You are on page 1of 12

SPE

SPE 19009 Boundary-Dominated Flow in Solution Gas-Drive Reservoirs


by R, Camacho, iMP-Mexico,
SPE Members

and R. Raghavan, Texas A&M U.

Copyright1989, Society of Petroleum Engineers, Inc. This paperwaa preparedforpreaentafionat the SPE Jointflocky MountainRegional/LowPermeabilityResarvoiraSymposiumand Exhibitionheld in Denver, Colorado,March 6-S, 19S This paper wes selected for presentationby an SPE Program Committeefollowingreview of informationCOntahW6d In an sibstracteubmittedby the author(s).Contenteof the paper as presented, have not bwn reviewed by the Society of Patroleum Englnaera and are aubjecl to correction by the author(s). The matarial. aa presented, does not necaaaarllyreffw any positionof the Societyof PetroleumEngineers,itaofficers,or members. Papara presentedat SPE meetingsare aubjactto publicationreview by EditorialCommitteesof the !30def of PetroleumEngineers.Permission to copyis reslrffited toan abstractof notmorethenS00words.Illuslratiis may notbe cqr+ad.The abstractafroufd containcorraplcuoua ~ of where and by whom the paper ia preaentad. Write Publlcationa Manager, SPE, P.O. Sox 833S36, Richardson, TX 75@6S-S8SS. Telex, 730SS9 SPEDAL,

ABSTRACT In this work the performance of wells in solution gas-drive rmervoirs during the boundary-dominated flow period is examined. Both constant wellbore pressure and constant oil rate production modes in closed systems are considered. For the constant wellbore preeaure production mode, the conditiona under which procedures in the literature can be used to anaIyze data are diacuaaed. Specifically, Arpaa equationa for performance prediction are examined and it is shown that predictions of future performance are strong functions of well spacing, well condition, and fluid properti-. The parameters, b (the decline exponent) and di (the initial decline rate), in the Arps equationa are expressed in terlils of physical properti=. The conditions under which these equations can be used are specified. An empirical procedure to predict production ratea is also presented. in the caae of constant oil rate production, an expression to correlate the pressure distribution in the reservoir is presented. The correlating function permits us to extend the definition of pemrdoetcady state flow to solution gas-drive aystema. Ita use also allows the simultaneous computation of average properties (pressure and aatumtion) during boundary-dominated flow from wellbore information. INTRODUCTION The intention of this work ia to document some theoretical results that are useful for predicting well performance from production data in solution gas-drive reservoirs during the boundary-dominated [low period. In the process of documenting these results it is also intended to furnish a theoretical support for empirical obaervationa that exist in this subject. Both constant wellbore pressure and conatarst oil rate production modes in circular closed systems are considered. The osrteomrx preaented in this communication take as a baeia the theoretical rrw iits presented in Refs. 1 and 2.* Specifically, the result. given here follv~ from our ability to correlate responses of solution gas-drive systems with the response of a slightly compressible liquid flow duri]lg (lie boundary-dominated flow period for botb constanl oil ram iII1d

conat,ant wellbore pressure

production

modes.

To accomplish our goala, this work is divided into three parts. In the first part, the theoretical results related to boundary-dominnted flow given in Refs. 1 and 2 are outlined to establish a framework for the findings presented in this communication. In part 11, the caae of a well flowing at a corretant preearrre during the boundary-dominated flow period ia enalyaed. For this case, Fetkovieh3 ahowed that the empirical family of curvu of Arps4 can be combined with tbe slightly compr=ible liquid flow solution (exponential decline rqmnaes) to obtain a family of curves that cars be used to predict future performance and estimate the reservoir pore volume. Refs. 1 and 6 report that during tbe boundary-dominated period the rate response plotted veraua time does not match a fixed value of the decline exponent, b, in the type curves of Ref. 3. An explanation for this observation ia presented. Refa. 3 and 7 emphaeiie that the decline exponent, b, must be less than or equal to unity. They also note that if transient data are used, then the value of tbe decline exponent, b, of the Arpaa solution can be greater than unity. Ueing the development given in Ref. 8 for transient flow, a theoretical justification for t.hia observation is provided. An empirical procedure to predict production rat= of welle produced at a constant pressure, over ehort time apana, ia also presented. Part 111is devoted to the situation when tbe production is held at a constant oil rate. For this case it ie known from Mrs. 1,2, and 9 that the reservoir does not achieve the condition of ~udoatdy-state, i e., the derivative of pressure with respect to time is not conatent and is also not independent of pcmition in the reservoir. In this work, a correlation for the preaacrre distribution in the mervoir during the boundarydominated flow period is developed. Thii correlating fnnction permits ua to obtain an extension of the peeudoeteady-atate concept to solution gas-drive reservoirs. Furthermore, thie timction also allows ua to compute aimultaneorieiy tbe vafuee of average pressure, ~, and average saturation, so, having wellbore information. . The numerical reaulte presented in thie paper wero obtained with a finite difference model described in Ftef. 1. Procedu~ followed b ensure the accuracy of the acdutione are given in Refa. 1 and 2.

* References

and Ilhtatraton

at the end of this paper.


A homogeneous closed cylindrical reservoir withs fully penetrating well located at its center is considered. The well u capable of producing at either a constant oil rate or at a constant wellbore ptessure. An annular region that is concentric with the wellbore, with a permeability ditferent from formation permeability is used to include the effect of a skin region l. The eflects of gKKWiLY, C@blY Pre~ure, and non-Darcy flow are not considered. where ~ follows

= *2 iDrW/A,

with the dimensiorrle.w time. %, defined as , (8)

0.006328k 90(t)T(~)d G = q$r$qo (t) I o a(v)

Here ~ and ~ represent the system mobility and system comprezaibility corresponding to the average reservoir pressure, j , (and average saturation, ~o ); respectively, thus F and m are givm by

Figures 1 and 2 show the PVT properties of tbc fluids ussrl in this work. Figure 3 presents the relative permeability data. The data mts shown in Figs. 1-3 are identical to the data wtr+ considered in Refs. 1,2,8 and 11 and are used here mainly to preserve continuity. The conclusions derived in this work do not depend on the specific data used in the simulations. Table 1 presents the information on the range of variables examined in this study. UND Following the development less pseudopressure sz follows IA
mD (r, t) =

