Académique Documents
Professionnel Documents
Culture Documents
Jl':-;E, 1961
a where X = total length of the reservoir, and
- Bo a~ (ugAo + uoABoRs) - (0 - BoRs) Po = initial gas-cap pressure.
The boundary condition at the original gas-oil
a
-(uoAB) - Bo(VoRs + V g ) - (0 - BoRs)Vo
contact is that the rate of gas entry at that point,
a~ i g • is determined by the expansion of the gas-cap
gas which has a volume V. The amount of gas in the
= cpAISg[Bo(o - BoRs)'- B~(o - BoRs)] + gas cap at any pressure is V 0 (P). and the rate at
[1- Swc] [Bo(BoRs)'+ B;(o - BoRs)]! ap which gas enters the oil zone is -v 0' ap = i g •
ap at at
At any production-well location, it is desirable to
cpA[D(P, Sg)] -' , . . . . . . . . . . . . . . (6)
at specify the' rate of stock-tank oil production qo
provided this rate does not exceed the productivity
where the function, D(P. 5 g ) is defined by the last of the well. This oil rate is related to the gas-
equality. production rate through the relative permeabilities
Next, u o • u g and 50 are eliminated from Eqs. 6 and fluid properties existing at that time at the
and 2 by use of Eqs. 3, 4 and 5, which results in location in question. The quantitative relationship
the following equations. is derived from material balances on the differential
element ,~ of the reservoir which contains a pro-
duction well. A material balance on the oil phase
yields
and
. . . . . . . . . . . . . . . . . . . .. (7)
Assuming ( as~
e:)1 (~:)2
= (az\ , the result IS
ax It aX/2
o krg flo . . . . . . . . . . . . . . (9)
gg go - - -
Bo k ro flg
claim is made that this is the only set of equations L iqu id Saturation, per cent
0
:;
to be evaluated at Xj and t n+l are replaced by terms 0 0.4
V>
of the form, I/)
0
<.:)
Jl''iE, 1961 95
the procedure to two or three dimensions is perfectly
straight-forward.
at 7j .
.
TIme d ··
envatIves suc h as (- - are approXImate d
aP
as
p.
1. n;t ,. n. 1 - p.
Production wells are approxi-
80 "'-
"-...," .....
~ .. ..
,,\ \
c:..
U
G;
"-
60
~
:ii
.."E
G;
a. 40
"> krg
.
.2
c:r
20
o 20 40 60 80 100
Liquid Saturation, per cent
50.-------------------------------------,
40
u
u
c' 30
o
u
::>
"'6
a: 20 wh-=re i g • qo. qg 2 0.
(5 There are (J + 1) grid points in the reservoir
corresponding to j == 0, 1, 2 . . . J-l, J. The reser-
10
voir may be divided into segments of length .... x.
centered about grid points. Eq. 11 as applied to the
point j represents a combination of the two material-
10 20 30 40 50 60 balance equations for the segment centered on j.
Time, min.
°
At the end points j == and j == J, there exists only
FIG. 4 - EXPERIMENTAL PRODUCTION HISTORY, one-half of a segment; hence, the right-hand side of
USED AS BOUNDARY CONDITION TO COMPUTE Eq. 11 should be divided by two when applying it
BEHAVIOR OF DISSOLVED-GAS DRIVE. to these points.
96 SOCIETY OF PETROLEI'M E'I'GI:\EERS JOI'R'I'AL
Maintaining the proper boundary conditions at the than the flow rate, this pressure is first corrected
points j = 0 and j = ] also requires that k - 1 and to allow for the pressure drop due to radial flow
k be set equal to zero in Eq. 11. The resulting into the well. Then this corrected pressure is used
J
form of the equation places a restriction on the in lieu of Eq. 11 at the grid point corresponding to
pressure such that the rate of flow at these points the well. Writing Eq. 11 for all points except those
is equal to that specified by the ig , qo' and q g at which pressures are specified and combining
terms of Eq. 11. If no production or injection occurs this set of equations with the specified pressures
at j = J, then ig =qo =qg =0 and the no-flow boundary results in a system of equations which can be
conditions of Uo =u g=O are satisfied. The boundary solved to yield a complete pressure distribution at
condition at the gas-cap contact is satisfied by time level n + 1 (assuming pressures and saturations
setting i g at j = 0 equal to were known at time level n). Then Eq. 11 written
for the points at which the pressure was arbitrarily
(P o,n+l - p o,n) specified will contain two unknowns, q 0 and qg,
- Vo'
and can be combined with either Eq. 9 or Eq. 10,
I1t
as appropriate, to yield qo and qg for the well.
