Professional Documents
Culture Documents
then all data should fall on the same trend. This is true for all of the cases though there is some scatter in the random rate case due ‘to the erratic nature of the rate function. 39Ideal. Constant: Rate Constant Pressure Rate Log Rate Log Rate Log Rate TABLE 2 ERROR ANALYSIS FOR WINESTOCK AND COLITIS METHOD: BASE CASES Dimensionless Dimensionless Dimensionless Dimensionless Slope Error Intercept Error ——Slope ___ ____Intercept ) eeeaeee (Semmes 1.151293 0.404539 --- --- 1.151290 0.399490 -0. 0002212 -1.248 1.138803 0. 584546 -1.519 44.45 1.169566 0.336053 1.587 -16.95 1 1.150893 0.615438 -1.772 52.13 2 1.165212 0.417676 1.209 3.247 5 1.165176 0. 3787s 1.082 6.379a wolgoung 6u1330[¢ uoLaLsoduadns aut, Ssa[voLsuaug 40 UOLZ9UNY e Se doURIMOJJag ansaid SSB[UOLsUaULG :SaSE) aSeg UMOPMEJG MOLY qUALsUes] - 2] o4nBL4 wm prot OE OT OT gt) gOT gO gor ort pe 0-0 Es: E os sase) LLY L [9° E s-or gratThe results and error analysis for Fig. 12 are given in Table 5. The absolute slope errors for Fig. 12 were all below 0.05% except for ‘the random rate case which was 3.6%. The absolute intercept errors for Fig. 12 were all below 1.5% except for the random rate case which was 49.2%. The high relative error for the randon rate case ocours in the attempt to use the least squares routine on a fairly scattered data set. This is because the optimization part of the least squares routine assumes a fairly smooth data curve. Though this may seem significant, Fig. 12 shows that all of the random rate data fall on or near the given trend. NUMERICAL INVESTIGATIONS - WELLBORE STORAGE CASES (SURFACE RATES) The six base cases were again simulated except that a dimensionless wellbore storage coefficient, Cp, of 1000 was added to the finite-difference model. ‘The addition of wellbore storage to the finite-difference model and verification of the solution is given in Appendix C. The wellbore storage (surface) rates for these cases are identical to those given in Fig. 9. Constant Rate Method ‘The wellbore pressure response for each of the simlated cases is shown in Fig. 15. Note that the wellbore storage distortion causes theCase Ideal. Constant Rate Constant: Pressure Ranion Rate Log Rate Log Rate Log Rate 43 ‘TABLE 3 ERROR ANALYSIS FOR SUPERPOSITION METHOD: BASE CASES Dimensionless Dimensionless Dimensionless Dimensionless Slope Error Intercept Error Slope _ __Intercept, oe ——(%) 1.181293 0.404589 --- 1.151290 0.399490 ~0.0002212 -1.248 1.180939 0.408284 0.03071 0.3103 1.192765 0. 208620 3.602 49.17 1 1.150823 0.404260 -0.04078 0.06901 2 1.181349 0.400910 0.004903 0.8971 3 1.181501 0.400157 0.007342 1.083g SQUBUOJ4ad BuNSSOud a4OqL LOM :S07ey BoRJANg ‘Sase) aBeA03S B40qL LAM UMOPMEJG MOLJ qWALSURAL - ET aunBLy ay faut Of Oi oe ot To 7.0t 2.0t gor popitiriitiiiitiiy roitiiiitiriitiiia o00e (1) 1S e# erey 607 (9) F 2# agey 607 (s) . T# azey 607 (p) [L oose Oy yey wopuey (ce) f : ery aunssaug queysuo) (2) F on yey quegsuoy (1) [7 (9) }~ 000% (2) 7 5 | o0sy ‘S) b (r) L 000s eisd) | f a3ey wopuey (€) -- Ov ff sanssoig uersuoy (2) aqey quezsuog (T) oot a/ay ‘>(quayeatnba Poy3aN S934d{09 pue x03SauM) ouyL ssajuoLsuaMLg Jo wol39uNj @ SP soURMLOJ4aq aunssaiy ssaluoysuauig :s92ey aoejpues ‘sase) abeio3s a40g] [ay UMOPMe4G MOL4 qUaLSUeA] - JT aunBLy a Gee cu se I I L L L {oy 1 i L oro fF sz i. i. a) e# 934 607 (9) E 2@ aaey G01 (5) [SE T# 03eY 607 (b) (8° (2)* (1), aqey wopuey (6) (5) aunssaug yueysuoy (2) [7 OT (9) ayey quezsuog (1) + L sardominated data where the rate is always increasing. Note that the familiar rate dominance associated with the Winestock and Colpitts Method is seen to start at about t,;10". ‘The results of the analysis performed on the straight-line portions of Fig. 17 are given in Table 6. Since the constant pressure case is unaffected by wellbore storage, this data had the lovest overall absolute error. For the cases that vere affected by wellbore storage, the minimum absolute Slope error was for the log rate +3 case, and the maximum was for the random rate case, with no error greater than 16% and with most less than 6%. ‘he minimum absolute intercept error was also for the log rate #3 case, and the maximum was for the random rate case, with all errors significantly greater than 40%. 4 possible explanation for this rather high error is that the Winestock and Colpitts solution does not account for an increasing rate function. This could also invalidate the use of the plotting technique even though the rates start on a decreasing trend. With this in mind, it is recommended that the Winestock and Colpitts Method should not be used for data with wellbore storage distortion except to estimate permeability. Superposition Method Figure 16 is a plot of the dimensionless pressure, pj, versus the logarithm of the dimensionless time plotting function, Xj), where both 55TABLE 6 ERROR ANALYSIS FOR WINESTOCK AND COLPITTS ‘METHOD: WELLBORE STORAGE CASES, SANDFACE RATES Dimensionless Dimensionless Dimensionless Dimensionless Slope Error Intercept Error (a) Case ——Slope___ ___Intercept __ ——_(#) ___ Ideal. 