Fracture Evaluation With Pressure Transient Testing in Low-Permeability Gas Reservoirs W.J. Lee, SPE, Teas A&M U. S.A. Holditch, SPE, Texas A&M U. Summary ‘This paper presents theoretical and practical aspects of methods used to determine formation per~ meability fracture length, and fracture conductivity in low-permeability, hydraulically fractured gas reservoirs. Methods examined include Horner ‘analysis, linear flow analysis, type curves, and finite- difference reservoir simulators. Introduction ‘The purpose of this paper is to summarize the theoretical background of methods that we have attempted to use to determine formation per- ‘meability, fracture length, and fracture conductivity in low-permeability, hydraulically fractured gas reservoirs. This summary is intended to emphasize the major strengths and weaknesses of the methods studied, These characteristics have not always been ‘emphasized in the original literature and, in some cases, have remained obscure to the practicing engineer. The paper also includes examples from 13. wells in which postfracture-treatment pressure buildup surveys have been analyzed in detail. ‘Test analysis methods discussed in the paper in- clude (1) a method applicable only after a pseudoradial flow pattern is developed in the reservoir, (2) a method applicable when linear flow dominates inthe reservoir, (3) published type curves, with emphasis on those that include finite- conductivity fractures, (4) 2 mo f flow techniques useful for fractures, and (5) use of finite-difference reservoir simulators ina history-matching mode. Pseudoradial Flow Russell and Truitt! pioneered application of methods ‘based on the assumption of pseudoradial flow in a ‘Sopriom 88 Sec of elem Engineer ot AME fractured reservoir for determination of formation permeability and fracture length. A working definition of pseudoradial flow is that sufficient time has elapsed in a buildup or drawdown test so that, bottomhole pressure (BHP) varies linearly with the logarithm of flow time (drawdown) or the Horner time group (fy + Af)/At (buildup), as expected for radial flow in &n unfractured reservoir. In an infinite-acting (unbounded) reservoir, the analysis technique is based on the use of skin factor, ‘can be caleulated from 151 [LBP tg +323], Oy and the observation that, for infinitely conductive vertical fractures, Ly = dye* +) Eqs. 1 and 2 can be combined to avoid the termediate step of calculating s: logit) -26]). In principle, we can plot buildup test data on a conventional Horner graph, determine the slope m, and thus estimate formation permeability (k = 162.6 dePagtalmh) and determin fracture bal-ength, , trom Eq. 3 (ee Fig. 1). In practice, there are tive serious problems wit his method: JOURNAL OF PETROLEUM TECHNOLOGYweeny Fig. 1 ~ Horner plot, pressure buildup test data, south Texas gas well 1, The time required to reach the required straight ine’ where the slope is related to formation perme- ability can be impractically long (months or years) in low-permeability gas reservoirs with long fractures, as demonstrated by Gringarten et ai.* and Cinco ef als 2. Implicit in the method infinite fracture conductivity, wi valid.’ 3. By the time the pressure transient has moved beyond the region of the reservoir influenced by the fractures, effects of the reservoir boundary already may have become important, preventing develop- ‘ment of the proper slope, Russell and Truitt developed a technique for ‘overcoming Limitation 3. While we have not found their proposal to be of direct value in dealing with low-permeability gas reservoirs, we have found that a related technique offers promise for estimating fracture length and formation permeability from limited-duration buildup test data. This technique involves a trial-and-error process of (1) determining, the maximum slope on a Horner plot; (2) estimating aan apparent formation permeability from this slope; @) calculating the ratio of true-to-apparent per- meability, ,/K,, from a theoretically derived correlation that” requires knowledge of fracture lengthy; (4) estimating fracture length using a square- root-of-time graph; and (5) iterating to convergence. Applicability of this method to finite-conductivity fractures has been demonstrated by Holditch