You are on page 1of 11
Single-Well Performance Predictions for Gas Condensate Reservoirs D.D. Fussell, SPE-AIME, Amoco Production Co, Introduction Production data for some gas condensate producing wells have shown that the productivity is severely curtailed when the flowing bottom-hole pressure is less than the saturation pressure of the in-place fluid. Itis generally accepted that this reduction is due to accumulation of condensed liquid near the wellbore and, for low producing rates, accumulation of con- densate within the tubing string or the annulus, or oth, depending on the type of well completion. juid condensate accumulation, hereafter referred to as condensate accumulation, is defined as the con- densed hydrocarbon liquid saturation within the for- mation in excess of the saturation given by the experimental volumetric depletion data. Kniazeff and Naville' and Eilerts ef al each de- veloped models for predicting the saturations and pressures in the vicinity of a gas condensate pro- ducing well. The method used by Eilerts ef al. requires data for the fluid system obtained from lab- oratory flow experiments in the porous media of interest. Both studies predicted high condensate satu- rations in the region of the well. ‘O'Dell and Miller’ presented a simple method based on steady-state flow concepts that can be used to estimate quickly the deliverability from the well. Results obtained using this method indicate that pre- dictions of producing rates will be pessimistic if the average reservoir pressure is below the saturation pressure of the in-place fluid, Roebuck er al. developed the first models that consider the flow of individual components and 860 account for component mass transfer between phases as dictated by phase equilibria data, One of these models! was used to predict the performance of @ producing well in a reservoir containing a rich gas ‘condensate. The duration of the prediction was very short (0.25 day), so no definite conclusions could be inferred from the study. The results did show that condensate accumulated very rapidly adjacent to the well as soon as the flowing bottom-hole pressure fell, below the saturation pressure. ‘The purpose of this paper is to describe the use of a modified version of the one-dimensional (1-D) radial model developed by Roebuck et al. to study long-term single-well performance in three gas con- densate reservoirs. This work was done to (1) show the effect of condensate accumulation on well pro- ductivity, (2) evaluate the applicability of a steady- state method similar to that presented by O'Dell and Miller, and (3) evaluate the effect of phase equilibria, data and relative permeability characteristics on pre- dicted performance. Description of Model Roebuck et al.‘ give detailed descriptions of 1-D linear, 1-D radial, and 2-D compositional models. Briefly, the models predict the flow of each hydro- carbon’ component® and the water phase as a func- tion of time and space. The components are allowed A hydrocarbon component oF component as used herein in Gudes" the. nonhyaracaryon ‘components, mtrogen carbon ‘vise ana hysrogen suite. Gas condensate wells, producing with a flowing bottom-hole pressure below the saturation pressure, suffer a more rapid decline in productivity than that predicted by the theory for dry gas wells. A compositional radial reservoir model has been developed and used to predict the productivity loss of three gas condensate wells. JOURNAL OF PETROLEUM TECHNOLOGY to distribute themselves between the hydrocarbon phases as dictated by the K-values for the fluid sys- tem. The phase equilibria data (K-values, phase densities, and viscosities) used by the models are dependent upon pressure and composition. ‘The 1-D radial compositional model used in this study was a modified version of the model” described by Roebuck et al. It was found that large material- balance errors occurred with their model when using unequal cell spacings to predict single-well perform- ance. ‘The finite-difference approximations to the partial-differential equations that describe individual component flow as proposed by Roebuck ef al. were modified to be applicable to unequal cell spacings The modified finite-difference equations are derived in Appendix A. ‘A schematic of the radial geometry is shown in Fig. 1. The derivation of the difference approxima- tions uses pressures, phase saturations, fluid compo- sitions, and fluid properties evaluated midway be- tween cell boundaries. Eq. A-5 of Appendix A shows that the flow of each component (and water) is de- pendent upon the radius terms r,s, and r,-»,. These radius terms account for the variable cross-sectional area with increasing distance from the well and are evaluated at the cell boundaries. These difference approximations are consistent with the derivation of the partial-differential equations that describe flow within a radial system. The fnite-difference approximations for the bound- ary conditions (Appendix A) are also consistent with the finite-dfference approximations of the partial differential equations ‘Three criteria were used to evaluate the appl bility of this model for performance predictions of 2 single gas condensate well, First, experience with compositional models has shown that their perform- ance is sensitive to the individual component material balance error. This model independently calculates the original mass, current mass, produced mass, and injected mass of each component. Using the differ- ence approximations derived in Appendix A, the ‘material balance errors were much less than 1 percent. Second, the permeability-height, kh, product used Fig. 1—Radial geometry. JULY, 1973, by the model was compared with the analytically de- termined kh product” based on predicted flowing bottom-hole pressures vs time. These comparisons were made for a single well within a homogeneous dry-gas reservoir during pressure drawdown and pressure buildup and for a two-rate test. The actual kh products and the analytically determined kh prod- ucts were in excellent agreement. The section entitled “Predicted Performance Results” will present com- parisons of actual and analytical kt products for gas condensate systems. Third, the transformation presented by Bruce et al.