Classic fractional flow theory is generalized, starting with the Buckley-Leverett theory for waterflooding. Physics and how the problems are treated are analogous to the multicomponent adsorption problem of chromatography. The alcohol flooding, enriched gas, and carbon dioxide cases are analogous in many respects to the micellar / polymer case.
Classic fractional flow theory is generalized, starting with the Buckley-Leverett theory for waterflooding. Physics and how the problems are treated are analogous to the multicomponent adsorption problem of chromatography. The alcohol flooding, enriched gas, and carbon dioxide cases are analogous in many respects to the micellar / polymer case.
Classic fractional flow theory is generalized, starting with the Buckley-Leverett theory for waterflooding. Physics and how the problems are treated are analogous to the multicomponent adsorption problem of chromatography. The alcohol flooding, enriched gas, and carbon dioxide cases are analogous in many respects to the micellar / polymer case.
Gary A. Pope, SPE, U. of Texas Abstract Classical fractional flow theory is generalized, starting with the Buckley-Leverett theory for waterflooding. The mathematics are based on the method of characteristics. The physics and how the problems are treated are analogous to the multicomponent adsorption problem of chromatography; thus, the methods of Helfferich apply. Enhanced oil-recovery (EOR) cases con- sidered are polymer flooding, carbonated water- flooding, hot waterflooding, hydrocarbon miscible flooding, low-tension flooding, and micellar/polymer flooding. The alcohol flooding, enriched gas, and carbon dioxide cases are analogous in many respects to the micellar/polymer case. Finally, three-phase flow problems are treated. Three-phase flow occurs in a variety of EOR processes. Introduction Fractional flow theory has been applied by various authors to waterflooding, 1-3 polymer 4 carbonated waterflooding,5,6 alcohol flooding, ,8 miscible flooding, 9 steamflooding) 10 and various types of surfactant flooding. 11 - fb Many of the assumptions made by these authors are the same and are necessary for obtaining simple analytical or graphical solutions to the continuity equations. Typically, the major assumptions, which are sometimes not stated explicitly, are: (1) one dimensional flow in a homogeneous, isotropic, isothermal porous medium, (2) at most, two phases are flowing, (3) at most, three components are flowing, (4) local equilibrium exists, (5) the fluids are incompressible, (6) for sorbing components, the adsorption isotherm depends only on one component 01977520/80/00067660$00.25 Copyright 1980 Society of Petroleum Engineers JUNE 1980 and has negative curvature, (7) dispersion is negligible, (8) gravity and capillarity are negligible, (9) no fingering occurs, (10) Darcy's law applies, (11) .the initial distribution of fluids is uniform, and (12) a continuous injection of constant composition is injected, starting at time zero. Several of these assumptions are relaxed easily. One of the most useful to relax is Assumption 12, continuous Injection. The principles of chromatography can be applied to analyze the more interesting case of injecting one or more slugs. Most of these processes require slug injection of chemical or solvent to be economical. In fact, a lower bound on the slug size necessary to prevent slug breakdown can be obtained from a simple extension of fractional flow theory. In this and other extensions the common new feature is the need to evaluate more than one characteristic velocity. A second example of this is the extension of fractional flow theory from simultaneous immiscible two-phase flow (the classical Buckley-Leverett water flood problem) to simultaneous immiscible three-phase flow (the classical oil/water/gas flow problem). A third example is the extension to nonisothermal cases. Here we need to consider the energy balance, mass balance, and velocity of a front of constant tem- perature. A fourth example is when one or more components are partitioning between phases. In all cases, mathematically, the extension is analogous to the generalization from the one-component ad- sorption problems (or two-component ion exchange problems with a stoichiometric constraint) to multicomponent sorption problems. 17 The latter theory has been worked out in a very general way for many com9onent systems using the concept of coherence. 1 Pope et al. 18 recently have applied this theory to reservoir engineering involving sorption problems. 191 Considering the classical Buckley-Leverett problem, I the continuity equation for water can be written asw + -.!L dfw asw =0. at . A(j> dS w ax ............. (la) It follows directly that the velocity of a front of constant saturation, the characteristic velocity, is ( dX) = -.!L dfw . . .............. (lb) dt SW A(j> dS w Direct integration of this single ordinary differential equation would solve this problem [give Sw =Sw (X,t)], except that, as is well-known, a "shock front" may form. This happens if saturation velocities, V sw , upstream are greater than those downstream. This is true for most oil/water frac- tional flow curves between certain limits of saturation, depending on the curvature of f W. Between such limits, an overall material balance gives the velocity of the shock, or saturation discontinuity. 192 1- Sorw Swr Sw (a) Fractional Flow Diagram 1- Sorw o Xo (b) Satu ration Profi Ie Before Water BT Np I-S orw - 5 wr o to (c) Oil Production Curve Fig. 1-Waterflooding. q fW2 -f wl v LlS w = A(j> Sw2 - Swl ................ (2) At the contact between the shock and the continuous saturation distribution, these velocities must be equal and given by fW2 - fwl = dfw I _ .............. (3) dSw SW- sw2 Waterflood For the typical waterflood case (Fig. 1), Swl is the known initial, uniform, irreducible water saturation, SWP and f wi' therefore, is zero. Thus, Eq. 3 yields the value for Sw2, or Sw2 Ofln be obtained graphically (Fig. 1) by constructing a tangent line to the frac- tional flow curve originating at Swr since Eq., 3 is a straight line with slope df wldSw and intercept (SwpO). Until breakthrough of the shock front 100070 oil is produced. The breakthrough (BT) time ex- pressed in pore volumes is just the reciprocal of Eq. 3. After BT, the Welge integration 2 gives the oil recovery implicity as at dfw )-1 tD=( dS w ' ....................... (5) In the above case the initial water saturation, SWi' was taken to be SWP but Eq. 5 holds for any value of Swi up to 1 - Sorw. N can be expressed explicitly as a function of tD for lhe case of straight-line relative permeabiIities, since for this case the f w curve does not have an inflection point so that no shock front forms and since Eq. 5 is simply a quadratic that can be inverted to give and used in Eq. 4. The velocity of the injected water is given by the slope of the tangent from the origin. This line intersects withf w at (s w3,f w3) in Fig. 1. Polymer Flood The next step up in complexity is polymer