1.181293 0.404539 --- --- Constant Rate 1.204586 ~0.020987 4.612 -105.2 Constant: Pressure 1.133803 0.584396 1.519 44.45 Random Rate 1.356819 0.621984 17.85 253.7 Log Rate 1 1.230825 -0.083800 6.908 108.4 Log Rate 2 1.204986 0.202141 4.664 -80.03 Log Rate 3 1.193084 0.204435 3.630 ~49.47uo132un4 64133014 UoLyLsoduadns ssazuoLsuauig 40 uorzouny & Se soueWiOJAdg aUNssaig SSB|uoLsuaUIG :59zey BDeJpURs ‘sase) 621035 B40G1 [aM UMOPMBUG MOL4 JUBLSUBAL - gt aunbLy ary ort «OE gO OT) gto, OT OTS T 1 1 L 1 1 | L | . 59se Liv C orovariables were calculated using saniface flowrates. Note that all of ‘the data fall on the same trend with the exception of some of the random rate data which lies very near the given trend. Also note that, Fig. 18 is exactly the same as Fig. 12, which is to be expected since each set of data used sandface flowrates in the caloulation of its respective superposition plotting function. The results and error analysis derived from Fig. 16 are given in Table 7. Again, the lovest overall error was for the constant pressure case since it is not affected by wellbore storage. For the cases that were affected by wellbore storage, the minimum absolute slope error was for the constant, rate case, and the maximum was for the random rate case, with no errors greater than 3% and with most less than 0.3%. The minimum absolute intercept error was for the constant rate case, and the maximum was for ‘the random rate case, with most errors less than 5%. The superposition method is the most rigorous method that can be used for variable-rate well testing. However, the sandface rate data may not be available. The use of surface rates (which usually are available) does reduce the accuracy of this technique, but as shown in the previous section, their use can still give excellent results. Therefore, it is recommenied that the superposition method be used for all variable-rate cases instead of the Winestock and Colpitts Method. This guarantees the most accurate results and eliminates the discrimination as to when the Winestock and Colpitts Method should be used.Case ‘Ideal. Constant Rate Constant Pressure Random Rate Log Rate Log Rate Log Rate 59 TABLE 7 ERROR ANALYSIS FOR SUPERPOSITION METHOD: WELLBORE STORAGE CASES, SANDFACE RATES Dimensionless Dimensionless Dimensionless Dimensionless Slope Error Intercept Error Slope ___ ___Intercept _ ____(%) ____. ____(®) __ 1.151293 0.404539 --- --- 1.181291 0.399479 -0.0001844 1.251 1.150989 0.405284 -0.03071 0.3103 1.188721 0.256525 2.990 86.59 1 1.183685 0.386079 0.2078 4.563 2 1.182206 0.396797 0.07954, -1.914 3 1.154520 0. 382864 0.2803 8.363SUMMARY The following points sumarize the major conclusions derived from this chapter: 1. The constant rate drawiom plot of pj» versus log t should only be applied to the constant rate case. No variable-rate case simlated could be analyzed using this technique. 2. Winestock and Colpitts Method is valid by an analytical definition for the constant pressure case. Though this may seem a trivial fact, the existence of this case as a bound of applicability does set a noteworthy precedent. 3. The Winestock and Colpitts Method can be applied to wellbore storage dominated data, regardless of whether surface or sandface flowrates are used to calculated Ap/g. However, care should be exercized in choosing the correct straight-line portion of the data. 4. Surface flowrates can be used to calculate the superposition plotting function with very little effect on the derived results. The use of surface flowrates in this calculation only significantly affects the wellbore storage dominated portion of the data. This result suggests that the added expense of obtaining saniface rate data may not be necessary in most situations.él GHAPTER IV ‘TRANSIENT FLOW INVESTIGATIONS ~ BUILDUP CASE In this investigation the derivation and application of the Horner constant rate solution, Odeh and Selig variable-rate approximation, Newton Iteration variable-rate approximation and the superposition solution will be studied. The effects of wellbore storage and buildup test duration will be included in this study. DERIVATIONS utd Horner developed the following equation to model the pressure buildup following a constant rate pressure drawiown for a well in an infinite reservoir. ee coe) - HAGE! oy Ge? where At, is the dimensionless tine after shut-in. For t,»100, Bq. 82 reduces to> Pancact grat) )-y ~ InG@at,)+ y) nal ‘po"ty’ ay t 2 yy mht at oy tee, eee e eee eee nee eee renee ene ee (83) Expressing Eq. 53 in field units gives t, PyrPyg 7 162.6 BH tog). boeeseeneeees 6M) Expressing Eq. 54 in terms of the shut-in bottomhole pressure, p,., gives t, Pag 7 Py ~ 162-6 TE og Mate) eee (65) Equation 55 suggests that a plot of p,. versus log (Gy, + At) /at) will