— — SPE/DOE Swfdy of PolmfsumEn@new Dopmtmonl of E-y U.S. SPE/DOE 16396 Observations and Recommendations Hydraulically Fractured Wells by M.J. Economies, in the Evaluation of Tests of Dowell Schlumberger SPE Member Copyright 1987, Society of Pelroleum Engineers Th!s paper was prepared for presentation at lhe SPE/DOE Low Permeability Reservoirs Sympomum held in Denver, Colorado, May 18-19.1987. Th!s paper was selected for presentation by an SPE Program Commiltee following review of information contained in an absfracl submitted by the author(s). Contents of the paper, as presen fad, have not been reviewed by the Society of Petroleum Engineers and are aubjecl 10 correction by the author(s). The malerial. GS presented, does not necessarily reflect any position of the Society of Petroleum Engineara, its officers, or members. Papers presented at SPE meetings are subject 10 publication review by Editorial Committees of Ihe Smiety of Petroleum Engineers. Permission to coPy is resmcted lo an abstract o: not more than 300 words. Illustfafions may not be copied. The abstract should contain conspicuous acknowledgment of where and by whom fhe psoer is presented. Wrile Publications Manager, SPE, PO. Box 833836, Richardson, TX 75083.3836. Talex, 730989 SPEDAL. ABSTRACT The pressure transient behavior of hydraulically fractured wells has been the subject of considerable study over the past few years. Several investigators have presented solutions of the fundamental equations, identified qualitative diagnostic trends and suggested interpretation techniques. This paper presents a systematic approach to the problem along with substantial observations on the potential of unique interpretations. Pre-treatment tests are considered here as necessary. Well tests in tight formations are often of very she, t duration, to allow the use of established methodologies. Hence, a technique to calculate the maximum reservoir permeability from a short well test is offered. In the case of post-treatment tests the data are treated using the convolution/deconvolution techniques and influence functions. The term “influence function” defines a relationship between pressure response and time at a constant unit ‘surface flow rate. Although drawdown well tests have advantages over buiidup tests because they are used while the well is producing, their interpretation has been hampered by varying flow rates. Conventional interpretation techniques assume either constant well flow rate or controlled variation of it. Pressure buildup tests are conducted with the well flow rate equal to zero and are, as a result, predominant. In the case of post-treatment well tess of hydraulic fractures a lengthy buildup test and the ensuing shut-in may result in severe damage to the generated fracture. Thus buildup tests are not always desirable for post-treatment evaluation. The convohrtion/deconvolution techniques and influel,:e “functions,” by normalizing the pressure response to a unit flow rate, permit the use of standard techniques for the analysis of drawdown tests. L The interpretations presented here utilize new versions of pressure and pressure derivative type cur~cs including tk dimensionless fracture storage coefficient, and the dimensionless fracture conductivity. Based on observations of the sensitivityy of the response to these parameters, three type curves have 3een developed, one for low, one for intermediate, and one f-r high conductivity fractures. The choice can be made on the basis of pre-treatment analysis and the fracture design. The storage “,nd the half-length of the generated fracture can then be talc dated with reasonable confidence. PRE-TREATMENT WELL ANALYSIS FOR TIGHT RESERVOIRS Tight formations are characterized by permeabilities less than 10 md. The permeability is usually less than 1 md and is often less than 0.1 md, Knowledge of the “undisturbed” reservoir permeabilityy is considered essential for the preparation of an appropriate fracture stimulation design. As will be demonstrated later on in this paper it is also essential in the post-treatment evaluation of job effectiveness. The standard and desirable methods of analysis for radial, infinite acting, reservoirs are usually not feasible in tight formations. Although the reservoir configurations would theoretically lend themselves to such analyses, the low permeabilityy results in a slow response to pressure perturbations. T1.: necessary pressure response patterns take a substantial amount of time to appear. Techniques for analyzing tight formations, which attempt to extract reservoir data from very early data by deconvolving wellbore storage effects have appeared in the literature. Kabir and Kucuk (1985) attempted to reduce the wellbore storage distortion period h a pumping oil well by using the convolu- 2 OBSERVATIONS OF TESTS OF HYDRAULICALLY FRACTURED SPE 1639 tion and deconvolution of test data. For wells not capable of flow to the surface they suggested the monitoring of the rise of the annular liquid level, Ahmed et ai. (1985) presented a method involving the measurement of transient rate and pressure for