flooding, where a very small amount by weight of water- soluble polymer is added to the water to increase its viscosity and increase efficient displacement of oil. Polymers also are used in micellar/polymer flooding, a more complicated process involving surfactants and typically applied as a tertiary process - i.e., one intended to mobilize capillary-trapped or residual oil. Polymer flooding is used to enhance the water- flooding process and is essential for an economic water flood when the oil viscosity exceeds a certain value. Polymer flooding greatly improves areal and vertical sweep efficiencies and displacement (or microscopic) efficiency. The improvement in all three is important and must be considered to evaluate properly the merit of a given polymer flood. However, displacement is the only efficiency con- sidered here, since it is evaluated from fractional flow theory. The polymer transports only in the aqueous phase (no partitioning to oil). Thus, the continuity equation SOCIETY OF PETROLEUM ENGINEERS JOURNAL for it or any other component of a similar nature can be written a(ct>iswCi) a [ ] q . at + at (l - ct P sA i + A a(cJw) . =0 ............... (6) ax Defining an adsorption C i in units of amount adsorbed per unit pore volume, C. = (l- ctPSAi I ct> , ..................... (7) and a(SwCi) ct> aC i q a(cJw) --"--'-- + - - + - = o. . .. (8) at ct>i at Act>i ax The continuity equation for water is a(swCw) q a(cwfw) ---"------'-'--- + - =0 ........... (9) at Act> ax Making the excellent approximation that the volume fraction of the polymer is negligible com- pared with the water (setting C w = l) and expanding Eq. 8 leads to this important simplification: ( Sw+ dCi ) aC i + qfw aC i =0. d ....... (10) C i at Act> ax Here we also have used Assumption 4 and have taken ct>i=ct> Now it again follows directly that ( dX) = q f w , ........... (lla) dt c Act> Sw+Di I where the general retardation term, D i , replaces dC;ldC i , since in the case of a self-sharpening polymer concentration front a polymer con- centration shock forms and the polymer con- centration jumps from zero (its initial value) to its injection value. Di for such a case is simply Cio Di= - .......................... (lIb) C io This is just the amount of polymer adsorption ex- pressed in pore volumes, and the polymer front lags by the factor D1 (see Eq. 11) over what it would otherwise be. Generally, two saturation shocks form during a polymer flood (Fig. 2). The first shock forms as the water saturation increases from its initial value, since (as in a waterflood) the saturation velocity is in- creasing going upstream. The second saturation shock forms at the polymer concentration front- i.e., where polymer water contacts initial (connate) water. These are completely miscible fluids. Using Assumption 6, the displacement between the two waters will be ideally sharp or piston-like. Therefore, at this front the velocities of the saturation shock, water, a front of constant saturation of water, and polymer must all be equal. Therefore, fW3 -fW]. _fW]. _ dfw I _ fW3 SW3 -SW3 - SW]. - dS w SW=Sw3- SW3 +Dp , ........................... (12) JUNE 1980 where f W2 is evaluated from the polymer fractional flow curve (the one on the right in Fig. 2), which differs from the water/oil curve only in that the aqueous phase viscosity is greater than that of water due to the polymer. The principle that velocities must be equal at the contact between a pair of miscible fluids recurs in subsequent examples. In this case, solution of Eq. 12 for SW.2 and SWJ gives the values of the water saturatIOn for the upstream side of the first and second saturation shocks, respectively. The solution can be obtained graphically by drawing a straight line from the point (- D p' 0) tangent to the polymer fractional flow curve (Fig. 2). The oil production is 100070 until BT of the first shock, then at constant cut f W2 in the connate water/oil bank until BT of Shock 2, and finally sweepout at decreasing oil cut as per Eqs. 4 and 5 applied to the polymer fractional flow curve, since the saturation velocities are given by q dfw vsw = Act> dS w ....................... (13) for Sw > Sw3. For cases where the initial water fw (Sw3,f w3l OIL AND WATER WITH POLYMER --..... OIL AND WATER -Dp 0 Swr Sw (al Fractional Flow Diagram I-S orw Sw I I Swr POLYMER WATER..L.. CONNATE WATER ,I , o I Xo (bl Saturation Profile Before Water BT Np I-S orw -Swr POLYMER BT o to (cl Oil Production Curve Fig. 2 - Polymer flooding. 193 saturation is greater than SwP the above result requires slight modification. The first shock is from Swi >Swr to Sw2' and the oil cuts change correspondingly. Low-Tension Flood Consider the application of a low-tension process to the case where S wi = S wr' This is generally not the starting point for such a process since it is meant to mobilize trapped oil after a waterflood or polymer flood; however, the analysis is a trivial extension of the polymer flood case, and so will be discussed briefly. Actually, there are technical and economic incentives to apply such a process as soon after primary as possible, but this generally is not done. By definition, for the purpose at hand, we will give the low-tension process the following characteristics. First, it is an aqueous process involving no transfer of chemical (the chemical is surfactant in practice) to the oil. Second, it is a low-concentration process involving negligible volume fraction of chemical in the aqueous phase. Third, the chemical lowers the interfacial tension enough to make the capillary number high enough to detrap oil, although not necessarily all of it. Fourth, from a practical point of view, polymer must be added to the chemical solution for mobility control. (When we consider slug processes, we will assume also that polymer has -os 0 I-Sore Sw lOW TENSION 194 Np So; SECON DARY TERTIARY o Swr Sw Swr Sw (a) Fractional Flow Diagram I-Sore.;t I-S orw Sw2 Sw Sw2 Xo I lOW TENSION I W A T E ~ ~ CONNATE WATER o I Xo (b) Saturation Profiles Before Oil Bank BT Np Soi to to (c) Oil Production Curves Fig. 3- Lowtension flooding. been added to the drive water or "mobility buffer.") Provided only that we have accounted properly for the resulting change in the chemical solution viscosity due to the polymer addition and that the amounts of polymer and surfactant have been selected such that ideally no chromatographic separation occurs between them (Ds=Dp), for the moment we can just consider a single low-tension fractional flow curve. This curve will be displaced to the right just as the polymer alone curve was, but now it will terminate at Sore rather than Sorw (Fig. 3). By precisely the same analysis as for the polymer flood case we find that two shocks form, followed by an oil "tail," but now the final oil saturation ends at Sore' not Sorw. The graphical construction is shown in Fig. 3. The secondary case is shown on the left, and the tertiary case is on the right. The only dif- ference between the two cases is that the water