* was used to convert the radial flow equations for 2 gas condensate fluid to equations of linear form, ‘The performance of a single gas condensate pro- ducing well was predicted with a 1-D linear? and the 1-D radial compositional models. The predicted performances were in excellent agreement, which indicates a negligible truncation error for the differ- cence equations presented in Appendix A. Other results presented will also assist in estab- lishing the validity of this model for predicting single- well performance within a gas condensate reservoir. Description of Predicted Performances ‘The 1-D radial compositional model was used to predict the single-well pecformance in three different ‘gas condensate reservoirs. These predictions, referred {oas A, B, and C, were all for the depletion process and, as discussed’ previously, did not consider the effect of flow within the producing string or the annulus. All predictions assumed that the reservoir formation properties were homogeneous, that gravity and capillary pressure effects were negligible, and that the composition of the original in-place’ fluid was constant. The three predictions each used two production schedules. Initially, the mass rate of uid production was held constant until the flowing bottom-hole pres- sure (Pg; in Table 1) reached a predetermined value. At that time, the model predicted the production rate as a function of time while holding the flowing bottom-hole pressure constant, The production rate declines with time as a result ofthe depletion process The rate of change of the production rate with time inall three predictions approached a stabilized value and the saturations near the well stabilized, ‘These Conditions were assumed to define a pseudosteady- state process. The predictions were stopped shortly TABLE 1 — RESERVOIR. PROPERTIES FOR PREDICTED PERFORMANCES Prediction Property a B = re ft) 21 5 558 na) 600. 1,364. 1,960. eh) 13.221 5. 134 Ih (mest) 194.58 200. 787. Sex (percent) 24.5, 110 27. . (sia) 2300.0 4,070. 4,600. ‘A (cell spacing) « 175 1.443 NN 12 10 16 Per (psia) 2208. 2,010, 2,700. “Cel spacing tor thi rn was chosen arta 861 thereafter, Steady state throughout the reservoir cannot exist since the predictions were for a deple- tion process. Formation Properties The formation properties for the three predictions are presented in Table 1. The kh products are nearly equal for Predictions A and B, The kh product for Prediction C is much larger. The well radii reflect the effect of fracture stimulation, which is a common ‘occurrence for these tight formations, The constant “A” was used to specify the cell spacing according to the formula rag = Ae INN, -@ where /is the cell number (NNV in number) and i+ ¥% is the radius to the outer boundary of the ith cell. Prediction A used an arbitrary cell spacing. The num- ber of cells into which the reservoirs were discretized ‘was variable, with a minimum of 10 cells for B and amaximum of 16 for C. No indications were observed nor were special tests conducted that would indicate the predictions used too few or too many cells. The model automatically seeks the largest possible time step; therefore, the effect of time-step size on pre- dicted performance was not studied. ‘The relative permeability ratio data for this im- bibition process are presented in Fig. 2 for each of the three predictions. These data have been normal- ized by removing the water saturation. They are presented as a function of the volume percent gas in the hydrocarbon pore space, Two of the predi tions (A and C) used data that are nearly identical. ‘The third used data that are more efficient with re- spect to ol displacement. The differences of these data will be used to demonstrate the dependence of well performance on relative permeability characteristics. Fluid System Properties ‘The three fluid systems were all gas condensates with ‘TABLE 2— PROPERTIES OF GAS CONDENSATE FLUIDS Fluid Component ComPenent Composition, Mo! Percent, A B c N 33.43 . 7677 78.146 30.95 co, 0.35, 317 o 877 673 HS 164 co. 4.95 51 ors 2473 5.04 ors 37 1.246 383 e 278 oss 2.62 BPI 1.00 2637 BP2 771 1.355 Bra os16 1225 BPs 0.560 1.133 PS. 0254 1.035 BPG 0.384 1.335 Maximum volume ‘Percent Liquid (Gepietion) 60 60 216 Saturation pressure, psia 2,270, 4,070. 4,320. Reservoir jemperature, °F 100. 