a short period of time and again using a deconvolution technique to remove much of the distortion caused by wellbore effects, for appropriate interpretations. In offshore locations where well completion, testing and fracturing are done back to back, the test duration is often brought into question. In the case of extremely tight formations (permeabilities belov 0.1 md) even these techniques may not be effective. The follo’ ( ing is offered as a method to calculate a maximum value . or the reservoir permeability. ; Hence, if /e,~.b, is known, then Eq. 4 may be rearranged to calculate k: Equation 4 may aid in the estimation of the reservoir permeability simply by utilizing the time at which wellbore effects are becoming less predominant. ~> 3ooocp - The evolution of the semi-logarithmic straight line (appearing, as the “rule of thumb” implies, i,5 eye@ away from the end of wellbore storage effects) would take a iervgthyperiod of time. tD = CD (60+ 3.5 S) (1) Assuming a nonzcro skin effect, and using the definitions of dimensionless time and the dimensionless storage coefficient, Eq. 1 becomes: 0,000264 &t dpc(rw2 5.615 C 2m$c(hr$ = 260 5.615 C (2) Simplifying, rearranging and solving for t: f (hrs) z 2 x 105 ~ hk (3) CONVOLUTION/DECONVOLUTION AND INFLUENCE FUNCTIONS IN WELL TEST INTERPRETATION which relates well, reservoir and fluid properties to the approximate time for the beginning of the semilog straight line. 2 x 105 70 Cp hk .3000~ TECHNIQUES There are practical problems associated with pressure buildup testing for tight formations both in the pre-treatment state as well as following a massive hydraulic fracture. ~rst, because of the very low permeabilities the shut-in times maybe extremely long. This would result in significant cost in both actual well test expenses as well as lost production. Second, in the case of tests following a hydraulic fracture the “drawdown” is often superimposed with the well and fracture “cleanup.” While this event is complicated by the presence of alien fluids, the ensuing buildup, if done, can result in significant fracture damage. Since the beginning of the semi-logarithmic straight line appears at 1.5 log cycles away from the cessation of wellbore storage effects (about a 70 fold increase in the value of time), Eq. 3 may be reduced to: te. w. b. 2 (5) fe. w.b. ‘n well tests of extremely tight formations, wellbore effects are ,lot oniy dominant but they could totally mask the entire test. Knowledge of te.~.bo, which is characterized by the end of the 45° line on the log-log diagnostic graph and is indicative of the end of predominant wellbore storage effects, would be sufflcient to calculate the maximum value of the reservoir Permeabdity. This interpretation technique is not intended as a substitute for rigorously conducted tests. Yet, it maybe considered a reasonable alternative when lengthy test duration would otherwise preclude the performance of any pressure transient test. (60 + 3.5 s) 2mpcthr~ h Equation 5 is significant since it may provide the maximum value of the permeability provided that the wellbore storage coefficient, C can be determined. With well test data, the wellbore storage coefficient can be calculated from the slope of a cartesian graph of pressure versus time as shown by Earlougher (1977). In the presence of a nonzero skin, the actual permeability will be somewhat smaller. For a skin less than 10 the permeability value computed from Eq. 5 wouId be no more than 1.6 times the actual value of the reservoir permeabilityy. This time may be calculated. The dimensionless time for the beginning of the semi-log straight line has been correlated by Agarwal et al. (1970): (4) (e.w.b. = end of wellbore storage) This expression provides the minimum time for the cessation of wellbore storage effects, This damage, which is often observed, could have several causes. Fracture face damage may be the result of unbroken polymer chains within the penetrated reservoir surrounding the fracture or of chemical reactions. “Choked” fractures could result because of insufficient cleanup and accumulation of fines and polymer chains in the vicinity of the wellbore. Both of these phenomena can be attributed to the shut-in. It is widely accepted that a continuous drawdown is preferable and that the detfimenal effects from a lengthy shut-in, in order to conduct a pressure buildup test, may far outweigh the benefits. Using sorr,e typical variables (for a gas well) such as C = 0.12 bbl/psi (0.27 msfbar), h = 100 ft (30 m) and p = 0.22 cp then Eq. 4 may provide the minimum time during which wellbore effects will be predominant. For a permeabilityy equal to 0,01 md this time is at least 8 hrs. The presence of any skin effect would add more time. Since the pressure versus time relationships are logarithmic, very long test times would be required L WELLS En M. J. ECONOMIES E 16396 Continuous measurements of wellbore pressure and downhole “ate would allow the use of the generalized form of the neasured wellbore pressure as presented by van Everdingen and Hurst (1949) and expounded upon by Stewart et al. (1983), Kucuk and Ayesteran (1983) and Kabir and Kucuk (1985). 