give a straight line of the following form + + at aCe) Pys ~ Py ~ mptos where63 a, - Bu 2. (87) Solving Eq. 57 for the formation permeability, k, gives x= 102.6 SF BeeEo cesses (58) ‘The solution for the skin factor was derived by van Everdingen”? and is @ combination of Eqns. 17, 54 and 57. This solution is given as Pigs) ar Pye At-O) s = 2.151 (wedbr wf _ jog | neg, t, 3.228 + log (PAA) ‘ph where P,,)nr 18 the pressure at At-1 hour from the straight-line portion of the Horner graph. Also, the log(t,+ 1)/ty, is usually assumed to be negligible These relations are illustrated graphically in Fig. 19. Figure 20 shows the pressure distribution as a function of the logarithm of radius for this case. Equation 23 was used to calculate the radii of investigation.dnp ting MoLJ qUBLSUeAL 87eY 3UE}SUO] 4oJ SLSALeUY [RUOLIUaAUOD - GT aunby4 Ww we ot got ot cor zot ot I Lu mn tu ut ut L 0008 [~ o0se L E L_ ooo += 00s [ epsd «SMaumopmeso MoL4 quatsuRdl a3ey 3Ue3SUO] eB BuLMoltod dnpLing MOL4 qUBLSURAL e 403 UOLINGLAZSIG auNSsatg 4LoAsasaY a3 Ssnipey = 02 aunbi4 3 elpel4 as ezesta ay c-162="3 (¥) 4y 9799-83 (e) ay ere=?3 (2) ay cets's (1) o08t oooz ooze O02 0092 eisd ‘aunssaagTt is necessary to note "Horner's Approximation" for variable- rate transient flow preceding the pressure buildup. This approximation is wirical, but it does give good results provided a long buildup is run, Horner's approximation is expressed as Ph Nest where q,,,; iS the last flowrate in STB/D, and @ is the cumllative produced volume in STB. The q,,., term is substituted for q,term in Eqns. 58 and 59 for purposes of analysis. Odeh and Selig modified By. 65 so that a constant-rate drawiown analog could be approximated for a variable-rate drawdown. This modification yields ve. CD where - (62)Gee feeteeteettsssseeeeeeseres (68) ‘The interpretation of this method is exactly the same as that for the Horner Method, therefore Eqns. 58 and 59 are used to estimate reservoir permeability and skin factor, respectively. This case is illustrated graphically for a linear logarithmic rate decline in Fig. 21. Note that only the very latest data fall on ‘the correct trend. This is because the method forces the pseudo-Horner data to "match" the last data point explicitly in the derivation Therefore, the earlier data will not lie on the correct trend. this nethod usually gives good results, but only the last buildup date should be considered for analysis. Newton Iteration Variable-Rate Approximation for Horner Method The Newton Iteration Method is derived in Appendix A, and its application is exactly the same as the Odeh and Selig Method. The Newton Iteration Method solves the time ani rate roots that force the Horner equation and the superposition solution to be equal at a single point. Tne logical point to “force match" is the last buildup data point since it is least likely to be affected by wellbore storage. However, the formulation is general, and any point can be used for a ergi ot uMopmesg Mol 4 quaISUEAL 23ey AULT 607 toy Sisfyeuy dnpiing mols quatsues. BE Las pue yapg - Tz aanBiy Or or u g poritiiritiis 3 Ot Ww we got poritioy el Oi 01 or 2 prritiiiitis rl 009 osay ory osey 000s esd Mdmatch. The advantage of this method over the Qich ani Selig Method is that it guarantees a more accurate solution because it is matched on ‘tthe superposition solution. The Qieh and Selig Method, on the other hand, is matched on the instantaneous point source solution. Since the Newton Iteration Method is applied exactly the same as the Odeb and Selig Method, Eq. 61 models the pressure behavior, and Eqns. 68 ani 69 are used for analysis. The function to be minimized for this method is given as eee ate t, +. - + m+ At = “j-1. F(t) = dy ME) ayaa at ty (ea) Since Newton interation is to be used, a closed form derivative of Bq. 63 is needed. This derivative is given as ‘This case is illustrated graphically for a linear logarithmic rate decline in Fig. 22. Note that as with the Qich ani Selig Method, only the last buildup data fall on the correct trend. Again, this is because the Newton Iteration Method was force matched on the last buildup data point. This method does give consistently better results70 UMopMeig Mo(4 quassueAL 93ey APOULT 607 40g sysfyeuy dmpiing mols 4UdtsuBAL UoLaesea] UoaMON - 22 ounbId i, wey 7 1 got Mi 03 or got rorrlriiitiiiitiiis rriitiiiit oosy [_ osov [7 oop b+ oszv ost eisd Manm than the Odeh and Selig Method, but it is also more complicated to evaluate. However, once the Newton Iteration Method has been programmed on a computer, the evaluation of the roots becomes trivial. ‘Superposition Solution The superposition solution for a buildup following a variable— rate drawiow was given by Earlougher® and Lee? and is reproduced in Appendix A. This solution is given as “yg 7 Py ~ 162.6 ah E a seat est .. . (66) where the summation term in Eq. 66 can also be expressed as %, 4 a ros A=) @ Combining Bmns. 66 