As stated earlier examination of Eq. 11 reveals Eq. 12 is solved explicitly for the saturation at
that, if conditions of saturation and pressure are. time level n + 1. As with all explicit equations,
known at time level tn, the only unknowns in Eq. 11 there is an upper limit on the size of the time step
are pressures at time tn+l' The equation may, there- that can be employed. This limit cannot be estab-
fore, be put in the form, lished precisely for the general case, but for the
case of constant coefficients (i.e., the pressure-
maintenance problem where fluid properties are
considered invariant), this limit is given by
where aj' hj' Cj and djareknownconstants for each
point j. The equation for each point involves three I1x¢ ]
unknowns, except for the firsd j =0) and last (j'" J)
IJ.t
A < [
- (u +u)f'
equations which contain only two. There are, al- . 0 g g minimum
together, ] equations and] unknowns. This system
of linear simultaneous equations can readily be The max1mum value of (u o + u g ) is employed in
solved by a method described in the literature 10 to f"g dfgg
yield a set of pressures at the new time level tn+l'
.
th1s .
equatlOn. Th e quantity
. 1S equa 1 to dS
These new pressures are used in a difference evaluated at the gas saturation existing at the gas
approximation of Eq. 8 to find the new saturation at front.
t n+l. The approximation used for Eq. 8 is This limitation was derived from a heuristic
stability analysis and verified by experimental
computing for the case of pressure maintenance.
While it was found that a slightly less stringent
time-step limitation could be applied for gas-cap or
dissolved-gas-drive calculations, this equation
serves as a useful guide in selecting the time step.
a. 32 .,-u 3
.=
E
.,c
::J
a;
0
:=
u 24 - - Computed Data C
"0 0 Experimental Data ] 2
0
a. .,c
~ 0 - - Computed Data
C
16 .,
11.
o Experimental Dato
., lit
:; "
..c
11.
.,
lit
lit
8 6
ct 0 O~ _ _L __ _L -_ _ ~ __ ~ __ _L~
0 0 o 10 20 30 40 50
0 8 16 24 32 40 48 56 Production of Oil, % of original
Production of Oi I, per cent of oriqinal
FIG. 6 - VARIATION OF THE DIFFERENCE IN OIL-
FIG. 5 - PRESSURE BEHAVIOR AT TOP OF COLUMN, PHASE POTENTIAL AT TOP AND BOTTOM
DISSOLVED-GAS DRIVE. OF COLUMN, DISSOLVED-GAS DRIVE.
-0c
40 - - Computed Data
- d-
III
0
~
~
120
0 Experimental Data 0
'Co
.~
0
-= c~
'7; Q)
30
Il'1
0
c' < 80
C
U c:
:> 0
II 20 U
a': GI
II> VI
40
.,"
t!I III
~
!
u. 10 U
0 o~----~----~----~----~
0
0 10 20 30 40 50 o 2 4 6 8
Production of Oil, % of original Distance from Gas Cap, thous. of ft.
FIG. 7 - FREE-GAS PRODUCTION HISTORY, DIS- FIG. 9 - VARIATION OF CROSS-SECTIONAL AREA,
SOLVED-GAS DRIVE. GAS-CAP RESERVOIR.