3 A simple method of data treatment can be employed using the generalized form of the pressure drop as shown in Eq. 6ii and by substituting the term Ap j (7) by @7)/d~ where F(7) is the drawdown for the well pro d ucing at a unit rate for a time /. This can be approximated for n intervals by: n I 4PW, (f) = \ q ~ (7) AP$~(f – 7) d~ o (6) = ~ 9D (f - 7) OiJ o (6a) Pi – tiwf)n =j~l (qn -j+ 1 - qn _j) F (/n – tn -j) (7) where F (tn – tn _j) = Fj (7) d7 + APWD (0 Setting n = j then where (for drawdown) n-1 Pi – f.Pw~n ‘j~, (9n-j+ 1 – qn -j) F (tn – tn-j) + ql Fn APW~(0 = Pi – Pw~ (f) @~~ (0 = Pi – Ps/ (/) A/J~ = pressure drop due to skin Psf = sandface pressure (8) and therefore n-1 qD(l) = qsf/qr qr = reference rate The primes indicate derivatives with respect to time. The convolution technique assumes a model (e.g., infinitely acting) and calculates Ap$j, Comparison with the measured A.owf and qs.. via Eqs. 6 and 6a results in the calculation of reservoir variables, The deconvolution technique calculates Ap~. which may then be used in the standard methodologies in search of an appropriate model and subsequent calculation of unknown reservoir and well parameters. ‘Ihe deconvolution technique and the use of influence functions would greatly enhance the applicability of drawdown testing and reduce the need for buildup. Several investigators worked on the subject. Hutchinson and Sikora (1959) used the principle of superposition in calculating a “resistance function” for a drawdown test with variable flow rate. Jargon and van Poollen (1965) developed an analytical expression for the influence function of a slightly compressible fluid. Katz et al. (1962) developed an expression for the influence function for the early part of a drawdown test, while van Everdingen and Hurst (1949) showed that the pressure drop calculated via the principle of superposition is generally correct regardless of the flow rate history of the well. pi – @wf)n– j~ ,(qn The simplest form of the influence function is dividing the pressure difference by the corresponding flow rate. In other words: Ap/q for oil and Ap2/q or Am(p)/q for gas. I – qn -j) Fj (9) ql Equation 9 shows that the influence function at any time tn can be calculated by knowning the pressure drawdown, pi – Owf)tr, the flOWrate and the value of the inffuence function at the previous time increments. For the first time step (10) In the case of gas reservoirs, the pressure difference functions in Eqs. 9 and 10 may be replaced by the real gas pseudopressure difference. Figure 1 represents pressure (wellhead and bottomhole) and flow rate history for well Travale 22 in Italy, These data were originally published by Economies et al, (1979). Figure 2 shows the calculated “influence functions” using two simple, normalizing expressions (pressure or real gas pseudopressures divided by the flow rate) and a deconvo[ved set of data. Any one of these could be used with the standard methods in search of an appropriate model for interpretation and calculation of unknown reservoir parameters. A STUDY ON FRACTURED Mannon ( 1977) has presented a form of the influence function of gas reservoirs using the real gas pseudopressure and Economies et al. (1979) presented a formulation for geothermal well testing. - j+ Fn = WELL TEST RESPONSE As with all types of well tests the analysis of fractured well tests aims towards the identification of well and reservoir variables that would have an impact on future well performance. However, fractured wells are substantially more complicated. The well-penetrating fracture has unknown geometric features (length, width and height) and unknown conductivity properties. There are certain presumptions that investigators in the field and well test analysts have used. These are mentioned here in order to identify a priori certain limitations: OBSERVATIONS OF TESTS OF HYC a) The fracture height is usually assumed to be equal or less (Raghavan et al,, 1978) than the reservoir height. Most of the interpretation work has modelled fully or partially penetrating fractures, However, in real fracturing operations there is much concern with uncontrollable propagating fractures in the vertical direction, above or below the targeted formation. Hence, interpretation of well tests must take into account containing layers, indications of fracture growth outside the interval, and especially the ‘connection of other productive horizons. Hence, reservoir engineering type analysis should always be accompanied by pressure analysis during the job execution and temperature or radioactive logging after the job. Needless to say, communication, via the fracture, with other Iayers would greatly complicate the analysis. b) The fracture permeability cannot be inferred from the laboratory-measured proppant permeabilityy, even at reservoirconfirring pressures. Theactual fracture conductivity is a c~mposite, bulk, variable taking into account phenomena such as embedment, reservoir fines migration and retention within the fracture and unbroken polymer chains that are either