Escolar Documentos
Profissional Documentos
Cultura Documentos
Summary and the oil zone formation and water expansion term, Efwo , is de-
An improved material-balance formulation is presented for deter- fined as
mining original hydrocarbons-in-place (OHIP) in waterdrive oil
and gas reservoirs. The improved formulation reduces the number p i * pc f ) S wo c w
E fwo + B oi . . . . . . . . . . . . . . . . . . (4)
of unknowns in the regression analysis through the definition of a 1 * S wo
combined aquifer/reservoir expansion term (CARET). Field exam-
ples analyzed with the CARET formulation are presented illustrat- The free gas expansion term, Eg , in Eq. 1 is defined as
ing the nonunique nature of waterdrive material-balance solutions
for OHIP. A method is presented for reality checking the best-fit B gi S og E o
aquifer parameters obtained from the regression analysis. The E g + B g * B gi ) . . . . . . . . . . . . . . . . . (5)
1 * S wg * S ogB oi
CARET material-balance formulation is shown to provide several
advantages over existing regression analysis techniques. and the free gas zone formation and water expansion term, Efwg , is
defined as
Introduction
Material-balance analysis of waterdrive reservoir performance to p i * pc f ) S wg c w
E fwg + B gi . . . . . . . . . . . . . . . . . . (6)
determine OHIP requires a method for estimating water influx. If an 1 * S wg * S og
analytical aquifer model is used, the aquifer description must be
known or determined as part of the OHIP analysis. Havlena and Eqs. 1 through 6 represent the general form of the material-balance
Odeh,1,2 Tehrani,3 and others have proposed material regression equation for an oil or gas reservoir. In a waterdrive gas reservoir Eq.
analysis techniques for determining OHIP. These methods work 1 reduces to
well when applied to volumetric reservoirs; unfortunately, they are
often less effective in waterdrive reservoirs. This paper presents an F + GE g ) E fwg ) US. . . . . . . . . . . . . . . . . . . . . . . . . . (7)
improved regression analysis technique that works well in both vol-
umetric and waterdrive reservoirs. In volumetric reservoirs, it sim- Similarly, in a waterdrive oil reservoir, Eq. 1 becomes
plifies to the solution proposed by Tehrani. Field cases are presented F + NE og ) US , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (8)
showing that regardless of the regression analysis technique used,
the potential for nonunique OHIP solutions must be addressed where Eog is a combined oil and free gas expansion term defined
whenever an aquifer is present. Finally, a method is presented for with the gas cap m ratio as
reality checking the analytical aquifer constant obtained from the
OHIP regression analysis to ensure that it corresponds to a reason- mB oiE g ) E fwg
able aquifer description. E og + ) E o ) E fwo . . . . . . . . . . . . . . . . (9)
B gi
F + N pB o * R s B g ) G p * G iB g ) W p * W iB w . E og E og
F + N ) U S . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11)
. . . . . . . . . . . . . . . . . . . . . . . (2) If the correct parameters are assumed for the aquifer model, a plot
of F/Eog vs. S/Eog should give a straight line with a slope equal to
The oil and solution gas expansion term, Eo , in Eq. 1 is defined as the aquifer influx constant, U, and a y intercept equal to the original
oil in place (OOIP), N. A similar solution technique is proposed for
E o + B o * B oi ) (R si * R s)B g . . . . . . . . . . . . . . . . . . . . . . (3) determining original gas in place (OGIP) in a waterdrive gas reser-
voir.
Copyright 1996 Society of Petroleum Engineers
Tehrani3 Regression Analysis. Tehrani advocated applying re-
Original SPE manuscript received for review Oct. 10, 1994. Revised manuscript received
March 4, 1996. Paper peer approved March 5, 1996. Paper (SPE 28630) first presented at the
gression analysis directly to Eq. 7 or 8 to determine values for OHIP
1994 SPE Annual Technical Conference and Exhibition held in New Orleans, LA, Sept. 2528. and the water influx constant, U. His approach requires multiple,
model, but may be easily extended to other analytical models. Rela-
tionships are also presented for sensitivity analyses of the aquifer
parameters obtained from the regression analysis. E CARET + 2c e S 1 ) m
1 * S wo 1 * S og * S wg
hh A
R
Aquifer Constant Relationship to OHIP. The VEH aquifer influx
constant, U, is a function of the aquifers effective compressibility,
inner radius, thickness, porosity, and angle-open-to-flow5,6:
)
mE g ) E fwg
B oi ) E o ) E fwo. . . . . . . . . . . (21)
U+ 1.1190c e r 2i h Af
q + 2c pr i h Af
360 e
2
5.615
q . . . . . (12)
360
B gi
Eqs. 18 and 20 are the equations of a simple straight line. If the correct
Eq. 12 may be modified to express the influx constant, U, in terms of values are assumed for the VEH4 water influx parameters a and rD , a
the reservoir pore volume and the aquifer/reservoir thickness ratio if plot of F vs. ECARET should give a straight line passing through the ori-
the reservoir is assumed to have the same geometry as the aquifer: gin with a slope equal to the OHIP as shown in Fig. 1. Simple linear
regression analysis can then be used to determine the best-fit values for
U + 2c eV pR hh .