saturation jumps from Swr to Sw2 in the secondary case and from 1- Sorw to Sw2 in the tertiary case. This results in much more recovery in the secondary case even though the final oil saturation is the same (Sore) in both cases. The production curves (Fig. 3) have been normalized by the initial oil saturation (Soi)' If the viscosity of the injected chemical solution is increased, the low-tension fractional flow curve shifts to the right (Fig. 4). Comparing Figs. 3 and 4 shows that the oil recovery efficiency continues to improve as the viscosity of the chemical solution is increased even beyond the point of theoretical hydrodynamic stability (no fingering). This sur- prising result becomes very important when con- sidering the practical slug case, especially in the tertiary application of the process. Stated another way, we have an incentive to increase viscosity, depending on shape of the relative permeability curves, to higher values than required for unit mobility ratio, the classical mobility control design criterion. At some sufficiently high viscosity the tangent line will intersect the low-tension curve at or very close to Sw = 1- Sore' and no additional im- provement will result for still greater viscosities. Hydrocarbon Miscible Flood Although its objectives are clearly the same, the hydrocarbon miscible process' close resemblance to the above cases with respect to fractional flow theory requires some explanation. There are important differences. We assume that a hydrocarbon solvent capable of fully displacing the oil will be something like propane; thus, adsorption will not be significant. More seriously, because such a solvent is not likely to be as viscous as the crude oil, it will tend to finger into it, resulting in a drastic reduction in sweep ef- ficiency. However, Claridge 19 recently has shown how this can be obviated by proper grading of viscosity (accomplished in practice by mixing the solvent and crude in carefully controlled proportions with a steadily decreasing fraction of the more viscous crude oil). In any event, we presently ignore this complication and consider just two idealized fractional flow curves. This time the curves are for the immiscible fluid pairs oil/water and SOCIETY OF PETROLEUM ENGINEERS JOURNAL solvent/water, both pairs assumed to have the same relative permeability curves but differ due to the different viscosities of the oil and solvent. If we plot the fractional flow of oleic phase, fo' vs. the saturation of oleic phase, So, rather than f w vs. S w' and consider the construction for the solvent displacement, the similarity to previous cases is striking when the solvent viscosity is greater than the oil viscosity (Fig. 5). However, this is not usually the case, as mentioned above, and so we do not treat the f w miscible case further. Carbonated Waterflood This is the last of the isothermal processes involving only a single characteristic velocity that we will consider. Although CO 2 flooding commonly is considered a miscible process, it does not meet the above criteria due to the appreciable solubility of CO 2 in both oil and water under reservoir con- ditions. In fact, its conditional miscibility with crude oil, if obtainable at all, depends mainly on suf- ficiently high pressure. In the carbonated water- flooding process, the water is saturated with CO 2 and injected. The CO 2 partitions into the crude oil, reducing its viscosity, swelling it, and increasing injectivity. -Os 0 Sw (a) Fractional Flow Diagram - I-S orw The continuity equation for the CO 2 (letting Cr and C? be the CO 2 concentrations in water and oil, respectively) is Sw Sw2 a( w 0) q a ( w 0)_ at SwCi +SoCi + Act> ax vwCi + foCi -0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (14) Given the equilibrium relationship between the CO 2 dissolved in the water/oil phase, C?= C?< Cr)' and again (this time less justifiably) neglecting the volume the CO 2 occupies in the water, dCC! acl'V q [sw + (1- Sw) dcr] a/ + Act> [rw +(1-fw) dCC! acl'V . a;- =0; ............... (15) I thus, ( dX ) q [ f w + (1 - f w ) (dC? I dCr ) ] N p dt Cj= Act> Sw+(1-Sw)(dC?ldCn Soi .......................... (16) For the self-sharpening CO 2 concentration case, the result is similar. A jump increase in CO 2 from zero to its injected value, occurs and has velocity. q [fw+(1-fw)K ] v t.Cj' = Act> S w + (1 - S w ) K ,......... (17) where o Xo (b) Saturation Profile a 1.0 to K . ......................... (18) (c) Oil Production Curve In either case the velocity of a front of CO 2 - saturated water and the aqueous saturation velocity must be equal where the carbonated water and initial water contact. Taking the simpler, self-sharpening case first, this means JUNE 1980 Fig. 4 - Low-tension flooding at high viscosity. 195 Swr SOLVENT AND WATER OIL AND WATER So Fractional Flow Diagram Fig. 5 - Hydrocarbon miscible flood. o Swr Sw (a) Fractional Flow Diagram Sw I Swr CARBONATED WATER+ WATER o I Xo (b) Saturation Profi Ie Before Water BT Np I-Sorw - Swr 1.0 to (c) Oil Production Curve Fig. 6 - Carbonated waterflooding. 196 [ fw+ (1-fw)K] dfw I S (1- S )K = _.. .... (19) w+ w w This is the equation of a straight line of slope dfw/dSw passing through the point (-K/I-K, -K!1-K), as in Fig. 6. Starting at Sw;=SwP first 100070 oil is produced until the first shock BT, then an oil/water bank free of CO 2 is produced at constant oil cut f02 until CO 2 BT at the second shock, and finally oil production at decreasing cuts along the carbonated water/oil curve until Sorw is reached. This result is qualitatively just like that for polymer flooding, but because the CO 2 generally will be retarded more due to its absorption into the oil than polymer will due to its adsorption on the rock, polymer generally will be more efficient. Hot Waterflood Coupling the energy and material balances, assuming heat convection is the dominant form of energy transport, and going through an analysis similar to the above (see Appendix A) gives the velocity of a temperature front for a hot waterflood: ( dX) q [A A A ] dt T= Act> PoCo+(PwCw-PoCo)fw I [Po Co + (PwCw -poCo )Sw + p S C R (1 - ct> ) let> ]. .......... (20) The oil recovery follows directly from this result. Slugs For simplicity and definiteness, consider the hydrocarbon solvent process first. The principles apply to the other slug processes as well. To be practical, we must follow the expensive solvent slug of pore volume Vps with a gas drive under sufficient pressure for the gas to miscibly displace the solvent. The velocity of the solvent at S wr is Vs = -.!L :!so I ..................... (21) Act> 0 Swr For convenience, we define a normalized velocity, Us=vs/(q/Act. Thus, its BT time in pore volumes is BT 1 ts = - ............................ (22) Us The velocity and BT time of the gas are 1 U g = -- .......................... (23) I-Swr and BT 1 tg = - + Vps' ...................... (24) U g Thus, the minimum slug size of solvent required to prevent deterioration of the solvent slug is Vps= - =[ :!s: ]:w: +Swr-l. ... (25) This and similar results are conveniently represented on a time-distance diagram (Fig.7). SOCIETY OF PETROLEUM ENGINEERS JOURNAL In Region I on Fig. 7, the saturations are just the initial values, So; and Swi. Region A is the oil bank, Region B is the solvent slug at irreducible water saturation, and Region C is the gas drive, also at irreducible water saturation. Between Regions A and B is a transition zone where both solvent and water flow. The slope of each line on the time-distance diagram is the velocity of a particular saturation. The solvent slug size, Vps is chosen so that the leading edge of the gas drive just intersects the slowest of these saturations at the outflow end (X D = 1). If the fractional flow diagram has zero slope at SwP the required slug size is infinite. However, for practical purposes, the slope at a slightly greater water saturation than Swr can be used to size the solvent slug. Of course, a small amount of oil will be trapped. Depending on the shape of the solvent/water fractional flow curve, very small to very large solvent slugs may be indicated. In the latter case, fingering and dispersion actually will be the limiting factors on economical slug size. The frac- tional flow result always will be a lower bound. The same tD -xD diagram can be used to represent the low-tension slug size requirement qualitatively if Band C are relabelled chemical slug and polymer drive and the transition zone is simultaneous flow of the two phases oil and chemical solution. However, we can influence the result by increasing the viscosity of the chemical slug and polymer drive by increasing their polymer con- centrations. This shifts the low-tension (surfactant and polymer in water plus oil) fractional flow diagram to the right on a (Sw' / w) plot. As previously discussed, at some high viscosity the displacement will be piston-like. Then in the absence of dispersion and adsorption (there will be no fingering since the slug mobility will be less than that of the oil bank and the drive mobility equal to or less than that of the slug), the slug size can approach zero. Without dispersion, the actual slug size required will be Ds. With dispersion, the results will depend on how the tension varies with surfactant concentration but always will be greater. Never- theless, a smaller slug size can be bought at the ex- pense of more polymer. Since the slug cost per barrel generally will be much greater than the polymer drive cost per barrel, this will be an attractive tradeoff up to a certain point. Three-Phase Flow Several circumstances involve three phases in reservoir flooding processes. The classical one in- volves oil, water, and free gas that has evolved from solution with the crude oil as the pressure declines below the crude oil bubble point. At some point water may invade from an aquifer by edge-water . drive or be injected as a water flood or polymer flood. In other cases, inert gas may be injected or natural gas reinjected. Among the EOR processes, surfactant and carbon-dioxide flooding are two examples that sometimes involve three or more phases. The case of oil/water/gas immiscible flow in porous media appears below, and Appendix B shows how other three-phase cases can be treated with the JUNE 1980 to Fig. 7 - Time-distance diagram. same approach. The conservation equations for water, gas, and oil are, in turn, as w q a/ w at + AI/> ax =0, ................... (26) ................... (27) and as o q a/ o - +-- =0 .................... (28) at AI/> ax Since we have Sw +So +Sg = 1, ...................... (29) only two of these are independent. Neglecting gravity and capillarity, the fractional flows are / = krw/Jlw w , (30) krw/Jlw +ko/Jlo +kg/Jlg ......... (31) Using the method of characteristics, 20,21 Appendix B shows that the two characteristic velocities for this problem are given by ( dX) q 1 [ u = dt = M"2 /11 + h2 .J (fll + h2)2 - 4(fl tf22 - /IV21 ) ] . . .. (33) The characteristics in saturation space are given by dS w -/12 = ............... ,. .. . (34) dS g u -h2 . Since the/ij are functions of saturation only, Eq. 34 can be integrated independently to give saturation 197 Fig. 8 - Schematic saturation paths of three immiscible fluids. o SoO SwO SgO to ...... Fig. 9 - Schematic timedistance diagram for threephase flow. z o t-
a::: ;:) t-
en o 50 B r-----j I 5 . L_o.!... 5 . Sor _ -- 5 ..... lIL. ______ - ,-__ ...;w;;.;r-j' __ -I : . -------------._._._._---------_ .. x Fig. 10 - Saturation profiles for oil/water/gas. ROUTE "I INITIAL SATURATION Sw So FI NAL SATU RATION Fig. 11 - Saturation paths for water displacing oil and gas at Swr. 198 paths (Fig. 8). The procedure is very similar to that for either two-component competitive adsorption or ion exchange with stoichiometry. Helfferich 2 has shown the equivalence for any number of invarient phases of arbitrary composition. One special feature of multiphase flow is the existence of residual trapped quantities of each phase at low saturations of the phase trapped. These are designated SWr> Sgr and Sor and shown in Fig. 8 as dashed lines parallel to the phase trapped. This results in a small triangle within the saturation triangle. Inside the smaller triangle all three phases flow; outside only one or two phases flow. In general, once Eq. 34 has been integrated for particular fractional flow curves, the saturations can be mapped in distance-time space (Fig. 9). Two saturation waves develop separated by a plateau of constant saturation. One or both of the saturation waves may be self-sharpening, in which case, for step changes in input, a single discontinuity or shock front forms and the wave collapses into a single line on the t D - x D plot. A discontinuity from the plateau followed by a diffuse zone is also possible. The latter case is illustrated by the polymer flooding case (second or upstream wave of Fig. 2b). Now saturation profiles can be obtained by crossplotting at fixed time (Fig. 10) and saturation histories ob- tained by plotting at fixed distance, usually at xD = 1. Actually, we usually are interested in production curves rather than saturation histories. To calculate production, we need the fractional flow of each fluid. These are calculated using Eqs. 30 through 32 and the known saturations. Suppose the initial water saturation is at its residual value and both oil and gas are mobile. Only water is injected. This problem was solved numerically by Gottfried et al. 23 The solution takes on particularly simple form for this case of three- phase immiscible flow. The two possible saturation routes are shown in Fig. 11. We must decide which is physical based on the first wave traveling faster than the second. Ordinarily, the gas will be much less viscous than the oil and water, so the first wave will be an oil/gas wave at residual water. This is followed by an oil bank at residual gas, a shock to some lower oil saturation, then slowly decreasing oil saturation until sweepout to Sw = 1 - Sor - Sgr (see Fig. 10). This is just what Gottfried et al. found