Jll:\,E, 1961 99
TABLE 2-PERTINENT RESERVOIR DATA-GAS-CAP DRIVE 8 0 0 r - - - - - -_ _ _ _ _ _ _--,
Original and Saturation Pressure (psig) • 2,700
Reservoir Temperature (OF). • • • • 181 >-
tI
Vol ume of Oil Zone (acre-ft). • • • • 20,986 ~
Volume of Effective Gas Cap (acre-ft) 10,883 :.;
...Q
600
Formation Dip (ft/mile). • •
Average Permeability (md) • • • • •
226
165
.,
.:
Average Porosity (per cent). • • • • 23 .!!
tI
Average Connate-Water Content of Oil Zone (per cent) 28 Q:
Average Connate- Water Content of Ga s Zone (per cent). 25 c 400
Original Oil in Place (STB). • • • • • 19,840,000 o£0
~
Original Gas-Cap Gas in Place (Mscf) • • • • • • 13,969,000
-g
tt
part of the experiment was estimated to be slightly 0 200
in excess of 1 psi. Inconsistencies in the material ~0
balance on gas of about 4 per cent prevented esti- I-
mation of the degree of supersatutation late in the
experiment. o~----~----~----~------~-J
In computing the behavior of this model, the rate o 2 4 6 8
of production of stock-tank oil at any time was Time, yrs.
derived from the data of Fig. 4. This rate schedule
was entered into the calculations as the boundary FIG. 10 - VARIATION OF TOTAL PRODUCTION RATE,
GAS-CAP RESERVOIR.
condition at the production end of the model, and
the resulting gas production and pressute behavior The difference in oil-phase potential measuted
of the model was calculated by the previously de- between the top and bottom of the column (Fig. 6)
scribed numerical technique, was relatively small, 2.5 to 3.5 psig, and the
Both the results of calculation and experiment pressure difference was of course smaller, 0.6 to
are shown in Figs. 5, 6 and 7. In all three figutes, 1.6 psig. The experimental data exhibit appreciable
the abscissa is th~ cumulative quantity of oil scatter; but, even so, the average deviation of the
produced. The ordinate in Fig. 5 is the pressute at experimental from the computed values is ± 12 per
the top of the column; in Fig. 6 it is the difference cent.
in the oil-phase potential (P + Po g L'l z) measuted The data of Fig. 7 indicate that the first free-gas
at the top and bottom of the column, and in Fig. 7 production occutred when about 15 per cent of the
it is the cumulative quantity of free (undissolved) original oil had been produced. Thereafter, the
gas produced. In these figutes, the solid cutves agreement of the experimental and computed data
represent the calculated data; the open' circles is reasonably good. In fact, the comparisons shown
are data points which were experimentally measuted. on Figs. 5, 6 and 7 indicate that the computed and
The compute'd pressute behavior at the top of the experimental reservoir behavior agree within the
column (Fig. 5) agrees quite well with that observed probable error of the experiment.
experimentally, especially early in the experiment.
Agreement was not as good late in the experiment GAS-CAP DRIVE
when the pressute was declining very rapidly, making Kirby, et al,12 published a limited amount of
it experimentally difficult to measute accutately production history for a gas-cap-drive reservoir in
the cumulative quantity of oil produced. Note, which there was no active water drive. Data for
however, that the final cumulative oil recovery this reservoir are given in Table 2. Table 3 gives
measuted after the final blowdown to zero pressute data on fluid properties as a function of pressute
agreed very well with the calculated ultimate for the res~rvoir fluids. Relative permeability
recovery. data for the reservoir are shown in Fig. 8. These
'"a.
'iii 2600
-
C>
C>
0-
0
2600 0
i ~
"
VI "
VI
VI
" ,-
VI '"
a':"
~ VI
a. a.
a. ~u
a 0
0
U 3=E
VI 2500 " g 0
a aU 2500
t:)
Ia.
E a
aU
-a aVI
cot:)
]() I
7. Shreve, D. R. and Welch, L. W., Jr.: "Gas Drive and 10. Douglas, Jim, Jr.: Jour. Assoc. of Computing Machi,,-
Gravity Drainage Analysis for Pressure Maintenance ery (1959) Vol. 6, No.1, 48.
Operations", Trans., AIME (1956) Vol. 207, 137. 11. Johnson, E. F., Bossler, D. P. and Naumann, V.O.:
8. Terwilliger, P. L., Wilsey, L. E., Hall, H. N., Bridges, "Calculation of Relative Permeability From Dis-
P. M. and Morse, R. A.: "Experimental and Theoreti- placement Experiment", Trans., AIME (1959) Vol.
cal Investigation of Gravity Drainage Performance", 216, 61.