and 67 gives Pye = ~ 262.6 x, - (68) Equation 68 suggests that a plot of p,, versus X will give a straight Line of the following form(69) where myo 168.6 Be see e esses CO) Solving Eq. 70 for the formation permeability, k, gives Gee mgt 162.6 sereees CL) ‘The solution for the skin factor, s, is a combination of Eqns. 41, 66 and 70. This solution is given as Pasdanr = Pug St-0) rast f = ~ 208 Gate? * 8-228) bene These relations are illustrated graphically in Fig. 23 for a linear logarithmic rate decline. Figure 24 shows the pressure distribution as a function of the logarithm of radius for this case. Bq. 23 was used to calculate the radii of investigation.78 UMopmesg MOL4 qUaLsUgu, a3ey UR9ULT 607 405 susfteuy dnpitng MoL4 juarsued UOL}1Soduadns - ez aunty Sx got “ ot vt a ot a ai 1 positirirtirirtisiitiriitisiitisiitisiit 9 0999 2 oy + ose ogy4 umopmesg Mold quaLsued agey 4eaUL] 607 e BuLMoL Loy dnpLing MOL4 JUBLSURAL B 4OJ UOLINGLAISIG BUNSSaug 4ALOAUaSEY - pZ aunBLy as ‘snipe got got oT T 7-01 wip ta aim pi over 3 2'2ee @ a3 peas" L~ oor (2) as gers" i" ©) LL oye 4 Elpsta Ovi () F ay T62=! es (») 1 orp au 979283 (£) ay ever=?a (2) r ay cet=4 (1) ose d Pisd ‘aunssau,It is also worth noting that @ conventional semi-log plot can be nade using the superposition plotting function if the following modification is mate Therefore, a plot of p,, versus log X} will give exactly the same relations as Bgns. 70 to ¥2 because X, - log X;. This convention is used so that the superposition plotting function can be plotted on the same graph as the Horner and pseudo-Horner data. NUMERICAL INVESTIGATIONS - BASE CASES (1O WELLBORE STORAGE) In this section the finite-difference numerical simulator described in Appendix C will be used to investigate the effects of variable-rate transient flow on the subsequent transient pressure puildup. The sane six rate cases used in the preceding chapter will be used for the drawdown, and the buildup duration will be either 10 or 1000 days. These resulting pressure buildup data will be analyzed using the methods previously described in this chapter. The six base cases used in the previous chapter were simlated with a 1000-day drawiown and either a 10-or 1000-day buildup. This was done to investigate the effect of buildup time on the Horner approximation, Odeh and Selig, and Newton iteration constant rateanalog methods. Odeh and Selig® state in their work that the buildup should be 1.5 times the drawdown for the assumptions made in their derivation to be accurate. This is an unreasonable constraint for most’ cases with the possible exception of drillstem tests or production tests early in the life of the well. Therefore, it would be useful to know how the applicability of this method is affected for gross violations of the Odeh and Selig criteria, hence, the 10-day buildup. For comparison and convenience, all plots the following sections are formilated in terms of the dimensionless pressure, pp. Horner Method Figure 25 is the dimensionless Horner plot for the 10-day buildup case. Note that all the cases have a similar slope trend, but only the constant rate case has the correct zero intercept. ‘he results and error analysis for Fig. 25 are given in Table 8. The minim absolute slope error was for the constant rate case and the maximum was for the randon rate case, with no error greater than 2% and most less than 0.2%. The minimum absolute intercept error was for the constant rate case and the maximum was for the log rate #3 case, with most errors less than 10%. Figure 26 is the dimensionless Horner plot for the 1000-day buildup case. Now the data focus on zero as an intercept indicating that the Horner approximation does approach the correct solution for a sufficiently long buildup. The results and error analysis for Fig. 26 are given in Table 9. The minimum absolute slopeor sisfeuy dauioy *sXeg OT = BULL dnpying ‘aouewiojag aunssaud SSa[uosuawig :9Se) aseg dnpiing MOLJ quaLsues, - 52 aunbLy Ww ug, aw My or OT OT oT ot or ot L iE L (v)a*(€) (2) (UZ # ayy 607 (9) 2# ayey 607 (5) T# 970y 607 (9) 97ey wopuey (¢) aunssaig 3ueqsuoy (z) yey quezsuo9 (T) or8 TABLE 8 ERROR ANALYSIS FOR THE HORNER METHOD: BASE CASES, BUILDUP TIME - 10 DAYS Dimensionless Dimensionless Dimensionless Dimensionless Slope Error Intercept Error Cease = _ Slope __Interoapt_ —)___ Ideal 1.151298 (0.000000 --- --- Constant Rate 1.151238 0.000187 0.008777 0.0187 Constant Pressure 1.151245 0.018237 0.004169 1.824 Random Rate 1.170842 0.098445 1.663260 -9.845 Log Rate 1 1.151335 0.064919 0.003640 6.492, Log Rate 2 1.151486 0.203323 0.015027 22.33 Log Rate 5 1.152009 1.1asa74, 0.062191 114.379 siskieuy 4ausoH #8420 OOOT = aul L dnpiing ‘aouewojuag aunssaug SSa[UOLsUaWIG :saSe) aseg dnping MOL4 juaLsueu] - 92 a4nBLy Ww aw + Wy got gO gO got, L { L i L (p)a*(e) (2) (I €# 832y 507 28 83ey 607 T# 9yey 607 yey wopuey aunssoug uUe3Su0) yey quezsuo9 (s) (9). ( ( i ( ( ( 9) §) Y € z 1 ) ) ) ) ) ) ot80 TABLE 9 ERROR ANALYSIS FOR THE HORNER METHOD: BASE CASES, BUILDUP TIME = 1000 DAYS Dimensionless Dimensionless Dimensionless Dimensionless Slope Error Intercept Error Case = __ Slope _Interospt — __(™)_ =H) Ideal 2.161298 0.000000 --- --- Constant, Rate 1.151238 0.000187 0.008777 0.0157 Constant, Pressure 1.151245 0.018237 0.004169 1.824 Randon Rate 2.160673 0.052893 0.8147 5.289 Log Rate 1 1.151335 0.064919 0.008640 6.492 Log Rate 2. 