~=[%+:l(,so,
and

(9)

ct.-&

(~)p-&j(fyj),,
B~ (p) +Fo dltj ~ (lo)

B. (II () dp For the constant

in Ref. 1, let us dcfim [he dinwwicm-

oil rate cazc, Eq. (7) can be simplified as follows fiD

(t) = 2~~,
defined as

(11)

141.2q. (t) {/[ ,


t + /[ o @(P! So) aP ~

a
-

ap

(P, So) ~

1
1

rfr

where ~

= ~r~/A,

with ~

1}
~t,

(1)

(12) Eq. (11) is an extension of the materials balance equ~t.ion for singlr= phase liquid flow (production at a constant rate); similarly Eq. (7) may also be considered to be a generalization of the rnnteriab bslance equation for production at a constant preaaure in solution gas-drive reservoirs. In Fiefs. 1 and 2, it waz shown that Eq. (7) may also be used if the production mode is changed from constant rate to constant pressure (or vice versa). For reference purposes, the definition of dimensiohaz time bawd on initial system properties given by the following ex]weeaion is int.rm duced (13)

Here the function a (p, S.) @ (P,

is given by k,. (S.)


so) =

(2)

PO (P) B. (P)

and F is the radius corresponding to the position iH the reservoir at which pressure p(r) is equal to the average re>,woir pressure, p . During the boundary-dominated flow period F cs 0.54928 ~~ D@ 0472 r. (see Fig. 5 of Ref. 2). Dimensionless distance is defined as rD = ~. rw In Ref. 1. it is established that during boundarjf-donlinated following reeults are valid m~ (rtt) = fiD (i) + }nr,D -:+s f (3) flow, the

For single-phatw liquid flow during boundary-dominated the dimensionless flow rate, qD, is given by1314: 141.29(t)@
9D (tD) = k~ (pi Pwj )

flow period

= ~exp
D

+ (

;-1 #

)[

1 r4D -L* 4 r~D

r,D

1[

hrD-

1 (l;, 1) 2 I(:D

(4)

where D.; [

4A In +28 e7C,4rW

fOr 1< rD < r@, and

[1 ~1
2tiDA

(14)

(15)

D(r)=fiD()+( k ( El )[

n%-:)+ 1 (r~D - 1) 2 r~~

Here T is Eulers constant, and C,4 is the shape factor. It can be shown that the average reservoir pressure and flow rste are related by the following equation,

l~_&+__ r~D Z r~D 1 (r~ 1) Z r~~

1
(5) (6)

kh (pi -P) D tD) = 141.29 (t) B/c -D[l-exfw

1)

In Ref. 1, it is also shown that for solution gas drive systerne the next relation is valid

for r#D ~ rD < r@. Here r#D is the dimensionless radius of the skin zone. The symbol fiD (t) in Eqs. (4) and (5) is the vohmnetric average of the pseudopressure, which can also be obtained as folbwa fiD (t) = ~
VT m~ (r, t)

=(=)-+-(%)]
Note that the right-hand side of ~ (7) ie independent results are expressed in terme of=, of r~D, where rcD = r.JrW.

(17)

of r.D when

W,

VT

whereas Eq. (1 i) u a function

where VT is the vofume of the reservoir. By using the Muskat12 Ref. 1 that kh D (t) = 141.2q. (t) / materials _, p(~) balance equation, a (p , r) dp = 2wE, we show in

(7)

Figure 4 exemplifies the use of the appropriate definition of dimensiordeas average pseudopressure, ~(t), and dimeneiordeee time for production at a constant wellbore preeeure and for botb data IWta used in this study. The tilled in data pointa eorreapond to Data Set 1 with a skin factor, s, of 10 and ro = 8000. The open data points correspond to Data Set 2 with u = 2 and r~ = 2000. h tbia figure

JI? 1ql-roq

R. straight data line represents represent Eq.

G.

Carnacho-V.

and

R.

Ra~havan Equation (21) can be written !u!2. @es follows (22)

the unbroken The circular

(7), ffiD (=)=,,=. I fJA

the dimensionless average pseu dopreeeure, fiD, versus the dimensionless time, ~, obtained from simulation runs. An excellent correlation with the h~prid solution is obtained for all values of producing time. Similar reau Ite are obtained for the case of constant rate productionlz. The triangular points represent ~Lr vahcee plotted as a function of tDAJ. Agreement with Eq. (7) is good for small values of tim~ however, during the bmrndarydominated flow period the correlation is poor. (Similar results me also observed for constant rate production]z.) The uubroken curves in this figure represent the right-hand side of Eq. (17) fortwovalum of reD. The square data points correspond to fi~>ahree obtsined from simulation runs and plotted eeafunction of fi. The results shown here suggest that Eq. (17) will predict ?iiD (=] Irr values with points

E Z~Ah _& ,lt 5.614

where NP is the cumulative oil production. Equation (22) shows the relative importance that parameters like relative permeability and PVT data, and pore volume havein theprediction. of future performance. Substituting theright-hand sideeof Eqs. (17) and (21) for the appropriate expremion in Eq. (20) we obtain the following exprmsion (23)

.,

reasonable accuracy better representation

until tDA x 3. of iED. (17) wehavethe

For larger timw

Eq.

(7) is a

From Eqs. (7)and

following rel;ltion

Equation (23) implies that if~~/?l is approximately constant with time, then we would obtain a straight line by plotting Iogqo vrtsus time. This observation may also be expected in intuitive gro!lnds, bsaed on single-phww flow theory. We can relate Eq. (23) with the Decline Curve Equations of Arps4 aa follows. Asiswell known the Arpsequations can rewritten as follows (24) q. = q.i w (Edit) , for the exponential decline, and

2E=D[exp(%11
(18) Expanding we obtain theexponential function intheright-hand G>%. explains why thedata points in~ernwof~ sideof Eel. (18) (19) ill Fig. This inequality 4 fall to the left of the data points in terms of & dominated flow. Theresulta presented inthissection findings presented in this work. forrnth during boundnry -

q. = qti(l + b dit)-lb ,

(25)

eframeworkforthe

Arpss equations are applicable only for for hyperbolic decline. boundary-dominated flow. In these expressions t represents timl: since the rate was qti. The parameters d and b in Eqs. (24) and (X) are considered to be constanta and represent the nomirrrd rate at which decline takea place and decline exponent, respectively. For expon,mtiid decline b = O and for harmonicdecline b = 1; in general b is in the range O < b <1. Considering Eq. (23) with Eq. (24), we can show that rameter di is given by the pa-

