This paper gives the theory and numerical implementation of a comprehensive acid-fracturing model. It solves for fracture geometry (2D or 3D), leakoff, heat transfer, and acid transport simultaneously. Acid fracturing currently is receiving renewed attention as a well stimulation technique.
This paper gives the theory and numerical implementation of a comprehensive acid-fracturing model. It solves for fracture geometry (2D or 3D), leakoff, heat transfer, and acid transport simultaneously. Acid fracturing currently is receiving renewed attention as a well stimulation technique.
This paper gives the theory and numerical implementation of a comprehensive acid-fracturing model. It solves for fracture geometry (2D or 3D), leakoff, heat transfer, and acid transport simultaneously. Acid fracturing currently is receiving renewed attention as a well stimulation technique.
Summary. This paper gives the theory and numerical implementation of a comprehensive acid-fracturing model that solves for frac- ture geometry (2D or 3D), leakoff, heat transfer, and acid transport simultaneously. The acid-transport model integrates a number of features not accounted for in earlier design models: multiple fluids (acids, gels, and foams); acidizing controlled both by mass transfer and rate of reaction; and leakoff, including the effects of matrix acidizing and wormholing. Examples of treatment design illustrate these features. Coupling with reservoir forecasting models gives the ability to optimize the job or compare it with a proppant treatment. Introduction Acid fracturing currently is receiving renewed attention as a well stimulation technique. Compared with advances in modeling tools for proppant fracturing, the models for acid-fracturing processes have not changed much in the last decade. During a period of in- tense study in the 1970's, the basic understanding of the reaction kinetics and acid transport in the acid-fracturing process was es- tablished by laboratory and theoretical studies. I - 3 The design models of acid-fracturing typically were decoupled from fracture- geometry models and use fixed fracture width and prescribed leakoff. 4 - 8 Also, it was recognized that quantitative prediction of wormholing was a major obstacle to theoretical predictions of acid penetration. It also was a principal reason for the discrepancy be- tween predicted and observed acid-fracture lengths. Very little work on acid-fracturing design was reported in the 1980's. The earlier models, with decoupled fracture-geometry cal- culations, were still being used. This was in sharp contrast to the great increase of sophistication in propped fracture design and made comparisons of these two alternatives difficult. Only recently was interest in acid fracturing revived. Ben Naceur and Economides 9 and Lo and Dean 10 presented improved models of acid transport, coupled with a specific 2D model of fracture geometry. Gdanski and Lee II described a more general model, but their work lacked mathematical details necessary to other modelers. Theoretical and laboratory work currently is being done to understand wormhole behavior l2 - 14 (which is the major factor responsible for large leakoft) and to develop fluids with better reaction and leakoff control. 15,16 This paper gives the theory of a comprehensive acid-fracturing model, the numerical implementation of the acid-transport and leakoff solution, and examples illustrating various features of the model. The model includes a number of new features essential for modern design. 1. Multiple fluids with different rheologies (reacting or not), in- cluding gel/acid sequences. 2. Acidizing process controlled by both mass transfer and rate of reaction. 3. Leakoff behavior, including a new method for calculating the increase in leakoff caused by matrix acidizing and wormholing ef- fects at the fracture wall. 4. Coupling with heat transfer by means of the thermal depend- ency of reaction kinetics and the heat of the reaction. 5. Implicit coupling with 2D or 3D fracture geometry and with reservoir forecasting models to optimize the job or compare it with a proppant treatment. This formulation is considerably more general than the model described in Ref. 10, which assumes infinite reaction rate and treats acid transport as steady-state. These features are illustrated by ex- amples of acid transport under different physical conditions (limited by mass transfer, reaction rate, effect of reaction order, etc.). The case study of a treatment with an alternating gel/acid sequence shows the model's ability to assess the importance of various design pa- rameters. In particular, the model can predict short penetration dis- tances in agreement with field experience. Equations for Acid Transport For simplicity, an element of a fracture with variable width but constant height is considered (Fig. 1). In the following, it will be Copyright 1993 Society of Petroleum Engineers 30 assumed that concentration, C, is defined as mass of acid per volume of solution. If one wishes to use concentration C in terms of moles, C must be replaced by MwC, The equations describing the flow of acid in a channel of unit height are given below. The equation of continuity within a channel of variable width is a(vz)/ax+O(vy)/iiy=aAlat-i, ........................ (1) where A = elemental area dxdy. The term aAlat represents the change of control volume caused by the change of fracture width, b, with time. The boundary conditions are Vyly=bl2=Ve, where vf= leakoff velocity. By integrating this across y and z tim- ing, we obtain the equation solved by the geometry module of a fracture simulator (except for compressibility): a(vA c )lax-2qe(x)=aA c lat-i, ....................... (2) where v(x,t) = average volumetric velocity, Ac(x,t) = local cross- sectional area, and qe(x,t)=leakoff rate on one fracture surface. If constant width vertically is assumed, Ac and qe are given by A c ='Y3h f b and qe=hffVe, where hf=total fracture height and hfe=leakoff height. If we neglect diffusion in the x direction, a a a ( ac) ac --(Cvx)--(Cv y )+- De- =--iC j , . (3) ax ay iiy iiy at where Vx and v y =2D components of the velocity field, C=acid concentration, and iCj=acid injection rate. The effective diffusion coefficient, De' is defined below. Two boundary conditions apply: (1) at the wellbore (x=O), C=C j and (2) at the fracture wall (y=b/2), the boundary condition states that the total flux of acid to the surface equals the acid spent plus acid leaking off with ve- locity ve' The total mass transport to the wall is the sum of diffu- sion and convection: -qA=(-CBVy+D e aC)1 , ..................... (4) ay y=bl2 where De = effective diffusion coefficient, which for laminar flow is equal to the molecular diffusion but increases in the presence of turbulence. The reacted amount is assumed to be r=k(l-q,)(CB-ceq)m, ............................ (5) where m=reaction order, k=reaction-rate constant, and C B and Ceq = boundary (wall) and equilibrium concentrations, respective- Iy. Ceq takes into account the effect of reverse reaction, which may be important for weak acids. 8 The amount lost to the formation is qAe=veCf' ....................................... (6) where C e = "loss concentration." Also, by continuity, the leakoff velocity, ve, must equal the boundary velocity, v Y ' which gives the general boundary condition as -(-CBVe+De aC)1 =k(1-q,)(C B -c eq )m+vf Ce' ... (7) ay y=bl2 SPE Production & Facilities, February 1993 c I CBvi y -------x I h I O ~ i ~ 7---_ -- / --- - -- 7------- Fig. 1-Physlcal system for acid-transport model. The only unclear issue in this equation is the relationship between C B and Ce. Note that this issue disappears if leakoff is zero. Most authors 5 - 7 ,lO assume that CB=C t . .......................................... (8) However, analysis of a control volume at the fracture wall, assum- ing that all reaction occurs on the surface, suggests that the acid concentration would have a discontinuity at the surface so that Ct<C B . The problem, as posed here, does not allow calculation of this extra variable without formulating and solving the equations for acid flow in the fracture wall. Eq. 8 is correct for a zero leakoff, very slow reaction, in which case the reaction term is negligible compared with the leakoff terms; and a very high reaction rate, which will force C B to be close to zero and consequently Ct"",CB"",O. Note that in the second case, the concentration will remain constant (even in a variable-cross- section channel), which follows from the continuity equation. This is analogous to the problem of heat transfer in a fracture with no leakoff and zero conduction and serves as a test for the model. Therefore, Eq. 8 is adopted, giving -D e(oCloY)!y=bl2 =k(I-q,)(C B -c eq )m . .............. (9) Application of the boundary condition in Eq. 9 requires the y- direction concentration gradient, which would force us to do a 2D solution (such as in Ref. 17) or to use some analytical approxima- tions. To avoid this, we use the parametric apparent mass-transfer coefficient, Kg, proposed by Roberts and Guin 7 : Kg(C-C B )= -De (oCloy) !y=bl2, .................... (10) where C=the average concentration across the fracture width. Note that Kg has the same units as velocity and can be related to meas- urements more easily than De' The value of Kif will depend on flow regime (thickness of the boundary layer) and IS discussed later. This coefficient is similar but not identical to the K coefficient de- fined by Domselaar et aI., 6 but they call it an apparent rate constant: KC=k(1-q,)C B
The equivalence of K and Kg follows from Eq. 9 if the reaction rate is high so that CB""'O. Eq. 3 can now be integrated across y and z to yield SPE Production & Facilities, February 1993 2.I,-----------------t----, 1.1 10.0 .. .. 1.5 e t.> ~ 5.' I:: 0 I 2.5 3 t.t t.M '.15 .51 CUM.A.ClD (LBJIPT2) Fig. 2-Measured and simulated leakoff with pad and acid fluids-Laboratory Experiment 1. where hfr = reactive height. Substi!!Iting the gradient at the bound- ary through Eq. 10 in terms of C yields the final form: o _ 0 _ _ --(vAcC)=-(AcC)+2[qeCB+hfrKg(C-CB)]-iC; . ... (12) Ox ot The boundary condition is then Kg(C-CB)=k(l-q,)(CB-Ceq)m, ................... (13) and Eqs. 12 and 13 must be solved simultaneously for C and CB' Eqs. 2, 12, and 13 are used for computer simulation. Note that the development is generalized easily to compressible equations that are solved in the model. (In fact, Eq. 2 is the result of fracture flow and geometric calculation.) The total loss of acid by reaction and leakoff in Eq. 12 is qAt =2[qeCB +hfrKg(C-C B )], .................. - ... (14) which is discussed in more detail later because it must include the effects of reaction on leakoff and worrnholing. Evaluation of Mass-Transport CoeHlclents The default method is based on the correlations for De' developed from laboratory experiments by Williams and Nierode 5 and Roberts and Guin. 7 Because De cannot be measured directly, Wil- liams and Nierode determined the overall mass transport and then used analytical solution to obtain oCloy to evaluate De from the integrated form of Eq. 10. If De is known, the overall mass transfer can be expressed by Eq. 10, and the coefficient Kg will be a function of De and the flow regime. Such solutions can be expressed in dimensionless form as follows (see Ref. 18). In turbulent flow (N Re >7,000): NSh =0.026(N Re )O.8(N sc ) II! . ....................... (15a) In transitional flow (1 ,800 <N Re <7,(00): NSh =0.0011038(N Re ) 1.1532 (N sc ) II!. . ............. (l5b) In Eqs. 15, NSc=plDe is the Schmidt number; N Re =2bV/p is the Reynolds number; NSh =2bKglD e is the Sherwood number; p=pJp is the kinematic viscosity; and v is the average velocity in the chan- nel. If De is known (such as from Ref. 5), Eqs. 15 allow the cal- culation K g =N Sh D e /2b. Older literature does not give the corresponding dimensionless equation for the laminar flow region. The numerical solutions from Ref. 13 seem to indicate that the Sher- 31 5 ,.--------------______ --, _ 4.' ! .. 3.5
