Beruflich Dokumente
Kultur Dokumente
variations in petrophysical properties are exhibited, parametric simple shapes of the individual transforms and can be well
regression often fails or leads to unstable or erroneous results, approximated by low order polynomials.
especially for multivariate cases. The statistical measures of the correlations used in this
Thus a non-parametric approach for estimating optimal work are the average relative error, ARE, and average absolute
transformations of petrophysical data was used to obtain the relative error, AARE, which are defined as follows.
maximum correlation between observed variables1. An
iterative procedure involving Alternating Conditional 100 N y measured − y calculated
Expectations, ACE, is an approach that does not require an a ARE = ∑
N i =1 y measured
................................. (4)
priori assumption of a functional form; the optimal
transformations are derived solely based on the set of data.
100 N y measured − y calculated
The ACE algorithm provides a method for estimating AARE = ∑
N i =1 y measured
.............................. (5)
optimal transformations for multiple regression that result in a
maximum correlation between a dependent (response) random
variable, y, and multiple independent (predictor) random ARE characterizes the accuracy (bias) and AARE
variables, x1, x2, …, xm. describes the precision (scatter) of predicted values obtained
A model predicting the value of y from the values x1, x2, with a particular correlation compared with experimental
… xm is written in the generic form values.
y= f −1
(z ) ................................................................(1) Surface Separation Calculations
The purpose of this work was to develop correlations which
where can be used when laboratory data are not yet available. Thus
the independent variables were limited to those data that are
m readily measured in the field. However, these data are not
z = ∑ zn ................................................................(2) measured in routine service company laboratory studies but
n =1
must be calculated using the compositions of the gas produced
at various reservoir pressures which are measured and
and
reported as a part of the reservoir fluid study. Thus surface
separation calculations were made for each composition
z n = f n (x n ) ...............................................................(3) reported in the 615 laboratory reports3.
Values of equilibrium ratios, K-factors, for the surface
The functions f1(.), f2(.), …, fm(.) are called variable separation calculations were taken from the GPSA
transformations yielding the transformed independent Engineering Data Book4. These equilibrium ratios provide
variables z1, z2, …, zm. The function f(.) is the transformation very accurate separator calculations, especially at the
for the dependent variable. In fact, the main interest is the moderate pressures and low temperatures normally
inverse: f-1(.), yielding the dependent variable y from the encountered in surface separation.
transformed independent variable z. This study was based on a three-stage surface separation
Given N observation points, it is desirable to find the best with conditions of 1050 psig and 90 oF, 65 psig and oF, and 0
transformation functions f1(.), f2(.), …, fm(.) and f-1(.), but first psig and 70 oF. The entire study was repeated with three-stage
not as algebraic expressions, rather as relationships defined separator conditions of 500 psig and 90 oF, 45 psig and 80 oF,
point-wise. The method of ACE constructs and modifies the and 0 psig and 70 oF with the resulting coefficients for every
individual transformations to achieve maximum correlation in correlation equation within two percent of those from the
the transformed space. original set of separator conditions. Thus the proposed
The resulting individual transformations are given in the correlations should be adequate for any reasonable set of
form of a point-by-point table. Thus in any subsequent separator conditions.
application (graphical or algebraic) interpolation is needed to
obtain the transformed variables. And interpolation is needed Prediction of Dewpoint Pressure
to apply the inverse transformation to predict y. Obviously, The dewpoint pressure correlation equation is based on field
the smoother the transformation the more justified and properties measured before the reservoir pressure decreases
straightforward is the interpolation, therefore some kind of below the dewpoint pressure of the reservoir fluid. These
restriction on smoothness is built into the ACE algorithm. properties are the producing gas-oil ratio from the first
That is, based on the concept of conditional expectation, the separator, RSP1d, scf/STB, the stock-tank condensate gravity,
correlation in transformed space is maximized by iteratively o
APId, and the reservoir gas specific gravity, γgRd. The
adjusting the individual transformations subject to a subscripts ‘d’ indicate that these properties are determined
smoothness condition. when the reservoir pressure is still above the dewpoint
The particular realization of the algorithm, GRACE, used pressure of the reservoir fluid and, thus, should be constant.
in this work consists of two parts2. The first provides the Producing gas-oil ratio from the first separator was
transformations in the form of tables and the second part selected rather than total producing gas-oil ratio as the first
allows the user to construct the final algebraic approximations independent variable because the producing gas rates from the
using curve fitting in a commercial spreadsheet program. second separator and the stock-tank are seldom measured. In
Fortunately, many physically sound problems have rather application of this correlation the reservoir gas specific gravity
IPTC 10320 3
must be determined with a recombination calculation using up in the stock tank, seldom builds high enough saturation in
RSP1d, oAPId, and the specific gravity of the first separator gas, the reservoir to have any effective permeability. So it stays
γgSP1d. This calculation procedure is very accurate and should behind in the reservoir and the resulting decrease in surface
not be a problem in application of the dewpoint pressure yield is often as much as seventy-five percent of the initial
equation5. surface yield.