178, 272. widely different phase behavior characteristics. Their propertics are summarized in Table 2, and the con stant composition expansion (flash) data are shown in Fig. 3. Fluid A, a simple five-component system, ‘was used during the initial testing of the model with fa gas condensate fluid. The maximum volume per- cent condensate during the flash process is approxi- mately 5 percent. Fluid B also yields a maximum volume percent condensate during the flash process of approximately ruin ¢ 0 | 100 100 PRESSURE, PSIA uu 8 20 190 a a) ‘VUE PERCE VAPOR, PERCENT HeAY Fig. 2—Normalized relative permeability ratio data. 862 150 — se Vo, VOLUME PERCENT LIQUID, PERCENT HCPY Fig. 3—Constant composition expansion data JOURNAL OF PETROLEUM TECHNOLOGY 5 percent, This fluid, being a naturally occurring gas condensate, is characterized by 12 components, 6 of which are heavy fractions, The latter are denoted by boiling-point fractions in Table 2, The saturation pressure of Fluid B, 4,070 psia, is much greater than that of Fluid A, 2,270 psia. Fluid C is a much richer gas condensate fui than Fluids A and:B, The saturation pressure of Fluid 4,320 psia, is slightly greater than that of B. Thi fluid, also a naturally occurring gas condensate, characterized by 14 components, 6 of which are heavy fractions Two predictions of the reservoir performance with Fluid A were made. One version of this model has merged with it the Amoco Redlich-Kwong equation of state.*"" This version was used for the frst pre- diction to include the effects of composition on K-values and, hence, om phase equilibria during the prediction. The use of the equation of state is advan- {agcous for this type of study since the predicted phase equilibria data as a function of pressure and composition are internally consistent. The second prediction with Fluid A and Predictions B and C were made with the standard version of the model that involves using data from tables and interpolation technigues for arriving at K-values and phase densi- ties. The equation of state was merged with a pro- gram that simulated the flash process for each fluid. Calculated K-values and phase densities are used by the model. These data are independent of composition. Predicted Performance Results ‘The predicted performance of a single gas-condensate producing well using the 1-D radial compositional ‘model gives information on pressures, phase satura- tions, fluid compositions, and other phase properties as a function of time and space. In addition, the ‘model predicts the composition and rate of produced fluid as a function of time. The pressures, produc- tion rate, and phase saturations. predicted by the model are of considerable interest ‘The predicted flowing bottom-hole pressures as a function of time during the initial drawdown periods were used with pressure transient analyses" to deter- mine the kit products for each prediction. These products should agree with the actual kf products used by the model, provided the pressure transient moves through the’ reservoir much faster than the condensate saturation develops. Condensate was ac- cumulating around the producing well for a portion of the drawdown period in all three predictions. Fluid B was initially saturated, so the condensate saturae tion in the cell adjacent to the producing well was increasing continuously during the drawdown, ‘The analytical and actual kh products for the three predictions are given in Table 3. The two values for each prediction are in excellent agreement. This proves that the accuracy of the finite-difference ap- proximations is not affected by changing relative permeability to the gas phase due to condensate accumulation, The agreement of the two kh products and the negligible material balance errors lend sup- port to the accuracy of the predicted performances. JULY, 1973, TABLE 3 SUMMARY OF RESULTS TO TEST FINITE DIFFERENCE APPROXIMATIONS. Property Prediction A Prediction B Prediction © Tr (maty* 195. 200. 797. kntmatye 194 201. 790. ‘The predicted condensate saturations are most im- Portant in fulfilling two objectives of this study; namely, the effect of condensate accumulation on the predicted productivity of the well and the evaluation ‘of the steady-state method. Figs. 4 through 6 present the “predicted” total liquid (Water plus condensate) saturations as a function of radius at the end of the three performance predictions; i.e, when pseudo- steady-state conditions exist within the reservoir. The connate water saturations are shown by dashed lines in these figures. Also shown on Figs. 4 through 6 are the depletion and the steady-state saturation profiles. This work compares the predicted single-weil performance with the “standard” well performance calculated by a depletion method to determine the productivity loss due to condensate accumulation; ie., the difference in the predicted and depletion saturations in Figs. 4 through 6. The predicted single-well performance is also compared with a steady-state method to evaluate the validity of that method, ‘Appendix B describes the depletion and the steady- state methods and presents the necessary equations for calculating single-well performance. Both meth- ods use a pseudopressure function similar in form to the pseudopressure function for a dry gas."*"* Figs. 7 through 9 present these functions vs pressure for the three gas condensate fluids given in Table 2. Note that the magnitude of the ordinate scale for m,(p) differs from that for m.