1.175321 0.025368 2.087 2.557 Log Rate 5 1.655762 0.176539 43.64 17.65error wes for the constant rate case and the maximum was for the log rate #3 case, with most errors less than 1%. The minimum absolute intercept error was for the constant rate case, and the maximum was for ‘the log rate #5 case, with most errors less than 6%. A it Figure 27 is the dimensionless Horner plot using the Qdeh and Selig constant rate analog approximation for the 10-day buildup case. This case was designed to be a gross violation of the Ouch and Selig criteria that the buildup time mst be at least 1.5 times the drawdown time. The results and error analysis presented in Table 10 show that only the constant rate case gives the correct trend for this case. the minimum absolute slope error was for the constant rate case, and the maximum was for the log rate #3 case, with all errors less than 52% and most less than 15%. ‘The minimum absolute intercept error was for the constant rate case, and the maximum was for the log rate #3 case, with all errors less than 57% and most less than 20%. Figure 28 is the dimensionless Horner plot using the Odeh and Selig constant rate analog approximation for the 1000-day buildup case. Though this case does violate the Odeh and Selig criteria, the results should not be significantly affected. All of the cases tend to "beni" towards the correct trend for (t,,, + At)/At less than 10, This proves that for a sufficiel tly long buildup, the Qieh ani Selig Method does force match the instantaneous point source and Horner solutions for the last 8182 susheuy 5119s pue yapo ‘skeq OT = aut dnpying ‘aouewiogag aunssoag SSaluoLsuauEG :S9Se] aSeq dnpi ing MOLs auaLsUAl - {2 a4nBL4 wW we cin ot ot got pot cot OL i i” i Ll 1 L 0 Lz A oe (9y e# ayey 607 (9) |_ 9 O 2# arey 607 (g) fs 1 arey 607 (pb) (x) a92y wopuey (cE) | 9 (h aunssauq uezsuoD (2) fe) ayey queqsuo) (1) or83 TABLE 10 ERROR ANALYSIS FOR THE ODEH AND SELIG APPROXIMATION METHOD: BASE CASES, BUILDUP TIME - 10 DAYS Dimensionless Dimensionless Dimensionless Dimensionless Slope Error Intercept Error Case = _Slope __Interoopt = _(m) Ideal 1.161293 (0.000000 --- --- Constant Rate 1.151216 0.coo197 0.00688 0.0197 Constant Pressure 1.122201 0.0aser7 -2.527 2.548 Random Rate 1.299909 -0.198514, 12.91 -19.83 Log Rate 1 1.068308 0.077186 “7.200 776 Log Rate 2 0.901662 0.225706 -21.67 22.57 Log Rate 3 0.560402 0.564383 -51.32 56.44pue yapo ‘Skea Oot sisfieuy Bias aut, dnpiing ‘aouewojiag aanssaug Ssaluolsuaujg :sase) aseg dnpjing molj quatsues, - gz eunby WwW ws got pot or OT or ot 01 or 1 ferereslayss [pepe ers eres renner ercrenrelererenend Ep) ef a3ey 607 (9) 2# eyey 607 (s) T# erey 607 (bp) F 23ey wopuey (£) aunssaug uezsu0) (2) aqey quegsuoy (1) F Otbuildup data point. The results and error analysis for this case are given in Table 11. The minimum absolute slope error was for the constant rate case and the maximm was for the log rate #3 case, with all errors less 7% and most less than 0.9%. The minimum absolute intercept error was for the constant rate case, and the maximum was for the log rate #5 case, with all errors less than 2% and most less than 0.3%. The important conclusions drawn in this section are that the Qieh and Selig Method does give fair results for cases where the buildup time is significantly less than the drawdown time and that only the last buildup data should be used for analysis. Neyton Iteration Approximation Method Figure 29 is the dimensionless Horner plot using the Newton Iteration Method constant rate analog approximation for the 10-day buildup case. All of the curves beni tovards the correct trend. This indicates that this method may not be as sensitive to buildup time as the Odeh and Selig Method. The results and error analysis for these cases are given in Table 12. The minimum absolute slope error was for ‘the constant rate case, and the maximm was for the log rate #3 case, with all errors less than 27% and most less than 6.4%. The minimum absolute intercept error was for the constant rate case, and the 85TABLE 11 ERROR ANALYSIS FOR THE ODEH AND SELIG APPROKIMATION METHOD: BASE CASES, BUILDUP TIME ~ 1000 DAYS Dimensionless Dimensionless Dimensionless Dimensionless Slope Error Intercept Error Case Slope Intercept, (%) —a)___ Ideal 1.151295, 0.000000 --- - Constant Rate 1.150974, 0.000182 0.02771 0.1082 Constant Pressure 1.147212 0.001796 0.3545 0.1796 Random Rate 1.184869 0.000279 0.2759 0.0279 Log Rate 1 1.141548 0.008011 0.8464 0.3011 Log Rate 2 1.115740 0.009059 3.088 0.9059 Log Rate 3 1.072451 0.020113 6.848 2.0110 1 sishyeuy uo1zeu93y uogman *sheq OT = amy dnpiing ‘aoueWtosiad ainssad Ssauorsuawig :sase9 aseg dnpling wol4 guatsued] - 62 aunbL4 w ne po a got a cot ot or 1 1 1 \ 1 ee a7ey 607, 2# 23ey 607 TF 232 607 yey wopuey aunssaug ua suo) yey URI SUD OL‘TABLE 12 ERROR ANALYSIS FOR THE NEWION ITERATION APPROXIMATION METHOD: BASE CASES, BUILDUP TIME = 10 DAYS Dimensionless Case = _Slope Ideal. 