numerically. Since only gas and oil flow downstream of the first wave and only water and oil flow upstream of the first wave, this case can be analyzed by the usual fractional flow diagrams. These are shown along with the appropriate graphical construction in Fig. 12; the oil production curve is in Fig. 13. Three-Component, TwoPhase Displacements This case can be applied to several processes. Examples are alcohol flooding and surfactant flooding involving only two phases. Also, displacements with hydrocarbon gas or carbon dioxide can be approximated provided the crude and gas can be modeled by three pseudocomponents and SOCIETY OF PETROLEUM ENGINEERS JOURNAL the pressure does not vary too much. If the gas and crude were miscible in all proportions, the previously considered miscible case would be sufficient. Generally, however, the phase diagram at fixed temperature and pressure will involve a two-phase region (Fig. 14). All the above processes involve significant interphase mass transfer. In the surfactant process, for example, except at very low surfactant concentrations, the aqueous or oleic phase or both will contain high concentrations of all of the three components (oil, water, and surfactant) and the low- tension case previously discussed will not apply. The differential material balance for any com- ponent, i, in two-phase flow is a( ~ ) q a - c-,S, +C,S2+C +-- at I I I Act> ax . (C il !, + C i2 h) = o. . .............. (35) Welge et al. 9 solved this problem for the enriched gas case. This is the problem that results when a gas of composition such as Point D in Fig. 14 displaces an oil of composition such as Point A. There is no mobile water and no solubility of the hydrocarbons in the connate water. W achmann 8 solved this problem for the alcohol case. Larson '5 recently solved the surfactant case. In the alcohol and sur- factant cases, only one component is necessary to represent the oil, but water is mobile and partitioning between the aqueous and oleic phases is significant volumetrically. Thus, a phase diagram such as in Fig. 15 is appropriate. The two cases are the same mathematically, except for the adsorption of the surfactant and the decreased oleic phase residual saturation if the interfacial tension is reduced suf- ficiently, which is usually the case for the surfactant case. Larson's analysis was restricted to the case of a self-sharpening surfactant front. The saturation and surfactant concentration profiles are shown in Fig. 16. The saturation profile is qualitatively the same as for the low-tension surfactant flood. As the oil is banked up, a jump in oil saturation occurs from Sorw to SoB (assuming, as is nearly always true, that the velocity of higher saturations are greater than those of lower saturations between these limits). Then at the sharp surfactant front, the surfactant concentration in- creases from zero downstream to aqueous phase and oleic phase values upstream corresponding to the tieline shown in Fig. 15, which passes through the injected composition, Point D. Without such a tieline extension, a jump from the oil bank composition, Point A, to the plait point, Point P, occurs, resulting in a completely miscible displacement. This is perhaps the most significant item of Wachmann's entire analysis. Compositions in that part of the diagram labeled miscible in Fig. 15 will displace all the oil miscibly regardless of interfacial tension, relative permeabilities, etc. This is true regardless of the sharpening behavior of the surfactant front. For the small slug case, such displacements will be limited only by the adsorption and dispersion of the slug and its hydrodynamic stability. JUNE 1980 o Sg = Sgr 5wr o Sw = Swr 1-5 0r -5 wr 5 w 5 g Fig. 12 - Fractional flow curves. to --... Fig. 13 - Oil production with initial gas. C2- C S Fig. 14 - Phase diagram for hydrocarbons. SURFACTANT OR ALCOHOL BRINE A B OIL Fig. 15 - Phase diagram for surfactant or alcohol. '99 fw Sw Fig. 16 - Surfactant concentration and oil saturation profiles. If the injected composition is in the miscible region of Fig. 15, all original water and oil are displaced, and with no adsorption the injected fluid has unit normalized velocity. At contact with the oil/water bank, the velocities of the water, oil, and injected fluid are equal, so that fw = fa = 1 Sw So . ..................... (36) Graphically, a straight line from the origin to the point (1,1) (Fig. 17) gives the values for SwB andf wB or their complements, SoB = 1 - SwB and foB = 1 - SwB' If the injected surfactant adsorbs Ds pore volumes per injected pore volume, all velocities are retarded to 1/(1 + D s ), and the line should be drawn from (-Ds' 0). Now return to the nonmiscible case. Given self- sharpening, only three regions of differing phase compositions exist. Far downsteam just oil and water flow, and far upstream the injected composition exists. Somewhere in between a surfactant shock must exist. The velocity of this shock, v Ac is obtained by making an overall material balance on a moving front (the shock) across which the concentrations and saturations discontinuously changed from Cijand SJ to Cij and SJ. The result is v = - f') - I'" - 1") ACi A</> II I 11 I tV I tV I +Ci2 - Ch J/[ (Cii Si' - Cil Si)-(Ci2Si' -ChS l)+Cjz-CiI +Cr-C;J. ..... (37) Since downstream of the surfactant (Component 3) C31 = C32 = C 3 = 0, .................... (38) then 200 z: a l- e:( a:: I- z: UJ u z: a u I- z: e:( I- U e:( l.&.. a:: ::::> en o C32 r-----' I I I C" I-_M-_-I I I SoB I
Sorw Sore r l Fig. 17 - Completely miscible displacement. o I z: a l- e:( a:: ::::> l- e:( en ...J a q fi'+b VAC3 = A</> Si' +a ' ................. (39a) where b= C 32 and C" c" 31 - 32 C" CAli ...................... (39b) a= 32 - 3 C 31 - C 32 . . ................... (39c) The C 3j are just the phase concentrations at the end of the tleline shown on Fig. 15. The value for Si' (and hencefi' from the low-tension fractional flow curve) can be obtained as follows. The velocity of any saturation in a two-phase flow region of constant composition is, from Eq. 35, q dfl vs! = A</> dS I . . ..................... (40) At saturation Si', v AC3' and vs! must be equal, so f'{ +b dfl S" =-dS I ................. (41) I +a I s! =sr The only unknown in this equation is Si'. Graphically, a line from (-a, -b) tangent to the low-tension fractional flow curve yields the value for S'{ (the total composition here is designated C' on Fig. 15). The oil bank saturation is given by applying Eq. 37 to a second component, say Component 2, and then eliminatingfi' and Si'. The result is l' +d VAC2 = +c ' ...................... (42a) where d=C22 -1-b(c21 -c22)' .............. (42b) and c=c22 -1-a(c21 -c22)' ............... (42c) SOCIETY OF PETROLEUM ENGINEERS JOURNAL We identify Si as the water saturation in the oil bank, SwB' and similarlY.!i =fwB' These values can be obtained graphically by a parallel to the above tangent passing through (- c, - d). All other up- stream saturations are obtained in the usual way by integrating Eq. 4 until the injected composition is reached. If the injected composition lies on the binodal curve, its composition is (Ci'I' C lI , C 3I ) and the final oleic phase saturation will be Sorc' However, if the injected composition is in the single-phase region such as Point D (Fig. 15), a second shock front exists upstream of the first surfactant shock. At this shock, So goes to zero since only a single phase exists. This can be thought of as a solubilization front. The velocity of this shock is given by making a material balance between the injected composition, say Point D, and a point in the two-phase region, say Point C. The result is q fa c +R U LlC2 = Act> So c + R ' ................. (43a) where