(p) Prediction A Jacoby and Yarborough have shown that composi- tion has a negligible effect on K-values during the depletion of most gas condensates. Their work is applicable to the depletion saturations shown on Figs. 4 through 6. The development of the steady-state ‘method? assumes that the phase equilibria data within the steady-state region (in the vicinity of the well) are those of the constant composition expansion process. Hence, the K-values are dependent only upon pres sure. The unusually high liquid saturations resulting from condensate buildup around the wellbore could conceivably make the composition dependence of K-values an important variable, Two simulations of single-well performance were made with this fluid system to determine whether the compositional dependence of K-values affected predicted well performance. The first used the Amoco Redlich-Kwong equation of state merged with the ‘model. The predicted saturations as a function of radius, referred to as R-K Predicted-K; = {(p, Z,), are shown in Fig. 4. The single-well performance was then predicted using K-values that are independent of composition. The predicted saturations are the lower of the two 863 ” ‘STEADY STATE PRD 1 fee atc to.2) anon i 1 ' \ LuaUIO SATURATION, PERCENT Raplus, FEET Fig. 4—Saturation profiles for Prediction A © STEADY STATE © preterion berction Or how, Rote e, =f! Oooo ong en Bx» er aan \, z aft \ Boe Sg = wit------—-; a fy = 5.0 Feet Swe fe > V0, Zo gee 7 Tet vanuus, Fer Fig, 5—Saturation profiles for Prediction 8 , STi stAte home. o prEictiow al! oe, © orrtion | 3H aa E lpm, aN. 2» 5 a a) we ° { a a a a) aus, FET Fig. 6—Saturation profiles for Prediction C. 864 or mg) x 10 ee 8B neo) indo) x 10 son a0 TOO Tan) 800 —toeo 708 Zeon 7o0 RESERVOIR PRESSURE, PSIA Fig. 7—Gas condensate pseudopressure functions, Fluid A. 6 gs melo) "2 2 nye) e ea an sates RESERVOIR PRESSURE, PSIA Fig. 8—Ges condensate pseudopressure functions, Fluid 8 10 9 e. 3 go melo) 1002700 PRESSURE, PSIA Fig. 9—Gas condensate pseudopressure functions, Fluid C. 3700700 JOURNAL OF PETROLEUM TECHNOLOGY saturation profiles referred to as Predicted-K,—f(p) in Fig. 4. They agree exactly with the. saturations that were dependent upon composition. The predicted pressures as a function of radius, the composition of the produced fluid, and the cumulative produced fluid (gas and condensate) were also in agreement. The results of these two predictions led to developing K-values and phase densities as a function only of pressure for the other two predictions. Omitting com- mal dependence reduces the cost of a predic- tion since the laboratory experiments required to generate data on compositional dependent phase equilibria" are not necessary. In addition, the num- ber of unknown variables during the prediction is reduced, which reduces the cost of simulation, The second simulation for Prediction A was allowed to continue beyond the first to establish a pseudosteady-state condition, The predicted satura- tion profile is the upper of the two predicted curves for Prediction A in Fig. 4. These saturations are in good agreement with the steady-state saturations near the producing well. The predicted and steady- state saturations then deviate considerably, with the steady-state method giving high condensate satu tions out to a radius of approximately 37 ft. This difference can be explained by the following analyses. The steady-state method prediets the volume of liquid that should exist within the reservoir for a given pressure differential. It does not, however, give any information to insure that sufficient liquid has been condensed to occupy this volume, A simple caleu- lation, described below, was performed to determine the maximum volume of liquid that could exist withi the reservoir at the completion of the prediction. If a reservoir cubic foot of Fluid A passes through a pressure drop from p: to py, the volumetric deple- tion data show how much of this original volume will be condensed. Based on the assumptions that (1) pe was the initial reservoir pressure, (2) p, was the final bottom-hole flowing pressure, and (3) no condensate flow occurred in the reservoir matrix during the prediction periods, an estimate of the maximum volume of condensate within the reservoir could be obtained. It was further assumed that the condensate saturation, 5,, was uniform and equal to an average steady-state saturation of 0.33 (33 per- cent, Fig. 4) throughout the region where condensate was’ present, The objective was to determine the radius of this condensate bank. The cumulative vol- ume (sef units) of fluid produced as well as the ap- propriate phase equilibria data at the completion of the prediction were known, Consequently, the radius, +, of this region of high saturation can be estimated by the equation © FEZ) (Com Prod (seH] rphS,, " @ Appropriate values were used in Eq, 2, from which thevvalve of r was found to be 14 fe Considering the assumptions of Eq 2, this radius is in closer ngree- ‘ent withthe rads of the predicted saturation pro- fle than with the radius of the steady-state profile JULY, 1973, One could propose that the difference between the lengths of the predicted and steady-state saturation banks is due to the prediction time's being insuffi cient to establish a steady-state condition. To estab- lish a condensate saturation equivalent to that pre- dicted by the steady-state method, the quantity of fluid that must be produced would cause the reservoir pressure at the exterior boundary to be less than the saturation pressure. At this time, the steady-state condensate saturations of approximately 30 percent would exist throughout the reservoir (see Prediction B as an example). A conclusion then is that the region of high condensate saturation is not neces- sarily the length specified by the steady-state method. This demonstrates one limitation of the steady-state method, The initial reservoir pressure (4,070 psia) for Pre- diction B corresponds to the saturation pressure of the in-place fluid, Fluid B. The steady-state method predicts that the condensate saturation for Fluid B in the steady-state region will be approximately 38 percent (Fig. 5). Our knowledge of phase equilibria and fluid flow within gas condensate