1.151293 Constant Rate 1.181216 Constant Pressure 1.145175 Randon Rate 1.224347 Log Rate 1 1.124360 Log Rate 2 1.064297 Log Rate 5 0. 849925 Dimensionless Intercept, 0.000000 0.000197 o.o171e4 ~0.156241 0.087243 0.192638 0. 708834 Dimensionless Dimensionless Slope Error Intercept Error eee eee eee 0.006688 0.0197 0.7051 1712 6.345 -15.62 2.339 5.724 7.565 19.26 26.18 70.83maximum was for the log rate #3 case, with all errors less than 71% and most less than 15.7%. Figure 30 is the dimensionless Horner plot using the Newton Iteration Method constant rate analog approximation for the 1000-day buildup case. Again, all of the curves bend towards the correct trend. However, since the Newton Iteration Method solves for the constant rate analog approximation for the last buildup data point, the t,,, for the 10- and 1000-day buildup cases are different. This causes Fig. 30 to be slightly displaced from Fig. 29, which means that the curves do not overlay as did Pigs. 27 and 28 for the Qich and Selig Method. The results and error analysis for Fig. 30 are given in Table 13. ‘The minimm absolute slope error was for the constant rate case, and ‘the maximum was for the log rate #3 case, with no errors greater than 2.9% with most less than 0.35%. The minimm absolute intercept error was for the constant rate case, and the maxim was for the log rate #3 case, with no errors greater than 1.4% and most less than 0.22%. The results for the Newton Iteration Method are considerably more accurate ‘than those obtained using the Qich and Selig Method. This advocates the use of the Newton Iteration Method to generate a constant rate analog for the variable-rate case. However, since the number of calculations for applying this method is many times that for the Odeh and Selig Method, it is suggested that the Newton Iteration Method be used only if it is programmed on a computer. 89StsA,euy uotze4aa1 uoaMaN ‘sAeG OOOT = awtL dnpLing ‘aoueWi0siag dunssadd SS9[UOLSUaUIG :5952) aseq dnpling MOL4 quaLsuesl - Of 24nbLy Ww ye aye? pot «= OF gt OT og OF zt piritiriitisiil Lovirtoiiitis (9) e# a7ey 507 2# aqey 607 T¥ a3ey 607 yey wopuey aunssaug Ue3Su09 aqey ueysu091 TABLE 15 ERROR ANALYSIS FOR THE NEWION ITERATION APPROXIMATION METHOD: BASE GASES, BUILDUP TIME = 1000 DAYS Dimensionless Dimensionless Dimensionless Dimensionless © Slope Error Intercept Error Case __Slope __ ___Intteroept (as) _%) Ideal 1.151295 (0.000000 --- --- Constant Rate 1.150974 0.001082 0.0877 0.1088 Constant Pressure 1.149750 0.001443 0.154 0.1448 Random Rate 1.152586 0.000543 0.1097 0.0543 Log Rate 1 1.147548 0.002175 0.3258 0.2175 Log Rate 2. 1.157418 0.005968 1.205 0.5968 Log Rate 3 1.118076 0.013851 -2.885 1.385‘Superposition Method Figure 31 is the plot of the dimensionless pressure, p,, versus ‘the logarithm of the dimensionless buildup tine plotting function, Xj, for the 10-day buildup case. Note that all of the data fall on the correct trend except for a few of the random rate data points. The results and error analysis for Fig. 51 are given in Table 14. The and the minimun absolute slope error was for the constant rate cas maximum was for the log rate #3 case, with no errors greater than 1.2% and most less than 0.26%. ‘The minimum absolute intercept error was also for the constant rate case, and the maximum was also for the log rate #3 case, with no errors greater than 1.75% and most less than 0.19%. Figure 32 is the dimensionless pressure, p, versus the logarithn of the dimensionless time plotting function, X,, for the 1000-day baildup case. Again, all of the data fall on the correct trend. The results and error analysis for Fig. 32 are given in Table 15. The minimum absolute slope and interoept errors were for the constant rate case, and the maximum were for the log rate #3 case. All of the absolute slope errors were less than 0.15% with most less than 0.036%, and all of the absolute intercept errors were less than 0.15% with most less than 0.11%.93 S{sf|euy uos3|sodaadns ‘shea of = auts dnpiing ‘aouewiossag aunsseug ssajuojsuawig :sese9 aseg dnpjing molj quatsueds - Te aunbiy Mx got OT gt OTT got 0 or 1 . i L 1 L 1 L Sase) Liv oOcase Constant Rate Constant Pressure Rate Log Rate Log Rate Log Rate ‘TABLE 14 ‘ERROR ANALYSIS FOR THE SUPERPOSITION METHOD: BASE CASES, BUILDUP TIME - 10 DAYS Dimensionless Dimensionless Dimensionless Dimensionless Slope Error Intercept Error Slope ——Intercept_ __() 8) 1.181298 0.000000 moe 1.151216 0.000197 