As explained in the Introduction, the remrlta ot this paper are presented in twoeections. We start by considering the ceae of con. stant wellbore preesure production mode. Inthisaectiun we intend to establish conditions under which procedurea available in theliterutnre can be used to analyze rate data are justified and also to furnish a theoretical support for empirical observations existing in this subject. A procedure to predict production rates ofrszwvoirs produced at a constant prewsure is presented. We then examine the case of production, at a constant oil rate and preaent acorrelation for the form of the preaaure distribution iu the reservoir. Thlsmrrelation provid- anextension of thepwudosteadystate concept toeolution gee-drive reservoirs.

(26) Thus from Eq. (26) we can deduc~ that the rate will decline in an exponential form only in the case Jt /i5t = constant. From Eq. (26) it is clear that the parameter di is dependent on both rock and fluid properties of the reservoir, the physical reservoir dimensions as w(4I as the wellbore condition (see also Ref. 15). Our simulation results i~dicate that, for the assumptions used in this study, the function ?~/A~ does not remain constant with time. This point is illustrated in Fig. 5, where rem+lts for both data sets are presented. It is clear that the variation in ?j/At with time is significant. These reaulte are typical of all simrilations we have conducted, and also appear to be typical of the results in Ref. 6. The parameter b can be obtained by the following exprearrioll 16 (27)

Constant P~ A na]vsis of A rDS4 eauations

Prock$kr.n M ode for OerfOrmanee Drrdictions average pseudopressure in

Beeed on the success in correlati~the

terms of both dimensionless timee, fi, and~, we willuee theee results to examine the aeeumptions involved in using the Arpa4 equations for performance predictions. Differentiating thedefinitions Eqs. (17) and(7)~ith reapect totime, considering

w%)=-b
From Eq. (23) it can be shown that

of~(Eq.

(12)) and tD (Eq. (8)), we obtain

% exp and kh@J dt respectively,

()
rr~ D

2iT0.006328k ~, 4A z

(20)

Thus, combining Eqs, (27) and (28) we obtain the following expression for the parameter b b= D~A 2s0.006328 d z, (29)

= _ 2rr0.006328k 4A
q. (o

x,
:,

(?1)

k ~ () ~

with t given in days.

From Eq. (29) we obee~e that the parameter b will be a long se the variable i+JA~ varies linearly with time.

constant as

BOUNDARY DOMINATED FLOW IN S The observations regarding di and b given above are important and to our knowledge have not been presented before. They demonstrate the assumptions that are inherent in using the Arpe relations to analyze data, or to predict future performance. Since both dj and b depend on relative permeability and fluid properties, a simple materials balance equation like the Muskatlz materials balance equation can be used to study the variation in ~~/Z1 for any specific situation to determine the consequences of using the Arps equations. hfore interesting and important is the fact that these equations clearly indicate that predictions of future performance are strong functions of well spacing, the well condition, and fluid properties, and thus they also furnish a theoretical support for the concerns expreeaed in Ref. 17 about the use of the production rate decline curves for flow rate predictions. For gas reservoirs, Fraim and Wattenbargerla aleo observed that the rate plotted versus time does not match a fix value of the decline exponent. Fig. 6 presents a match of data obtained from three simulation runs for Data Set 1 with the Fetkovich type curve. Here qdo and fdD represent dimensionless decline rate and dimensionless decline time defined by Fetkovich (see Eqs. 19-22 of Ref.3). The ~bjective of this plot is to show the consequences of the variation in &/A~. For solution gas drive systems Refs. 3 and 7 sugg=t that b should be in the rmrge 0.333< b < 0.667. The responses shown here tit tbe range 0.4 < b < 0.8. The match shown here follows the 6 = 0.7 curve at early times. At later times the responses cut acrosa several curv.m oince for these simulations Zt/~t is not a linear function of time. Am alrendy mentioned, similar behavior is reported in Refa. 6 and 18. The fact that the value of b is not constant in most of the thcoretical studies reported in the literature to date deserves comment. First one must consider the question: Can one obtain a constant value of b for constant pressure production? In general, we found that the parameter b is not a constant for the constant preaeure production ceee. Second, what is the nature of the wellbore prwure reaporree whel! the production is forced to folfow a specific value of the parameter b in the Arps equations? Computations suggest that the production mode must be a variable pressure-variable rate mode if the rak ie to folbw a specific b value. in some cases we have noted that in ode? to Mow a specific value of b, the wellbore pressure must irrcreeee with time (assuming ihe skin zone properties and drahrage.area remain constant with time). Third, if we eesume constant pm.ssure production, then Eq. (29) indicates that t~e only possibility to obtain a constant b is for the product DAd (ZJJ~) /dt to be a constant. Since D includes the skin factor, this line of reasoning leads to the argument that variable skin factors may account for a constant value of b. One obvious possibility to obtain a variable skin factor is non-Darcy flow. Mr. hi.J. Fetkovicblg has also suggested this p-ibility to tm. In Ref. 20, baaed on an examination of Inflow Performance relatiom we noted that non-Darcy flow may very well be the norm in solution gruwfrive reservoirs. Baeed on the observations given here it appears that nonDarcy flow may yield constant value of b. This observation, however, doea not imply that if field data suggest that b is a constant then this result is a consequence of non-Darcy flow since there are other factom, like changes in skin zone properties and drainage area, that could be the cause of this behavior. The above observations are important in understanding the consequences of using the Arps curves. Let us now consider the computation of pore volume for the results shown in Fig. 6. Fetkovich and coworkere7 suggest that it is possible to obtain reservoir pore volume from the match ahown in Fig. 6 by the relation BO(F) t-t q.(tt) (30) vp=_ . qdD Q(P-PWI)V----- where VP ia in barrels and i is the time at which the initial rate, qo(t) = qm. .%$bstituting appropriate vahrea (~ = 5344.94 psi , PWI = 3389 psi, S$ = 0.0336, Et = 1.4468x 10-5 psi-l, qo(f) = q. (t - t ) /qdD = 220.86 STB/D, t t*= 100 days, t,%D= 1.3) in Eq. (29), we obtain } = 1.039x 10s bbls. In simulations% was mumed to be 1,124x 10 bbls. Although msultad idnotmatcha curvewith a specific vafue of b, it is interesting the fr.ct that for this epecific example, a good estimate of the pore volume is obtained. Nok that Eq. (30) indicates that fluid propertied should be computed at the average

UTION GAS DRIVE RESERVOIRS pressure, ~. Justification for th~choice and ~

SPE es correlating Concc@ variablea.