reservoirs has shown that the condensate saturations in the exterior region of the reservoir should, as a function of pres- sure, equal saturations determined from constant vol- ‘ume depletion experiments. ‘The predicted saturation profile is shown in Fig, 5. A region in the vicinity of the producing well has been affected by condensate accumulation and can be considered to have steady-state characteristics. ‘The steady-state method, on the other hand, shows that high condensate saturations exist throughout the reservoir because the pressure in the exterior cell is Jess than the saturation pressure of Fluid B. The depletion saturation profile merges with the predicted saturations at a radius of approximately 200 ft. Be- ‘ween the steady-state and depletion regions exists a “transition region”. The predicted saturations agree with the depletion, saturations in the outer region of the reservoir but are less than the saturations calculated by the steady- state method around the producing well. The di agreement of the predicted and steady-state satura tions near the well is due to the change, with time, of the fluid composition entering this region, The steady-state method assumes that a fluid of constant composition, Z;, is entering the steady-state region. ‘This was true for Prediction A where the steady-state and predicted saturations did agree near the well since the reservoir pressure was greater than the saturation pressure of Fluid A. The depletion process in the exterior region of the reservoir during Prediction B will cause the com- position of the fluid (principally vapor) entering the steady-state region to vary with time. The process within the region adjacent to the producing well is now, at best, a pscudosteady-state process; we shall continue to fefer to it as a steady-state region, Tt was assumed that the fluid entering the steady- state region had the composition of the vapor in equilibrium with the condensate as given by the 865 volumetric depletion data. ‘Three pressures (4,000, 3,500, and 3,000 psia) were selected, and constant ‘composition expansion (flash) data wore determined for each vapor composition using the Amoco Redlich- Kwong equation of state. These flash data and the relative permeability characteristics were used to cal- culate the saturations as a function of pressure according to the steady-state method. These satura tions along, with the saturations for the original uid are shown in Fig. 10. ‘The comparison of the data shown in Fig. 10 demonstrates that as the vapor phase becomes leaner (depleted), the saturation at a given pressure de- creases. This partly explains why the predicted satu- rations (Fig. 5) lie below the steady-state saturations. At the completion of Prediction B, the average reser- voir pressure was 3,860 psia and the fowing bottom- hole pressure was 3,270. Using the data in Fig. 10, one would estimate a total liquid saturation in the fist cell of approximately 37.0 percent. This is in Rood agreement with the predicted saturation of 36.8 percent (Fig. 5). ‘Additional calculations were made to investigate the effect of composition dependence of phase equi libria data on the steady-state saturations shown in Fig. 10. The equation of state, used to calculate these. saturations, accounts for the dependence of each initial vapor composition on the phase equilibria data for that vapor. The K-values for the original in-place fluid were now used to calculate the steady- state saturations for the three vapor compositions obtained from the depletion predictions for Fluid B. ‘These saturations, in essence, duplicated the satura- tions shown. in, Fig. 10. Therefore, the assumption that composition dependence of phase equilibria data can be neglected is, atleast for this prediction, valid. The neglect of this variable would most likely result in larger errors as the average reservoir pressure con- tinues to decline, Generally, the high-pressure region Fu QUILL GASY AT MAX, VOL, % QUID 14010 PSIA (ORIGINAL FLUID) 5. 2 4000 PSIA 476 3 3500 PSIA 32 4 300 PSIA CF “GASES IN EQUILIBRIUM WITH CONDENSATE AT PRESSURES GIVEN DURING NORMAL DEPLETION & x, ye * HL ml a a a ) RESERVOIR PRESSURE, PSIA Fig. 10—Etfect of composition on steady:state saturations TOTAL LIQUID SATURATION, PERCENT is of greatest interest since the economics of con- densate production are highly dependent upon early performance. The results of these calculations show that the steady-state method may have the capability of cor- rectly predicting the saturations in the vicinity of the Well, provided the volumetric average reservoir pres- sure can be estimated. However, its use is limited since the radial extent of the steady-state region and the composition of the fluid entering this region are not known. The compositional model, on the other hand, predicts the length of the highly condensate~ saturated region, the change of the saturations with this region due to the variation of the incoming fluid ‘composition, and the depletion saturations in the exterior region of the reservoir. Prediction © ‘The saturation profile for Prediction C is shown in Fig, 6. It is in excellent agreement with the steady- state saturation profile, The agreement is evidence that this prediction met, in essence, all the assump- tions of the steady-state method. The composition of the fluid entering the two-phase region was con- stant during the prediction since it is the original in-place fluid. The exterior pressure and drainage radius are large, This results in a two-phase region, with its inherent liquid accumulation, that