You are on page 1of 17
Interpretation of Well-Block Pressures in Numerical Reservoir Simulation D, W, PEACEMAN MEUBER SPESAINE ABSTRACT Examination of grid pressures obtained in the mumerical simulation of single-phase flow into a Single well shows that the well-bluck pressure is essentially equal to the actual flowing pressure at @ radius of 0.2 Ax. Using the equation for steady State .adial jlow then allows calculation of the flowing botiom-bole pressure. The relation between pressures measured in @ buildup test and the simulator well-block pressure is derived. In particular, the buildup pressure and tbe welleblock pressure are shown equal at a Shatin time of 67-5 uc, Ax2/k, This is about one= third the shutein time stated by previous authors, tho derived their results from an erroneous ‘assumption conceming the significance of the well-block pressure, When only a single buildup pressure is observed at a different shut-in time, an adjustment 0 the observed pressure can be made for matching with the simulator well-block pressure INTRODUCTION When modeling reservoir behavior by numerical methods, inevitably he horizontal dimensions of ny giid block containing a well are much larger than the wellbore radius of that well, It long has been recognized that the pressure calculated for a well block will be greatly different fom the flowing bottomhole pressure of the modeled well, but the literature contains few specific guides as to how to rake the correction. In this study, we confine our attention to single phase flow in two dimensions. Consider che five blocks abstracted from a regular grid system ( 1) with the center block containing a well producing at rate q. Schwabe and Brand! proposed the selationship tte June piblucaion Dee 20 boo) eatcprevented athe SPE-AIMEs2nd Ann Fall (GBsts Sachets oe Pewrotou Engineers of AIME EXXON PRODUCTION RESEARGH CO. HOUSTON 2ukh Pe Pub sa) ay Tn fe) +5 where 7, is taken equal to Ax, and p, is an effective pressure at the “drainage radius,” rg, obtained > e fot to imply that it be taken as zero. Thus, in the absence of a skin effect, Eq. 1 reduces to (oy ~ PQ) q= 2mh Po Pee u inGax/r,) ‘The most significant treatment of this subject until now was that of van Pollen et al.? They stated that the calculaced pressure for a well block should be the areal average pressure in the portion of the reservoir represented by the block, Assuming steady state, the pressure distribution 4 ax FIG. 1 — BLOCK 0 CONTAINING A WELL AND ITS FOUR NEIGHBORING BLOCKS, about a well is given by : Integration over a citcle with an area equal to that ‘of the block (assuming negligible r,) gives, for the average pressure, au m2]... .. Pug + p 2-4 where 2 wey = Oxeay- Let Ax = Ay, Then, ay = Axl Assuming 7 equal t p,, in accordance with the basic assumption of van Poollen et al. that the well-block pressure is the same as the areal average pressure, then cee se OD 2rkn Po Pee u met vt w Coats ct al.3 used a productivity index to relate the block pressure w wellbore flowing pressure in theie steam simulator. For single-phase flow, an equivalent relation would be GAPE (R= yg oO Theis specification for PL (neglecting the dimensional constants that they used) was ee en in"Oxayla _ 4 t 7 Since Eqs, 7 and & combine to give Eq. 6, we see that the Coats er al. approach also is equivalent to assuming that the Block pressure equals the areal average pressure, When deriving the finive-difference equations for transient flow, which come from material balances written for each block, it is natural co regard the accumulation tem, Ap,j/At, as the change in the’ average pressure for the block (i, j). This leads «0 the association of #j; with the average pressure of the block. And, indeed, for the majority of blocks, those which contain no wells, such an association is appropriace, For a block containing a well, however, we show that the block pressure is not an average pressure, This study presents che proper inter pretation of the well-block pressure, and shows how it relates to the flowing bottom-hole pressure. Finally, we show how the buildup pressures measured in a shut-in test may be related to the simulator well-block pressure. While the buildup analysis is similar to that of van Pollen ef al.,? the cesult is significantly different because of the new interpretation of well-block pressure. EQUIVALENT