et al.? In summary, the major limitation of the standard Russell and Truitt method (Eq. 3) is that it is rarely applicable to low-permeability, fractured gas reser~ voirs because of the long times required to establish the straight line where the slope is related to the for~ ‘mation permeability. the assumption of is not always Linear Flow Millheim and Cichowicz® showed that when linear SEPTEMBER 1981 flow into a fracture dominates (at earliest times) in a drawdown test, pressure/time behavior is modeled by 1.054 bee (Hala) s Pi-Pug iy @) For a buildup test, with linear flow continuing to time 1, + At, Pi~Pug =m (Nig Ft ~VBR). ese eeeeee Further, when f, >> at, © ‘When these equations adequately model reservoir behavior, a plot of BHP vs. a square-root-of-time funetion will result in a straight line with slope m’ related to fracture half-length and formation per meability: Ly Vk Prs~Pup =m’ Vai. @ Psa (ta) O86) Ge)” Fig. 2is an example graph for a buildup test. ‘The assumptions on which Eq. 7 is based limit its applicability in many eases. These limiting assump- tions include the following. 1. High (but not infinite) fracture conductivity so that fluid entry into the fracture isthe same per unit ‘cross-sectional area near the wellbore and near the tip of the fracture (uniform flux*). In many cases, fracture conductivity is simply not ths large. 2. An independent estimate of formation per- meabilty, k, must be available if we wish to estimate Ly. In principle, this is possible using a prefracture pressure buildup test; in practice, such a test may be tnavailable, may complicate the testing program, or may be virally impossible to obtain because’ of operational problems in unfractured, low: wmFig. 2 ~ Millhein-Cichowicz plot, pressure bulldup test data, south Texas gas well permeability gas wells, 3. Application of the method to earliest-time data, which may be dominated by linear flow, requires absence of wellbore storage distortion. "Unfortu- nately, wellbore storage can distort data for a signifi- cant time period in some low-permeability fractured gas wells. Type Curves Several type curves?” have potential application to analysis of transient tests in low-permeability, fractured gas reservoirs. Particularly important are Cinco er al.'s? curves for finite-conductivity frac- tureg, constant-rate case (Fig, 3) and Agarwal ef ai.’s® curves for finite-conductivity fractures, cconstant-bottomhole-pressure case. (Fig. 4 shows ilar curve we have generated for use in our research work.) Even these curves, however, are based on assumptions that are sometimes limiting, such as negligible wellbore storage distortion and infinite-acting reservoirs. To include these variables and finite-conductivity on a single type curve probably would be impractically complex, yet these ‘effects can be important and can be misinterpreted. In principle, if we place data in the position of best fit on the Cinco et al. or Agarwal et al. type curves, we can determine simultaneously (and uniquely) reservoir permeability, k, fracture half-length, Ly, and fracture conductivity, why. Unfortunately, the position of best fit is not always obvious; this can lead to a multpliciy of possible formation and fracture descriptions.® As Agarwal etai.® point out, the uniqueness problem is diminished considerably if we know formation permeability independently (Grom a prefracture buildup test, for example) but, as we noted earlier, a prefracture buildup may be most difficult to run’ or it may be too late when we recognize the need for the information. ‘Another problem with application of type curves m8 to buildup tests is that most type curves (including those of Cinco et al. and Agarwal ef al.) were developed for use with drawdown tests in which we plot log (p; — Py) ¥s. log 1. They are applicable to buildup tests if We plot log (Pies — Pys) ¥S. log AF ‘when and only when producing time, fp, before the ‘buildup testis large compared with maximum shut-in, time analyzed in the buildup test. Unfortunately, production and shut-in periods are often of com- parable magnitude in testing programs for low- permeability gas wells, making conventional type curves inapplicable unless the values of Ar are corrected for fp, as suggested by Agarwal!” Modified