A
R
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (13)
OHIP with the voidage minimization technique proposed by Tehrani.3
Aquifer Property Relationships. Once best-fit values for OHIP
and the aquifer influx constant, U, have been determined, it is im-
In a gas reservoir, the reservoir pore volume is given by portant to confirm that the U value represents a realistic aquifer de-
GB gi scription. Each OHIP solution represents specific values for the
V pR + . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (14) VEH dimensionless time constant, a; the dimensionless aquifer ra-
1 * S wg
dius, rD ; and the VEH aquifer influx constant, U. The VEH aquifer
Eqs. 13 and 14 may be combined to relate the aquifer influx time constant, a, is defined in terms of aquifer properties as6
constant, U, to the OGIP, G:
a + 2.3092 k . . . . . . . . . . . . . . . . . . . . . . . . . . . (22)
U + 2c e B gi h A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (15) fmc e r 2i
G 1 * S wg h R
Using the definition of the aquifer constants U and a given by Eqs.
A similar method for expressing U in terms of G was first proposed 12 and 22, we can show that the brine mobility for the VEH un-
by McEwen7 as a way to reduce the number of unknowns in the wa- steady-state radial aquifer model is related to the aquifer thickness
terdrive OGIP regression analysis. This relationship may be ex- and angle open to flow by
tended to an oil reservoir with a primary gas cap by defining the res-
ervoir pore volume as the sum of the oil zone and gas cap pore k 139.32Ua . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (23)
volumes. With the gas cap m ratio, the total reservoir pore volume m+ qh A
may be represented by
V pR + NB oi 1 *1S wo
) m
1 * S og * S wg
. . . . . . . . . . . . (16)
U + 2c B
N e oi
1 ) m
1 * S wo 1 * S og * S wg
hh .
A
R
. . . . . . . (17)
s
7,446 0.5804 1.0529
7,400 0.5825 1.0530 FE CARET
V+ 100, . . . . . . . . . . . . . . . . . . . . . . . . (24)
F
TABLE 3WATERDRIVE GAS RESERVOIR PRESSURE AND where the standard error of estimate of F from ECARET is
PRODUCTION HISTORY
2
F * F fit
t p Gp Np Wp
s FE + . . . . . . . . . . . . . . . . . . . . . (25)
(years) (psia) (MMscf) (MSTB) (MSTB) CARET n*1
0.0 8,490 0 0 0 and F is the arithmetic average of the n values of F used in the regres-
0.5 8,330 1,758 2 0 sion analysis.9,10
1.0 8,323 5,852 3 1
4. Repeat Steps 2 and 3 for a range of rD values to determine the
1.5 8,166 10,410 66 3
2.0 8,100 14,828 98 4 rD value giving the minimum regression error, V, and the best-fit val-
2.5 7,905 21,097 138 7 ues for G or N. Use Eq. 15 or Eq. 17 to calculate the water influx
3.0 7,854 26,399 180 9 constant, U, from the best-fit G or N value.
3.5 7,858 30,042 215 10 5. Reality check the solution using Eq. 23 and the best-fit OHIP
4.0 7,900 32,766 237 11 and U values to calculate the required aquifer qhA value. If this cal-
4.5 7,971 34,548 257 11 culated product of the aquifer angle open to flow and thickness
5.0 7,883 37,590 282 12 matches the actual aquifer description, the OHIP solution is reason-
5.5 7,728 42,446 314 16
6.0 7,550 51,117 375 54 able. If it does not, re-evaluate the aquifer permeability and inner ra-
6.5 7,446 57,697 420 153 dius used to calculate the time constant or try another aquifer model.
7.0 7,400 63,678 465 433
7.5 7,600 65,432 475 715 Field Example
8.0 7,675 65,613 475 753
8.5 7,600 67,593 477 1,042 The field example considered here is a wet-gas waterdrive reservoir
9.0 7,600 70,688 484 1,237 with reservoir and aquifer data as summarized in Table 1. As shown
9.5 7,615 72,226 488 1,575 in Fig. 2, the fields production rate averaged 10 MMscf/D during the
10.0 7,623 72,943 489 2,383 first 6 months of production and reached a peak of 47 MMscf/D in
Year 6. The peak gas rate declined sharply after water breakthrough.