The correlation equations for dewpoint pressure are The surface yield is a function of initial stock-tank oil
gravity, APId, and original reservoir gas specific gravity, γgRd,
ln p d = 8.48 + 0.32239 z + 0.00477 z 2 .......................(6) both determined while reservoir pressure remains above the
dewpoint pressure of the reservoir fluid. A third independent
where variable is reservoir temperature, TR. All three of these
variables are constants for a particular gas condensate
3 reservoir. Yield is calculated at a particular reservoir pressure,
z = ∑ zn ................................................................(7) p, where p is a selected reservoir pressure less than dewpoint
n −1
pressure. The original reservoir gas specific gravity must be
and
determined with a recombination calculation as described for
the dewpoint pressure correlation above.
z n = C 0 n + C1n VAR n + C 2 n VAR n ...........................(8)
z
The correlation equations for surface yield are
N VARn C0n C1n C2n ln YSP1 = 3.684 + 0.61967 z + 0.015359 z 2 ................ (9)
1 ln RSP1d 9.8895 -0.87528 -0.01691
2 API d 11.7 -0.29709 0.00151 where
3 γg Rd 3.5202 -2.9145 -0.81744
4
required, values of two-phase gas deviation factor, z2PH, condensate ratio, Rd. In this case R is the total surface gas;
maybe obtained with a correlation equation which requires again the subscript d indicates a property measured while the
reservoir gas specific gravity as input11,7. reservoir pressure is above the dewpoint pressure of the
The specific gravity of the gas in a gas condensate reservoir gas. Fig. 4 shows that the underlying relationship
reservoir decreases by as much as fifty percent as condensate between composition of heptanes plus in percent and initial
is released from the gas into the reservoir as reservoir pressure producing gas-condensate ratio, Rd is a power function.
declines. This decrease will have a significant effect on
reservoir gas properties discussed above. ⎛ Rd ⎞
−0.86391
The downhle volumes, represented by the subscript d, are These correlations have application in the analysis of
measured at the pressure and temperature of the production production logging as well as in reservoir engineering.
logging tool. The formation volume factors and solution gas-
oil ratios in the equations are estimated at the same Nomenclature
temperature and pressure. ARE = Average Relative Error, Eq. 4
This comparison is much harder to perform in gas- AARE = Average Absolute Relative Error.
condensate wells due to the retrograde phase behavior. Any Eq. 5
such conversion process should not involve the use of APId = gravity of stock-tank liquid, oAPI,
equations of state which require hard to find input parameters determined when p≥pd
and ideally should be driven by the availability of readily Bg = gas formation volume factor, bbl/scf
found parameters such as those that may be obtained from the Bo = oil formation volume factor,
gas sales report. bbl/STB
The process to perform a downhole to surface rate Bw = water formatioin volume factor,
comparison using the new correlations is as follows: bbl/STB
a. Obtain the (1) initial producing gas-condensate ratio, C0n, C1n, C2n, C3n, C4n = coefficients in the transform
(2) specific gravity of the gas from the primary equations for variable zn
separator (initial), and (3) the gravity of the stock- C7+ = composition of heptanes plus, mole
tank liquid (initial). These data can be from either %
initial production tests or early production (sales fn(xn) = variable transform yielding the
data). transformed independent variable zn
b. If the data from (1) above are from production (sales) f-1(z) = inverse transform function of the
records also obtain the operating pressure and dependent variable z
operating temperature of the primary separator. p = reservoir pressure, psia
c. Use Gold, McCain, Jennings equations5 to calculate pd = dewpoint pressure, psia
the specific gravity of the reservoir gas at reservoir pr = reservoir pressure divided by
pressures at and above the dewpoint. dewpoint pressure of initial reservoir
d. Estimate the dewpoint pressure of the reservoir gas gas, Eq. 13
using Eq. 6, Eq. 7 and Eq. 8. Qos = surface oil volume, STB/day
e. If the downhole pressure measured with production Qws = surface water volume, STB/day
logging toolstring is above the estimated dewpoint Qgs = surface gas volume, scf/day
pressure from the step d above, the surface gas- Qod = downhole oil volume, bbl//day
condensate ratio is estimated to be the same as the Qwd = downhole water volume, bbl/day
initial producing gas-condensate ratio obtained in Qgd = downhole gas volume, bbl/day
step a. Rs = solution gas-oil ratio, scf/STB
f. If the downhole pressure measured with the PL tool Rsw = solution gas-water ratio, scf/STB
is below the estimated dewpoint pressure from Rd = producing gas-condensate ratio, total
calculations in step d (the usual situation), the surface surface gas, scf/STB, at p≥pd
yield YSP1, is estimated using Eq. 9, Eq. 10 and Eq. RSP1d = producing gas-condensate ratio,