R= , .................... (43b) C lZ -ClI where the phase concentrations at Point C are just C lZ and C lI ' and is the injected oil composition. The oleic phase saturation at Point C, sg, is given by the requirement that the velocity of this saturation also must be the same as given by Eq. 40. Thus, , .fb +R dfo I c =-- . So +R dS o
.............. (43c) In summary, the composition route starts with the given initial overall composition, A, and going upstream, develops as follows. 1. A jump to Point B, the oil bank at C 3 = O. Use Eq.42a. 2. A jump to Point C' on the tieline passing through Point D. Use Eq. 41. 3. A continuous decrease in So from sg' to sg occurs at fixed phase compositions, Cij, until Point C is reached. Use Eq. 40 for continuous change and Eq. 43c to get Point C. 4. A jump from Point C to Point D, the injected composition, which is a given known. The- oil recovery is given as follows. Before oil- bankBT, Np =foitD; ......................... (44a) for t Dl :5, t D :5, t DZ (during oil-bank production), Np =foitDl + foBt D ; .................. (44b) for tD3 <t D <tDZ (after surfactant BT), Np= (Ci'l -Ci'z)(Sw+tWo) +Ci'z -Swi' . (44c) and _ ( dfl )-1 tD - dS I , ...... (44d) where JUNE 1980 tDl = SwB -Swi , ................... (44e) fwB-fwi Si'+a tDZ = fi'+b ...................... (44f) and SC+R tD3 = lo +R' ....................... (44g) Np above is not corrected for the oil injected, t D' These equations reduce to the previous water flood, polymer flood, and low-tension flood cases when Ci'l = 1 and Ci'z = 0 (no mass transfer of oil into water and vice versa). Consider the slug size required for these surfactant flood cases that involve significant partitioning at high surfactant concentration. For the miscible case, the required slug size is affected only by dispersion and adsorption. Thus the adsorption in pore volumes, D s ' is a lower bound on the slug size. With dispersion, a somewhat larger slug size will be required. The slug size for the nonmiscible case will depend on whether we fix as our criterion the displacement of all oil in the swept volume or the displacement down to Sorc only. Several con- siderations enter here. First, if the capillary number is high enough Sorc will be zero and the required slug size will be the same for both criteria. However, if Sorc is not zero" the decision will depend primarily on the velocity of the solubilization shock, which often will be very slow, and on the oil content of the residual oleic phase. If the slug size is to be large enough to recover all the oil, then the velocity, us' of' its back side will be given by Eq. 42a and the required slug size is 1 Vps= - -1. ........................ (45) Us If the slug size is to be large enough to reduce So to Sorc' then the velocity of its back side will be Us = -.!L dfo I , ................ (46) Act> dS o So = sore and the required slug size is dfo -1 Vps=(-dS) +Sorc-1, ............ (47) o sore provided Us is greater than U Llc from Eq. 43a. Just as in the low-tension case, if Uie injected viscosity is high, a step decrease from SoB to Sorc will occur. Us will be given by l+b Us = , ..................... (48) I- S orc+a and the required slug size is I-Sorc +a Vps = 1 + b +Sorc -1. . ........... (49) Here, as before, we have assumed that the adsorption of surfactant is irreversible and the same as that of polymer if both are expressed in pore volumes. The physical significance of Eq. 49 is emphasized by ZOI defining a partition coefficient, K, and an adsorption is pore volumes at concentration C 31 ' Ds: C" K = ~ (50) C 31 ' ......................... . and C" D = _3 (51) s C 31 ......................... . Simplifying Eq. 49 gives Vps=Ds+KSorc . ..................... (52) This result implies the slug must be sized to account for both adsorption on the rock and absorption into the residual oleic phase. The latter greatly increases Vps for large K. However, this is an unduly pessimistic result since K can be reduced by decreasing the electrolyte concentration in the polymer drive to K values as low as 0.01 to 0.1, depending on numerous other factors. Then dispersion and ion exchange will play an important role since an electrolyte gradient will be set up which depends on these phenomena. These con- siderations 24 ,25 are beyond the scope of this paper. Discussion Among numerous limitations to the fractional flow theory presented, the most basic is the uncertainty in the relative permeabilities upon which the entire theory of multiphase flow in porous media is based. This limitation also applies to more general models of flow in porous media such as the three- dimensional, three-phase, compressible reservoir simulators used by reservoir engineers.. This is because the basic differential equations, which they solve numerically by finite difference or finite element, also have the relative permeabilities in them. The concept of relative permeability has proved useful for the simultaneous flow of oil and water at low Reynolds number. 26 This is amply documented by various laboratory experiments and by the analysis of field-scale waterfloods. But for more complicated circumstances the basis for its use is more tenuous. For example, only a few three-phase experiments have been done, except those where one of the phases was immobile (trapped). Dietrich et al. 27 recently reviewed these experiments and the attempts to describe three-phase relative per- meabilities analytically. Gravity has been accounted for by several authors. 1,28-31 In some cases gravity is the most important factor in the entire process. The criteria for gravity stabilization are well-known and widely applied. Another major limitation of the theory is its restriction to one dimension. However, its extension to two dimensions through streamline models 32 ,33 can and has been done by numerous workers. Patton et al. 4 did this for polymer flooding, for example. Additional permeability layers also are added easily, provided there is no cross flow .34 As the mobility ratio deviates significantly from one this becomes a poor approximation, however. It may even be a poor approximation at unit mobility ratio if transverse 202 dispersion is significant. This may be true for l a r ~ e permeability contrasts, thin layers, and small slugs. 