After 10 years on production, the reservoir pressure had declined to
Eq. 23 may be used with the best-fit estimates for U and a to calcu- 7,623 psia, a drop of less than 900 psia, while the produced water rate
late the corresponding value of the aquifer qhA product required by had climbed to over 4 MSTB/D. The reservoirs PVT data and pro-
the OHIP solution. duction and pressure histories are tabulated in Tables 2 and 3.
OHIP Solution Technique. The following CARET regression
analysis procedure is recommended for determining OHIP and the Constrained OHIP Solution. A CARET regression analysis for
aquifer influx constant in waterdrive reservoirs with radial geome- OGIP was performed with the VEH4 unsteady-state radial aquifer
tries. The procedure assumes the reservoirs production and pres- model and the calculation procedure outlined previously. To ac-
sure histories are known and calculates water influx volumes with count properly for condensate production in the material-balance
the VEH unsteady-state radial aquifer model. analysis, the produced condensate volumes were converted to a gas
equivalent and added to the produced gas volumes with a conver- vestigated with Eqs. 23 and 22 by calculating and plotting the re-
sion factor of 0.708 Mscf/STB. quired aquifer thickness, hA , and inner radius, ri , as functions of the
Regression Analysis. Eq. 22 was used to calculate an aquifer time assumed aquifer permeability, k. These results are shown graphically
constant of 6.893 years*1 from an aquifer permeability estimate of in Fig. 4. Entering the plot with the estimated aquifer permeability of
22 md and an aquifer inner radius of 3,300 ft. CARET regression 22 md, the required aquifer inner radius and thickness are shown to
analyses were run for 50 different values of the aquifer dimension- be 3,300 ft and 110 ft, respectively. This plot can be used to evaluate
less radius, rD , spaced logarithmically between 2 and 25. An infi- uncertainty in the aquifer permeability estimate by entering the plot
nite-acting aquifer response was assumed for rD values greater than at a different permeability value and reading off the aquifer thickness
25. The regression error and OGIP estimate for each assumed rD and inner radius required by the OHIP solution.
value were plotted against the rD value as shown in Fig. 3. A best-fit Reservoir Pressure Solution. As an additional check on the va-
rD value of 7.7 was picked at the minimum regression error of 1.6%. lidity of the OHIP regression analysis, the OGIP and aquifer param-
The CARET voidage vs. expansion plot for this best-fit rD value eters from the best-fit solution were used to backcalculate a materi-
shown in Fig. 1 yields an OGIP of 225 Bscf and an aquifer influx al-balance reservoir pressure history from the reservoirs
constant, U, of 1,973 RB/psi. production history. Fig. 5 shows the agreement obtained between
Aquifer Constant Reality Check. To confirm the validity of the the calculated and observed reservoir pressure histories. The peri-
OHIP solution, the best-fit aquifer constant value, U, was reality ods of increasing pressure correspond to the production rate de-
checked with Eq. 23. Assuming an aquifer angle open to flow of creases in Fig. 2. These reductions in the produced voidage rate al-
360, the required aquifer thickness was calculated to be 110 ft by lowed the aquifer to repressurize the reservoir partially. Once the
solving Eq. 23 for hA . This calculated thickness agrees reasonably produced voidage rate again exceeded the water influx rate, the res-
well with the map-based aquifer thickness estimate of 100 ft, con- ervoir pressure history resumed its decline.
firming the validity of the OHIP estimate. Additional sensitivities to
the assumed aquifer permeability, inner radius, and thickness were in- Unconstrained OHIP Solution. The previous OHIP regression
analysis was constrained by an aquifer time constant, a, calculated
from estimates of the aquifer permeability and inner radius. Nonlin-
ear material-balance regression analysis techniques available com-
Fig. 5Actual vs. calculated waterdrive gas reservoir pressure Fig. 6CARET regression error vs. VEH4 aquifer time constant
histories for OGIP+225 Bscf, a+6.893 years1, and rD +7.7. and dimensionless radius for waterdrive gas reservoir example.
mercially allow simultaneous determination of the overall best-fit permeability is high, only finite-acting solutions may be found for
values for OHIP, the aquifer influx constant, U, the dimensionless low rD values, as shown in Fig. 6 for rD values of 7.7 or less.
aquifer radius, rD , and the dimensionless time constant, a. The con-
sequences of attempting this kind of unconstrained regression anal- Comparison With Existing Solution Techniques. For comparison
ysis were investigated with the CARET method. Regression analy- purposes, the waterdrive gas reservoir example was also analyzed
ses were done for rD values of 5, 7.7, 10, 25, and infinity for 50 with both the Havlena and Odeh1,2 and Tehrani3 solution techniques.