RADIUS OF A WELL BLOCK Icis convenient ro associate an equivalent radius, ‘oy With the well block, This is the radius at which the steady-state flowing pressure for the actual well is equal to the numerically calculated pressure for the well block, This definition for r, gives per tgp inbe sO and 2rkn Po” Put + 19 u InCes/e,,) EVALUATION OF EQUIVALENT RADIUS OF WELL BLOCK By examining the aumerical solution for Laplace's equation near a single well, we can get the cortect value for r,. This is done by solving for the steady-state pressure distribution in a repeated five-spot pattemn using the 10% 10 square gid shown in Fig. 2. Details of the calculation are given in Appendix A, The numerical solution for the various blocks is plotted as a function of radius in Fig. 3. On this semilog plot, a steaighe line with slope ifm fits very well through all the points up to a radius of 6 Ax, Similar runs were made for larger grids (20 20 and 30 x 30) and che results are practically identical. Ik is probably not a new observation thar the numerical solution near a Single well behaves so tnuch like the solution to the radial flow equation. What appears to have escaped notice is the ttt 1012345678 9101 i FIG. 2 ~ 10 x 10 COMPUTING GRID FOR REPEATED FIVE-SPOT PATTERN. significance of extrapolating the straight line of Fig. 3 to the horizontal line p ~ p, = 0. From Eq. 9, it follows that this intersection occurs where rye 50 we have the important result from Fig. 3 0.2 xe eee OD eens isthe new inverpettion for wellblock pressure; it equals the steady-state flowing pressure ata radius of 0.2 Ax. This result contrasts with that obtained by assuming that the well-block pressure equals the areal average pressure, Comparison of Eq. 6 with Eq. 10 gives ox 1, dn oh = BE exp (-1/2) = 0.342 ax +» 012) Thus, that assumption leads co the cesule that the well-block pressure is equal to the flowing pressure fat a radius of 0.342 Ax. This ratio of 4/Ax also is shown on the ? ~ py = 0 line of Fig. 3. The deviation from the we intercept is quite apparent. APPROXIMATE CALCULATION OF EQUIVALENT RADIUS ‘Another observation whose significance appears to have escaped notice is that the pressure calculated for the block adjacent to the well block (that is, Py,, in Fig. 2, oF py in Fig. 1) is practi= cally on the’ straight Line. Thus, we can substitute Ax into Fg, 9 to obtain GH py AX 0 * Tek © ry In addition, we have available the difference equation for the well block (see Appendix A) of - 2 PLP, 03) (kh/u) (Py + Pp + Ps + Py ~ 4B.) = a ~ . (4) 102 FIG, 8 — NUMERICAL SOLUTIONS FOR PRESSURE PLOTTED VS RADIUS. By symmeu Py = Ps = Py. as) Combination of Eqs, 13 through 15 gives In(x/r,) = 1/2" rg = Ox exp(-n/2) = 0,208 ox 6) result depends only on the assumption thar the pressure of the adjacent block lies exactly on the straight line. Fig. 3 indicates tha this assumption is nor exactly correct; hence, Bq. 16 is only an approximation, EXACT CALCULATION OF EQUIVALENT RADIUS For aay square gtid of arbitrary size, we can, in fact, calculate 7, exactly without’ using the graphical construction of Fig. 3, This can be done by solving numerically for the steady-state pressure distribution (as outlined in Appendix A) and then using the equation for the pressure drop between injection and producing wells in a repeated five-spot pattern given by Muskae of op = $5; [inca/e,) = 0.6190], « «07 where d is the “diagonal” distance between wells. Ire take Ap to be the difference in pressure Between the injection and production well blocks, then the yy of eq. 17 should be replaced by 15. Farther, we have a= /2M ox: Then, Eq. 17 may be rewritten as kh _ ae (PM,M 7 Poyo) = In(/2 HM ax/x,) = 0.6190 (yf) = VE M exp [ 0.6190 ich, ~ BE au “Pie > 0,0): 78) Calculations for grids ranging in size from 1x 1 to 32 x 32 are shown in Table 1. The second column lists the dimensionless pressure drop, (bh/qp) (2M. ~ ojo» obtained in the numerical calcula tion for each grid; the thied columa lists the values of ro/Av calculated using Eq. 18. As expected, the 1x Land 2% 2 grids are anomalous. The 1 I grid has no diagonal point between the wells, while che 2 2 grid has only one iaterior diagonal point. But for M > 3, we see that the value of r,/Ax is berween 0.194 and 0.198, with 0.1982 being the apparent limit as M+-0, The rule of thumb that rg = 