Millheim-Cichowicz Method We have found a mo cation ofthe square root-of. time plot, as suggested by Millheim and Cichowicz,® to be helpful in estimating fracture properties; we call this technique the modified M-C method. The method is based on a plot of dimensionless pressure Wa the square root of dimensionless time for finite- Conductivity vertical fractures. For large values of fracture conductivity (characterized by the dimen- Sionless_fractureconductviy parameter, Cr = rs ye pay aca function ot with an fneredp near 20, For lower values of C, (more pressure drop inthe fracture), 2 plot of P; ~ Py =. Yris nonlinear at earliest times. There is a later linear portion; however, the intecep far from zero. We Fave found thatthe intercepts related uniquely t0 dimensionless fracture conduct sy lstrates the character of this type of plot. ‘This figure js simply a plot ofthe data published by Cinco et a/,* for constant producing rate in the form Of pp vs-vip, where KM(D;—Pug) 1412 ag Beata JOURNAL OF PETROLEUM TECHNOLOGY Po sere ®“powyeutoW pelpow‘suojejodesyx9 uojbou se6Ur ob juss ~ O14 £ SEPTEMBER 1981. ad and pe 0.000264 kt a PO batalf Notice in Fig. 5 thatthe true linear portion of the early-time data occurs for 0.15 .0.34 the flow pattern is phasing into pseudcradial. Notice also tht there is a family of eqright lines” with ciffeent slopes and intercepts, the values of which depend on dimensionless fracture conductivity. “The dashed lins in Fig. $ are extrapolations of the lines inthe linear region. Fig. 6 presents these same Straight lines, and ital shows the solution proposed ‘by Millheim’ and Cichowicz® that was derived by fssuming constant ux into the fracture. Note that nly the uniform-fux solution has exactly 2er0i- {eroept and that the slopes of the curves for fixed values of G, (shown in Fig. 6) vary from case to case and also differ from the slope of the uniform-fhux Tine. Beeause we take this variable slope into account, fur intercepts. differ from those presented bY ‘Agarwal er al. "The modified M-C method proposed inthis paper user the solutions derived. for Tinite-conductivity fractures and, therefore, should be more applicable to field data than te original M-C method (uniform flux solution). However, an important point con- cerning the two methods is that at very early times Gig <0.) the constan-flux solution and th $neBitesconduetivity solution are essentially identi Cal, as Fig 7 ilustrates, Therefor, the standard M-C fechnigue should be accurate at early times when C; is grater than 100 1780 7 ~ Modified and original MCourves. “To use the modified M-C method, an independent estimate of formation permeability, kis required. Given this value of F, we then plot drawdown test data a8 Pp vt, Wor. fora buldup tes, pp = (Daeg ~ Pop) 181-2 dy Bgattg VS. VA, find the linear portion ofthe graph, exttapolateto-Vfp = 0, and Fecord the value ofthe intercept. The intercept fixes Gye which characterizes the intercept onthe feheralized graph (Fig. 6). Using the value of the Intercept and the slopes in Fg. 6, we can establish a match point by noting values of Vip (Fig. 6) and Vt from the linear portion ofthe actual test data plot for some selected value of pp. Fracture length is then sven by arnt) ty z ee Hale oe 1p 7 match point” In our experience, this method has proved to be the most accurate hand method for estimating fracture properties. The major limitation is the need for an independent permeability estimate ‘The problem of wellbore storage distortion also remains for this method. However, we have found that, in most cases, wellbore storage effects will diminish before reaching the desired linear portion of the square-root-of-time graph. Therefore, in wells Tong fractures that are shut in for a sufficient time, the modified M-C method can be used to obtain a good estimate of both fracture length and fracture conductivity. Simulator History Matching In our experience, simulator history matching using the model reported by Holditch and Morse* has proved to be the most general method for deter- JOURNAL OF PETROLEUM TECHNOLOGYTABLE — INPUT DATA, FOR BUILDUP TEST SIMULATION Original reservoir pressure, psi 4,000 Net gaspay. ft 50 Gas porosity, % in ‘Wellspacing, acres 0 Reservotr temperature, "F 