permanent or very slowly d.isirtegrating remnants of the fracturing fluids. Recent laboratory measurements (Roodhart et al., 1986) have shown substantial proppant pack permeability 10SS(from 15L70to 75VO) after treatment with various fracturing fluids. c) The fracture storage coefficient and especially the transition from wellbore storage to one of the discernible patterns, (hi-linear, linear or other flow) may give misleading results. A visual pattern of pressure response may appear similar for a number of fracture conductivity and fracture storage factors. In general, a sma!l fracture conductivity and a small storage coefficient would give similar looking patterns at early and middle times of a well test. Hence, real data of a particular shape may prompt investigators (especially when they have type-curve generating c,~pacity) to use a set of type curves (with a combination of fl acture conductivityy and storage coefficient) that are inapp: ~priate. More on this will appear later in this paper, The classic fractured well test interpretation papers were published by Gringarten and Ramey (1974) for the ‘“infinite conduct ivity fracture” and Gringarten et al. (1975) for the’ ‘constant flux” fracture. Traditionally, natural fractures have been interpreted by the infinite conductivity model (Kazemi, 1969, Kucuk and Sawyer, 1980, Cinco and Samaniego, 1982, Samaniego and Cinco, 1983 and others) while acid fractures have been interpreted via the constant flux model (Gringarten, 1978, Cuesta and Elphick, 1984), Alagoa et al. (1985) have used these two models to interpret the behavior of apparently firrite conductivity fractures, In their paper, in which they incorporated the use of the pressure derivative, !hey had to use very large storage coefficients to accomodatc ne trends of the data. The definitive description of the flow behavior of a hydraulically induced propped fracture (hence of finite conductivity) was presented by Cinco et al. (1978) and Cinco and Samaniego AULICALLY FRACTURED WELLS SPE 1639 (1981 a). They offered the concept of hi-linear flow, which, for certain values of the fracture conductivityy forms a very distinctive slope equal to 0.25 on a log-log graph, Two commonly appearing deviations from the “ideal” fracture response (in addition to the aforementioned partially penetrating fracture) are: 1) damaged fractures, where a damaged zone, extending in a normal direction into the res. toir, enci~cles the fracture and, 2) choked fractures, where the fracture permeabilityy just away from the wellbore is reduced. Damaged and choked fracture behavior was described by Cinco and Samaniego (1981 b). Wong et al. (1984) applied the pressure derivative to a fracture with fluid loss damage. Cinco (1982) has described qualitatively the successive pattei ns that emerge during the flow from a finite conductivity fracture well. The following study should provide a more focused approach to the problem by presenting a methodology (and a rationale) within the ranges of well, reservoir and fracture variables that are likely to be encountered in practice. Fundamental variables and choice of axes in graphing, The variables to be employed here are the usual dimensionless groups i.e. Pressure: Ap/141.2qBp for oil kh (11) PD = PL) = kh Am(p)/ 1424qT for gas(12) Time: tDxJ = 0.000264k At/@c# (13) Storage coefficient: CDy = 5.615C/2x~c[/rx; (14) Fracture conductivity:FcD = klw/kxy (15) As shown by Bourdet et al. (1983, 1984) the use of the pressure derivative greatly improves the analysis and the uniqueness of the analysis of well tests. Furthermore, as shown by Grirrgarten et al. (1979) the graphing of pressure against fD/CD collapses repeating families of type curves that would appear for each value of CD into one family of curves manv of which would ~hare a sin~le wellbore stor~ge portion. Thi; will be employed here, Hence, all type curves to be presented are graphed with fDti/CDf on the abscissa. This forces all solutions to share a single wellbore storage-dominated straight line (with a slope equal to unity). The derivative used here is perforrr,ed with respect to the natural logarithm of the time function. This met hod as shown by Bourdet et al. (1983) and extended by Alagoa et al, (1985) to fractured wells, provides characteristic shapes for wellbore storage, linear, hi-linear and infinitelyacting radial flow. Using a basic algebraic principle: dpD/d ([0 tD,f) = tDP ‘D (16) it can be said that if PD - J’xj (17) . -.-.”. L.*. .