0.2 Ax is, therefore, well substantiated by these numerical experiments. EXTENSION TO UNSTEADY STATE ‘The validity of the rule of thumb that r, = 0.2 Ax so far has been demonstrated for steady-state conditions. However, a numerical calculation shows that this rule also is cue for unsteady state, Appendix B outlines the numerical solution for the unsteady-state problem of production of a compressible fluid from a single well ia an infinite region. Results of the calculations are summarized in Table 2. The numerically calculated solution for the well-block pressure drawdown is listed in the second column of Table 2 for various values of the dimensionless time, tp = kt/(que, Ax2). The third column gives the “radius of investigation," defined by Thay ~ Wet! Gue,] 2 = axCiey)® - a9) The last column gives the equivalent radius of the well block obtained by comparison with the analytical solution to this unsteady-state problem (also given in Appendix B). For tp greater than 1.0, ry = 0.2 Av is indeed @ satisfactory rule of thumb. (Although Table 2 shows 1, to be considerably greater than 0.2 \v for ip less than 1.0, this does not detract significantly from the utility of this rule, For this short period of time after a rate change, the radius of investigation is less than 2 Av, and well-block pressure cannot be expected to relate to bottom-hole pressure in a simple way. Where a numerical calculation must reflect realistically the pressure changes occutting so TABLE | — NUMERICAL CALCULATION OF PRESSURE ‘DROP FOR REPEATED FIVE-SPOT PATTERN AND OF ‘THE WELL-BLOCK EQUIVALENT RADIUS in $9 domi Pod “ a tolldx 1 os oes 2 Sete? our 3 oreert oe a aaron ov t858 5 Saccen O88 3 proce oro ; {sees Syers 3 Lone oars : ‘seer ous? * Seu otere 2 aot ore ‘ eess 2.00 8 Suen e360 ® tae oct 2» ‘aeezo out 2 Varost our Fa Vor our a Views tee: = ‘ever tase = vee O98 2 {Sone Some soon after a rate change, a smaller value of Ax must be used.) The equation, 1, = 0.208 Ax, previously derived for steady state by the approximate theoretical argument leading to Eq. 16, also can be derived for pseudosteady state, which is done in Appendix C. Because most transient situations can be described well by pseudosteady-state behavior near a well, this derivation provides additional support for the conclusion that ry = 0.2 Ax also is true for transient conditions. EXTENSION TO GAS FLOW While the presented derivations and calculations for the equivalent radius of a well block aze for incompressible or slightly compressible fluids, identical results would be obtained for highly compressible liquids and gases. Further, situations where the Darcy permeability is a function of pressure (as in the case of the Klinkenberg effect) also can be included, It suffices merely t0 define a quantity, m(p), by the integral P m(p) -| Pref analogous 1 the eal gas potential defined by Al-Hussainy and Ramey. If one uses m (p) for the dependent variable, then the differential and difference equations for both steady state and pseudosteady state have exactly the same form as the differential and difference ‘equations for incompressible ot slightly compressible fluids using p as the dependent variable, although with different constants. Thus, the numerical calculations for steady state,” the derivation Teading t0 Eq. 16, and Appendix C go through in ‘exactly the same way, leading to the conclusion that rq ~ 0.2 Ax also is true for highly compressible Uguids and gases. o(p) k(p’ ue)? COMPARING PRESSURE BUILDUP DATA WITH SIMULATOR RESULTS In theory it would be sufficient, for history match- ing, to celate the simulator well-block pressure to the flowing well pressure by the equation TABLE 2 — NUMERICAL. CALCULATION OF PRESSURE [DRAWDOWN FOR AN ISOLATED WELL IN AN INFINITE! FEGION AND THE WELL-BLOCK EQUIVALENT RADIUS. See Pale fl ra rr) 0.89 0.281 02 oar oi oar 05 0.2003 sat oy 19 0.3002 200 oats 20 oar 2a 208 50 baat oar oer 109 2's000 62 0.200 200 0.5508 a 99 =o, t pe in pees + 20) Put @akh Ir which was obtained by combining Eqs. 10 and 11. However, the flowing well pressure is difficult to measure, and Fa. 20 requires exact knowledge of the effective wellbore radius, r,,, which also can be difficult to obtain, Normally, the well is shue in and pressure at the bottom of the well is recorded, tither at a single point in time or as a function of