0.006688 o.o1g7 1.151510 0.000240 0.01885 0.0240 1.aaavia 0.001218 -0.8714 0.1218 1 1.182486 -0.000260 0.09928 0.0260 2 1.154210 0.000617 0.2554 -0.0617 3 1.164141 0.017502 1.16 1.750siskyeuy uolatsodiadns ‘seq OOOT = aut dnpying ‘souewosiag aunssoag SS9|uoHsuauiLg :S9Se) Seq dnpling MOL4 qUa}sUeAL - Ze a4nBLy sasey Ly otTABLE 15 ERROR ANALYSIS FOR THE SUPERPOSITION METHOD: BASE CASES, BUILDUP TIME ~ 1000 DAYS: Dimensionless Dimensionless Dimensionless Dimensionless Slope Error Intercept Error Case ——Slope___ ___Interoept_-- ___(®) _____ ____(®) __ Ideal 1.151293 0.000000 --- --- Constant Rate 1.160974, 0.001032 0.02771 0.1082 Constant Pressure 1.151120 0.001088 -0.01503 0.1048, Random Rate 1.150864 0.000969 -0.035525 0.0969 Log Rate 1 1.151288 0.001077 0.000454 0.1077 Log Rate 2 1.151998 0.001185 0.0612 0.1185 Log Rate 3 1.152950 0.001505 0.1459 0.1505‘The results for the Superposition Method show that buildup time is not as critical a factor as it was for the other methods. This is to be expected since the Superposition Method accounts exactly for the variable-rate drawlown regardless of the buildup time. Therefore, the Superposition Method is recommended, for accuracy, over the other approximate methods. ‘NUMERICAL INVESTIGATIONS — WELLBORE STORAGE CASES (SURFACE RATES) The six base cases were again similated using a 1000-day draiiown and either a 10-or 1000-day buildup. This time, however, a dimensionless wellbore storage coefficient, Cy, of 1000 was added to both the drawdown and buildup. In addition to studying the effect of buildup time, the effect of wellbore storage on the Horner Approximation Method, the Odeh and Selig and Newton iteration constant rate analog methods and the Superposition Method will be studied. Horner Method Figure 33 is the dimensionless Horner plot for the 10-day buildup case. Note that all the curves have a similar slope treni for (t+ At)/at<10°. The results and error analysis for Fig. 33 are given in Table 16. The minimum absolute slope error was for the constant rateStsfteuy sausoy *sfeq oF = auLL dnp, ing ‘aouewioyag ainssalg ssa[uoysuaug :sa7ey aoegang ‘sasey aBesoys auogi lam dnp|ing mos yUaLsuedl - ce aunBiy W ST we) eS mip) tga ee a vt > me 6 ie f These relationships are shown graphically in Fig. 53 for various fracture coniuctivities. Fracture conductivities greater than 100 are generally assuned as modeling the infinite fracture conductivity case. Note that only the very lowest fracture conductivity case (Cr-1) does not follow the correct trend. This will be discussed later. 17168 Suponuasay paunzoey A|[eo1q4eq ‘UMOpMEsG MOL quaysuedy. Ol X OT 380d @32Y 3UE}SUC] 40} SLSKLeUY [eUOLQUaAUOD - Eg aunbLy ay Samy aosz ove ‘o00r isd «Mg159 CONSTANT PRESSURE SOLUTIONS ut: ‘ivatiion ir The post-transient constant pressure solution was originally 33,54 derived by Russell and Prats’ for a bounded circular reservoir. This solution was later extenied by Ehlig-Economides and Ramey,*°+26 using the Dietz’© shape factor, to account for a bounded reservoir of any shape. ‘he general shape solution is given as + (90) vhere y is Buler’s constant and has a value of 0.877216, and C, is the dimensionless reservoir shape factor as defined by Dietz. Expressing Bg. 90 in field units gives kay Pye) = 0.005513 kt. . ER ( ee te I" 70.6 Ba oes $A —A A x, Or, Equation 91 suggests that a plot of log q versus t will give a straight Line of the following form160 log q = 10g dy, + 2-805 Dt ac) where Solving Eqns. 95 and 94 for the reservoir drainage area, A, gives 4, 2 oi A = — 0.10160 - (95) Dy they (Py Pye? Solving Bq. 9% for the reservoir shape factor, C,, gives iE eelarageva Sn mee 10 0 (96) C, v Deuce Solving Bq. 94 for the reservoir shape factor, C,, gives(97) HCP “Pye HAC; “Pyp) oa G56 ams ‘These relationships are shown graphically in Fig. 54 Solution Derivation - Vertically Fractured Reservoirs (Infinite Fracture Conductivity) ‘The post-transient constant pressure flow solution for the vertically fractured reservoir case is given by Blasingame and lee” as ~ ty ance eee (98) 2 where y is Buler’s constant and has a value of 0.877216, ami Cp is the dimensionless reservoir/fracture shape factor as defined by Gringarten.