“ .I . JJbu,. ”,v,, 10-3 and 10-6. Hence, these are realistic fracture Storage coefficients for which solutions should be generated for both oil and gas wells. then: dpD/d (h tD$) = t@ ‘D - mt;’f ”JJo (18) The first task was to run a simulation for the two bounds for the fracture storage constant i.e., for cDf = IO--6 and ~Df = 10-3. Figure 3 represents pressure and pressure derivative response for the first value and for dimensionless fracture conductilities equal to 0.1, 1, 10, and 100. Figure 4 represents the same solutions but for a fracture storage constant equal to of time was selected to show 10-3, For both cases the range realistic and appropriate response features. This implies that for the first three cases the derivative graph parallels the pressure graph, displaced in the vertical direction by log m. In wellbore storage-dominated flow t he two curves are superimposed since log m is equal to zero, for m equal to unity. For infinitely acting reservoirs the derivative is equal to 0.5 (which is its limiting value) as shown by Bourdet et al. (1983, 1984) and Alagoa et al. (1985). For both cases, the infinitely acting behavior appears at fDx. approximately equal to unity. Hence, assuming a reasonable value for the porosity (0.25) and the viscosity compressibility product for both oil and gas (5 x IO-6 cp-psi -1, 7.3 x 10-5 cp-bar-1), the time of the infinitely acting behavior (in hrs) is given by: The above are significant not only as diagnostic and interpretive tools but also, for the purposes of this study, they are invaluable in the identification of the occurrence (or lack thereof) of certain flow regimes and their duration. For example, bi-lineai flow can be identified only, if (and only if) parallel portions of both the pressure and pressure derivative curves form slopes equal to 0.25, t-— The ranges of the dimensionless variables for which solutions are generated are also important. Of these, the dimensionless fracture conductivity and the storage coefficient are of particular importance, The dimensionless fracture conductivity as given by Eq, 15 is, in essence, a measure of permeability contrast. Obviously, the higher the fracture conductivity, the better the performance of the well will be when compared to the pre-treatment state, Hence, for very tight formations (low reservoir permeability) even narrow, and without particularly high permeability, fractures could result in significant performance improvements. Furthermore, it shows in elegant form why usually dramatic results are commonly obtained in tight formations (with properly executed fractures). The values of the dimensionless fracture conductivity range from around unity for very poorly producing fractures to 10 (for moderate) to over 100 for highly conductive fra:!ures. Cinco and Samaniego (1981) have shown that for values over 300 the finiteconductivity solution is indistinguishable from the Gringarten and Ramey (1974) infinite-conductivity solution. In fact, even for fracture conductivities far below 300 the differences are practically indistinguishable, xf2 200 k (19) where I in hrs, xf in ft and k in md. Of particular interest here is the development of the flow regimes. For the low storage coefficient there is a significant region of hi-linear flow shown with the large shaded area in Fig. 5. The smaller shaded area is for linear flow where the slope is equal to 0.5. This appears only on the larger fracture conductivityy (~CD = 100) as one would expect. No half slope of appreciable duration was observed in the smaller fracture conductivities. A much more interesting observation was noted for the larger storage coefficient as shown in Fig. 6. No hi-linear flow was observed (except perhaps a small portion for FCD = 1 around tD#cDf = 102). Linear flow was observed, again at FCD = 100. Obviously, for storage coefficients larger than 10-3 it would be unlikely for hi-linear flow to appear. Linear flow, though, is evident especially for very large dimensionless fracture conductivities. These observations lead to the generation of a new set of useful type curves, one for low (FCD = 1), one for intermediate (FCD = 10) and one for high (FCD = 100) conductivityy. These will be described in the next subsection. The fracture storage constant, as defined by Eq. 14 can be calculated in the manner shown by Earlougher (1977). He has provided techniques to calculate the wellbore storage constants for both full wellbores and falling liquid level wellbores. These calculations take into account the density and the compressibility of the wellbore fluids. The dimensionless fracture storage constant as defined in Eq. 15 is similar to the standard wellbore storage constant, Instead of the wellbore radius, the variables in the definition are divided by the fracture half-length. A NEW SET OF PRESSURE AND PRESSURE DERIVATIVE rYPE CURVES FOR HYDRAULICALLY FRACTURED WELLS Figures 7, 8, and 9 are the new set of type curves in which pressure and pressure derivative are graphed for values of the dimensionless fracture StOrage coefficient, CDf equal to 10-3, 1o-4, IO--5, and 1o-I5, A range for CDf has been established for this study. Using typical reservoir and fluid variables and allowing the fracture half-length to vary between 100 ft (30 m) (rein) and 1500 ft (460 m) (max), values of 5 x 10- t(max) and 10-6 (rein) were obtained for a gas filled wellbore. For oil, and using both a full wellbore or a falling-level model, these values ranged between The graphing of the ratio tDXf/CDf in the abscissa merges all curves into a single wellbore storage portion. For an FCD value equal to unity (Fig., 7) there is an extensive hi-linear flow region (slope equal to 0.25) and no linear flow W! “. 