200 Gas gravity 08 Constant tow rate, Met! 2000 Producing ti 35 Bur at shut 1.980 mining formation and fracture properties simultaneously. Uniqueness problems common to simulator history matching are minimized when we ty to match not only the data from a given test, but all test data obtained on the well and the long- production history of the well. We assume that the only unknowns are k, Ly, and why. Ifother rock and fluid properties are unknown,” the solution can become nonunique. ‘Simulator history matching is an_ excellent technique for handling simultaneous effects of wellbore storage, reservoir boundaries, and finite fracture conductivity. It also offers the prospect of taking into account possible effects of fracture and reservoir heterogeneity, although the chances of obtaining a unique reservoir description diminish when heterogeneities become important. To illustrate this concept, consider the data presented in Tables | and 2. The data in Table 1 were held constant for each computer run. All these data ccan be measured easily or can be estimated us Togs, cores, laboratory analysis, ete. The data in Table 2 represent four computer runs with different combinations of fracture length, fracture con- ductivity, and formation permeability. Each com- bination of parameters in Table 2 resulted in a lowing BHP after $5 days of 1,990 10 psi. Fig. 8 presents the computer-generate buildup ‘TABLE? - PERMEABILITIES AND FRACTURE PROPERTIES USED IN SIMULATIONS Fracture Fracture Formation Length Conductivity Permeability case (met) (rma) 1 ‘so 500 220 800 3 1000500 oorg 41000 cy ons ‘curves for these four cases. Notice that the four ‘cases, which had identical production characteristics, have completely different buildup characteristics. 1f ‘one analyzes the data in Fig. 8, it becomes apparent that (1) the early-time slope of a fractured well is dominated by fracture conductivity, (2) the late-time slope is controlled by formation permeability, and (3) the fracture length influences the path of the buildup ‘curve between the two end points, Py and. Cinco et al.> presented a graph of stabilized flux distribution along a vertical fracture for various fracture conductivities (Fig. 6 in Cinco ef al.’s paper). This graph shows that, as dimensionless fracture conductivity and fracture length are changed, flow geometry in the reservoir also changes. In radial flow cases, the effects of skin, permeability, etc. will not change the flow geometry, only the pressure distribution. The difference between the flow geometries of the two cases does affect the ‘uniqueness ofthe history match. ‘Our experience has led us to the conclusion that fone cannot history match a long-term pressure buildup survey from a fractured gas well unless the correct fracture length is input into the model. Referring to Cases 2 and 3 in Fig. 8, buildup data toa Horner time of approximately 10 should be measured to differentiate accurately between the 250- and the 1,000-fe fracture. For this example, a 6-day buildup Fig. 8 ~ Pressure bulldup plots gen SEPTEMBER 1981 ted using computer medel and data in Tables 1 and 2 ve“TABLE 9 ~ FORMATION PROPERTIES FOR EXAMPLE FRACTURED WELLS Reservoir Resewoit NetGas Gas Frac Formation Depth Pressure Temperature Pay Porosity Gradient wet i) tosi. ia} i wail) + Timestone 74500 0 ‘080 2 limestone 14,000 % 078 3 limestone 13,000 52 075 4 limestone 13,000 2 075. 5 limestone 13,000 0 075, 8 limestone 16,400 Ea 075 7 sandstone 11,500 6 082 8 Sendstone 11/500, 2 1.00 © Sandstone 10,466 2 090 30 Sandstone 13.480, 2 080 M1 Sandstone 12.250, es 0.80 12 sandstone 6,780, “6 08s 13 sandstone 8100, 190 2 080 would be required. For most cases, a 14-day buildup will provide sufficient data to obtain a unique match. By assuming that the finite-difference history match provides the correct answers, it is possible to compare the results from the’ various hand calculation methods and to compare the calculated fracture length with the design fracture length. The remainder of this paper presents a comparison of these different methods using field data that have bbeen analyzed during the past 2 years, Analysis of Field Data ‘The data from 13 low-permeability, hydraulically fractured gas wells have been included in this paper. ‘The