0.0"'0 S-Q-+:-TC-M-IN-) 71----+----1' Fig. 3-Measured and simulated leakoff with pad and acid fluids-Laboratory Experiment 2. wood number is fairly constant, but Roberts and Guin'sl work in- dicates strong influence of free convection (see their Fig. 6). Un- fortunately, such results were not correlated with the Reynolds number. Lo and Dean 10 recently presented a correlation for the laminar flow region. They define the Nusselt number as NNu = -(bIC)(aClaY)!y=bI2 ........................ (16) and assume that C B =0. Their definition can be generalized for the case of C B 0, which makes it equivalent to the Sherwood number: NNu = - _ b ac I =bKglD e =N Sh12. . ........ (17) (C-C B ) ay y=bl2 Ref. 10 then correlates N Nu with the Peclet number, N Pe = v f bl2D e , by NNu =4.10+ 1.26Npe .................. " ... (lSa) if NPe < 10 and NNu =2N pe .................................... (lSb) if NPe where the constant a2 (quoted as 0.04 in Ref. 10) must be a2 =0.02675 to maintain continuity. Eqs. 15 and IS are used in the model by default. However, they predict a large (an order of magnitude) discontinuity in Kg at the boundary between lami- nar and transitional flbw, which makes them far from satisfactory. Also, the correlating dimensionless groups change at this point. Because of the difficulties of a priori prediction of mixing, the best method of determining the mass-transport coefficients is by simulating the experiments for acid flow between parallel reactive surfaces and matching the measured acid concentration profile. For this purpose, a separate model was developed. This model simu- lates results of a laboratory experiment with a fracture of constant width and with prescribed leakoff. The matched values of Kg and k can then be used for field predictions. The model also can be used for field-size simulation with a fracture of constant width and prescribed leakoff, which is comparable to earlier design models. Leakoff and Wormhollng The increase of leakoff from the combined effect of matrix acidiz- ing and wormholing is believed to be a primary reason for the gener- ally short penetrations of acid fractures. The increase of leakoff in laboratory cores has been measured by a number of investiga- tors, 15,16,19 and an example of actual data is presented below. In the field situation, the increase measured in the laboratory will af- fect primarily the invaded (C l ) region. Therefore, one would ex- pect field leakoff to increase to the same degree in gas reservoirs 32 . , 00 : or - . . ,. , ' . . ,Vl m
o ,f o o. , ., . . \ ; C8 ------ .... -.--- ---- --'- c Fig. 4-Representation of wormhole leakoff in the model. and to a somewhat lesser degree in oil reservoirs. Nierode* and King** found that the increased leakoff was necessary for agree- ment between designed and measured fracture lengths. Hucka- bee,20 however, reported that field testing indicated no increase in the overall leakoff coefficient in an oil reservoir. The wormholing effect is particularly difficult to predict analyti- cally. Hung'sl3 method allows calculation of single wormhole growth and simulation of the competition of several wormholes, but it makes simple assumptions about the pressure field. The cal- culations also depend on the knowledge of statistical distribution of pore sizes, which requires laboratory work. Daccord et at. 14 used fractal concepts to characterize the wormhole pattern. While theoretical models of wormholing may become predictive in the future, the laboratory leakoff measurement currently is the means of quantifying measured leakoff for modeling. The method here is proposed to express the increase in reacting fluid leakoff com- pared with leakoff of an inert fluid. Then acid leakoff is simulated with the general model of Ref. 21 and increased in this ratio. Method implementation is as follows. 1. Generate a theoretical prediction ofIeakoff velocity, vf, vs. time from the leakoff model used in the fracturing simulator, with realistic data for the acid fluid but without acidizing effects. Such a forecast can be obtained by running a laboratory leakoff test with inert fluid properties and matching it with the model developed spe- cifically to simulate leakoff data. 2. In the laboratory, measure the leakoff velocity, (v f ) A' for the acid. From the comparison of (vf)A measured in the laboratory with the theoretical inert fluid leakoff ve for the same fluid, deter- mine the ratio RAf = (ve)A IVe=f(MA)' ............................. (19) where MA = some measure of the cumulative amount of acid con- tacting the fracture wall. 3. When the function RAe has been established, calculate the leakoff velocity in the fracture-geometry model: (ve)A(x,t)=R Ae ( ,)ve(x,t), ........................ (20) where ve is computed the usual way (without the effect of acidizing) . In the model, RAe can be correlated with one of two quantities that are internally computed as part of the solution: RAf = f(M Af ) , where M Ai is the cumulative mass of acid loss to the formation per unit area; or RAf=f(M At ), where MAt is the cumulative mass of acid contacting the wall per unit area (Le., acid spent and leaked off). Note that, although the method of representing the data through RAf was motivated by an increase in leakoff, it also can be used to model a decrease that may result from pore collapse or solid precipitation with some acids. The function RAf can be determined by matching two laboratory measurements. An example is shown 'Personal communication with D.E. Nierode, Exxon Production Research Co., Houston (1988). "Personal communication with G.E. King, Amoco Production Co., Tulsa (1987). SPE Production & Facilities, February 1993 TABLE 1-THERMAL REACTION DATA FOR COMPARISON WITH LEE AND ROBERTS18 DATA (LIMESTONE AND DOLOMITE FORMATIONS) Acid type Mass-transfer coefficient, Kg Reaction constant, em/sec Reaction order Equilibrium concentration, % Dissolving power, kg/kg acid Density of soluble rock material, kg/m 3 Activation energy, kJ/kgmol Heat of reaction, kJ/kg Reference temperature for reaction data, C Key to add loss velocity to Kg in Fig. 2, where a laboratory run on a low-permeability gas reser- voir core with the pad fluid was followed by a displacement with acid. The measured and simulated leakoffvolumes from these two runs are shown. The simulated match was obtained with the func- tion RAi shown. After the match, a second laboratory test on a different core was simulated and matched with the same function R Ai . This result, shown in Fig. 3, supports the validity of the con- cept expressed by Eq. 19. For field application, however, reduc- ing the effect in liquid-filled reservoirs may be appropriate if the leakoff can be calibrated from good minifracture data. Further im- provement of this method is possible by applying the correction only to the C 1 region in the field situation. Relation Between Leakoff and Mass Transfer. When the model was tested with increased leakoff, it was found that, for certain low values of mass transfer,' the computed acid concentrations in the fracture would increase above the input concentrations, which is physically impossible. This problem (also reported in Ref. 10) is caused by ignoring the relationship between Kg and Vi and by het- erogeneity of leakoff (wormholing). In the homogeneous-Ieakoff case (i.e., no wormholing), the phe- nomenon can be explained as follows. As discussed, the concen- tration in the fracture will remain constant if the total amount of acid leaving the fracture face is equal to the leakoff rate times the in-situ concentration, qAi=ViC. If we now have a reaction that is mass-transport-limited, C B will be close to zero (or Ceq), and the total amount of acid leaving is according to Eq. 14, viCB+ Kg(C-C B ), which potentially can be smaller than ViC. However, the mass-transfer coefficient is physically related to boundary-layer thickness, and the boundary layer will diminish with increasing leakoff. Therefore, Kg must increase with leakoff. To the best of my knowledge, this dependency has not been measured and reported in the literature; however, this atgUment can be used to give alimit- ing condition on the value of Kg: .......................... (21) The worst case is if CB=O; therefore, the above condition will al- ways be satisfied if ........................................ (22) Because most of the reaction data are obtained for low leakoff rates, one could argue that the leakoff velocity should be added to the measured data-i.e., one should use an effective coefficient, Kge=Kg+vi' .................................... (23) In the model, an option is provided either to increase Kg auto- matically to always satisfy Eq. 22 or to add leakoff velocity to Kg according to Eq. 23. In the presence of wormholing, one must also consider the het- erogeneity of leakoff. Once a wormhole develops, the leakoff into the wormhole is at a concentration close to C instead of C B' Also, the leakoff rate is nonuniform because most of the leakoff rate is through the wormhole (Fig. 4). Therefore, for calculation of the loss term, a weighted average of the two concentrations should be taken, which will restore physically correct mass balance. Then SPE Production & Facilities, February 1993 Limestone Formation HCI Correlation 4.129x10- 4 0.441 o 1.37 2710 0.23000 x 10 5 0.10910x 10 4 37.770 o Dolomite Formation HCI Correlation 1.252x 10- 3 0.669 o 1.37 2710 0.40480 x 10 5 0.10910x 10 4 93.330 o Kg can remain at a value consistent with the much lower matrix leakoff rate. The effect can be quantified by the use of the leakoff increase function, RAe, defined by Eq. 19. We assume that part of the in- crease results from increased matrix leakoff and part from flow into the wormhole. We can calculate the two components of leakoff. The total leakoff velocity is (vi)A =RAeve, .................................... (24) where Vi is the inert fluid leakoff. Therefore, the total increase is (R Ai -l) = (RAi -1)ve = . (vik ................... (25) RAe The wormhole leakoff velocity, V l'w' and the matrix velocity, vim' are (R Ai -l) vew=Fw!!.ve=Fw (Ve)A .................... (26a) RAi and Vim =(Ve)A -Viw' .............................. (26b) We can now reformulate the boundary condition for total flux of acid to the fracture surface (Eq. 14) if we assume that the leakoff to the wormholes is with the concentration C and the cross-sectional area of the wormholes is small so that the decrease in the reactive area ofthe matrix is negligible. Then Eq. 14 can be generalized to QAt=2[QimCB+qiwC+hjrKg(C-CB)]' ............... (27) Implementation of Eq. 27 requires only additional input of the wormhole fraction, F w' If F w is high, use of this equation elimi- nates problems with nonphysical concentrations. To guarantee con- sistency under any conditions, we must apply the argument given above to the matrix component. The conditions of Eqs. 22 and 23 change to ....................................... (28a) and Kge =K g +vim' ................................ (28b) Thermal Features of the Model The fracturing simulator that hosts the acid-transport model can be run in an isothermal or thermal mode. Correspondingly, the for- mulation of the acid model includes several features that are cou- pled with heat transfer. These features become active if the fracturing simulator is run in the thermal mode. Mass Transport. The internally generated mass-transport coeffi- cient, Kg, is a function of temperature. This is reflected in the cal- culation method of Eqs. 15 or 18 because De and the fluid PVT properties and viscosity that enter the dimensionless groups are func- tions of temperature. An alternative is to use an Arrhenius type of equation, as is customary for the reaction rate. The reaction-rate constant, k, is assumed to be a function of tem- perature according to the Arrhenius equation: k=ko exp(-EaIRT), .............................. (29) 33 "r---------------------------------------, 51 110 UMBSTONB 6DOLOMITB x NO BUT 01' .uenON Fig. 5-Temperature profiles with and without heat of reaction-limestone and dolomite formations. where Ea = activation energy constant. The value of k entered in the input data is assumed to be k, at the reference temperature, T,. Then the value at any other temperature can be computed as k(T) =k,exp[ Ea ( - . ..................... (30) R T, T Heat of Reaction. The heat generated from the reaction can sig- nificantly change the temperature and therefore the acid spending along the fracture. 9 ,18 The heat energy generated is related to the mass of acid reacted by Q,=k(1-cP)(c-c B )mH rxn =rHrxn' ................. (31) where Q, = rate of heat generation per unit area and H rxn = heat of reaction (energy/mass of acid reacted). In the model, the amount of acid reacted in each gridblock during timestep n, Wp, is com- puted and stored. During the solution of the heat balance (which follows the solution of acid transport), the heat generated in the fracture gridblock is W,H rxn' and the accumulation side of the energy equation is modified to include this term, which is essen- tially a "heat source" in the energy equation. The feature was tested in the model with Lee and Roberts'18 data. They solved two example problems assuming a limestone or dolomite formation, with other common data about the treatment shown in Table 1. Unfortunately, the reaction data are incomplete, and their model appears to be decoupled from fracture geometry (constant fracture width). Consequently, we cannot expect to cate their results. Using the interpretation of the reaction data in Table 1, we obtained the temperature distribution in the fracture shown in Fig. 5. The results are similar to Fig. 2 of Ref. 18, but the temperature peaks are closer to the wellbore. Numerical Implementation The above theory of acid transport and leakoff was implemented in three separate codes: stand-alone leakoff (LOSS), stand-alone acid-transport model (ACID), and a fully integrated fracturing simu- lator (FRACANAL). The two auxiliary programs are important parts of the system because they can be used to characterize the data necessary to make field predictions, as shown above for fluid- loss behavior. Acid-Balance Equation. The acid-balance equation is discretized with the same basic technique as for propp ant transport. In fact, the two problems are very similar because the equations for prop- pant concentration in suspension and acid concentration are analo- gous, and it is possible to maintain parallel structure of the two modules. The major difference is that there is no equivalent of set- tlement in acid transport. 