6 OBSERVATIONS OF TESTS OF HYDRAULICALLY FRACTURED SPE 153% WELLS } I regime. Even for an FCD value equal to 10 no linear flow appears (Fig. 8). However, for an FCD value equal to 100, the hi-linear flow vanishes almost entirely while, as one would expect, the linear flow becomes much more prevalent. storage constant may be calculated. The FCD value, and the newly calculated fracture half-length would lead to a rough estimate of the kfw (fracture permeability-fracture width) product, There is a compelling by-product of the study of these type curves. There is an obvious, visual correlation between fracture conductivityy and storage coefficient. The higher the fracture conductivity the lower the curvature is, while the higher the storage coefficient the higher the curvature of the solution is. Field Application As a result, if one were to observe Figs. ? and 9, a set of data that would look analyzable with FCD = 1 and CDf = 10-3, would require a much larger CDfvahre on the FCD = 100 type curve to produce the same visual trends. This, in fact, was done in the Alagoa et al. (1985) paper in which infinite-conductivity solutions were used to analyze the posttreatment data of hydraulically fractured wells. In their paper, much larger wellbore storage constants were necessary. (CDJ values graphed were between 3 x 10–s and 3 x 10– 1.) In the field examples that they analyzed, the fracture half-lengths calculated were roughly equal to 100 ft. Analysis with the type curves offered here would result in much larger fracture half-lengths. What is required to use the type curves offered in this paper is for the analyst to decide whether a low, intermediate or high conductivity type curve is indicated. This is usually not a proMem. From the definition of the fracture conductivity (Eq. 15), it should be obvious that irr very tight formations (k < 0.01 md) even a moderav: fracture would result in a very high conductivity type curve. For high permeability formations (k > 5 red), even a reasonably sized fracture would result in a low conductivity. Hence, a generalized, order of magnitude clasWication (as is required) can be done before the analysis is initiated. Well KAL-5 in the Pan.lonian Basin in Yugoslavia (a gas condensate well) was fractured with a rather modest treatment because of the limited availability of proppant on location. The chronicle of the operation was described by Economies et al. (1986). Table 1 contains well, reservoir and fluid data from a post-treatment test at well KAL-5. The pressure buildup lasted for 332 hrs, The actual pressure data were presented by Economies et al. (1986), Real gas pseudo-pressure differences and their derivatives are used for the analysis. Since the ordinaf e is set (the reservoir permeability was obtained by a pre-tret tment test and found equal to 0.0035 md) then for any value c f Am(p) (eg. 107) the dimensionless pressure can be calculate: (0.0035) (216.5) (10’) khAm(p) ‘D = 1424qT = (1424) (2020) (814~ = s z x ~O_3 ‘ Thus, with Am(p) = 107 and PD = 3.2x 10-3 superimposed, the data and their derivatives are moved from left to right until they match with a type curve. Ordinarily, for this well a proper fracturing treatment would result in a very high conductivity fracture. However, the treatment done was rather small and thus an attempt to match the data with an intermediate conductivity fracture is attempted here. Figure 10 shows a match which appears quite reasonable. The dimensionless storage coefficient CD$ is roughly equal to 10–4 and the time match results in: tD~f/CDf = 8 and At = 100 hrs. The method of analysis itself is straightforward. Since the permeability of the undisturbed reservoir is obtained via a pretreatment test, then the vertical match in a type curve rilatch attempt is de facto set. (All variables multiplying the dimensional pressure difference are constant and known.) Hence, the only flexibility allowed by the type curve is movement along the time axis. Simultaneous pressure and pressure derivative match should be done. The data are moved U! the selected type curve from left to right along a constant pressure match line. This is, of course, necessary only if wellbore storage data disappear very rapidly (before they can be definitely recorded). If wellbore storage data exist then the match has no degree of freedom and the knowledge of the undisturbed reservoir permeability should serve as corroborating evidence. In moderate to high conductivity fracture wells, wellbore storage effects often disappear rapidly, hence the methodology outlined above would be normally employed. From the time match and the value of the dimensionless storage coefficient, the fracture half-length and the dimensionless Using the gas physical properties published by Economies et al. (1986) and the data given in Table 1, the time match results in a calculated value of the fracture half-length equal to approximately 800 ft (240 m). From the value of the dimensionless storage constant and its definition, a dimensioned wellbore storage constant equal to 0.075 bbl/psi (O.i 7 m3/bar) is calculated. Considering that this is a 11,200 ft (3480 