Trans., AIME (1951) Vol. 192, 285. 12. Kirby, J. E., Jr., Stamm, H. E., III andSchnitz, L. B.:
9. Wooddy, L. D., Jr. and Moscrip, Robert, III: "Perform- "Calculation of Depletion History and Future Per-
ance Calculations for Combination Drive Reservoirs", formance of Gas-Cap Reservoir", Trans., AIME (1957)
Trans., AIME (1956) Vol. 207, 128. Vol. 210, 218.
DISCUSSION
The authors have considered a very difficult the multivalues given by Buckley and Leverett occur
problem and one in which many questions remain to because they made a mathematical transformation
be answered. The partial differential equations are which is not valid in the vicinity of the saturation
very complex, and the finite-difference method of discontinuity. Thus, there is some disagreement
solution has not been established rigorously. It is concerning the Buckley-Leverett solution. For this
known that finite-difference methods can fail to reason, it is discussed in the following paragraphs.
approximate the true solutions of partial differential The Buckley-Leverett solution can be obtained
equations. Thus, there is reason to doubt the validity by well established mathematical methods of the
of the method from a mathematical viewpoint. type employed in Ref. 2. The partial differential
The results of reservoir analysis may lead to equation for the gas saturation is
large capital investments and may be a prime factor
in determining the amount of oil recovery. Thus, ag(po - Pg) G(S ~ = _¢~
there is an economic incentive for the results of flo g ~ at
reservoir analysis to reflect the true nature of the . . . . . • • . • . • . (1)
reservoir. Misleading results .can be worse than no where a = the dip angle,
results at all if they lead to poor' investments and
the waste of important national resources. Thus, V(t) = a known function of time,
there is a considerable justification for being very 1
F( Sg) '" -------"'--- , and
critical of the validity of new methods of reservoir 1 + flg ko
analysis. flo kg
The method of analysis presented by the authors
invol ves the solution by finite-difference approxi- G(Sg) '" kolk F(Sg)'
mations of two simultaneous partial differential Eq. 1 is a first-order partial differential equation
equations containing the fluid pressure and one of of a type for which there is a well known method of
the fluid saturations as dependent variables, and obtaining solutions (see, for example, Chapter 4 of
distance and time as the independent variables. Ref. 3). The analytical solution of Eq. 1 is
Since two dependent variables can vary with dis-
tance and time, it seems reasonable to expect that tV(t) ag(p -p)
x F'(Sg) f -dt 0 g G't I(S)
the authors would present examples where the pres- o ¢ ¢ flo +t;' g
sure and the saturations have considerable varia-
• • • . • • . • . . • . • . . • . . (2)
tions with both time and distance. Also, one would
expect that numerical results would be shown to where tjJ(Sg) is the initial gas saturation. For the
case considered by the authors where the initial
approach the results obtained by analytical methods
as the increments in time and distance are reduced. gas saturation is zero, tjJ(Sg) is zero. That Eq. 2
In the example of complete pressure maintenance, satisfies Eq. 1 can be verified by direct substitu-
the numerical results do not appear to converge to tion. Eq. 2 yields the same multivalue saturations
the analytical solution of the partial differential profiles as obtained by Buckley-Leverett. Thus,
equations. The analytical solution is contained in the Buckley-Leverett solution is the true analytical
the well known paper by Buckley and Leverett. 1 It solution, and it contains multi valued saturations.
is maintained by some that the Buckley-Leverett Furthermore, the analytical solution does not con-
solution is not the true analytical solution and that tain a discontinuity.