extends for only 30 ft into the reservoir. The use of Eq. 2 with the phase equilibria and formation data and the cumulative fluid produced predicts the volume (radius) of the reservoir that would be occupied by aan average condensate saturation of 40 percent. The radius was determined to be 100 ft. Unlike Prediction A, sufficient condensate was available to generate the saturations in the proximity of the producing well as calculated by the steady-state method. ‘As mentioned previously, one might expect good agreement between the predicted and steady-state profiles. As the depletion process continues, the com= position of the fluid entering the steady-state region will vary. This will cause the saturations in the steady- slate region to change with time, much as they do in Prediction B. The steady-state ‘method would then fail to predict the wells performance accurately. Effect of Relative Permeability Characteristics The three predictions were used to investigate the effect of relative permeability characteristics on the ‘magnitude of the predicted condensate saturations in the steady-state region, The pressures given below, unless stated otherwise, are the pressures for each prediction in the cell adjacent to the well. ‘The volume percent liquid, V., obtained from Fig, 3 for Fluid A at 2,220 psia is approximately 2.5, percent. For Prediction C, which used nearly identi- cal normalized relative permeability characteristics, Fig. 3 shows 18 volume percent liquid at a pressure of 3,165 psia. The predicted condensate saturations shown in Figs, 4 and 6 will be converted to percent of hydrocarbon pore volume (HCPV), For Preic- tion A, the condensate occupied 44 percent HCPV, whereas for Prediction C, the condensate occupied JOURNAL OF PETROLEUM TECHNOLOGY ‘TABLE 4— PRODUCTIVITY RATIOS FOR PREDICTED PERFORMANCES Prediction (ela) (Bal) (Mset/D) A 2aeh. 2208, 13 8 = 378,201. sae € 570. 2700. 10,082 54 percent HCPV. Although the ratio of V for Pre- diction C to V. for Prediction A was 7.2, the ratio of the percent HCPV occupied by the condensed liquids was only 1.2. This comparison shows that the condensate accumulation for a lean gas conden- sate as required by radial flow effects may be nearly ‘equal to the condensate accumulation for a rich sys- tem. Of course, the cumulative fluid produced before the required condensate saturation is achieved will be greater for the lean gas condensate system. For Prediction B, the average reservoir pressure (G,860 psia) was less than the saturation pressure of the original in-place gas condensate, The calcula tions performed to generate the results shown on Fig. 10 included constant composition expansion data for each vapor composition, These data were used to estimate that V, was 3 percent at 2,270 psia for the vapor composition entering the steady-state region. ‘The predicted saturation from Fig. 5 is equivalent to 29 percent HCPV. Prediction A, on the other hand, occupied 44 percent HCPV, with the value of Vs being 2.5 percent. The relative permeability charac teristics (Fig, 2) for Prediction B are more efficient in terms of liquid displacement by gas and, hence, require a smaller saturation before a flowing con- densate phase is established. This results in a lower retained condensate saturation, Predictions using one gas condensate and two dif- ferent sets of relative permeability characteristics (see Predictions A and B on Fig. 2) have not been made, but the previous results indicate that the condensate saturations in the steady-state region will differ sig- nificantly. The steady-state method, Appendix B, can be used to show that the well deliverability using the relative permeability characteristies for Prediction B will be greater than the deliverability calculated using the relative permeability characteristics for Prediction A. Effect of Accumulated Liquid on Productivity The accumulated condensate in the vicinity of the well should affect the productivity of the well. The productivity ratio, defined as the ratio of the pro- ductivity index of the actual well to the productivity index of the standard well, can be used to estimate the productivity loss. In this study, the standard well (defined in Appendix B) for each prediction has fluid and formation characteristics identical with those of the actual well. The only dilference is that the phe- nomenon of condensate accumulation due to radial flow effects is igored. Instead, condensate satura- tions as a function of pressure are the saturations calculated from the volumetric depletion data for the fluid; ie, the depletion saturations given in Figs 4 through 6. The ratio of the productivity predicted by the L-D radial model to the standard well productivity JULY, 1973 4% a (sef/D) —(sef/D) PR 261 966 05327 1960 17.098 = 030.2 11291 -32,880 031084 (Appendix B) is defined as the productivity ratio, PR=4/a. @) where the evaluation of q and g, must be for given values of p. and po. This productivity ratio will show the effect of the condensate accumulation on the well performance. It is also, of course, an estimate of the error resulting from the use of m(p) to account for radial flow effects during areal predictions. Second, the steady-state productivity ratio is defined as PRe=Q/Q, 2 0 2 @ where q. is the steady-state production rate (Appen- ix B). It is implied that the values of py and p, are the same for evaluation of q. and ¢,. ‘The two productivity ratios will be useful for com- paring the predicted performance with the steady- state performance. The predicted flow rates and pressures at the conclusion of