19009

follows directly from the use

of the variables fiD, ~,

on of Peeudo-time

Equation (23) suggests a procedure to take into account variations in ~l/.?,. It suggests that the b = O solutions can be used for performance predictions if ~ is used. This observation from integration of Eq. (23), follows directly

(31)

Nere F

must be interpreted as a paeudotime since the rate was q.i. A similar conclusion for single-phase gas flow is presented in Ref. 18. Note that the exponential decline type can be approximately warranted be rewriting Eq. (23) aa follows (32)

Thus if we use the definition of dimensionlcs time G instead of just t in the semilogarithmic plot of qc versus time, we would approximately obtain a straight line for engineering purpcaea, es mentioned in Ref. 21. It is important to realize that Eqs. (31) and (32) are approximate. For some of the simulations we have conducted these expmions are very good approximations; in other cases they are not adequate. Fig. 7 demonstrates one C- where Eq. (31) dmm not do M well as one would hope. Tbe circular data points are plots of rate in krrruurrf tDf and the square data points are results expressed in krme of tD. The unbroken line is the b = O solution, The reaponata in krrne of tDi (or f dD) intersect several vahrea of b. In this case, however, the application of the pserrdotime concept does not resolve matkm Complekly. Eq. (31) is, however, extremely useful if a plot of q. vemua time intweect eeveral=valuea of b, in this cm extrapolations baaed on a plot of q. veraus t would result in a much better estimate of Row rate with lime. In a paper to be reported we use the idea srrgg=kd by Eq. (31), and combine it with the materials balance to develop ratetime relationship for oil and gas condensate remrvoirs.

Curves

to

Refs. 3 and 7 note that if transient data are used to cornpul e the value of b, then such data may suggest that the value of b is grrwter than unity. The basis for this observation can be seen in the following development. In Ref. 8, it is shown that during the transient flow period, the skin factor of a well produced at a constant pressure can be obtained by the following relations

6 = ~
1 or If we combine expression for b, Eq.

-1

[ dhrqa/dint

+,-,nl$i , 1
+1-in 4NpiDi fqo (t) e? 1 Eq. (27) to obtain

(33)

1 -q. (t) t s = ~ [ NPdlnqJdlnt where NP is the cumulative form of Eq. (33).

(34)

oil production.

(34) is the general

(33) with Eq. b=2s+ln~.

the following (35)

This expression leads to two observations. Fimt the paramekr b is a function of time during the tranaient flow period. Second, if b takes values greater than unity then Eq. (35) yields (36) This inequality can be easily satisfied for large reaervoim and large vrdues of fDi, in fact even for small reSCKV&S and Smd} vafues oft IX

R. G. Camacho-V. SPE 19009 Y the skin factor can in practice satisfy this inequality for moat carrea. For example for tD~ = 10s, the inequality in (36) would be aahsfied for s >-5.66, which is a high negative skin factor. Thus in most of the cases the parameter b would be greater than one if transient data ia used. A similar conclusion obtain results if Eq. (34) is used. [n this case we

and

R.

Raohavan Fetkovichs procedure23 tipredict

Noteagain, that Eq. (43)resemblea future performance given by qo,maz,f


90.mar,p

Z = =Pp

(44)

Ifweaseume

IVP-iq~,

then w~obtaih

~=_

[ln(4tDi/e7)+2S]2 1- hl(4tDi/t?7) - %

(38)

Equation (38) again leads to the conclusion that the parameter 6 is a function of time and that it would be in moat of the cases greater than one wheh transient data are used to compute b. The above development provides theoretical justification for the observations in Ref. 3thatb~l forthe Arpsprocedure to reapplicable. ~te det-ation of rxodr@kxr rates

Figure 8 presents the derivative of rate data with respect to time versus average pressure for a case where Data Set 1 is used. Theunbroken line represents simulator values of the function dqe/df verau~ ~. The circular and square data points correspmd to the rate derivative evaluated from Eqs. (41) and(43) reapective1y. The computationsara done by starting with the simulation value of -dqO/dt at tD~i = 0.1. Toevaluate theright-hand sideofboth Eqs. (41) and(43), sir@ation values have been used. Close agreement with actual valueniaobtaiued with both Eqs. (41) and (43). Thercamlta obtained byusing Eq. (41) were slightly better than the reaults obtained by using Eq. (43). In other simulations, similar results were observed. Themain advantage of Eq. (40) isthatgivea ccsan opportunityto predict future production rates. Eq. (40) can berewritten ss follows

=%-(%9 [b-:+)
(45) +(:-1)(%-39 Integrating Eq. (45) from~ito using theaseumption ti=@2, ~pandto PJ(withpi>PJ>PP >Pfj, in Eq. (45) we arrive at thereeull (46) Eq. (46) permits us to predict flowrate at afuturevalueof~ and in this serrse this function is similar to Eqs. (42) and (44). Once the future vahreofqo,~a. isdetermined then an IPRcurve can bedeveloped using the relations of Vogel*G or Fetkovichz3. Thcsedelivembility equations are given, respectively by

The development given above also suggests a procedure to predict flow rates over short time spans without the use of relative permwrbility data or decline curves. Here we present a simple proccdurc to predict flow rates over a short time spans, which may also be used to and is similar to the Inflow Performance Predictions of Stand ing22 and Fetkovich23. Numerical duction computations d /{, Q(P?sO)~ dt [ suggest that forconstant tlp 1} t dr % O, prcasure pro(39)

during the boundary -dominate dflowperiod. Intuitively, atlate times this result could becxpected. DVferentiatingE q.(4) with respct to time, evaluating the resulting expression at rw, and considering Eqs. (7) and (39), we obtain the following expression for the case of eon,t.ant pressure production 1

q. =
90,maz forVogel, and 90 =
9.zm.2

1- 0.2~

0.8~,
P

(47)