aquifer time constant values spaced logarithmically between 0.01 The CARET formulation was found to be considerably more tolerant
and 10,000 years*1. The associated regression errors were plotted of error in the early-time pressure data than the Havlena and Odeh ap-
vs. a and rD as shown in Fig. 6. proach owing to the origin constraint required by the CARET
Nonunique OHIP Solutions. A closer inspection of Fig. 6 reveals straight-line plot. No pressure data editing was required for the CAR-
the nonunique nature of the unsteady-state radial aquifer OHIP solu- ET analysis; however, the first two points in the pressure history had
tion. At least one OGIP solution (regression error minimum) may be to be omitted to obtain a successful Havlena and Odeh solution.
found for any assumed value of rD . The OGIP estimates correspond- Constrained OGIP Estimates. When constrained with an aquifer
ing to each of these regression error minimums are plotted as the solid time constant of 6.893 years*1 calculated from aquifer properties,
curve in Fig. 7. Each point on the curve represents the best-fit OGIP the Tehrani method gave an OGIP estimate of 221 Bscf, very close
estimate and associated minimum regression error value for a given to the CARET estimate of 225 Bscf. Constraining the Havlena and
rD value. Although the rD +7.7 solution clearly gives the overall Odeh analysis in a similar manner resulted in an OGIP estimate of
minimum regression error for this waterdrive gas reservoir, in reser- 177 Bscf, 25% lower than the CARET and Tehrani estimates.
voirs with lower hydrocarbon compressibilities the best OHIP solu- Regression Error Plot Comparison. Next, unconstrained OGIP
tion may be more difficult to identify. Fig. 8 shows a CARET regres- regression analyses were made with the Havlena and Odeh and Teh-
sion error plot for an undersaturated oil reservoir illustrating this rani methods. Fig. 9 shows the regression error plot obtained with
point. Each rD curve minimum represents a different potential OOIP
the Havlena and Odeh solution technique. A comparison of Figs. 6
solution and aquifer description. The OOIP solutions range from un-
and 9 shows that the regression minimums obtained with the CAR-
der 10 to over 30 MMSTB despite similar regression errors. OHIP
ET method are more pronounced and better defined than those ob-
solutions with low aquifer time constants correspond to smaller, less
tained from the Havlena and Odeh technique. Also, while the over-
permeable aquifers and vice versa. Because both smaller and larger
all best-fit (minimum error) time constant obtained with the CARET
aquifer descriptions must match the same pressure history, smaller
aquifer descriptions give larger OHIP estimates. This points out the method agrees with the time constant calculated from rock proper-
dangers of unconstrained regression analyses, because any of these ties, the same cannot be said for the Havlena and Odeh solution. The
OHIP solutions might potentially be identified as the best-fit given the Havlena and Odeh minimum error solution occurs on the infinite-
range of uncertainty associated with the reservoir pressure history. acting aquifer curve at a time constant corresponding to an aquifer
Fortunately, some of these solutions can be ruled out by inspection. permeability 1,000 times lower than the field estimate. The regres-
In both Figs. 6 and 8, the OHIP solutions corresponding to rD +25 sion error plot obtained with the Tehrani method was very similar
give an OHIP estimate less than the cumulative production to date. to Fig. 9, except the regression error in the Tehrani plot was reduced
As might be expected, the aquifer time constants for these clearly un- by a factor of two to three and the OGIP estimates were higher.
realistic OHIP solutions represent aquifer permeabilities hundreds or OGIP Comparison. Fig. 10 plots the Havlena and Odeh regression
thousands of times larger than the actual field value. These invalid error minimums from Fig. 9 vs. the corresponding best-fit OGIP val-
OHIP solutions are easily avoided by constraining the regression ue they represent. Similar results obtained with the Tehrani method
analysis with an aquifer time constant value calculated from field data are plotted in Fig. 11. These best-fit OGIP vs. minimum regression
with Eq. 22. error curve shapes are distinctly different from that obtained with the
Aquifer Flow Regime. Multiple OHIP solutions also occur be- CARET formulation shown in Fig. 7. The curve shapes differ because
cause the aquifer flow regime may be either finite or infinite acting. the CARET formulation fixes the relationship between OGIP and the
In Fig. 8, each rD curve has two a values that minimize the regres- aquifer constant, U, with Eq. 15. As a result, for all OGIP solutions,
sion error, one finite acting and the other infinite acting. This occurs the CARET formulation requires a constant U/G ratio determined by
because even small aquifers can exhibit infinite-acting behavior the specified aquifer compressibility. In the Tehrani and Havlena and
early in the fields life if the aquifer permeability is low. However, Odeh methods, the relationship between U and G is not constrained
if the reservoir has been on production for several years or the and the U/G ratio varies between solutions.
Acknowledgments
The methodology presented in this paper was developed and tested
with the support and encouragement of several people within Arco.