m) well with approximately 180 bbl (29m3) capacity and using a compressibility equal to 4 x 10-4 psi-1 (5.9 x 10-3 bar-1, at wellbore conditions, results in a storage constant equal to 0.072 bbl/psi (O.166 bbl/psi) which is very near the one calculated from well data. The value of the dimensionless fracture conductivity (FCD) chosen for this analysis (and its definition in Eq. 15) results in a kfl product equal to approximately 30 md-ft. This value based on the methodology outlined earlier is the one likely to exhibit the largest error. However, the forecasting of future well performance is not particularly sensitive to the kfl product, (within reasonable ranges). Furthermore, experimental studies done by Roodhart et al. (1986) has shown that the M, J. EC I 16396 kacture permeability may vary by a factor of three to four iepending on the fracturing fluid used. Hence, its calculation s intended to provide an approximate value. CONCLUSION A methodology to interpret fractured well tests has been presented. This methodology is based on observations of the belxa:’iorof fractured wells. Their slow response in developing the well kr.own patterns led to a technique for the estimation of the maximum reservoir permeabilityy. The latter is crucial for both the design as well as the post-treatment evaluation phases of a fracturing job. The convohttion/deconvolution techniques and influence func. tions have been suggested in order to allow the interpretatim of drawdown (instead of buildup) data. This should be of par. ticu!ar use in sensitive formations where shut-ins may resull in fracture damage. New type curves for pressure and pressure derivative allow tht calculation of the fracture half-length and fracture storage coef ficient. A reasonable value of the fracture permeability-fractur{ width product may be extracted as’ well. ACKNOWLEDGEMENT The author wishes to thank C Ehlig-Economides and M Karakas of Flopetrol-Johnston-Schhsmberger for their help il the development of the new type curves. IMIDES ? 7, = fracture half-length f = gas gravity (to air) I = viscosity b = porositjj = gas deviation factor ?EFERENCES 4garwa1, R. G., A1-Hussainy, R. and Ramey, H. J., Jr.: “An investigation of Wellbore Storage and Skin Effect in Unsteady tiquid Flow: 1. Analytical Treatment,” Sot. Pet. Eng. J (Xpt. 1970) 279-290. 4fmed, U., Kucuk, F., Ayesteran, L.: “Short-Term ‘ransient Rat: and Pressure Buildup Analysis of Low-F .imeability Wells,” p~per SPE/DGE 13870, presented at the 1985 Low Permeability Gas Reservoirs of the SPE/DOE. Alagoa, A,, Bourdet, D., Ayoub, J. A.: “How to Simplify the Analysis of Fractured Well Tests,” World Oil, Ott. 1985, 17-102. Bourdet, D., Whittle, T. M., Douglas, A.A. and Pirard, Y.M.: “A New Set of Type Lurves Simplifies Well Test Analysis,” World Oil, May 198. . Bourdet, D., Ayoub, J.A. and Pirard, Y.M.: “Use of Pressure Derivative in Well Test Interpretation,” paper SPE 12777, presented at the 1984 California Regional Meeting of the SPE. NOMENCLATURE B= formation volume factor c~ = totai system compressibility c= welliwre storage constant FCD = dimensionless frature conductivity h = reservoir thickness hp = perforated interval k= kf m= permeability = fracture permeability slope m(p) = real gas pseudopressure P= pressure Pi = initial pressure Pwf = bottomhole pressure P’ q = pressure derivative . flow rate rw Sw = well radius = water saturation t = time T= absolute temperature w= fracture width Clnco-Ley, H., Samaniego-V, F. and Dominguez, N.: “Transient Pressure Behavior for a Well with a Fi- . Conductivity Vertical Fracture,” Sot. Pet. Eng. J. (Aug. 1978), 253-264. Cinco-Ley, H. and Samaniego-V., F.: “Transient Pressure Analysis for Fractured Wells,” Jour. Pet. Tech., (Sept. 1981) 1749-176ri. Cinco-Ley, H. and Samaniego-V., F.: “Transient Pressure Analysis: Finite Conductivity Fracture Case Versus Damaged Fracture Case,” paper SPE 10179 presented at the 1981 Anuual Fall ltieeting of the SPE. Cinco-Lejj, H. and Samaniego-V., F.: “Pressure Transient Analysis for Naturally Fractured Reservoirs,” paper SPE 11026 presented at the 1982 Annual Fall Meeting of the SPE. Cinco, H.: “Evaluation of Hydraulic Fracturing by Transient Pressure Analysis Methods,” paper SPE 10043 presented at the 1982 International Petroleum Exhibition of the SPE, Beijink, China, Cuesta, J.F. and Elphick, J. J.: “Pre- and Post-Stimulation Pressure Test Analysis and its Role in the Design and Evaluation of Hydraulic J%acture Treatments,” paper SPE 12992 presez:ed at the 1984 EUROPEC. Earlougher, R. C., Jr.: Advances in Well Test Analysis, SPE, Dallas, 1977. I 8 OBSERVATIONS OF TESTS OF HYDRAULICALLY Economies, M. J., Brigham, W .E., Cinco-Ley, H., Miller, F.G., Rarney, H. J., Jr., Barelli, A. and Manetti, G.: “Influence Functions and their Application to Geothermal Well Testing,” Geoflr. Res. Corm. Trans. v. 3, 1979,177-180. FRACTURED WELLS SPE 163% Raghavan, R., Uraiet, A. and Thomas, G. W,: “Vertical Fracture Height: Effect on Transient Flow Behavior,” SoC. Pet. Eng. J. (Aug, 1978), 265-277. Roodhart, L., Kuiper, T.O. and Davies, D. R.: “Proppant Rock Impairment