the three predictions have been used to calculate the productivity ratios given by Eqs. 3 and 4. The results are given in Table 4. ‘The magnitude of the productivity ratios for these predictions indicates the severe loss of productivity ‘due to liquid accumulation. Prediction A indicates a production rate of one-half that of the standard well. The other two predictions that are for real fd systems show that the wells will produce only about one-thinl the fluid produced by the standard wel. Had the standard well been a dry-gas well, the loss of productivity would appear to be more severe. The saturation profile for Predictions A and B did not agree with the steady-state profile, The compari- son of PR and PR,, for these predictions shows that unrealistically low production rates are calculated by the steady-state method. These low rates are caused by the excessive lengths of the region of high con- densate saturations. Prediction C showed excellent agreement between the predicted and the steady-state saturations. The productivity ratios are also in excel- lent agreement. They demonstrate the advantages of the I-D radial compositional model over the steady- state method. In addition, the model can accurately predict performance when, for example, the forma- tion properties are inhomogeneous, or the initial in-place fluid composition is variable, ot when there are gravitational and capillary pressure effects. The predicted performances could be used during the fareal simulations to account accurately for radial flow effects. Pressure Buildup ‘The reservoir engineer is constantly searching for a method of correcting phenomena that reduce the productivity of a producing well, One might expect that the accumulated condensate would revaporize if the well were shut in and the pressure allowed to 867 build up. In other words, parametric pulsing would increase the net production from the reservoir. This appears to be most economical for the conditions in Predictions A and C, but pressure buildup should also have merit for the conditions of Prediction B. The well in Prediction A was “mathematically” shut in at the completion of the depletion perform- ance, The shit-in condition was predicted until the pressure throughout the reservoir was, in essence, constant. These particular depletion and subsequent buildup predictions were obtained with the model that had the equation of state merged as an integral part — ie., R-K predicted — K}=f(p, Z,) on Fig. 4 ‘Therefore, the compositional dependence of the K- values was accounted for. The predicted condensate saturations were for all. practical purposes unaf- fected. Some revaporization was noticed near the outer edge of the two-phase region (r > 30 ft). The condensate saturations in the cells nearest the well increased very slightly. The pressure buildup of Prediction B gave similar results, Some revaporization occurred in the exterior region of the reservoir (r > 200 ft), The condensate saturations increased very slightly in the vicinity of the well, ‘The revaporization in the exterior regions of the reservoir during pressure buildup has been demon- strated by 1-D linear models. There is sufficient in-place vapor to hold the vaporized liquid as the pressure increases, In the vicinity of the producing well, the ratio of volume of condensate to volume of vapor is considerably larger. The in-place vapor can- not hold any appreciable amount of condensate. The vapor flowing into this region, as required to cause the increased pressure, is also saturated. These re- sults show that parametric pulsing will not be of any value as a method of improving the productivity of a gas condensate wel. Summary A L-D radial compositional model has been modified and used to predict the single-well performance in three gas condensate reservoirs. The finite-diflerence approximations to the partal-diferential equations describing component flow within a reservoir are presented. ‘The predicted performances showed that the pro- ductivity of the well compared with that of a stand- ard well can be reduced by a factor of three, owing to condensate accumulation in the region of the pro- ducing well, The condensate saturations inthis region are much greater than those measured experimentally during the constant volume depletion process for the fluid When the average reservoir pressure falls below the saturation pressure, the composition of the uid entering the steady-state region varies with decreas- ing pressure. The varying composition affects the magnitude of the condensate saturations within this region. The composition dependence of phase equi- libria data, on the other hand, can be neglected for ‘most gas condensate fluid systems during the predic- tion of single-well depletion performance. The relative permeability characteristics for the 868 formation significantly affect the magnitude of the condensate saturations in the steady-state region. Predicted performances using the model were com- pared with performances calculated by a steady-state method. The comparison revealed that the steady- state method may not accurately predict the distance within the reservoir in which high condensate satu- rations exist. Two predictions obtained with the ‘model were used to demonstrate the incorrectly pre- dicted lengths. In both cases, low production rates were calculated by the steady-state method. Shutting in (for pressure buildup) of a gas- condensate producing well in which condensate accumulation has severely restricted the productivity was shown to have a negligible effect on the con- densate