+%(%) [(ll++s)
(40)
(9)(%-*)1 Eq. (40) can also be written Mfollows
(11)

()

#?iJ ,
P

(48)

for Fetkovich. in Eq. (48), theexponent nisinthe range 0.5 <n< 1, and is intended to account for the existence of non-Dercy flow; however, in Ref. 20 it issbown that n can be different from mlity even if rion-Darcy flow is non existent. Note that Eq. (46) reqllires information at two prcasure levels. In this regard it is similar to the method of Kelkar and COXZ7and the Pivot-Point metbod2*. Since Eq. (46) is baaed on the assumption that a (~) = @*, baaed on the results of Ref. 20, we would intuitively expect Eq. (46) to be applicable over short time spans. This limitation also applies to Eqs. (42) and (44) (see Ref. 20). The main advantage of Eq. (46) is thnt predictions can be made with rate and pressure data alone and it does not require reIative permeability data. Figures 9 and 10 show the use of Eq. (46) to predict flow m-. Results of flow rates versus average reservoir pressure with both Jata sets are presented. The filled in data in Fig. 9 correspond tn Dnta Set 1, and the data in Fig. 10 to Data Set 2. The unbroken Iinea correspond to simulation values. The initial pair of values (Q., ~)i was taken at the onset of the boundary-dominated flow period, fn ~t,i = 0.078 (~ = 5662.59 psi) and tD~i,i = 0.093 (P = 1438.54 psi). for Data Set 1 and Data Set 2, respectively. Other details are given in Figs. 9 and 10. Results are normalized by the initial pair of value8. The circular data points reflect the predictions made by using F~. (46) with the second pair of data points taken at tDAi,p = 0.124 and ~D~i,p= 0.117 for Data S&s 1 and 2, respectively. The equ~e (Iata points represent reau Its by using valuea of q. and F at the same time tiasthe circular pointa but the other pair of valuea of q. and T at
1

where the subscripts respectively.

f and p refer 10 future and present

conditions,

Nofc that Eq. (41) rescn]blc sStandings22 proccduretopwdict future prwformarrce,given l)ythefollowing exprmsion
%mar,J qo.m.r,p ~ p,a] (42) j5pEp

where q~,ma~ is the maximum flow rate (rate corresponding to P,, J = O). It should be noted, howeier, that Eq. (42) can be obtained by assuming that the function o(p) varies linearly with pressure2i1a42s, but nothing has been assumed regarding the shape of the functiou a(p) in thedcrivationof Eq. (41). If we now assume that E % 0~2, where 6 is a constant, by Fet.kovich 23, into Eq. (41) yields Ok#)f ~F;(@/d~)/ (dqO/dt)P P; (@/dt)P
(@3/d0,

as suggested

(.!3)

= (dP3/dOp

BOUNDARY DOMINATED FLOW IN SOLUTION GAS DRIVE RESERVOIRS timetr greater than !hat for the circular p0int8(tDA{,p =4.4095 for Data Set l,arrdtDA@=l .3435forData8et2). Agreement wil.hlhe numerical solution is excellent in three of the four cases considered here. For the case corresponding to the circular data points in Fig. 9 agreement is good only for a short time span beyond the second lest vafuea ($DAi = 0.1243, P = 5638.89 psi). Unfortunately, as with other empirical methods, it irr not passible to deduce definite rxmclllsions regarding the accuracy of this method. As shown in Fig. 9, if test conditions are different results may be substantially different. The principal limitation appears to be the assumption &(P) = f@2. The only observation we can make is that Eq, (46) yields good results iu all cases for short time spans; this limitation also applies to the methods given in Refs. 22,23,27 and 28. This limitation can be avoided il the method of Ref. 20 is used. f {r~klk/exp fOr 1< rD ~ r.D, and ~rrfi -T + ~ Hl}=(x$,

SPrr Iqnna
... .

4)

(,t)*{/ePd=+(%

)l}=f(xD)$)$

for r.D ~ rD < rtLr. Differentiating Eq. and t and combining the resulting equations expression for the pressure gradient

(54) with respect t.o r we obtain the following

%=(%)

%/[-2(000M28$a

Crrnstrmt

Oil Rate

Production

Mode

In this section we show that by correlating the pressure d i-( ribution in the reservoir during the boundary-domir, ated flow I,criod we can extend the pseudosteady-atate concept to solution gws (Irive systems. Using Eq. (11), Eqs. (4) and (5) can be written mD (r)t) ~ 2~~]Ilrrrk~+~g lk () rfi-1 ~ + se follows

for 1 ~ rD ~ r,D. If we differentiate Eq. (11) with respect to Iime using the definitions of firJ and ~ for constant rate we obtai II Ihe following expression for & ~ = -5.614qox/~ &4hdj7/dt Muskats materials balance equation]2 Gzi ~= ap ()7EF

(57)

is given by (58)

+[nreD-:+)+(:-~$ fOr 1< r ~ r#D, and


@-l)+ lnre - D rt) = 2tiDA - rD + 2 r~D .

[i%-*1*

19)

where B=%.
o

(.53

D_:+

k ( k,

1)[

;*_*+A,D

r~D

1
,

(50)

@ (p, ~o)
where
go

Irr Refs. variables

1 and 2 it is shown that @(F, t) = /3 (jZ ~~), where the with a bar indicates volumetric average values. . Denot.i ng by ~, we have 5.614qo -w
is the production

d~ = z

((iO)

for r,D < ?D ~ rCD. Eqs. (49) and (50) imply that mD (r, t) =
{ [&k/exp (4~~ + & (~))] = f(x,D),

rate. k,h

The production

rate for s # O, is

given by,

(51)

9@=~

()