During Hydraulic Fracturing,” paper SPE 15629 presented at the 1986 Annual Fall Meeting of the SPE. Economies, M, J., Cikes, M., Pforter, H., Udick, T,H. and Uroda, P.: “The Stimulation of a Tight, Very HighTemperature Gas Condensate Well,” paper SPE 15239 presented at the 1986 Unconventional Gas Technology Syrnposium of the SPE. Samaniego-V F. and Cinco-L,ey, H.: “Pressure Transient Analysis for Naturally Fractured Gas Reservoirs: Field Case,” paper SPE 12010 presented at the 1983 Annual Fall Meeting of the SPE. Gringarten, A.C. and Ramey, H. J., Jr.: “Unsteady State Pressure Distributions Created by a Well with a Single-Infinite Conductivity Vertical Fracture,” Sot. Pet. Eng. Y. (Aug. 1974), 347-360. Stewart, G., Meunier, D. and Wittmann, M. J.: “Afterflow Measurement and Deconvolution in Well Analysis,” paper SPE 12174 presented at the 1983 Annual Fall Meet;ng of the SPE. Gringarten, A. C., Ramey, H. J., Jr. and Raghavan, R.: “Applied Pressure Analysis for Fractured Wells,” J. Pe(, Tech, (Juiy 1975), 887-892. van Everdingen, A.F, and Hurst, W.: “The Application of Laplace Transform: to Flow Problems in Reservoirs,” Trans., AIME (1949), 186, 305. Gringarten, A. C.: “Reservoir Limit Testing for Fractured Wells,” paper SPE 7452 presented at the 1978 Annual Fall Meeting of the SPE. Wong, D. W., Barrington, A.B. and Cinco-Ley, H.: “Application of the Pressure Derivative Function in the Pressure Transient Testing of Fractured Wells, ” paper SPE 1305 presented at the 1984 Annual Fall Meeting of the SPE. Gringarten, A. C.. Bourdet, D. P., Landel, P.A. and Kniazeff, ~?.J,: “A Comparison Between Different Skin and Wellbore Storage Type Curves for Early-Time Transient Analysis,” paper SPE 8205 presented at the 1979 Annual Fall Meeting of the SPE. TABLE 1 RESERVOIR, WELL AND FLUID INFORMATION FOR KAL-5 Hutchinson, T.S. and Sikora, V, J.: “A Generalized WaterDrive Analysis,” Trans. AIME (1959) 216, 169. (FROM ECONOMIES Jargon, J.R. and van Poollen, H, K.: “Unit Response Function from Varying-Rate Data,” J. Pet. Tech. (Aug. 1%5), %5. Kabir, C.S. and Kucuk, F.: “Well Testing in LowTransmissivity Oil Reservoirs,” paper SPE 13666, presented at the 1985 California Regional Meeting of the SPE. = 1.94 9C = 117.6 B/D Qeq = 2.02 Katz, D. L., Tek, M,R. and Jones, S.C,: “A Generalized Model for Predicting the Performance of Gas Reservoirs Subject to Water Drive,” paper SPE 428, presented at the 1962 Annual Fall Meeting of the SPE. Kazemi, H,: “Pressure Transient Analysis of Naturally Fractured Reservoirs with Uniform Fracture Distribution,” Sot. Pe(. Eng. J. (Dec. 1969) 451-461. 103 Mscf/D 9g X X 103 Mscf/D ET AL,, 1986) hp = 36 ft 4 = 0.062 rw = 0.54 ft rg = 3,94 (to air) Sw = 0.3 T = 354° F k = 0,0035 md h = 216.5 ft Fhsid Composition (percent mole fractions) Kucuk, F. and Ayestaran, L.: “Analysis of Simultaneously Measured Pressure and Sandface Flow Rate in Transient Well Testing,” paper SPE 12177 presented at the 1983 Annual Fall Meeting of the SPE. Kucuk, F. and Sawyer, W. K.: “Transient Flow in Naturally Fractured Reservoirs and its Application to Devonian Gas Shales,” paper SPE 9327 presented at the 1980 Annual Fall Meeting of the SPE. Mannon, L. S.: “The Real Gas Pseudo Pressure for Geot herrnal Steam,” M.S. Report, Department of Petroleum Engineering, Stanford University, 1977. H2S = 0.006 i-C5 = 0.442 N2 = 1.452 n-C5 = 0.379 co~ = 10.931 C6 = 0.516 cl = 72.613 CT = 0.644 C2 = 6.24 C* = 0.541 C3 = 1.63 C9 = 0.388 i-C4 = 0.553 n-C4 = 0.693 cc C,O+ = 2.979 Pwi or 10’ 750 - q (MSCF/d) Ptf 2000 600 - F psi 10 1 1000 800 3ooo~ Time (days) 1 lU 10- 10 . Time (days) Fig. !-Pressure and Ikw me dst, 1.xTrw,I. 22. 101 [ FIS. 2-cvcOnvOWcdand nowM2&.11 influence2imtlon8 f- I I I I I I I Fc -0 ~ -~ 100 ?2 Q 101 ,.., .0 102 1(J3 Io’f Fb. 3-W$W* 101 I -g 101 I M PI.uw. 105 t~x, 1CD, Mvatlvo ruww I I 101 100 106 fw bw.aarw 107 108 frutwed wdl (CC+. to - *I. I 1 I I ,03 104 I t ,“ 10-3 ~o-z @ 102 %xf I %ff M. 4-P-WI. -i pm-w. duhwiw c.wmtm fix hkh+mw 57 mwd inn ICW. 10-31. Trwmb 22. 101 100 ~-n -an a n 101 ~10-2 102 1 t 103 1 105 104 I I 106 107 I lfj8 tox’ I CD’ Fig. 5-B[.linear and linear flow regimes for low.stor~e wells. 101 100 ‘i 0’ 102 q n3 I .“ 10-3 I 102 I 1()-1 Fig. 6-Llnenr 1 t 100 101 tDx’I cDf flow regime in high-slorage se wells. 1 I 102 103 104 10’ “ m 1 1 O( 10” 10: -31 I of 1 (-J-3 1(-J-1 102 100 1(y 1fj2 101 [FCD 7-Pressure I 104 103 lo6 105 c~f t~xf / Fig, I and pressure derivative type curve for low-fracture conductivity well. = 101 I lt’fFINITELY I ACTING+ 1~ t’ 100 \ . , . I “an & =10-5+ ~c.f 102 ‘u Q 10-3 10-4 I (J-4 10-3 102 101 101 100 t~xf / 102 103 104 Cof Fig. 8-. Preseut8 and preaeure derivative type curve for intermediate fracture conductivity well. 59 1(J5 106 101 1 -- II 100 1 (J-3 104 I 0-2 @ I00 101 102 103 104 105 106 ‘Dxf / cDf Fig. 9-Pressure and preesure derivative type curve for high-fracture conductivity well. 1(’)1 IF A. I-bu =10[ I I I I I 101 --’’’’’=.{OO 101 102 103 I I-AF41TELY ‘ACTING+ 100 n 1 ()-1 1 I I I A 10-2 1(’J-3 10-4 10-4 10-3 102 cDf Fig, 10—Type-curve match for example application. t~xf 60 104 105 106 / I