results, once, as reported in the literature (Liu, 2001; Hildebrand et al., 2001; Li and Johns, 2006), it was expected that a reduction in the number of fluid components would lead to a decrease in the simulation time.

Figure 13 CPU time obtained for simulations with water or CO2 injection.

7. Conclusion
Despite the fact that the saturation pressure presented a low deviation in all cases, the phase diagrams indicated that there was something different in cases 7C and 4, predicting that maybe the simulation results would not be the same. After the analysis of the simulation results, it can be affirmed that most of the cases used lumping clusters more adequate than cases 7C and 4, as the results were closer to the more complete case (25), reproducing correctly the fluid behavior observed previously, confirming the envelope indication. The simulations with CO 2 injection pointed out these differences among the lumping cases. Comparing cases 7A, 7B and 7C proves that the choice of which components should be lumped together is very important, as it can influence direct in the results. Case 7C, for example, is not adequate for this fluid. An important decrease in the simulation run time, of five times, was obtained by reducing the number of components to five, maintaining the reservoir production performance. 9

Rio Oil & Gas Expo and Conference 2012 As a suggestion, among the lumping choices tested, the case 7B can be used with reasonable number of components, with a significant reduction in the simulation time, without loss of confidence in the results obtained. For future works, it has to be study if the EOS adjusted for the initial fluid is still valid for the fluid after the CO2 has been injected and altered its initial concentration; another point is if the fluid lumping and EOS adjustment should be tested first in simulations involving CO2 injection and then water injection, as the results in this article showed that the solvent injection evidences the differences among the lumping schemes.

7. Acknowledgements
The authors would like to thank Petrobras, CEPETRO / UNICAMP and CMG for their financial and technical support.

8. References
CHRISTIE, M. A., BLUNT, M. J. Tenth SPE Comparative Solution Project: A Comparison of Upscaling Techniques. (SPE 66599). In: SPE Simulation Symposium, Houston, Texas, USA, 2001. DANESH, A. PVT and Phase Behavior of Petroleum Reservoir Fluids. Elsevier, 1998. HILDEBRAND, M.A., LIU, K., MEZZATESTA, A.G. A Study of the Influence of Number of Components on Compositional Reservoir Simulation Results. (SPE 69575). In: SPE Latin American and Caribbean Petroleum Engineering Conference, Buenos Aires, Argentina, 2001. LEIBOVICI, C.F. A Consistent Procedure for the Estimation of Properties Associated to Lumped Systems. Fluid Phase Equilibria, v. 87, p. 189 197, 1993. LIGERO, E. L. Modelo de Simulao Comunicao Interna, Abril, 2011. LIU, K. Fully Automatic Procedure for Efficient Reservoir Fluid Characterization. (SPE 56744). In: SPE Annual Conference and Exhibition, Houston, Texas, 1999. LIU, K. Reduce the Number of Components for Compositional Reservoir Simulation. (SPE 66363). In: SPE Reservoir Simulation Symposium, Houston, Texas, US, 2001. LI,Y., NGHEIM, L.X., SIU, A. Phase Behavior Computations for Reservoir Fluids: Effect of Pseudo-components on Phase Diagrams and Simulation Results. (SPE 85-06-02). Journal of Canadian Petroleum Technology, v. 24. n. 6, p. 29 - 36, 1985. LI, Y., JOHNS, R.T. Rapid Flash Calculations for Compositional Simulation. (SPE 95732). SPE Reservoir Evaluation & Engineering, v. 9, n.5, p. 521 529, 2006. PEDERSEN, K.S., CHRISTENSEN, P.L. Phase Behavior of Petroleum Reservoir Fluids. CRC Press, 2007. SHEALY, G.S. Equation-of-State Characterization of C7+ Fractions without Using Critical Properties or Complex Regression. In: CHORN, L.G.; MANSOORI, G.A. (Org.). C7+ Fraction Characterization. New York: Taylor & Francis, 1989. p. 197 219. THOMAS. J.E. et alli. Fundamentos de Engenharia de Petrleo. Ed. Intercincia, Rio de Janeiro, 2001. WHITSON, C. H., BRUL, M.R. Phase Behavior. Richardson, TX: SPE, 2000.

10

Vous aimerez peut-être aussi