34 TABLE 2-ACID-TRANSPORT DATA FOR LO AND DEAN'S EXAMPLE 10 Pumping Schedule Treatment volume, STB Pumping rate, bbl/min Acid concentration, oAl Reaction Data Acid type Type of formation Diffusion coefficient, De' cm 2 /sec Reaction constant, cm/sec Reaction order Equilibrium concentration, GAl Dissolving power, Ibm/lbm acid Density of soluble rock material, Ibm/ft 3 Isothermal Run Reference temperature for reaction data, OF 200 20 15 HCI limestone 1 x 10- 4 0.1 1.0 o 1.37 169.1853 60.000 Time-Independent Fracture-Fluid Rheology Newtonian viscosity, cp 100 Constant Fracture-Geometry Data Fracture height, ft Depth to fracture center, ft Fracture width, ft Fluid density, Ibmlft 3 Average leakoff velocity, ftlmin Wall porosity 100 3,936 o 62.4 0.0010 0.1500 Fracture grid in x direction: 35 blocks, Ax= 10 ft The problem is solved numerically as hyperbolic because main- taining the fronts between different fluids with no dispersion is de- sirable. This requires logic to track the position of a front within a gridblock and to subdivide the gridblock with the front into two elements for the solution of mass balance. The details of this proce- dure are given in Ref. 22 and apply directly to the acid model. The mandatory additional term is the rate of decrease of acid mass in a block caused by spending (reaction) and by loss, expressed by Eq. 14 or 27. These equations contain two unknowns, C i and C Bi' for each gridblock, but the boundary concentration can be eliminated. For example, the implicit discretization of Eq. 14 is (qAt)t+ 1 = 2 [qe; Cr 1 +Ai,Kg(C-CB)t+ 1 ) .. (32) The reaction boundary condition, Eq. 13, is discretized as Kg(C-CB)r+ 1 =k(l-cP)[(CB)r+l -ceq)m . ........... (33) If m = I, C B can be expressed as K k' (CB)r+ 1 =---g-Cr+l +-----Ceq=FbCr+l +F e , ... (34) Kg+k' Kg+k' where k'=k(l-cP). After elimination from Eq. 32, the final ex- pression is (qAt)r 1 = 2 [qfiF b +Ai,Kg(l-Fb)Cr 1 +2(qi -Ai,Kg)Fe' .................................... (35) This term can be incorporated into the finite-difference equations as an implicit sink/source term. For n I, the elimination cannot be carried out. However, the following linearization (by Ben Naceur*) solves the problem. Find an equivalent value of k' so that the reaction rate would be the same with m=l: k;(C B -Ceq)=k'(C B -ceq)m . ....................... (36) If C B in Eq. 36 is evaluated explicitly at the beginning of the timestep, the effective value can be computed and the elimination carried out as before. Personal communication with K. Ben Naceur. Dowell Schlumberger, Nigeria (1989). SPE Production & Facilities, February 1993 t.1. OWiDm oAVG.CONC. 0." xWALLCONC. .... i 7 '.tT i , ... , Z
.. Q
5 0.esS! ..
.... ! u De ..... 1 emll.ee .. Z .. ... U ,
'.11 0.'1
....
25 225 250 Fig. 6-Acid-transport solution-comparison with Ref. 10 (simple reaction data from Table 2). Integration of Acid Transport Into the Fracturing Simulator. The acid transport described here was integrated into a general simu- lator that includes options for 2D and 3D fracture geometry, 23,24 generalized leakoff treatment,25 and a number of other sophisti- cated features. 26-28 In the overall solution structure, the acid trans- port is solved sequentially after the geometry and leakoff solution have been obtained and before the heat-transfer solution has been updated. The acid module is essentially interchangeable with the proppant-transport module and can be integrated into the structure as follows. . 1. During the fracture-geometry/leakoff solution, the elastic frac- ture width, b, is increased by the acidized width, b:/. This is done only to calculate flow in the fracture (transmissibilities and flow velocities) and not for the fracture-opening equations. Also, the acid- ized width is not added to the mass-balance equations (to fracture elemental volumes); this corresponds to an assumption that the prod- ucts of reaction occupy the same volume as the dissolved rock. 2. The computed leakoff velocities are increased by the factor RAf determined by table look-up from the input data; on the basis of the amount of acid leaked off at the beginning of the timestep. The leakoff treatment of foams is different and accounts for the separation of the liquid and vapor components in porous media. 3. After the geometry/leakoff solution is completed, new acid flow velocities are computed (with a no-slip assumption) and the acid transport is solved. 4. After the new acid solution is obtained, the increase in acid- ized volumes of all blocks is determined and the cumulative acid masses lost and reacted for each gridblock are updated. The acid- ized volumes and widths b A are computed on a 2D grid with horizontal and vertical definition and are used for coupling with reservoir simulators as described in Ref. 26. 5. Temperature in the fracture is updated by solving the timestep of the energy-balance equation, including the heat of reaction com- puted from the acid-transport solution. At the end of the simulation, the geometry of the acidized frac- ture [i.e., bA(x,z)] is transferred to the reservoir simulator, and its conductivity is calculated as a function of the original acidized width and the local effective stress by Nierode and Kruk's 19 method or by tables obtained from laboratory measurements. Validation of Acid Transport Solution To illustrate the various aspects of the physics, a number of cases were simulated with the stand-alone ACID model and Lo and Dean'slO data. Table 2 shows the common data. Because the model of Ref. 10 assumes an infinite rate of reaction and C B = 0, a high value of reaction constant, k=O.1 cm/sec, n=l, and Ceq =0 are used here for comparison. Also, a constant value ofleakoffve- locity, vf =0.001 ft/min, was used. Fig. 6 shows the result with SPE Production & Facilities, February 1993 1 ...-------------------,0.11 OWlnm OAVG.CONC. XWALLCONC. 0." . ... '.tT 0."