time. Previous analyses of buildup curves (e. Miller e¢ al.,© Diecz,” van Poollen et ai.,? and Kazemi®) focused on the average ptessure in some bounded region, or at least required the specification of some drainage radius or external radius. Here, we attempt a more straightforward approach chat avoids specification of rg oF ree Fig. 4 shows @ typical plot of pressure vs the logarithm of + — first at shutin time, ¢ (the straight line), and then at ewo subsequent times after shut-in, After shut-in, the slope of the pressure curve is zero at the wellbore radius, ‘4, ‘which explains why buildup pressure is insensitive fo Ty: Ifa horizontal line is projected from the shut-in pressure, Pq, (0 the straight line, ic intersects at Some radius) rq (which we call the “probe radius"). This probe radius is the radius at which the steady-state flowing pressure (before shut-in) uas egual to the current well pressure after shut: Note that the probe radius, rq, is a function of shut-in time. By two independent derivations, ‘Appendix D shows that the probe radius and the shutein time are related by the equation katy 2) 2 gue, ty Now, the shut-in pressure of the actual well will be equal to the simulator well-block pressure at a shutin time, Ar, when the probe radius is equal to the equivalent radius of the well block. Thus, we have k at? Sm 0.445" oe 2) Tw ma tma FIG. 4 — PRESSURE DISTRIBUTION BEFORE AND AFTER SHUTAN and Substitution of Eq. 23 into Eq. 22 for this shut-in time gives atQ = 0.445 gue, (0.24%)/k» = 0.0178 gueztx’/ke +. -- - 4 In practical units, Eg. 24 becomes (atQ)" = 67.5 puch (axt)?/at © + +25) This may be compared with the corresponding ation reported by Earlougher? (which he chained directly from the work of van Poollen et al.2), (ap = 200 guet(ax')*/et » - - 26) Comparison with Eq, 25 shows that this time Es too large by a factor of three, As this is the time at hich the observed buildep pressure is expected to agree with the simulator well-block pressure, usiog this overly lagge time will resule in an ertor in the history-matching process because there will be an terror in the matching pressure equal to A PS 7 P = [au/ rea) Jn(r8/r,) « But, from Eq, 22, anGeA/e,) = F inl (At) 4/ (ato) 4 1n(200/67.5) = 0.543 + ‘Then, in practical units, = 0.543 (141) (q"y/kth") = 76.6 gtu/kth'> ‘An example calculation given in van Poollen e¢ i.19 shows how large this ettor can be. In that ‘example, (k/1) = 235 md/ep, b”= 12 ft, g” = 1,350 B/D. The error in the pressure to be matched would be 36.7 psie USE OF SINGLE-POINT BUILDUP PRESSURE Previously, we assumed that the buildup curve was available so that a shut-in pressure equal to the simulator well-block pressure could be measured at the shut-in time given by Eq. 25. However, as Earloughec® pointed out, if only one pressure is observed, Eq. 25 cannot be used, and an adjustment must be made. If the observed pressure, p23, is measured at shut-in time, Argbs, then ‘the corresponding probe radius is given by Eq. 21 as (908)? = ie 4e2P5/(0.645 gue,) » 27) ‘ s By Eq. 9 obs | Bay (x98 Pas 7 Po * Sp 1 y/o) Substitution of Eqs. 22 and 27 into Bg, 28 then gives obs , au © obs. pops + 7aH- fan ac? - in aeoPSy, In practical units, Eg. 29 becomes + ee a [pesroees> + G0) which is consistent with the standard well-test equation that relates well pressure with the logarithm of time. Substitution of Eq. 25 into E 30 gives the more convenient form be gor yobs ph = (ph) 67.5 gues (ax")* 4 162.6 4" yoy 13 buce eh 20) aes) nin +. 6D ‘This is the equation reported by Earlougher, except that his constant, 200, in the argument of the log, is replaced by the correct value of 67.5. As Earlougher pointed out, the chief requirement for using Eq. 31 ts that the observed pressure be on the straightline portion of the semilog buildup plot. Further qualifications ought to be made. If there is only one measurement, at time 1, + Arg#S, then 1488 obtained from Eq. 27 should not be too small of too large. Not being too small means that rg should be considerably larger than r,,, so that the effect of near-well inhomogeneities will not be imporant. ‘The more likely situation is that *gbS may be oo large, relative to either Ax of to some neighboring well or fault that may cause interference. Unfortunately, it is frequently the practice to report aly the 72-hour buildup pressure with the