5 Capillary pressure plays an analogous role for the immiscible case. Non-Newtonian rheology is another factor that may be significant in three-dimensional flow. For example, polymer injectivities are significantly higher in radial flow than in linear flow if the polymer solution is shear thinning. These and similar limitations are best removed by going to a fully numerical simulator. Even in one space dimension these can be quite complex. An example for the surfactant process is the compositional simulator of Pope and Nelson. 24 See Pope et al. 25 for additional applications of this simulator. Even the numerical simulators have major limitations. One of these is the inadequate representation of fingering. This is particularly serious for the CO 2 process, which is inherently unstable from the onset. Fingering is perhaps the most serious limitation to describing qualitatively the unstable processes with either fractional flow theory or numerical simulation. The best approach we have to the problem is that of Claridge. 19 Particular limitations in addition to the above also apply to each particular process discussed. A good discussion of these for the surfactant process is presented by Larson. 15 The restriction to self- sharpening systems is an example. The complete neglect of electrolyte changes and effects is another example. Even so, the theory sheds a lot of light upon how the process works qualitatively, which is its primary purpose in all cases. An extension to include the three-phase surfactant problem should be useful in this regard. In many circumstances CO 2 is only partially miscible with crude oil, and if mobile water is present, three-phases will flow. This process is complicated also by the need to consider the crude oil as more than one component, the pressure depen- dence of the phase behavior and fluid properties, and fingering. Thus, an extension to the tertiary CO 2 process is both one of the most important and dif- ficult extensions of fractional flow theory yet to be made. Particularly important are the issues of retrapping of CO 2 -mobilized oil by water drive, the required amount of CO 2 , and the efficacy of alternate water and CO 2 injection. A novel extension of fractional flow theory to the determination of the fractional flow function itself using multiple partitioning tracers was recently presented by Deans. 36 It seems safe to assume still other applications will be discovered. Nomenclature A Ai = area adsorption of Component per unit mass of rock defined by Eqs. 42b and 42c concentration of Component i in fluid per unit pore volume concentration of Component i on stationary phase (rock) per unit pore volume of rock SOCIETY OF PETROLEUM ENGINEERS JOURNAL concentration of Component i in Phasej specific heats of oil, rock, and water D; general retardation factor for Component i, adsorption of Component i in pore volumes per pore volume injected eval- uated at injected concentration for self-sharpening cases f = fractional flow, the flux of a given phase divided by the total flux f ij the derivatives off with saturation as defined in Appendix B k permeability k r relative permeability K = partition coefficient defined by Eq. 18 or Eq. 50 k T = thermal conductivity _ L = length of porous medium M = mobility ratio Np oil production, pore volumes q flow rate Sg gas saturation So oil saturation So; initial oil saturation Sw water saturation Sw; initial water saturation Sj saturation of Phase j t time T= u= v Vps = X= X D ~ P. P Ps (1 e/> e/>; non dimensional time or injected pore volumes of fluid, qt I ALe/> temperature normalized frontal velocity, v divided by qlAe/> frontal velocity, qlAe/> for single phase flow in porous media pore volume of porous media, ALe/> slug size, pore volume distance or space dimension nondimensional distance, XI L difference operator viscosity density gain density Of rock (or matrix) characteristic directions porosity effective porosity accessible to Componenti Subscripts and Superscripts JUNE 1980 BT breakthrough time of injected fluid e = chemical or carbon dioxide D nondimensional g gas gr gas residual component number j phase number 0 oil oB oil bank ore oil residual to chemical or sur- factant orw oil residual to water p polymer r relative value or residual value s surfactant w water wr water residual References 1. Buckley, S.E. and Leverett, M.C.: "Mechanism of Fluid Displacement in Sands," Trans., AIME (1942) 146, 107-116. 2. Welge, Henry J.: "A Simplified Method for Computing Oil Recovery by Gas or Water Drive," Trans., AIME (1952) 195 91-98. 3. Craig, F.F. Jr.: "The Reservoir Engineering Aspects of Waterflooding," Monograph Series, Society of Petroleum Engineers, Dallas (1971) 3. 4. Patton, J.T., Coats, K.H., and Colegrove, G.T.: "Prediction of Polymer Flood Performance," Soc. Pet. Eng. J. (March 1971) 72-84; Trans., AIME, 251. 5. de Nevers, Noel: "A Calculation Method for Carbonated Waterflooding," Soc. Pet. Eng. J. (March 1964) 9-20; Trans., AIME,231. 6. Claridge, E.L. and Bondor, P.L.: "A Graphical Method for Calculating Linear Displacements With Mass Transfer and Continuously Changing Mobilities," Soc. Pet. Eng. J. (Dec. 1974) 609-618; Trans., AIME, 257. 7. Taber, J.J., Kamath, I.S.K., and Reed, R.L.: "Mechanism of Alcohol Displacement of Oil from Porous Media," Soc. Pet. Eng. J. (Sept. 1961) 195-212. 8. Wachmann, C.: "A Mathematical Theory for the Displacement of Oil and Water by Alcohol," Soc. Pet. Eng. J. (Sept. 1964) 250-266; Trans., AIME, 231. 9. Welge, H.J., Johnson, E.F., Erving, S.P., and Brinkman, F.H.: "The Linear Displacement of Oil From Porous Media by Enriched Gas," J. Pet. Tech. (Aug. 1961) 787-796; Trans., AIME,222. 10. Shutler, N.D. and Boberg, T.C.: "A One-Dimensional Analytical Technique for Predicting Oil Recovery by Steamflooding," Soc. Pet. Eng. J. (Dec. 1972) 489-498. 11. Fayers, F.J. and Perrine, R.L.: "Mathematical Description of Detergent Flooding in Oil Reservoirs," Trans., AIME (1959) 216,277-283. 12. Davis, J.A. and Jones, S.C.: "Displacement Mechanisms of Micellar Solutions," J. Pet. Tech. (Dec. 1968) 1415-1428; Trans., AIME, 243. 13. Larson, R.G. and Hirasaki, G.J.: "Analysis of the Physical Mechanisms in Surfactant Flooding," Soc. Pet. Eng. J. (Feb. 1978) 42-58. 14. Larson, R.G.: "The Influence of Phase Behavior on Sur- factant Flooding," Soc. Pet. Eng. J. (Dec. 1979) 411-422; Trans., AIME, 267. 15. Foster, W.R.: "A Low Tension Waterflooding Process Employing a Petroleum Sulfonate, Inorganic Salts, and a Biopolymer," J. Pet. Tech. (Feb. 1973) 205-210; Trans., AIME,255. 16. Hales, H.B. and Odeh, A.S.: "An Improved Method for Simulating Ideal Low Tension Flooding Processes," Soc. Pet. Eng. J. (April 1976) 53-56. 17. Helfferich, F.G. and Klein, G.: "Multicomponent Chromatography," Marcel Dekker Inc., New York City (1970). 18. Pope, G.A., Lake, L.W. and Helfferich, F.G.: "Cation Exchange in Chemical Flooding - Basic Theory Without Dispersion," Soc. Pet. Eng. J. (Dec. 1978) 418-434. 203 19. Claridge, E.L.: "A Method for Designing Graded Viscosity Banks," Soc. Pet. Eng. J. (Oct. 1978) 315-324; Trans., AIME,265. 