8p r~~ ,-1

(01)

fOr 1< rD < r#D, and mD (r}t) = ~ [r~/exp for r8D ~ rD ~ r,D. Since the pzeudopreasure is a function of pressure and saturw! ion, it is possible that either one or both of these variables, pressure aml/or saturation, is a function of the variables XD and X,D in the correspending range of radial distance. This posaibllity was explored a d it wss found that pressure is a function of X,D (for 1 ~ ~D < S$D ) u [Id (fOr r#D s rD s re~). Saturation; however, is not a unique fulIction of these variables. Figure 11 illustrates the point that preasu w is a function of X,D or XD. Irr this figure the pressure ratio p (r, t) /p, has been correlated for tDAi ~ 0.1. The data points correspond to Data Set 1 with s = 10, r, = 4.105; and r, = 8000. The dimensiolllms rate, qoD ,e, irr this figure is defined ss follows
XD

where k, is the skin zone permeability. Substituting the right-hand side of Eq. (56) in Eq. (61) for 8p/8r, neglecting (rW/r# and using Eq. (57) we obtain the following result (52) (X) Zg = .w~. (02)

(4ti2 -

(%))1=

Eq. (62) is, of course, also valid for the case ofs = O. The import ~lnce of this expression to solution gas-drive systems is that it is a gelwral ization of the paeudosteady-state concept at the peaitions rw a ml F. The question that needa to be addressed at this point ia if the g,weralization of the pzeudosteady-state concept is valid at other Pcsi! ions in the reservoir. Irr order to answer this question, let us consider for simplicity the region outside of the skin sone. Differentiating Eq. (55) with respect to r and t combining the resulting equations we obtain the next expression

((i3) 141.2 (AOBJi qo


9oD,e =

:=(hw[-2d0006328a (53) The oil rate at any position r (r, < r < r.) is given by

kk,rnhpi

For clarity seven locations within the skin zone and five locations nntside the skin region have been considered. The simulation was stopped when an abandonment preasu re of 245 psi was reached, which orcmw . at 450 days (tDAi = 0.95, tDA= O.?). The pressure responses frmn a single curve for the appropriate variable. These results are typim I of all simulations conducted in this work. Thus based on this observa Iion we can write the-following relations p(r, t) a

kh
90(r) = ~

()

r~~

ap

(64)

Substituting the right-hand side of Eq. (63) into the right-hand of Eq. (64) using Eq. (57) we obtain rz ~ ~ 8pj8t . @/dt

side

90(r)=

~qo(r.

) 1[1

(65)

SPE

19009

R.

G.

Carnacho-V. IIOW

and

R.

Raghavan

It is shown in Refs. 1 and 2 that during the boundary-dominated period the next expression is valid

90(r) *q.

(rw)

1-$ [1

.
e

(66)

Thus from Eqs. (65) and (66) it follows (67) for any position r (r, < r < r,). Tbe same result can be obtainol for pceitions inside the skin zone. Thus Eq. (67) represents a generalization of the pseudosteady-state concept to solution gas-drive syst rms for all positions in the reservoir. Equation (67) also, provides us an insight in the behavior of soiution ga-drive systems produced at a constant oil rate. Differential Iing Eqs. (4) and (5) with respect to time and considering the definil ions of pseudopressure and average pseudopressure given respectively by Eqs, (1) and (7), we obtain the next result

Set 2. The value of rcD is 100 and the skin factor is a-cro. The nnbroken and dashed lines correspond to simulator vahses of se and ~ respectively. These lines also represent the solution of Eqe. 69) and (70). The circular and square data points correspond to d . and ~ respectively, kth computed by solving simultaneously Eqe.. (60) and (62) using the fourth-order Runge-Kutta procedure. Vafua of ~f and rYWfwere used from simulation runs. The maximum ralatk differences in the values of ~e and ~ are 1.1 and 4.5 percent, reapactively. Similar results are obtained in other CU. These results aleo indicate that the approximation given by Eq. (62) is reasonable. Thcomputations also suggest that it is possible to compute P and ~e using wellbore data (pressure and gas and oil rates).

CONCLUS1ONS In the past 20 years, it has been shown that decline curve analysis involves the same basic principles as well test analysis7. Unfortunsrt.cly, like well test analysis extensions to multiphsae flow situationa and the development of theoretical foundations for multipha.w. flow eorrditiorra have been based on mralogy and empiricism. Most disconcerting in the fact that it is not normally possible to reproduce the decline curvm of Arpa using numerical models, altbougb these models reproduce behavior of wells very well for other conslitionsze. This fact suggests that an important gap exists in our understanding of the physical mechanisms that influence well behavior in solution gas drive systctne. This study bsz attempted to provide a general framework for studying ~$ell performance for multiphsse flow in general and solution gae drive in particular and fills an important gap by highlighting the baeic assumptions in the use of Arps equations. It is somewhat surprising that the basis for the Arps equations have not been examined until now. This rigoious examination also permits us to evaluate the conaequerrof predictions based on the Arps equations and has also provided us with an improved analysis procedure for predicting well performance wning the exponential decline curve of Arps. In addition, the role of various parameters such as well spacing, well condition and fluid and formation properties (SW Conclusions 1 and 2 given below) on performance predictions are exalni,t d. Iksides examing decline curve anafyrcis a simple expressio] m -d,ct.ing Row rates over short time eprmn is presented. The ad V* gc of the method is that relative permeability and fluid property data are not needed. This empirical procedure is particularly suited for predicting future Intlow Performance relations. flawd on the results presented sions arc warranted:
J ,_ A... ..,-_..

Comparing Eq. (67) with Eq. (68) we can conclude that the in(t~rtil term in Eq. (68) can be neglected compared to the other terms. [his observation has been confirmed by numerical results. In fact, ii has been found that the integral in Eq. (68) may even be consi{lwvd negligible toward the end of the transient flow period. The same r, suit is also true for constant wellbore pressure production for the tran~iol]t flow period as repor[ed by Camacbos. Fig. 12 is a piot of the variation in sap/c% with distance rD. Ihe ordinate is normalized in terms of the value of at3p/at at r = E tha I is, (crdp/dt)r Our objective was to tesk the reasrmableness of the apl,r,,.imation given by Eq. (67) during the boundary-dominated Row Iwriod. Results for tbe transient period are also presented (filled in !lii(~ points). This case represents a severe test of our developrmmt since Ihe vahre ofs is large (k/k, = 8.08). For the case tDAi = 0.089 the vi\riution in crap/& k significant slthough for small distances (rD < 1000) the variation in rrap/r9t is negligible. This behavior during tramimt Row can be expected by noting that during this flow period, the IMXtdopressure (Eq, (1) with P = pi) may be expressed in term Or I IIC Exponential Integral function. Thus if the ratio tDi/r~ h blg en: 111:11 for the Sernilogaritbmic approximation to be valid (tDi/r& > 10). I he derivative of the pseudopressure with respect to time would be imlrpendent of r. For the case presented in Fig. 12 at ID = 100II iIII(l = 17.89, which expiains the con+l:mt tDAi = fl.089, tbe ratio tDi/r& behavior of rx9p/r% for rD <1000. For the boundary-dominated llmv period the overall wwiation is less than 25 percent. Outside th~ ski II zone the maximum variation is less than 19 percent. The vario I ion decreases with time. These results verify the validity of Eq. (67). The validity by simultaneously comparing these by solving known oft he expressions derived above may also be pj, Ivcrl solving Eqs. (60) and (62) to compute ~~nd ~,,. aIId computations with the values of F and S@ obt .Iii 1A material balance equations like