mistakea notion that the object of the buildup tese is to measure a stabilized pressure corresponding to aa “average’’ pressure near the well. In a hi permeability reservoir containing slightly compressible fluid, such a long shut-in time can be excessive, resulting in a ptobe radius larger than Ax and possibly larger than the distance to the nearest well or fault, or the distance to the reservoir boundary, DISCUSSION White the derivations in this study are based on single-phase flow, the concept of an equivalent radius of a well block equal to 0.2 Ax has been useful ia more complicated reservoir simulations involving multiphase flow. here the well block is not square, the empirical relation yy = 0.2 V Ax: Ay hhas been used, Further theoretical and numerical study has been conducted for ‘x not equal to Ny, as well as for the case of anisotropic permeability. ‘This research will be reported later, CONCLUSIONS 1. In numerical reservoir simulation, the pressure calculated for a well block is the same as the flowing pressure at an equivalent radius, r,. For a square grid, r, = 0.2 \x. This equation is valid for both steady “state and transient conditions. 2. The conventional equation for steady-state radial flow then can be used to relate the well-block pressure (at radius r,) to the flowing bottom-hole pressure (at radius 1,.). 3. Buildup pressures should be measured at a shut-in time equal t0 67.5 gue; (Ax’}2/k to obtain 4 pressure for matching with simulator well-block pressure, 4, The previously reported value for this shut-in time is too large by a factor of three and use of the coresponding buildup pressure can result in a significant error in history matching of pressures. 5. If one buildup pressure is measured at a time different ftom that given in Conclusion 3, an adjustment can be made to obtain the value to be marched with simulator well-block pressure. NOMENCLATURE otal compressibility of rock and fluid, atm? compressibility in field units, psi constant of integration diagonal distance between injection and production wells in five-spot patter, cm a9 exponential integral function reservoir thickness, cm 1” = reservoie thickness in field units, fe J = Bessel function k = permeability, darcy A’ = permeability in field units, md A = number of blocks on each side of computing grid p= pressure, atm p= pressure in field units, psi average pressure, atm Ap = pressure drop between injection and produc tion wells in five-spot pattern, atm dimensionless pressure = (kb /qy) 0 pressure at external boundary, atm initial pressure, atm simulator well-block pressure, atm flowing bottom-hole pressure, atm bottom-hole pressure after shut-in, atm productivity index 9 = production rate, ce/see 47 ~ production rate in field unies, B/D 1 = radius, em 77 = radius in field units, fe radius of circle whose area Ae = Ay, om adius of extemal boundary, em fadius of investigation, defined by Eq. 19 probe!" equivalent radius of well block, cm wellbore radius, em us from buildup test, cm ime, seconds dime in field units, hours 1 A@ue, Sx?) in Appen- imensionless time dix B dimensionless shutin time = kA‘g/Agucy) To Appeadix E me at shutin, seconds time since well was shut in, seconds shutein time when ty fos seconds neh zero of Bessel function, J, grid spacing i x direction, em arid spacing in field units, fe {id spacing in y digection, em constant of integration, defined by Eq. D-17 Euler's constant = 0.5772157 ... seosity,

then Eg. A-L simplifies to (py) ga at Oy) 41,5 * (Pp) 5,51 + pg gar 7 Mp gs = Bye OD here for i,j # 0,0 or MM > Sua ‘The following reflection conditions are used: Pag" Payg for 0 + 63) chen Eq, Bel simplifies to Cpsa,y * Py asa * Pr s,j-1 20D) a4 * Ops ga > MPD as 7 855 + GE bees BH where bij 77} for 4,5 = 0,0 845 O for i,j # 0,0- The initial condition is @y)yg 70 for all a5, while the boundary conditions are given by the reflection conditions of Eqs, A-3 and Af, The resulting system of difference-ifferential equations was solved with a sufficiently large umber of grid poiats (M = 20) that, for ip < 20, (p)y,4 is" less than 0.001. Thus, the Solution obtained near the well is the aame’as that which would be obtained for an infinite region. Further, 2 small enough tine step was used that che time truncation error was negligible. (Both dicect solution and altemating. direction were used