pertinent reservoir properties for these examples are presented in Table 3. The data in Table 3 were obtained from log analyses, core analyses, and static pressure surveys. The fracture gradients were calculated from the fracture treatment data. ‘The first six wells were completed in massive, high- temperature, carbonate reservoirs in three different areas of Texas. The seven sandstone examples are from Texas, Louisiana, and Canada. In the analyses of these wells, the properties listed in Table 3 were held constant. No sensitivity studies were performed to determine the effects of changing the basic formation characteristics. We recognize that if the data in Table 3 were varied, the calculated results also would be altered. However, our ex: perience indicates that the basic well evaluation ‘would not be altered significantly by minor changes inreservoir data. Table 4 is a summary of the fracture treatment design data for the 13 wells. Included in these data are estimates of created fracture height and propped. fracture length. The values of created fracture height were estimated using the open-hole logs and, when available, postfracture production logs. The design Fracture lengths were calculated using the estimated fracture height and the actual treatment data. ‘Those experienced in fracture treatment design calculations recognize that the estimate of created fracture height is as much art as science. Therefore, considerable care was exercised in selecting the values presented in Table 4, We have treated each well consistently, and the design lengths are our best ‘estimates for each well. For 12 of the example wells, postfracture pressure buildup test data were analyzed. Well 6 was the only well for which pressure buildup test data were not available, This well was evaluated by history match- ing the postfracture production data. ‘The pressure buildup data were analyzed using the methods of Russell and Truit," Gringarten et a. Cinco ef ai.,? the modified M-C method, and finite- difference computer history matching. ‘The results of these calculations are presented in Table 5. To calculate the fracture length using the modified M-C technique, the permeability from the history match of each well was used to calculate the ‘TABLE4 ~ SUMMARY OF FRACTURE TREATMENTS Estimated Fracture Proppant Design Height Volume engin o Fracture Fi (a o 350 “paluralpolymer - B0iem “Yoo,00 210.000 ~ 400, 300 aynthetie polymer 110'bm 184000 196,000 700 420 “natural polymer —s0lbm 100000 135,000 640 218 natural polymer - 60ibm 1801000 280,000 900 30 aturalpalymer — e0tbm 30000 410.000 1,600, 300 ‘gelled 5% acid 5000 96,000 ‘500 100 synthetic polymer ~ t001bm 260,000 170.000 1.500 S0 Synthetipobmer = So.em 10000 110090 1.000 400 40000 1,200,000 1500, 300 naturalpolymer ~ elm $50,000 250.000 "700 250 aturalpolymer - 7O1bm 200,000 850,000 1,000, 200 180% qualty foam 190,000 230,000 800 200 naturalpolymer = dibm 100,000 $40,000 600 172 JOURNAL OF PETROLEUM TECHNOLOGYdimensionless pressures. ‘As discussed earlier, the results from the finite- difference history match are considered to be the best estimates of the in-situ reservoir and fracture parameters. Therefore, the results for each calculation in Table S were compared with the results, from the history match. It can be noied in Table $ that only a few of the wells were evaluated using type curves. The problem, also reported by others,® was our inability to obtain ‘unique match for these wells. Despite our lack of success in quantitative appli- cation of type curves to these wells, we still recommend a log-log graph of test data for every well. We have found these graphs to be very useful in ur inital, qualitative evaluations. The shape of the TABLES ~ SUMMARY OF BUILDUI log-log graph helps determine if wellbore storage and/or fractures are dominating the buildup test data, Further, we have found that the use of effective time!® and pseudotime!? significantly diminishes the uniqueness problem. Formation Permeability To optimize the economic development of a field, it is vital that in-situ gas permeability!" be known accurately. When formation gas permeability is known, in addition to gas porosity and net gas pay, the optimum fracture length and well spacing can be calculated. Prefracture pressure buildup tests can and should be run; however, we have found that the permeabilities determined ‘from these tests are sometimes misleading. Problems such as low initial P TEST ANALYSIS. RESULTS Formation Fractute Fracture Permeability Length Conductivity we ma) wo 7 bose 3 08 50 409.