Q o.os5! 5 o ... ! o.J
'.12 O.tl Fig. 7-Effect of increased leakoff, with n=0.5, Ceq =0.2% and k = 0.01 cm/sec. internal generation of mass-transfer coefficients. Because Ref. 10 used a constant value of De=O.OOOI cm 2 /sec, the corresponding solution with this value is also shown in comparison with Lo and Dean's results. The solution here is very close to the 2D solution of Ref. 10, which validates the method used to calculate mass trans- fer from De' The same agreement was obtained for the case without leakoff. To test other features (not included in the model of Ref. 10), the same basic set of data was used with modifications. The use of a more reasonable value for the reaction constant, k=O.OI cm/sec, results in higher boundary concentration and therefore smaller acid- ized width and demonstrates that the formulation of the model can handle acidizing controlled by mass transfer, reaction rate, or a combination of both. The use of nonzero Ceq results in unspent acid and much smaller acidized width. This may be important for treatments with weak acids or low concentrations. Having a reac- tion order lower than 1 increases the rate of reaction. This makes the result for k=O.OI and n=0.5100k more like the infinite-reacting system of Fig. 6. The feature is important because the measure- ments often show low values of n. The above parameters, although changing the fracture widths, do not have a large effect on penetration. The final, more realistic example shows the effect of increased leakoff. The reaction data were k=O.OI, n=0.5, and C eq =0.2, and the laboratory leakoff increase function shown in Fig. 2 was used. The result is a signifi- cant decrease in fracture penetration (Fig. 7). In higher-permeability reservoirs, the increase in leakoff can cause the fracture to stop growing or even to collapse, as shown later. The above cases show how the ACID model can be used to de- termine the sensitivity of acid penetration to various parameters. However, its most important use is to obtain the best reaction data for the field design by matching acid flow experiments on slabbed cores. This procedure, together with matching the leakoff data de- scribed previously, determines the key parameters for simulation of the performance of the fracture job. Example of Fracture Design As an example, consider treating a formation with properties given in Table 3. Two treatments will be compared here: (1) a job con- sisting of a gelled pad followed by straight 28 % acid and (2) a treat- ment consisting of gel/acid sequence. In both treatments, the total amount of acid used is the same (1,200 bbl), as shown in Table 4. Alternating the gel and acid has been claimed to reduce acid loss 29; here the effect is quantified through use of the model. The difference in the leakoff of straight acid as opposed to alternating gel and acid was investigated first on a laboratory scale. Fig. 8 shows the hypothetical laboratory runs simulated with the LOSS model. In both runs, the first pad was assumed to have a spurt loss and 35 TABLE 3-RESERVOIR DATA FOR THE FIELD EXAMPLE Rock-Mechanics Vertical stress, pSia Horizontal stress, psia Young's modulus, psi a Poisson's ratio Biol's constant Critical stress-intensity factor, K c ' psialJit Reservoir Properties Average pressure, psia In-situ horizontal permeability, md In-situ porosity Reservoir fluid viscosity, cp Total in-situ compressibility, pSia- 1 k, to reservoir fluid at Swc k, to filtrate in invaded zone In-situ water saturation, Swc Residual reservoir fluid saturation to filtrate Thickness, ft Depth to center of pay, ft Thermal Properties Initial reservoir temperature, of Density of fluid-saturated rock, Ibm/ft3 13,000 6,141 10,100,000 0.2500 0.6000 1,368.00 2,680.00 1.2000 0.0800 0.02060 0.000330 0.1830 0.5000 0.2500 0.1000 130.000 11,877.00 Heat capacity of fluid-saturated rock, Btullbm-oF Thermal conductivity of saturated rock, Btu/ft-D_oF Drained coefficient of thermal 167.0 130.00 0.21 50.00 expansion, 1/ o F OAOOx10- 4 a wall-building coefficient of C m =0.002 ft/min l->. As acid comes in contact with the fracture wall, the filter cake previously deposit- ed is assumed to be effectively destroyed, as shown experimentally in Ref. 15. Subsequent exposure to gel will build a new filter cake, but without spurt loss. Also, because the gel will enter the worm- holes preferentially, the decrease in filtrate viscosity is smaller and thus the effect of C 1 is larger. This was modeled by assuming a 90% viscosity screenout factor for the pad but only 50% for sub- sequent gel stages (see Ref. 21 for the formulation of viscosity screenout). Only the initial pad leakoff follows the classic theoretical curve. During the acid stage, leakoff accelerates; in the gel stage leakoff is retarded. As a result, the loss with the alternating fluids is about one-half that with straight acid. Similar improvement would be ex- pected in the field case, as Fig. 9 shows.
1. / ACIDONLY",/ / / / / '" /
/ / I / GBUACID / -< --./
./ ./ 0 ./ - -
./ - " ./ - 1i i 1,3
11 17 \\)(B (.flmn.d1 U 71 Sl ,. PAD ACID I GBI. I ACID GBI. I ACID Fig. 8-Slmulated laboratory leakoff experiment-comparison of continuous acid and alternating gel/acid schedule. 36 TABLE 4-TREATMENT DATA FOR FIELD EXAMPLE Pumping Schedule-Alternating Acid and Gel STC Volume Rate Concentration Fluid Acid Step (STB) (bbllmin) (%) Number Number --- 1 200.00 20.0000 0.00 1 - 2 400.00 20.0000 28.00 2 1 3 200.00 20.0000 0.00 3 1 4 400.00 20.0000 28.00 2 1 5 200.00 20.0000 0.00 3 1 6 400.00 20.0000 28.00 2 1 Pumping Schedule-Continuous Acid 1 200.00 20.0000 0.00 1 - 2 1,200.00 20.0000 28.00 2 1 Assumed fracture height, ft 400.00 Depth to fracture center, ft 11,877.0 Downhole injection temperature, of 85.0 Acid type HCI Type of formation Limestone Mass-transfer coefficient, cm/sec 1.0x 10- 3 Reaction constant, cm/sec 2.3x10- 3 Reaction order 0.7 Equilibrium concentration, % 0 Dissolving power, Ibmllbm acid 1.37 Density of soluble rock material, Ibm/ft3 179.17 Activation energy, kJ/kgmol 0.23 x 10 5 Heat of reaction, kJ/kg 0.10910x 10 4 Reference temperature for reaction data, C 60.00 Key to add loss velocity to Kg 1 Rheology for Fluids 1 and 3 (Gel) Maximum Minimum Time Viscosity Viscosity (minutes) K' n' (cp) (cp) --- --- 0.00 0.04000 0.5000 625.50 44.23 29.95 0.03000 0.5200 490.87 38.59 60.05 0.02000 0.5500 350.22 32.28 120.10 0.01000 0.6200 205.03 27.38 360.00 0.00100 0.8500 34.30 15.49 Rheology for Fluid 2 (Plain Acid): Constant Viscosity of 1.508 cp Fluid Number 1 2 3 C III coefficient, ft/min 112 0.002000 0.000000 0.002000 Filtrate viscosity reduction 0.900 1.000 0.500 Spurt loss/unit area for C III 0.01330 0.00000 0.00000 t.51 us 1M gl i!! 1.35 " lift .3t "" ... I
75
.. It S .- ." i!I y '" 45 15&1
3t '.10 15 "'5
. ...