to solve the equations at each time step, with essentially “identical results.) ‘The numezical solution for the dimensionless pressure drawdown of the well block is given in Col, 2 of Table 2 for Selected values of dimensionless time. Toobtain an expression forthe equivalent radius, we need to compare the numerical solution for the Sell-block pressure with the exact solution forthe radial tansient problem. Tor an infinite reservoir initially ae pressure py, producing at a constane fate, g, fom live sink at = 0 fom time t~ 0 to timer, the solution satisfies (Muskat* and Matthews and Russel!) the equation oue, 2 so vay ate [-ee(- SEY} (B-5) For sufficiently small r, the E function can be approximated by ~ Bi(-x) = - y - In x + «(B-6) ‘Thus, the wellbore pressure is approximated well by 2 oue,r, wet = 74 aby [= y= m(E)] = (B-7) From the definition of the equivalent radius of a well block (Eq. 9), we have - iy. . (8) Pus) = Poo * Baek MF Substitution of Eq. B-8 into Eq. B-7 and rearranging gives 2 duet - - fey. Fo]. Poxo its [ ya (“ae J Using Eqs. B-2 and B-3 simplifies this to 2 x 1 2 poo ta | »(z 25 ) : y and this may be solved for the equivalent radius as r= bx [*o exp {-y- Sr(eo,ot fr? Sees + + (B9) using (pp)o,o listed in Col. 2 of Table 2, The zatio of ra t0 Ax % listed in the last columo of Table 2. APPENDIX APPROXIMATE CALCULATION OF EQUIVALENT RADIUS FOR PSEUDOSTEADY STATE Under transient conditions, the following differential equation describes the tlow of a slightly compressible fuid in a radial system: 2a(, ap). fe ap, rar (Fae) * Eo 3e ‘The usual definition of pseudosteady state is that the rate of pressure decline throughout a closed region is a linear function of cime.!! A less restrictive definition can be used here, namely that dp/dr simply be independent of r. Eq. CI then can be integrated to cy 2 gue,’ yet PB ar eat (2 Use of the boundary condition, 2nkhe, «- E (H) gives 2 ¢ aM eEw ap - cs ‘1 2th 2k at “ Dividing Eq. C-2 through by and integrating once more gives gue? P= The at 1 int ° Cq can be eliminated using the boundary condition P= Pp AEF = rye They 2_ 42 - 12) - SN wap ak P* Pug * 4k at +o ry bees Ac-4) Bq. C-4 and noting that 's negligible (except in the Substituting Eq. C-3 the contribution of 7 log term) gives ouey? ue Po Pus + Te 22 4 Hy ae * Trek ny (C5) In Appendix B, the numerical solution for unsteady-state conditions was outlined. Take \v Ay and use the notation of Fig, 1. Then, Eg. B-l written for the well block, and the ‘symmetry conditions of Eg. 15 lead to gue Ax 4p, - PL) = at € ae (C-6) ‘To derive the equivalent radius of the well block, we now make two assumptions. First, py is assumed to be the same as the radial solution given by Eq. C5 evaluated at r= Av, (This corcesponds co the assumption that was used in the dezivation of Eq. 16.) Thus, 2 gue ,ax = Sc’ 3p 4 gh 4, oe i a a SE (C7) Second, we assume that the rate of decline of the well-block pressure, dp,/d, is the same as the uniform rate of decline, 4/di, in the pseudosteady- state radial solution. Combining Eqs. Bol, C-6, and C-7 then yields Eg. 16. Thus, Eq. 16 can be derived noe only for the steady-state conditions used in the text, but also for pseudosteady state, which adequately describes most transient situations near a well. APPENDIX D DERIVATION OF PROBE RADIUS AS A FUNCTION OF SHUT-IN TIME ‘We seek che solution to the following differential equation for flow of a slightly compressible fluid in a radial sy stem: 2 Le apy. ap _ SYS ap. (1) metro" k 3t Two approaches using different initial and boundary conditions are presented, USING Assume an infinite reservoir initially at pressure pj producing at constant mte 4 from a line sink at 7'20 from time ¢ = 0 to shut-in time f,. Then, the solucion at 1, satisfies (Muskat+ and Matthews and Russell!) the equation 2 gue ,r post) = py ~ 7s [- a(- | - (D2) i FUNCTION After shui pressure is Pye (ty + Ata) =P. by superposition, the bottom-hole duce - |e ee (- ete Sukh Tk(e Hae) 2 oue,t - w (D-3) ve (ie) By combining the definition of ry P(ry.t,) = pyg(t, + at.) o-» with Eqs. D-2 and D-3, we obtain overs vat (- Ge) et (- ( gue xr ) +e (- (D5) teats For sufficiently large ¢, and At, the Ef fonction can be approximated by - Bix) = - y- In x where y = 0.5772157 is Euler's constant, Then Eq. wy (UE ener? gue, x2 -In Setw Me rw ¥ (acte5) tye an