** Expressing By. 98 in field units gives EHC) Pye) T° 70.6 Be 348 Cae rap (= 0.005015 kt bey aa re ++ (99) 161162 S4LoAdasay SnosuaBowoH ‘umopmedg MOL} qUaLsUe41-7Soq auNssaid 24091 [3M 3UeISUOD 4O} SLSAL BUY [2UOLIUBAUOD - pg a4nBLY ay uy, ooosz ooo0z ooost 00001 000s 0 pore tara ir teria tira i tirig pr @ TTT ot Or a/ais *b163 Bquation 64 suggests that a plot of log q versus t will give a straight line of the following form Log G = 10g yyy + 2-808 Dygt veer where + (2) Solving Eqns. 101 ani 102 for the reservoir drainage area, A, gives tives ee) D, xe. cesses (208) ‘ye $80, (Pi Pye) = 0.10160 Solving By. 101 for the reservoir/fracture shape factor, C,, gives 2.246 A (04) LOO BT poonbose apace onnaoded DO A fSolving Bq. 102 for the reservoir/fracture shape factor, C,, gives 2.246 A eats ee C105) , 70.8 GiyeBe ExP(; i These relationships are shown graphically in Fig. 55 for various fracture conductivities. Fracture conductivities greater than 100 are generally assumed as modeling the infinite fracture conductivity case. Note that only the very lowest fracture conductivity case (Gr=1) does not follow the correct trend. This will be discussed later. VARIABLE RATE SOLUTIONS The general solution for variable-rate flow in a homogeneous bounded reservoir is derived in Appendix B. ‘his use of this solution 16 for any shape reservoir is made possible by the use of the Dietz’ shape factor. The general shape varieble-rate solution is given as 164165 satoasasay paungoeuy £1,291349, ‘uMOpMesG MOL qUa{sUe4|-7504 Bunssag a40q{ [2M 3ULISUO) 40 S{SK[euy [eUOLIUAAUO] - gg aunBLy ay fue gol X OT 80 a0 0 20 L 1 1 L L L oro TIT 201 G/ais ‘>where y is Buler’s constant and has a value of 0.577216, C, is the dimensionless reservoir shape factor, and ¢ is the superposition Plotting function which replaces t in the ty, term in the constant rate solution (Bq. 74). ‘The superposition plotting function, ¢, is defined (or) ‘Though this solution has only been derived and verified for the homogeneous reservoir case, the only rigorous constraint is that the reservoir is undergoing post-transient, radial or pseudoradial flow conditions. This suggests that the @ substitution (1.e., the superposition plotting function) for variable-rate flow conditions can be extended to other reservoir types, so long as the required flow condition is met. Expressing By. 106 in terms of field units gives PLP, ap . “ASE _ 20.6 Be an(—4-4 5 + 0.2880 52 oe (aos) q q %G,2,) fho,A 166Equation 108 suggests that a plot of Ap/q versus t will give a straight-line of the following form ap. an an + Mant - (109) where _B_ My 7 0-2539 Be, » G10) » Gu) qa) Solving Eqns. 110 and 111 for the reservoir shape factor, C,, gives (as) lerThese relationships are chown graphically in Fig. 86. Figure 87 shows the pressure distribution as a function of radius for the variable-rate case. Note that (ap/gr), is essentially constant after the pressure response has reached the outer boundary of the reservoir Recall that this assumption is implicit in Dietz’s derivation for the constant rate case. Since the constant rate solution (Bq. 76) and the variable-rate equation (Eq. 106) are exactly the seme, except for the t substitution in the variable-rate case, then it can be assumed the t substitution can be extended for other reservoir types. Though this development is not rigorous, it will be shown through simulated examples to be correct The limitations of the new variable-rate reservoir limits test method are the sane as for other well testing methods. Some of the violations of this method are vater influx, milti-thase flov, solution gas evolution and reservoir heterogeneities, if any of these violations exist, alone or in combination with others, then the method my give inaccurate results. Solution Derivation - Vertically Fractured Reservoirs (Infinite Fracture Conductivity) The variable-rate post-transient flow solution for the vertically fractured reservoir case is given as 168169 S4LoAdasay snosua6ouoH ‘uMopmeag MOL4 qUaLsuedL-350q BUNSsadd B4OGL [aM 3UeISUOD OJ SLSKLeUY UOLZLSOduadns - gg aunBLy 449 00008, 00009 0000¢ 00002 0 L ! 1 | . | . eT 02 Ov 09 08 oot G/au/tsd *b/dy170 Saponsasay snoauaGouoy ‘uMopmeig MoLj 3Ua}sue4J~3S0g ounssoug ‘2409 | La 3Ue3SUO) Bul.ing WoLgNgL43s1g aunssaig uLon4asay - LG BUNBLy 008 ag ‘snupey 009 00% 002 0 i . ree : o0sy ay setat=*s ay 6sus="3 ay e6ztely eisd ‘aunssaug1 wy - Ban 445 oMogxe + Babyy ee sees Gay where y is Euler's constant and has a value of 0.577216, and C, is the dimensionless reservoir/fracture shape factor as defined by Gringarten.**aiso = is given by Eq. 107 and is again substituted for t in the tp, term for the constant rate solution (Bq. 82). Expressing Bq. 114 in terms of field units gives PLP, ee ey eee eee fT A a 6 ee ‘Sox? 0. Sic, (ais) Equation 115 suggests that a plot of Ap/q versus = will give a straight line of the following form 4p . = @ 7 Purves * Bypyg® veers estes eesceeeeseees (116) where B Barve ~ 0-2589 5.0.4 seeeeees (117) qva Bu Dyrye ~ 70-6 BE an sees (218) Solving Bq. 117 for the ir drainage area, A, gives a- 0.2850 5B seeeenes (119) cf Meee Solving Eqns. 117 and 116 for the reservoir/fracture shape factor, Cy, gives Sis Se 70.6 Bu £ These relationships are shown graphically in Fig. 88 for various fracture conductivities. Fracture conductivities greater than 100 are generally assuned as modeling the infinite fracture conductivity case. Note that again, as with the constant rate and constant pressure cases, only the very lowest fracture conductivity case (Cr=1) does not follow the correct trend. NUMERICAL INVESTIGATIONS In this section the finite-difference numerical simulator ave