—_—sistory maten 0.0380 7% 50 moditleg MC 2 098 75 se —-RussellandTrutt 007 400 400—_—lstory mateh 0.005 427840 modiiea Mc 3 0058 27 = Russelland Truitt 0.066 cry =. Gringarten type cure 018 137 3900 Cinco eral type curve 04 150 307 istory mater 0.04 167-2087 modifled MC 4 0029 183 w= ——Russelland Trult ‘007 550227 Astor match 0.007 781572 modified ac 5 0089 183 Gringarten type curve 0.028 22 a71—_Cincoeral. ype cune 0.08, 200 © 540 —_—story maich 0.076 as = Russelland Truitt 0.05 245 3800 modified MC 6 002 400 400 _istory match of production 7 028 136 Russell and Truitt 028 184 1620 Ginco eta type curve 0.98, ore "548 moaiied MC. 008, 1900800. hstory match 8 0x26 25 = Ruscelland Tultt Oe 500 50 history mater 04 404 102 modifiea Mc 2 on 200 = Russell ana Truitt oot 180 5900 modifieg Mc oot 500 '500_istory mate 10 0139 1 Fussell and Truitt 0.08 5 35 Cincoetat. typecurve 0.08 500 20 history maten 0.04 618 15 mealfed Mc 1 008 50 = Russell and Truitt 03 150 148 Gincoeraz type cune ‘9.003 382 «803 modified MC 0.003 600 © 300——istory match 2 oor 100 c Russell ang Truitt 0.045 6 © Gringarten type curve 0.082 150 5 Gincoetal.typecurve 001 70 20 history match 8.001 32260 moaified MC 13 082 16 = ——Russelland Truitt or 400 280—_—istory match O01 521640 —modiflea MC SEPTEMBER 1981producing rates, wellbore storage, formation damage, and natural fractures tend to complicate the analysis of prefracture buildup tests in low- permeability reservoirs. Also, many of these Teservoirs are highly stratified; therefore, estimation of net pay ‘with the perforations i ficult to determine. Also, in most cases, the radius of investigation of a prefracture buildup testis quite ‘small and permeability estimated from the test may not be representative of the entire drainage area of the well. Despite their limitations, however, prefracture tests are of considerable value. If a zone is broken down before running the buildup test, itis probable that acceptable values of in-situ formation per- ‘meability will be obtained. Knowledge of formation ‘TABLE6 ~ COMPARISON OF CALCULATED FRACTURE LENGTHS Design Length mitre’ — muito Lmeltro* ay “th ca ‘ea 400 95 125 190 70 107 a 610 402 234 250 23 eu 730 38 133 163 200 a4 687 632 25 500 204 50 1009 1250 10 Tha 280 50 00 850, 125 os 1020 20 500 950 Average” 108 Ba az ‘average 2t 78 ee 78 ne Emprical corection curve for Horner analysis of pressure bulldup data in fractured reservoirs, permeability that is at least approximately correct, ‘can be quite useful in analyzing postfracture pressure Duildup test data, regardless of the analysis technique used. ‘The two preferred methods for determining permeability from a postfracture pressure buildup survey are the Russell and Truitt method and finite- ifference history matching. The Russell and Truitt estimate, however, must be corrected to reflect the ‘duration of the shut-in period. To analyze the calculated permeabilities reported. in Table 5, the results from the Russell and Truitt ‘method were compared with the results from the istory match. The data were correlated using the jimensionless maximum shut-in time for each well. (0.000264 k ar PH SatalF ‘The correlation is presented in Fig. 95 theoretical verification has been provided by further research, ‘The Russell and Truitt method is based on ob- serving the maximum slope on a Horner graph of the buildup data, However, as pointed out by Cinco ef ai.,3 a dimensionless shut-in time ranging from 210 5 required to reach the semilog. straight line Therefore, the correlation in Fig. 9 is required to estimate permeability for shorter test ‘The correlation in Fig. 9 contains several im- portant implications. If a well intercepts a short fracture, has a high formation permeability, or is shut in for a very long period, a Horner analysis of the buildup test data should’ provide an adequate estimate of formation permeability. On the other hand, for low-permeability wells with long hydraulic fractures, the permeability calculated from a Horner graph can be too optimistic by an order of Magnitude, especially if only 3 to 7 days of buildup data are obtained. Even if a well is shut in long enough to eliminate the linear flow period (.