a) ak at, t m2 Hrm,) 7 em 8 Then, k Ot, Be tote oot ee , 4 eee 8 = 0.44527(t, + ot, )/t, cee ses (D6) For sufficiently large t, (geeater than 2,000 At), Eq, D-6 is the same as E USING BESSEL SERIES Here, assume a bounded reservoir with internal dius »,, and external radius r,. The pressure at the external boundary is maintained constant at Pe: (The nature of the external boundary condition iS not important, as van Pollen et a/2 found when they compared” solutions obtained by assuming constant pressure at ry with solutions obtained by assuming no flow across rz. Further, we show that the solution is, in fact, insensitive to the choice of res) Suppose that steady-state radial flow exists juse before shut-in, Then by Eq. 3, we have Pg = Pye * Cau/(2nkh)] Ine, y/e.) , vce t ete e eee eee OD while the initial condition can be written as pOrst.) = py + Lau/(2nkh)] InGe/e,) + eee (ws) The solution to Eq, Del subject to chis initial condition and to the boundary conditions =. -(D-9) D-10) ap (D-10) is given by Muskat# as + plese, + ae.) oo Bee "Ee Te) : > Io Gege/e exp -xet,) ol x IEG, 2. Dan) where Jy and Jy are Bessel functions of the first kind, x's ate che zetos of J,5 that is, Tj) = 0 met, 2,... and tp is a dimensionless shut-in time defined by k ary - (P12) The use of the limiting condition ry +0 at the interior boundary (Eg, D-10) is justified by the observation that Py (P44 At,) is insensitive to the value of ry. To avoid the apparent dependence of the solution on ry and pyy, we substitute Eq, D-7 into Bg, Dell to obtain py 7 Prt, + Ot.) = 2 o > J, Ox,t/t expe ty) kh ~~ 32Rzy OT ot x YG) In particular, by letting ++ r,, +0, and noting that Jo (0)= 1, ve get eo ~ Pyglty + Ot) = Lau/(nkh)} a(t)» ee, +. 3) where Fy exp oxze) ote) = Dy eee... coy 22 ot 2 ha) By substituting r» into Eq. D-8 and combining with the definition ofr, Eq, Dad, me get Pulty + at) = py + [auf @nkh)] tnGefe.) «+. = D9) Substitution of Eq. D-15 into Eq, D-13 then gives the simple relation 4g In(x fx) = oft)... +. (D6) Ic seems reasonable to assume tentatively that JUN 9 tm is independent of ¢,, at least for some range of tp. Ie follows that a % =0> fe @- sine) =0 e (@o/ty) (atylaz.) = Ur.) Bar, - 3) 2 dey/ar= -2kat,/(dueye2) = = 2ty/r,- Thos, (a9/aty) = -1/(4ty)> a= % Anlalty)y eee «(D-17) where « is an undetermined constant. ‘The assimp- Tion that ry is independent of re is validated if « is indeed found constant. Substitution of Eq. D-17 into Eq, D-16 yields 2,2 2 Fifre = tyfa = k dt,/(atue,t.) > k At, Sra eee eee (D-18) ee ue ge ‘The constant, a can be evaluated precisely from the series of Eq, D-14 by rewriting Eq. D-17 as a= ty exp(do)- Table 3 summarizes the calculations. For each value of tp, the series was terminated when the last term was less than 10-8, Clearly, for fp < 0.1, a may be taken constant at 0.445. Eq. D-18 is, therefore, the same as Eq. 21. RANGE OF VALIDITY We have established chat Eq, 21 is valid only if tp <0.1. Combining Eqs. 21 and D-12 gives TABLE 3 ~ EVALUATION OF THE CONSTANT, «, FROM BESSEL SERIES Nuber of ty ree Akt) x0 2 ro on 2 0120 os 3 0.03561 03 3 onia24 oe 4 20249 on 5 osras 0.03 8 ores 001 8 o.sea07 0.009 B 1.2600) o.001 2 18007 0.0008, 8 reas .0001 8 21003 00008, 208 aor 2.00001 8 2erss by 7 06445 (ry/e,)? < 0.1 nm sB< 0.47. Thus we can use the buildup curve to “probe'” to a radius up co one-half chat of an external boundary. APPENDIX E FURTHER COMPARISON WITH van POOLLEN ET AL. van Poollen ef al2 obtained the following equation for the shut-in time at which buildup pressure should equal the simulator well-block pressure: °. 2 (ata 7 06167 guc,ry /k « Substitution of Fq. 5 gives °. 2 (at, = 0.0532 gue,ax7/ke We can derive this resule directly by substituting Ea. 12, A 0.342 ax» into Eq. 22 t0 get (eg), 0.0520 ouepax?/e- - - (ED [the difference between Eqs, E-1 and F-2 results from van Poollen ef al, using graphical methods based on the graphs of Miller et al.