20. Aris, R. and Amundson, N.R.: "Mathematical Methods in Chemical Engineering," First-Order Partial Differential Equations With Applications, Prentice-Hall Inc., Englewood Cliffs, NJ (1973) 2. 21. Courant, R. and Hilbert, D.: "Methods of Mathematical Physics," Partial Differential Equations, Interscience, New York City (1962) 2. 22. Helfferich, F.G.: "General Theory of Multicomponent, Multiphase Displacement in Porous Media," paper SPE 8372 presented at the SPE 54th Annual Fall Technical Conference and Exhibition, Las Vegas, Sept. 23-26, 1979. 23. Gottfried, Byron S., Guilinger, W.H., and Snyder, R.W.: "Numerical Solutions of the Equations for One-Dimensional Multi-Phase Flow in Porous Media," Soc. Pet. Eng. J. (March 1966) 62-72; Trans., AIME, 237. 24. Pope, G.A. and Nelson, R.C.: "A Chemical Flooding Compositional Simulator," Soc. Pet. Eng. J. (Oct. 1978) 339- 354. 25. Pope, G.A., Wang, Ben, and Tsaur, K.: "A Sensitivity Study of Micellar/Polymer Flooding," Soc. Pet. Eng. J. (Dec. 1979) 357-368. 26. Jones, S.c. and Roszelle, W.O.: "Graphical Techniques for Determining Relative Permeability from Displacement Ex- periments," J. Pet. Tech. (May 1978) 807-817; Trans., AIME, 265. 27. Dietrick, J.K. and Bondor, P.L.: "Three-Phase Oil Relative Permeability Models," paper SPE 6044 presented at the SPE 51st Annual Fall Technical Conference and Exhibition, New Orleans, Oct. 3-6, 1976. 28. Claridge, E.L.: "Prediction of Recovery in Unstable Miscible Flooding," Soc. Pet. Eng. J. (April 1972) 143-155. 29. Hirasaki, G.J.: "Sensitivity Coefficients for History Match- ing Oil Displacement Processes," Soc. Pet. Eng. J. (Feb. 1975) 39-49. 30. Hawthorne, R.G.: "Two-Phase Flow in Two Dimensional Systems - Effects of Rate, Viscosity, and Density on Fluid Displacement in Porous Media," Trans., AIME (1960) 219, 81-93. 31. Hagoort, J.: "Oil Recovery by Gravity Drainage," Soc. Pet. Eng. J. (June 1980) 139-150. 32. Caudle, Ben H.: Fundamentals of Reservoir Engineering, SPE videotape course textbooks (1967, 1968) 1,2. 33. Le Blanc, J.L.: "A Streamline Simulation Model for Predicting The Secondary Recovery of Oil," PhD thesis, U. of Texas, Austin (1971). 34. Jewett, R.L. and Schurz, G.F.: "Polymer Flooding-A Current Appraisal," J. Pet. Tech. (June 1970) 675-684. 35. Koonce, K.T. and Blackwell, B.J.: "Idealized Behavior of Solvent Banks in Stratified Reservoirs," Soc. Pet. Eng. J. (Dec. 1965) 318-328; Trans., AIME, 234. 36. Deans, H.A.: "Using Chemical Tracers to Measure Fractional Flow and Saturation In-Situ," paper SPE 7076 presented at the SPE Fifth Symposium on Improved Methods for Oil Recovery, Tulsa, April 16-19, 1978. 37. Lake, L.W. and Helfferich, F.G.: "The Effect of Dispersion, Cation Exchange, and Polymer/Surfactant Adsorption on Chemical Flood Environment," Soc. Pet. Eng. J. (Dec. 1978) 435-444. APPENDIX A Derivation of Energy Balance Equation In addition to Assumptions 1 through 12 in the text, assume (1) constant specific heats (Co' C w , and C R ) , (2) no kinetic or potential energy con- tributions, (3) negligible radiation energy transfer, (4) no phase changes, (5) no work terms, and (6) no significant changes in density with temperature. An energy balance then yields a [ at (PwSwCw+PoCo (1-Sw)ct>T 204 a [ + (1-ctp S CR T J + ax pwiwCw +PoC o (l-fw)T] + :; )=0 ....... (A-I) Expanding the derivatives and using the continuity equation for water, as w q dfw as w at + Act> dS w ax = 0, ............ (A-2) we get [PoCo + (PwCw -PoCo )Sw +Ps (1- ct CRIef>] aT q ] . at + Act> PoCo + (PwCw -PoCo )fw aT a aT .-+-(k T -) =0 ........ (A-3) ax ax ax At high flow rate we can neglect conduction (k T = 0), so from Eq. A-3 the velocity of a front of constant temperature is ( dX) dt T=- (:;) -poCo )fw ]/[PoC o + (pwCw -poCo )Sw ,ps(1-ctC R Ief>]. .............. (A-4) From this we can calculate the position of the hot water front, after which the fractional flow and, hence, oil recovery are given by the hot-water/oil fractional flow curve. The construction is analogous to the polymer flood case without adsorption or inaccessible pore volume. APPENDIXB Derivation of Characteristic Directions Eqs. 26 and 27 can be expanded to give space and time derivatives of saturation only, since fw =fw (Sw' Sg) andfg=fg(Sw, Sg). The result is asw -.!L [a f w asw afw _ _ at + Act> as w ax + aS g ax - 0, ... (B 1) and -.!L[ af g asw _ at + Act> as w ax + aS g ax -0 .... (B 2) Using the notation fll =afw1asw, etc., and nor- malizing t and X, asw asw _ !l +f11!l +f12 -0, ........ (B 3) ut D uX D aX D and asw at D +121 aX D +122 aX D -0 .......... (B-4) SOCIETY OF PETROLEUM ENGINEERS JOURNAL The characteristic directions or velocities, dXDldt D for this pair of equations are given by the deter- minant fll f12 0 hI h2 0 dX D 0 dt D 0 =0. ... (B-5) 0 dX D 0 dt D The solution for dX Dldt D is given by Eq. 33. Next we show how the above solution method for three-component immiscible flow in porous media can be used to solve the general three-component problem. There are three cases to consider: single-, two-, and three-phase flow. The single-phase case is trivial unless adsorption and/ or dispersion are considered; this case is treated elsewhere. 17,18,35,37 For two-phase flow Eq. 35a can be written ac a __ I + __ (Citfl + C i2 h) =0, ....... (B-6) at D aX D where i= 1, 2, or 3. Noting that the term Cilft + C i2 h can be expressed as a function of any two of the total component concentrations, say C 1 and C 2 , and expanding the space derivative, aC i a aC I a - +-(Ctfl +C"2h)- +- at D aC I 1 1 aX D aC 2 ac, . (Cilfl +C i2 h)-'" =0, ......... (B-7) aX D Formally, this equation is the same as either Eq. B-3 or B-4 if the appropriate substitutions are made, or the characteristics can be written down directly. For three-phase flow the equation of continuity is ac a -a 1 + -- (Citfl + C i2 h + C i3 h) = 0, ... (B-8) tD aX D JUNE 1980 where i= 1,2, or 3. With three phases all Cij are fixed. Then expanding the derivatives are ( aSI afl) (C ll -C 13 ) - + -- + (C 12 -C l3 ) atD aX D ( aS2 + ah ) =0, ............... (B-9) at D aX D and ( aSI afl) (C 21 - C 23 ) - + -- + (C 22 -C 23 ) at D aX D ( aS2 + ah ) =0, .............. (B-lO) at D aX D Noting that afj = Y.J... aSI + Y.J... aS2 , aX D aS I aX D aS 2 aX D ........ (B-ll) we get aS I afl aS I aft aS 2 -+--+--=0, ...... (B-12) at D aS I aX D aS 2 aX D and aS 2 + ah aS I + ah aS 2 =0. . ..... (B-13) at D aS I aX D aS 2 aX D Eqs. B-12 and B-13 are identical to Eqs. B-3 and B-4 if Phase 1 is equivalenced with Phase wand Phase 2 equivalenced with Phase g. Thus, the previously discussed three-phase solution applies directly to any three-phase case regardless of the nature and composition of the three phases. Relative per- meabilities and all phase properties, of course, will vary among different systems. SPEJ Original manuscript received in Society of Petroleum Engineers office May 1978. Paper (SPE 7660) accepted for publication April 6, 1979. Revised manuscript received Jan. 29,1980. 205