in this work the following conclu-.,___

~,- m,.. . . . . ... .. .. 1 ne paramew u m nr[]ss equwwns


~/ti.

wpenrls on llre Iunct ton, This function must be approximately constant with time in order to obtain an exponential decline in rate.

J_.

must vary linearly with time in order for the 2.- The function ~/~ decline exponent, b, in the Arps equations to be independent of time. 3.- Exponential decline behavior can be used to approximate production rate if a pseudotime, ?, based on ?t and ~~, is used (see also Ref. 21). 4 ,- If transient data are used with Arpss decline curves, then III* reticid justification exists for the conclusion that the deelinl exponent b will be a function of time and in most of tbe cesea wonld take values greater than unity. 5.- For constant oil rate production the pressure distribution ill the reservoir can be correlated es a function of distance and tiwr.

(Ii!))

where NP and GP are the cumulative oil and gas productions rc+lctively. Note that the system of equations (69) and (70) is indepen, kwt of relatite permeability data; also these equations are valid f,,r ~tll times. The data that is required to so~e simultaneously Eqs. (60) t,ml (62) is an initial pair of values of P, SO, relative permeability dal ;,. :in eetimat.e of pore volume, and the value of the function a (rW). Iiu 1110 13 allows computations of average pressure and saturation for Ik,l n 747

can be extended to eolut.ion 6 ,. The concept of peeudosteady-state gas-drive systet& The product aap/& is approximately independent of distance during boundary-dominated flow.

J$IOMENCLATURE A = drainage area, ft2

8 b = decline exponent Br= Bo= gas formation

BOUNDARY DOMINATED FLOW IN SOLUTION GAS DRIVE RESERVOIRS in Arps equations ,


X,D =

SPE rate production, (Eq. (52))

19009

dominated correlating dominated

flow and constant

volume factor (RB/SCF)

oilformationvohrmefactor(RB/STB) shape factor compr-ibility at initial conditions, ataverage conditions, parameter (psi-]) (Eq. (5)), in f\rl~s

variable inside the skin zone, during bound ttry flow and constant rate production, (Eq. (51)) (Eq.(2)). (Eq.(59)).

CA = geometric
Cti =

a = function of prww+ure and saturation P = function of pressure ~ = Eulers constant, and saturation

totelsystem

~=

total system compr=ibitity (psi-l )

0.57721 ... cp-l (Eq. (9)), cp-l

A~i = total mobility at initial condtilons, T = total mobility at average conditions, p = gae viscosity, cp

di = nominal rate atwhich equations D= GP= h= k= dimensiorrlesrr constant cummulative formation absolute

decline takwplace, (Eq. (15)) MSCF

PO = oil viscosity, cp 4 = porosity SU!2@l& D = dimensionless

gsaproduction,

thickrress, feet permeability, md togss, tooil, fraction fraction (l)) (Eqs. (6)w =

kr~ = relative permeability k,o = relative permeability rn~(r, t)=


fiD (t) =

e = external i = initial conditions s = property wellbore of skin region or shut in conditions

dimensionlezz dimensionlezz (7)).

paeudoprezzure(Eq. volumetric

average pseudopressure curve STB

n = exponent

of backpressure oilproduction,

NP = cummulative p= pressure, psi

Acknowledgements Computing ime was provided by The University of TuIaa.

pi = initial pressure,

psi

p~j = wellbore flowing prezsrrre, psi F-z average reservoir pressure, psi rate azdefinedin Ref. 3 1. CamacheV, R. G.: ~ce University Un cler SQMQtI S&L?.@, of Tulsa, (16S7). R.: Some Theoretical Results Ph. D. Dissertation, 2. CamarJro-V., REFERENCES

qd,L) = decline curve dimensiorrless qf = gas flow rate (MSCF/D) q. = oil flow rate (STB/D)
fk.D,c =

dimensionless

rate (Eq. (53)) feet radial distance


radius, feet

R. G. and Raghavan,

r = radial distance, rD = dimensionless

Useful in AnaIyzing Well Performance Paper zone, feet 16580 submitted

Under Solution Gas Drive,

to SPE of AIME. Using Type Curves, AIME, 269. ~,

r~ = externrd drai.lage r, = radius of altered

permeability

3. Fetkovich,

M. J.: Decline Curve Analysis 1980) 1065-1077; k,

rw = wellbore radius, feet F = radius where p(r) = F, feet R = producing . gas-oil ratio, SCF/STB

J. Pet. Tech. (June

4. Arps, J. J.: Analysis of DecIine Curvea~ 160, 228-247. 5. van Everdingen, A. F. and Hurst, W.:

AIME (1945)

R, = anlution gas-oil ratio, SCF/STB s = mechanical skin factor

The Application in Reservoirs,

of the k,

S~ = gas saturation Sf = volumetric average of gas saturation

Laplace IYansformation

to Flow Problems

AIME (1949) 186,305-324. 6. Jones, J. and Raghavan, R.: (Interpretation Wells: Paper of Flowing Well ft.e SPE 14204 preaerrted and Exhibition 22-25, 1985). of tbe

Sg. = critical gas saturation SO = oil saturation SO = volumetric average of oil saturation

sponses

in Gas Condensate

So, = residual oil saturation


.%i =

at the 60th Annual Teclniical

conference

initial and inmobile water saturation

SPE of AIME, Las Vegas, Nevada, (Sept.