1. 1t 3t (JtNUTdf 7. 10 ,. 1" PAD I ACID I GBI. I ACID I GBI. I ACID Fig. 9-Fracture hydraulic lengths and widths for acid Jobs with continuous acid and alternative gel/acid schedule. SPE Production & Facilities, February 1993 In the job with continuous acid injection, the fracture width dur- ing the acid stage decreases until the fracture closes. During this period, fracture length is stationary. Only about 470 bbl of acid could be pumped with the 200-bbl pad. This volume then presents the natural upper boundary for the optimum treatment size (which can be determined more accurately by coupled production simula- tions and economic analysis). For alternating fluids, fracture width increases in each gel stage, and the entire job can be pumped. The fracturing pressure follows the same oscillating pattern as the width (similar to results reported in Ref. 9). It is apparent that the oscillations caused by alternating the fluids tend to average out with time. This suggests that smoother behavior could be achieved with a larger number of smaller stages. Fig. 10 shows the resulting acidized fracture profiles. The solid shape is the acidized width at the end of the job; the line is the hy- draulic width (without the acidized contribution) at the end of pump- ing. The alternating-fluid design gives greater penetration and larger width, mostly because of the ability to place a larger volume. Even- with the large treating volume, however, the penetration depth is only about 100 ft or less. This is in general agreement with field experience. The model was used subsequently to simulate field cases where both pre fracture and postfracture pressure data were available and was found to be in excellent agreement with the fracture lengths derived from pressure-transient analysis. Conclusions A general model of acid fracturing and a method to characterize the leakoff in acid jobs were developed. The novel model features follow. I. The acidizing process can be controlled by both mass transfer and rate of reaction and allows arbitrary order of reaction and equi- librium concentration. 2. The model allows multiple fluids with different rheologies, particularly alternating gel/acid sequences. 3. The increase in leakoff caused by matrix acidizing and worm- holing effects at the fracture wall is an important factor for realis- tic calculation of acid penetration. A convenient method was developed for quantifying the effect experimentally and including it in the numerical model. 4. A method was developed to account for the influence of in- creased leakoff on mass-transfer coefficients. The modeling system developed has three features: (I) auxiliary programs for modeling leakoff and reaction laboratory experiments, used to derive the necessary data by history matching; (2) the main acid-fracturing model, implicitly integrated with 2D or 3D frac- ture geometry and heat-transfer calculation; and (3) coupling with reservoir forecasting models, which allows optimization ofthe de- sign or comparison with a proppant treatment. With proper data, the design model predicts small acidized lengths consistent with field observations. Future model extensions will in- clude the treatment of viscous fingering for ungelled acids and com- bined acidlproppant designs. Nomenclature A = elemental area, L2 Ac = cross-sectional area of fracture, L2 Air = reactive area in Block i, L2 b = fracture width, L C = concentration, mlV C r = leakoff coefficient for reservoir, LIYt Crn = leakoff coefficient for filter cake, LIYt De = effective diffusion coefficient, L2 It Ea = activation energy, Elm Fb,Fe = terms defined by Eq. 3 F w = fraction of ieakoff increase resulting from wormholing h f = fracture full height, L hfe = fracture leakoff height, L h fr = fracture reactive height, L SPE Production & Facilities, February 1993 .... ACID ONLY ' 1 .... .11 ... .... GEL/ACID
1 == "" .... .::.
.::. .11 III toOl is 0 -I.H ..: ... -.... DISTANCE ALONG FRACTURE 1FT) Fig. 10-Fracture acldlzed width profiles for acid jobs with continuous acid and alternating gel/acid schedule. H rxn = specific heat of reaction, Elm i = injection rate for one wing, V It k = reaction-rate constant, Lit K = apparent rate constant from Ref. 6, Lit Kg = apparent mass-transfer coefficient from Ref. 7, Lit L f = fracture half-length, L NNu = Nusselt number, bKgID e NPe = Peclet number, v[b/2D e N Re = Reynolds number, 2bV/p Nsc = Schmidt number, plDe NSh = Sherwood number, 2bK g ID e q = rate, Vlt R = gas-law constant RAe = ratio of leakoff velocity increase caused by acid t = time T = temperature ve = leak-off velocity, Lit Vx,y = flow velocity components, Lit v = average, velocity in fracture, Lit 'Y3 = shape coefficient for fracture cross-sectional area p, = fluid viscosity, cp p = kinematic viscosity of fluid, p=p,lp, L2/t p = density, mIY cf> = porosity Subscripts A = acid B = boundary (fracture wall) D = dimensionless e = effective eq = equilibrium i = injection = leakoff (loss) m = matrix r = reference t = total w = wormhole Superscripts m = reaction order n,n+ I = beginning and end of timestep = average 37 Author r ~ Antonln Sattarl, president of Simtech Consulting Services in Calgary, develops computer simulation techniques for reservoir modeling and hydraulic fractur- ing. Before establishing Simtech, he was with Intercomp Resource Development & Engineering Ltd. in Calgary, London, and Houston. He has been involved in a wide range of research and engineering applications in reservoir simulation, en- hanced recovery, hydraulic fracturing, in-situ thermal processes in oil sands, and completion engineering. Coauthor of Petroleum Reservoir Simulation, Settari also was a 1989 Distinguished Lecturer. He holds a BS degree from the Technical U. of Brno and a PhD degree from the U. of Calgary, both in mechanicai engi- neering. Acknowledgments This research was done in the course of Simtech's maintenance of the FRACANAL software and was supported by the FRACANAL User Group. I express my appreciation to a number of people who contributed their ideas, particularly Kamel Ben Naceur,Dale Nier- ode, Curtis Crowe, and M. Tambini. References 1. Roberts, L.D. and Guin, J.A.: "The Effect of Surface Kinetics in Frac- ture Acidizing," SPEJ (Aug. 1974) 385-95; Trans., AIME, 257. 