°) to obtain Eq. E-L, while the more exact methods of Appendix D were used to obtain Eq, 21 and, from this, Eq. 21 Ia practical units, Eq, E-1 is essentially the same as Eq, 26, which gives a time too large by & factor of three, Similarly, Eq. E-2 is too lage by 2 factor of three, compared with our Eq. 24 ‘Another error that can occur from using Ref. 2 is when calculating the difference between the vellblock pressure and the flowing well pressure. ‘The relative exror in this difference iv (by Eqs. 10 dhrough 12), Pep) ~ (Py ~ Pye? > Pup) Cp. an(eA/e) = In(eg/r,) in(z,/r,) 1n(0.342/0.2) in(0.2 ax/r,,) Assume Ax/r,, canges from 200 to 3,000. Then, this relative error will range from 15 to 8 percent. SPE 6988 Discussion of (SPE 6893) Interpretation of Well-Block Pressures in Numerical Reservoir Simulation by Donald W. Peaceman Discussion by I. Mrosovsky A very interesting treatment of well block pressures has been presented by Peaceman in the above paper. In particular, he demonstrates that in applying the usual formule z = pws + Su in FO Po = Put * Oh ry to well productivity, rg should be set equal to 0.24x, where ax is the length of the side of a block. (The same symbols as used by Peaceman will be employed here as far as possible.) Another approach to this problem indicates that this rule is not universally true, not even for single phase flow in square blocks. For semi-steady state compressible flow, the average pressure 5 in a closed reservoir, of area A and producing through a single well, is given by the equation? + au fin A + .809 PS Pw a E CaEwe | where C, is a shape factor. For a circle Cy is 31.6 and the equation reduces to the familiar form 5 + fin-b -3 But * ookeh j Ty Al For a square C, = 30.9 and the term in square brackets becomes [Bazp/rw) - 7385] where rp = ax. If, nevertheless, well productivity is to be described by the formula r + Ww in 2 Pwr Sikh” Ty 5 then, in order to obtain correct answers, one must put an 22 = in axAe - .7385 Tw This readily yields rp = 0.274x, a result midway between that of Peaceman and van Poollen?. § is here the average block pressure in the sense that it is equal to the fully built-up static pressure that would be obtained if the well were closed in. 5 is also the pressure of the block that would be obtained in numerical reservoir simulation. The case just discussed is not only a special one, but a trivial one as far as simulation is concerned. Nevertheless, it is closely related to cases of practical importance. In modeling a field with a large number of wells, it is frequently necessary to have wells in all of many adjacent blocks (figure la). For the purpose of well productivity, each block may be treated as a sealed cell. Of course, the block boundaries are not sealed, but in general, it is unknown in which directions the flows across these boundaries will occur, or in what quantities. It seems therefore reasonable to take the middle condition in which flow is in neither one direction nor the other, and apply as an average rule of thumb ry = 0.27sx to all blocks. A block pressure is the average pressure of a block in the sense that it is the static pressure that would be obtained if the well were closed in and the block were simultaneously sealed at its boundaries. Suppose, however, that some more detail can be afforded in the simulator, as in figure lb, The boundaries of the small blocks in which the wells are situated, cannot even approximately be considered as no flow boundaries. If the well block is small enough, the pressure distribution over it can be described by P= Pye + ean 2ekh ry The average pressure is obtained by integration over the block square, but the result is close to the formula used by van Poollen: Beye + ou ps Ax/ST L 1 2akA ty 2} However, Peaceman has shown that the block pressure is not the average pressure of a block when there is a well in it. This is disturbing since a simulator is set up to maintain material balance. If the block pressure is not the average pressure of the material in the block, then a material balance error is incurred. -4- Peaceman's three demonstrations that a well block pressure is not the average pressure of the block, all are based on incompressible fluid flow. No material balance error is therefore involved since the amount of fluid in a block is then independent of pressure. Some comment on this problem would be welcome. lyatthews, C.S., and Russell, D.G., ‘Pressure Buildup and Flow Tests in Wells'. SPE Monograph 1967. 2van Poollen, H.K., Breitenbach, E.A., and Thurnau, D.H. “preatment of Individual Wells and Grids in Reservoir Modeling". Soc. Pet. Eng. J. December 1968. . . . . . . | . . . Fig. la. | Adjacent Well-Blocks . . . . . . . . . T Fig lb. More finely gridded simulator

You might also like