t= time , hours, days td,f) = decline curve dimensionless


tDi = dimensionless t~A~ = dimensionless ~ ~ = dimensionless = dimensionless time as defined in fief. 3 time based on initial conditions (Eq. (13) )

7. Fetkovich,M .J.,@ ~.: Decline Curve Analysis Using Type Curves Case NistoriesV SPE Forma tion Evaluat ioil, (1987), 637-656. R. G.:Constant Wservoirs: hnsient Pressure Flow, Production SPE Paper in $loIution 17118 srcb-

time based on A and initial conditions time (Eq. (12)) time (Eq. (8))

8. Camacho-V., Gaa-Drive mitted

for publication

of SPE of AIME. ear Ei7ects Du rimr llans ient Fluid F low in

XD = correlating

variable outside the rikin zone, during boundary748

9. Aanorrzen, S. L: w

.
.

SPE er t!m Well-~ Dr. Scient., University 10. Hawkins, hf. F., Jr.:

R.

G.

Carrtacho-V.

and

R.

Raghavan M. J.: The Isochronal Testing of Oil Wells, Paper

, Dissertation

23. Fetkovich,

of Bergen, Norway, (1985). A Note on the Skin EfTeet~ ~, AIME

SPE 4529 presented Nevada, 265.) (Sept.

at the 48th Annual Fall Meeting, Las Vegas

30- Oct. 3, 1973). (SPE Heprint Series No. 14

207,356-357. 24. Handy, L. L.: Effect of Local High Gas Saturations 11. CarnachmV., in %iution R. G., and Ritghavan, Gu-Drive Technical Reservoirs, Conference R.: Performance of Wells at the 25. Whitson, C. S.: Reservoir Well Performance and Predicting De tivity Indiccn, ~. and P~ Paper 16745 presented and Exhibition on Produc-

, API (1957) 111-122.

62nd Annual (Sept. 12. Muskat,

of the SPE,

27-30, 1987), Dallas, Tx. M.: The Production Histories of Oil Producing Gaa

liverability,

Paper SPE 12518 submitted

for publication. for solution Ga

26. Vogel, J. V.: Inflow Performance Drive Wells, J. Pet. Tech. (Jan.

Relationships 1968), 83-92.

Drive Reservoirs, Jour. of Armlied Phvsirs (March, 159.

1945), 147-

27. Kelkar, B. G. and Cox, R.: Unified Relationship 13. Uraiet, A. A. and Haghavan, R.: Unsteady F1OWto a ~#m1803,. \ l&sient Pressure, 28, Uhri, D. C. and Blount, Predicts and Well Performance: E. M.: Pivot ture IPR Curves for Solution 14239 pr=ented Exhibition 1985). Gas-Drive

to Predict

FU

Reservoirs,

Paper SPE and 22-25

ducing at a Constant 1812. 14. Ehlig-Economides,

Iressure~

J .Pet. Tec h. (Oc~~O), \

at the 60th Annual

Technical

Conference (Sept.

of the SPE held in Las Vegas, NevAa,

C. A. and

Ramey,

H. J. Jr.: at Constarx

Rate Decline Analysis for Wells Produced Sot. Pet. Enc. J . (Feb. 15. Gentry, Fluid (Sept. 16. Muskat, 1981), 98-104.

Point Method

Quickly

World 0,[, (hfay 1982) 153-164.

R. W. and hfcGray, A. W.: The EtTect of Ibervoir Properties on Production Decline Curves,

J, Pet. T ech ,

1978) 1327-1342. M.: ~ of Oil Pu&&R, Develcoment corporation, Interna(1981), 460-

tional Human Resources 461. 17. Nind, T. E. W.: ~ Book Company,

of Oil Well Prnd-

, McGraw-Eilf
Drainasc Radius,
r., feet

3s!2

(1981 ), pages 43-44.


2024.672, 16404 0.3 10. 0.3280S 5704.78 0.3 1.0s5x10-~ 0.20s so., 15,55 ZooO.,so. 0.119 10. 0.5 15C0 0.0 t.334xlo-~ 1.704s 25, so

18. Fraim, M. L. and Wattenbarger,

R. A.: Gaa Reservoir

Dexlin~

Porosity, A fraction Permeability, k, md

Curve Anafysis Using Type Curves with Bcal Gas Psmrdoprcasure and Normalized 682. 19. Fel,kovicb, M. J.: Personal 20. CamarhwV., lationships presmtted Communication R.: Inflow Performance ReTlme~ ~ Formation Evaluation, (1987), 671-

\vell tad,,,

,W , fet

Initial Pressure, pi, psi Initial Waler Saturation, SW; Initial Compressibility, ctl, psi-L Initial 0S Viscosity, Ma, cp Thicknw,

h, ft

R. G., and Haghavan, for Solution

Pressure Product ion Ratio,

p., /p;:

0.2971-0,s941 40- Imll

Omof- 0,8 10 Iro

Gas Drive Hme.rvoira, Paper SPE 16204


Production rate, .?., (sTB/D):

at the 1987 SPE Production

Operations

Sympmium,

(March 9-11, 1987), Oklahoma 21. Chen, H. Y. and Poston, fhrnction b Permit presented Better S. W.:

City, Oklahoma. Application of a Paeudct-Time Paptw 17051 held in Pit&

Decline Curve Anafysia, Regional Meeting

at the SPE

Eastern (Oct.

burgh, Pennsylvania, 22. Standing, formance


M. B.:

21-23, 1987). the Calculation Solution of Inflow PerR.ae.rvoir,

Concerning Producing

of Wells

C;aa Drive

J. Pet. Tech. (Sept.

1971), 1141-1142.
749

SPE

19009

?? -1

$$
I I

70

! *
il
~Q

%?

?2
a3zIlVW0N

!%/%

3m

e
w

1!

Ii
aE ?5
1 1.

Q
(*\) Ob/(l)Obs@b 31W ss31t401W3wKl

(ti
01

gp/%p.

3Auvwm

3.WM

,.,

sPE

19009

. ,

AVERAGE OIL SATURATION, ~

!d/(+J)d OIIW

341TSS3Ud

(w(+w~~

You might also like