In formulating the partial differential equation, it
lReferences given at end of Discussion. IS assumed that the saturations are continuous
102 SOCIETY OF PETROLEUM E:-.IGINEERS JOURNAL
functions of distance. The presence of a discontin- that approach the solid line in Fig. 2 as the incre-
uity violates this assumption and across the dis- ments in time and distance approach zero.
continuity the partial differential equation breaks The Buckley - Leverett flow considered by the
down. If it is assumed that a discontinuity exists authors is a very simple flow compared to what the
at a given point and a given time, continuity con- partial differences can represent. There may be
siderations yield the relation for the velocity of many discontinuities present in the porous media.
the discontinuity. Buckley-Leverett, followed by a They can arise within the body, and they can begin
number of investigators, have used the procedure at the boundaries even after the flow has been in
of first determining the analytical solution and then progress for some time. Since the authors have
eliminating the presence of multivalue saturations failed to present a completely satisfactory demon-
by employing discontinuities which satisfy condi- stration that their method yields the correct results
tions. for a very simple case, how can one be sure the
method will be applicable in a very involved case
The results presented in Fig. 2 do not approxi-
which may arise in the analysis of a natural reser-
mate the true analytical solution but appear to ap- voir?
proximate the combination of the analytical solution
and the discontinuity which satisfies material REFERENCES
balance. The authors should explain why the re-
1. Buckley, S. E. and Leverett, M. C.: "Mechanism of
sults do not approach the true analytical solution Fluid Displacement in Sands", Trans., AI ME (1941)
since the finite-difference relations are represented Vol. 146, 107.
as being approximations to the partial differential 2. Martin, J. C.: "Some Mathematical Aspects of Two-
equations. Furthermore, in the vicinity of the dis- Phase Flow With Applications to Flooding and Gravity
continuity the numerical results are a very poor Segregation Problems", paper presented at 1st Vene-
zuelan Regional Meeting of AIME, Caracas (Nov.,
approximation to the combination of the analytical
1956); also Prod. Monthly (April, 1958).
solution and the discontinuity. Thus, it is not clear
3. Miller, F. H.: Partial Differential Equations, John
that the finite-difference method would yield results Wiley & Sons, Inc., N. Y. (1941).
Many of Martin's comments concern the example the case of complete pressure maintenance, it does
case of complete pressure maintenance. The ana- not approximate what he calls the "true analytical
lytical statement of this problem was formulated by solution" of the differential equation. The solution
Buckley and Leverett! as a first-order quasilinear to which he refers was given by Buckley and Leverett
hyperbolic differential equation. This class of and contains multiple values of saturation at some
differential equation often arises in simplified locations in space. Obviously, this is not a physi-
theories which ignore mechanisms of dissipation cally significant solution of the problem and so
such as viscous stresses, heat conduction, ohmic must be rejected. The generalized solution which
loss or, as in the present case, capillary pressure. does occur in nature is formed by replacing the region
It is the rule, rather than the exception, that the of multiple-valued saturations by a saturation dis-
physically meaningful solutions of these problems continuity so located that the resulting solution
develop singularities (discontinuities) after a finite satisfies the law of conservation of mass. The
time. 2 Lax 3 notes that, while the classical solution authors selected the difference analogue of the
of the differential equation cannot be continued differential equation with the objective of approxi-
after this time, it is possible to continue it in a mating the solution having physical significance.
generalized sense. This kind of generalization is The comparison of Fig. 2 gives a measure of the
dictated by the integral version of the conservation degree of achievement of this objective. Martin is
law. Lax refers to this generalized solution as a critical of the accuracy of the approximation in the
weak solution and gives a formal definition of it. It vicinity of the saturation discontinuity. He states,
develops that these weak solutions are not deter- ct • • • It is not clear that the finite-difference method
mined uniquely by their initial values. Therefore, would yield results that approach the solid line in
an additional principle is needed for developing a Fig. 2 as the increments in time and distance ap-
relevant subclass. In this aspect, the problem is proach zero". The authors take issue with this state-
similar to the solution of a quadratic equation. A ment. For a perfect solution, the length of theregion
unique answer cannot be obtained from the equation of the reservoir containing saturations between zero
alone; additional information is necessary to reject and the frontal saturation should be zero. The data
tqe extraneous root. of Fig. 2 show that when a distance increment of
With this background,-we are prepared to consider 1,000 ft was used in the numerical procedure, this
Martin's comments. One of his objections is that, region occupied 22 per cent of the reservoir (from
when the authors' numerical procedure is applied to 7,000 to 9,200 ft). For a distance increment of 500