2. Williams, B.B. et at.: "Characterization of Liquid-Solid Reactions. Hydrochloric Acid-Calcium Carbonate Reaction," Ind. & Eng. Chern. Fund. (1970) 9, No.4, 589. 3. Nierode, D.E. and Williams, B.B.: "Characteristics of Acid Reaction in Limestone Formations," SPEJ (Dec. 1971) 406-18; Trans., AIME, 251. 4. Nierode, D.E., Williams, B.C. , and Bambardieri, C.c.: "Prediction of Stimuiation From Acid Fracturing Treatments," J. Cdn. Pet. Tech. (Oct .-Dec. 1973) 31-4l. 5. Williams, B.B. and Nierode, D.E. : "Design of Acid Fracturing Treat- ments," JPT (July 1972) 849-59; Trans., AIME, 253. 6. Domselaar, H.R., Schols, R.S. , and Visser, W.: "An Analysis of the Acidizing Process in Acid Fracturing," SPEJ (Aug. 1973) 239-50; Trans., AIME, 258. 7. Roberts, L.D. and Guin, J.A.: "A New Method for Predicting Acid Penetration Distance," SPEJ (Aug. 1975) 277-85. 8. Williams, B.B., Gidley, J.L., and Schechter, R.S.: Acidizing Fundamen- tals, Monograph Series, SPE, Richardson, TX (1979) 6. 9. Ben Naceur, K. and Economides, MJ.: " Design and Evaluation of Acid Fracturing Treatments, " paper SPE 18978 presented at the 1989 SPE Rocky Mountain Regional /Low Permeability Reservoirs Sympo- sium, Denver, March 6-8. 10. Lo, K.K. and Dean, R.H.: " Modeling of Acid Fracturing, " SPEPE (May 1989) 194-200; Trans., AIME, 287. 11. Gdanski, R.D. and Lee, W.S.: "On the Design of Fracture Acidizing Treatments," paper SPE 18885 presented at the 1989 SPE Production Operations Symposium, Oklahoma City, March 13-14. 12. Hung, K.M., Hill, A.D., and Sepehrnoori, K.: "A Mechanistic Model of Wormhole Growth in Carbonate Matrix Acidizing and Acid Frac- turing, " JPT (Jan. 1989) 59-66. 13. Hung, K.M.: "Modeling of Wormhole Behavior in Carbonate Acidiz- ing," PhD dissertation, U. of Texas, Austin, TX (May 1987). 14. Daccord, E., Touboul, E., and Lenormand, R. : "Carbonate Acidizing: 38 Toward a Quantitative Model of the Wonnholing Phenomenon," SPEPE (Feb. 1989) 63-68; Trans. , AIME, 287. 15. Crowe, C.W., Hutchinson, B.H. , and Trittipo, B.L. : " Fluid-Loss Con- trol : The Key to Successful Acid Fracturing, " SPEPE(Aug. 1989) 215- 20; Trans., AIME, 287. 16. Crowe, C.W., McGowan, G.R. , and Baranet, S.E.: "Investigation of Retarded Acids Provides Better Understanding of Their Effectiveness and Potential Benefits," SPEPE (May 1990) 166-70. 17. Chang, Ch.-Y. Guin, J.A., and Roberts, L.D.: "Surface Reaction With Combined Forced and Free Convection," A IChE J. (March 1976) 22, No. 2, 252-59. 18. Lee, M.H. and Roberts, L.D.: "The Effect of Heat of Reaction on Temperature Distribution and Acid Penetration in a Fracture," SPEJ (Dec. 1980) 501-07. 19. Nierode, D.E. and Kruk, K.F.: "An Evaluation of Acid Fluid Loss Additives, Retarded Acids, and Acidized Fracture Conductivity," paper SPE 4549 presented at the 1973 SPE Annual Meeting, Las Vegas, Sept. 30-0ct. 3. 20. Huckabee, P.T.: "Carbonate Stimulation Optimization Using Hydraulic Fracturing Field Testing, " paper SPE 18224 presented at the 1988 SPE Annual Technical Conference and Exhibition, Houston, Oct. 2-5. 21. Settari, A.: "A New General Model of Fluid Loss in Hydraulic Frac- turing," SPEJ (Aug. 1985) 491-501. 22. FRACANAL-Technical Description, Simtech Consulting Services Ltd., Calgary (1983-88) Chap. 1, Sec. 3. 23. Settari, A.: "Quantitative Analysis of Factors Influencing Vertical and Lateral Fracture Growth," SPEPE (Aug. 1988) 310-22. 24. Settari, A. and Wong, S.K. : "Modeling of Fracture Propagation in Layered Reservoirs," paper SPE 18958 presented at the 1989 SPE Rocky Mountain Regional/Low Permeability Reservoirs Symposium, Denver, March 6-8. 25. Settari , A.: "General Model of Fluid Flow (Leakoff) From Fractures Induced in Injection Operations," paper SPE 18197 presented at the 1990 SPE Annual Technical Conference and Exhibition, Houston, Oct . 2-5. 26. Settari, A., Puchyr, P.J., and Bachman, R.C.: "Partially Decoupled Modeling of Hydraulic Fracturing Processes," SPEPE (Feb. 1990) 37-44. 27. Settari , A., Bachman, R.C., and Morrison, D.B.: "Numerical Simu- lation of Hydraulic Fracturing Treatments With Low-Viscosity Fluids," J. Cdn. Pet. Tech. (Sept.-Oct. 1987) 31-40. 28. Settari, A.: " Modelling of Fracture and Deformation Processes in Oil Sands, " Proc. , Fourth UNIT ARIUNDP Conference, Edmonton (1988) paper 43. 29. Coulter, A. W. et at. : "Alternating Stages of Pad Fluid and Acid Pro- vide Improved Leakoff Control for Fracture Acidizing," paper SPE 6124 presented at the 1976 SPE Annual Technical Conference and Ex- hibition, New Orleans, Oct. 3-6. SI Metric Conversion Factors bbl x 1.589873 E-Ol m 3 Btu x 1.055056 E+OO kJ cp x 1.0* E+OO mPa's ft x 3.048* E-Ol m ft2 x 9.290304* E-02 m 2 ft3 x 2.831 685 E-02 m 3 OF (OF-32)/1.8 =oC in. x 2.54* E+OO cm Ibm x 4.535924 E-01 kg psi x 6.894757 E+OO kPa psi- 1 x 1.450377 E-Ol kPa- 1 Conversion factor is exact. SPEPF Original SPE manuscript received for review April 15, 1991. Revised manuscript received June 24, 1992. Paper accepted for publication Aug. 10, 1992. Paper (SPE 21870) first presented at the 1991 SPE Rocky Mountain Regional/Low Permeability Reservoirs Sym- posium held in Denver, April 15-17. SPE Production & Facilities, February 1993