€., tpy, ‘= 0.125), the permeability calculated using a Hornet graph can be high by a factor of two. ay JOURNAL OF PETROLEUM TECHNOLOGYFracture Length Table 6 presents a comparison of fracture lengths calculated using the different analysis methods. All calculated lengths have been normalized with respect to the design length. Two averages have been presented for each technique. Average 1 includes the Tatios for every well. Average 2 was prepared by climinating the values for Wells 1, 3, and 5. These three wells obviously contained’ short fractures, probably caused by poor sand transport; they are discussed in more detail later. Therefore, Average 2 ig believed to be more representative of a typical treatment. “The fracture lengths calculated using the Russell and Truitt method averaged only 5% to 11% of the designed fracture lengths. The history-match fracture Tengths averaged about 68% of the designed lengths. ‘The fracture lengths calculated using the modified MC technique averaged about 79% of the design lengths. It must be remembered, however, that the fracture lengths calculated using the modified M-C method are based on the formation permeability as determined from the history match. ‘An analysis of the data in Table 6 leads to several important observations. If one assumes that the history match is correct, the average treatment only achieves about 70% of the design length. This ob- servation implies that (I) the actual fracture is wider ‘and shorter than predicted by the conventional ‘design equations, (2) sand transport in the fracture is not as efficient as expected, (3) fluid loss in the fracture is larger than predicted using conventional techniques, (4) the barriers that control fracture height are routinely underestimated, or (5) there is some other, less obvious, reason. ‘The important point is that for the examples inthis paper, ony 70% of the design length was achieved during the fracture ‘eatment. Three of the examples, Wells 1, 3, and 5, ap- parently do not contain long hydraulic fractures. In Tables | and 2, it can be seen that all three wells were completed in deep, massive, high-temperature limestone reservoirs. For these three wells, detailed analyses of the fracture treatment data indicated that two problems existed. First, good barriers to fracture growth were not obvious from the logs and second, the gel concentrations were low, which may have allowed the sand to settle below the gas pay during the treatment. ‘On the basis of these observations and similar ‘observations by various operators in these same areas, recent treatments have used higher gel con- centrations, smaller-mesh proppants, and density- ‘controlled treatments." Tt was apparent from the analysis of Wells 1, 3, and $ that long fractures were being created, and large volumes of proppant were ‘being pumped; but the fracture opposite the pay zone was not being propped. The emphasis on sand transport that resulted from the analyses of these three wells has helped to define a problem and, it is hoped, future wells in such reservoirs can be stimulated more effectively, SEPTEMBER 1981 TABLET ~ BUILDUP DATA FOR WELLS fp = 10648 minutes up) = 0.02609 B= 8800psi 7x 10-8 psi-* a = 1800MeID 541 DIM = son y= 08 5 = 008 328° at (ints) SESERESSENSER ESET SE 3 31900 4020 3 Example 1 — Short Fracture ‘To illustrate the analysis of the field data, three of, the wells have been selected for further discussion. Well 5 in Table 3 represents the case in which the correct reservoir and fracture parameters can be ‘calculated using any analysis technique. ‘The buildup data for Well $ are presented in Table 7, Figs. 10, 11, and 12 illustrate the type-curve, Horner, and square-root-of-time graphs for these data, The type-curve graph indicates the half slope that is expected from a fractured reservoir. In Table 5, it can be seen that use ofthe log-log graph resulted in calculated fracture lengths and formation per- ‘meabilities that correspond quite closely to the results from the history match. The Horner graph (Fig. 11) indicates that the correct straight line has been reached, The last few data points can be extrapolated. to the correct static reservoir pressure. The square- root-of-time graph for Well 5 (Fig. 12) is typical of a es1786 Fig. 10 ~ Pressure buildup data for Wel, logiog graph E . Fig. 11 ~ Pressure bulldup data for WellS, Horner graph with ult from history match, Fig. 12 — Pressure buildup data for Well 5, modified M-C graph using a permeability of 0.05 md. JOURNAL OF PETROLEUM TECHNOLOGY‘well that contains a short, highly conductive fracture. ‘The data form a straight line that can be extrapolated toa dimensionless pressure of approximately zero, Each analysis technique applied to this case led to ‘consistent results, Using the history-match values of 0.05 md, a 200-ft fracture, and 4,000 minutes of shut-in time, the dimensionless shut-in time was calculated to be about 2.6. Therefore, the reliability Of the Russell and Truitt analysis is confirmed for this well. Tn Table 4, it can be seen that the fracture treat- ment for Well $ was designed for 110 ft of created fracture height. The actual treatment consisted of 320,000 gal of 60-Ibm gel carrying 410,000 Ibm of propping agent. This size treatment should have resulted in a propped fracture of 1,500 ft. It becomes obvious that (1) the created fracture height was probably much greater than the 110 ft used in the design calculations, and/or (2) the 60-Ibm gel in the 325°F reservoir did not maintain enough viscosity to transport the proppant as required. These con- 2,198 6. Milbein, KK. and Cichowice, Lz “Testing and Analyzing LLow-Permenbity Fractured Gas Wels,” J. Pet. Tech (Feb, 1968) 193-198; Trans, AIME, 28, 7, Gneo, H. and Samaniego, F "Effect of Wellore Storage tnd Damage onthe Transent Pressure Behavior of Vertically Fractured Wels,” paper SPE 6732 presented atthe SPE 52nd ‘Annual Technical Conference and Exnbition, Denver, Ot 2,197 8, Azarwai, R.G., Carter, RD, and Pollock, C.B: “Evaluation find Prediction of Performance of Low-Permeabiiy Gas ‘Welle Stimulted by Massive Hydraulic Fracturing.” paper SPE 6638 presented at the SPE 52nd Annual echnical Conference and Exhitition, Denver, et 9-12, 1977. 9, Barker, Benjamin J. and Ramey, HJ 2 "Transient Flow to Finite Conductivity” Vercal Fractures,” paper SPE 7689 presented atthe SPE S8ré Annual Technieal Conference and Exhidion, Houston, ct. 13,1978 10, Agarwal, R.G:""A New Method to Account for Producing Time Elfess When Drandown Type Curves Are Used To Analyze Pressure Buildup and Other Test Data,” paper SPE 45289 presented a the SPE SSth Annoal Technical Conference land Exhibition, Dallas, Sept. 2-24 1980, 11, Holditshy S.A, Jennings, J.W., Nease, S.H., and Wyman, RAE: “ine Optimization of Well Spacing” and Fracture ‘Lenath in Low-Permesbity Gas Reserv,” paper SPE 796 presented atthe SPE sird Annual Technical Conference and Exhibition, Houston, Ot 13,1978. 12, Misaly MED.,Atiebery,"RD., Vendito, JJus_and Fredrickson, E.° "A. Fracturing’ Technique to) Minimize ‘Water Production.” paper SPE 7563 presented atthe SPE {31d Anuual Techtitl Conference and Exhibition, Heuston, oat. 13,1978. 13, Agarwal R.G.: “Real Gas Pseudo-Time’ ~ A New Function for Pressure Buildup Analysis of MHF Gas Wells,” paper SPE 4279 prevented a the SPE Sth Annual Technical Conference find Exhibition, Las Vegas, Sot. 23-26 1979 1791APPENDIX Summary of Analysis Equations Written in Terms of Pseudotime and Pseudopressure!? This Appendix contains a restatement of major equations from the body’of the paper, written in terms of pseudotime, /g, and pseudopressure, m(p).. Use of these variables can lead to signi provements in accuracy of pressure buildup test analysis, particularly when using type curves. However, use of computer programs to calculate m(p) and t, is required for these variables. Eq.1 m{p) —mi kay, 1s 22 = (up) on Be 43.28 z srog(f te)... (any Fad _ $093 o,7an (fe) Vela 6 mPa) = 21) secre (AD) £q.8 Khlea (pj) =m (Prep) ae (3) p= | 422 dT on £q.9 0.000264 kt as 1 ore SI Metric Conversion Factors acre x 4.046 873 E+03 = m? bbl x 1.589873 E-Ol = m? ep x 1.0" E-03 cuft x 2.831 685 E-02 "F CP=32/1.8 ft x 3.048" B-o1 gal x 3.785 412 E-03 Ibm x 41535924 E-O1 psi/psia x 6.894 757 E+00 JOURNAL OF PETROLEUM TECHNOLOGY