saturation in the vicinity of the producing well, Consequently, the cumulative production over a period of time would not be improved by para- metric pulsing, The radial model has the advantage over the simple steady-state predictions for many predictions, especially when the average reservoir pressure is less than the saturation pressure of the original in-place uid or when the formation is inhomogeneous. The model described in this paper is accurate, stable, and simple to use. Nomenclature variable to control cell boundary ra in terms of rey C = 14.65 1/520 pin Eq. 2 Cy; = mass fraction of Component j in the gas phase, Ib,/Ib gas .,; = mass fraction of Component j in the con- densate phase, Ib;/Ib condensate kh = permeability-height product for the for- ‘mation, md-ft ing = relative permeability of the gas phase, 1 Ky = relative permeability of the condensate phase, 1 K, = equilibrium ratio (K-value) for compo- nent j, mol j gas/mol j condensate = flow cociticient for the gas phase, Ip = Kegkhpy/ uy 1, = flow coefficient for the condensate phase, 1, = kedkhpel te ‘m(p) = pseudopressure function for gas conden- sate fluid map) = 2 § Ke | . al dda i ws A mp) = pseudopressure faction fo standard well map) = 2 f Ke aan le M, = mass of component j per unit pore volume, Ib, /unit PV My = (pS.Cn + pySsCr.) NN = number of cells into which the reservoir is discretized P= reservoir pressure, psia tial reservoir pressure, psia productivity ratio— production rate of actual well/production rate of standard well JOURNAL OF PETROLEUM TECHNOLOGY steady-state productivity ratio— steady- state production rate of actual well/ production rate of standard well PV = cell pore volume, PY = agh(risey, — ro), cuft q = production’ rate in surface’ units unless otherwise defined, Mscf/D radius, ft condensate saturation, percent of pore volume connate water saturation, percent of pore volume time, days temperature, °R volume percent of HCPV occupied by the gas or condensate phase compressibility factor for the gas or con- densate phase, 1 composition of the fluid mixture, mole fraction variable of integration phase viscosity, ep porosity-height product, ft angular dimension in’ radial geometry, radians refers to steady-state conditions and total ‘condensate production exterior boundary of reservoir gas phase cell location condensate phase standard well well location refers to final pressure at the well Superscript n= time level References 1. Kniazef, V. J. and Navill, 8, A.: “Two-Phase Flow of YVolatie Hydrocarbons." Soe. Pet Eng. 1. (March, 1965) Medd: Trans, AIME, 234. 2. Biles, CK” Sumner, B. F. and Potts, N. Ls “Integra: tion of Partial Diferential Equation for Transient Radial Flow of ‘Gas-Condensate Fluids in Porous Structures” Socn Pet. Eng. J (ne, 1968) 141-182; Trans, AINE, Fas 3. ODell, H. G. and Mille, R. N.: “Successfully Cycling 8 Low, Permeability, Hish-Vield Gas Condensate Res: exvott" J. Pere Tech (Jan, 1967) S1-44, 4, Roebiick, LF, Ir, Ford, W. T., Henderson, G. E. and im, Jez “The Compositional Revervoir Simu- i The Linear’ Model" Soe. Pet. Eng. 1969) 115-130: Trans, AIM, 246. LP Sts, Ford, W. Henderson, G. Band Douglas, Jim, y.2 “The Compositional Reservoit lator: Case TH — The" Radial Geometry," unpublished paper, avalable from Core Laboratories, ic, Dallas, Tex, 6. Roebuck, LF. Jr Ford, W. Tr, Henderson, G. E, and Doualas, Jim, r.s"“The Compositional Simulator: Case M1——"The ‘Two. Dimensional Model” paper SPE 2235 presented (at SPE-AIME. 43rd Annual Fall Mecting, Hiouston, sept. 29-et. 2, 1968, 7. Matthews, ©. S. and Rissell, D. G.: Pressure Buildup ‘and Flow’ Tesis in Wells, Monograph Series, Society of Petroleum Engines of AIME, Dallas (1967) 1 8. Bruce, G. Hl, Peaceman, D. W., Rachford, H.H., and Rice, J, b.: “Calculations of Unstendy-Staie Gas Flow ‘Fyigueh Porous Media” Trans, AIME. (1983) 19% JULY, 1973 9. Fussell. D. D, and Yarborough, Lyman: “The Eifect of Phase Data on Liguids Recovery uring Cycling of a Gay" Condensate Reservoir,” So. Pet. Eng. J. tAptl, 1972) 86-102. 10,Zudkevitch, D, and Joffe, 1, “Correlation and Predic- ton of Vapor Laud Equitia wih te Relishwong Equation of State.” A/CKE J. (190) 16, No, 1. 112. 1. AkHussainy, R., Ramey, H. J. Jr, and Crawford, P. B.: “The Flow of Real Geses Though Porous Media." J Pet. Tech, (May, 1966) 625-636; Trane, AIMB, 237. 12. AL-Hussainy, R, and Ramey, H. J. Jn: “Application of Real Gas Flow Theory to Well Tesiing and Deliverability Forecasting” J. Pet. Tech. (May, 1906) 637-642: Trans, AIME, 237, 13, Jacoby, R. H. and Yarborough, Lyman: “PYT Measure. ‘ments on Peiroleum Reservoir Fluids and Their Uses, Ind. and Eng. Chem. (Oct, 1967) 89, 48 14. Craft, B.C, and Hawkins, M. F., Je? Applied Petroleum Reservoir Engineering, Preniog Hall, Ine. Englewood Cite N.. (1955). APPENDIX A Development of Difference Equations ‘The logic of the 1-D simulator used in this study is similar to that described by Roebuck et al. The finite-difference approximations of the nonlinear pactial-differential equations (PDE) that describe individual component and water-phase flows within the reservoir are diferent, Our difference approxi- mations for @ component are derived below. ‘The difference approximations for the water-phase PDE. and the pressure equations‘ are obvious variations of the equations presented below, ‘The PDE describing the 1-D radia flow of the jth hydrocarbon component within the reservoir is +46) 2]. an oe ‘The solution of this equation for each component at time # can be obtained by integrating Eq. A-1 over the limits 0 << 2y and re

You might also like