11. scf/STB, of first separator at p≥pd
TR = reservoir temperature, oF
Therefore the newly developed correlations can be used for VARn = independent variable n
production log interpretation in gas-condensate wells. It is xn = independent variable n
important that the PL toolstring has the necessary sensors to y = generic dependent variable, Eq. 1
measure the downhole gas and liquid flowrates with sufficient ymeasured = measured value of dependent
accuracy, such as described by Rounce, et al.12. variable (used to create correlation)
ycalculated = calculated value of dependent
Conclusions variable (calculated with correlation
The three correlations presented above, dewpoint pressure, equation)
surface yield, and reservoir gas specific gravity, can be used YSP1 = surface yield, STB/MMscf,
with confidence. All three have negligible bias, i.e., ARE calculated with gas rate from first
equal 0.0, and have scatter, i.e., random error, about the same separator
as experimentally measured values of the same properties. z = the sum of transforms of the
The forth correlation, heptanes plus composition, was independent variables
presented primarily to illustrate that the previous correlations zn = transformed independent variable n
can be used with confidence for various reasonable separator γgR = reservoir gas specific gravity at p<pd
situations. However, if needed, this correlation can be used γgRd = reservoir gas specific gravity at p≥pd
with confidence due to the large data set used in its γgRr = specific gravity of reservoir gas at
preparation and the conformation by the Eastern Venezuela p<pd divided by specific gravity of
correlation of Marruffo, et al.6. reservoir gas at p≥pd, Eq. 12
6 IPTC 10320
Subscripts
d = at the dewpoint
g = gas
n = identifies an independent variable
r = “reduced”, i.e., variable divided by
value of variable at dewpoint
R = reservoir
SP1 = first stage separator
2PH = two-phase in reservoir
References
1. Breiman, L. and Friedman, J. H.: “Estimating Optimal
Transformations for Multiple Regression and Correlation,”
Journal of American Statistical Association (September 1985)
580-619.
2. Xue, G., Datta-Gupta, A., Valko, P. P. and Blasingame, T. A.:
“Optimal Transformations for Multiple Regression: Application
to Permeability Estimation from Well Logs,” SPE Formation
Evaluation (June 1997) 85-93.
3. McCain, W. D., Jr.: The Properties of Petroleum Fluids, 2nd Ed.,
PennWell Books, Tulsa (1990) 374-385.
4. Engineering Data Book, Vol. II, Gas Processors Suppliers
Association, Tulsa (1987) 25-1 – 25-112.
5. Gold, D. K., McCain, W. D., Jr. and Jennings, J. W.: "An
Improved Method of the Determination of the Reservoir-Gas
Specific Gravity for Retrograde Gases," Journal of Petroleum
Technology (July 1989) 41, 747-752; Trans., AIME, 287.
6. Marruffo, I, Maita, J., Him, J. and Rojas, G.: “Statistical
Forecast Models to Determine Retrograde Dew Pressure and C7+
Percentage of Gas Condensates on Basis of Production Test
Data of Eastern Venezuelan Reservoirs,” paper SPE 69393
presented at the SPE Latin American and Caribbean Petroleum
Engineering Conference, Buenos Aires, 25-28 March 2001.
7. Piper, L. D., McCain, W. D., Jr., and Corredor, J. H.:
"Compressibility Factors for Naturally Occurring Petroleum
Gases," Gas Reservoir Engineering, Reprint Series, SPE,
Richardson, TX (1999) 52, 23-33.
8. Hall, K. R. and Yarborough, L.: “A new equation of state for Z-
factor calculations,” Oil and Gas Journal (June 18, 1973) 82-92.
9. Standing, M. B. and Katz, D. L.: “Density of Natural Gases,”
Trans., AIME (1942) 146, 140-149.
10. Lee, A. L., Gonzalez, M. H. and Eakin, B. E.: “The Viscosity of
Natural Gases,” Journal of Petroleum Technology (Aug. 1966)
997-1000.
11. Rayes, D. G., Piper, L. D., McCain, W. D., Jr., and Poston, S.
W.: "Two-Phase Compressibility Factors for Retrograde
Gases," SPE Formation Evaluation (March 1992) 87-92; Trans.,
AIME, 293.
12. Rounce, J., Lenn, C., and Catala, G. “Pinpointing Fluid Entries
in Producing Wells” paper SPE 53249 presented at the SPE
Middle East Oil Show, Bahrain, 20–23 February 1999.
IPTC 10320 7
Table 1 – Data and calculation results from 615 reservoir gas dewpoint compositions used in
developing the dewpoint pressure correlation, Eqs. 6-8
Table 2 – Data and calculation results from 851 constant volume depletion steps used in developing
the surface yield correlation, Eqs. 9-11
Table 3 – Data and used in developing the correlation for composition of heptanes, plus, mole
percent, Eq. 17, 862 lines of data
12,000
10,000
8,000
pd calculated
6,000
4,000
2,000
0
0 2,000 4,000 6,000 8,000 10,000 12,000
pd Measured
Fig. 1 – Comparison of measured and calculated values of dewpoint pressure indicates the accuracy and precision of the new
dewpoint pressure correlation, Eqs. 6 through 8, 615 data points
300
250
200
YSP1 Calculated
150
100
50
0
0 50 100 150 200 250 300
YSP1 Measured
Fig. 2 – Comparison of measured and calculated values of surface yield indicates the accuracy and precision of the new
surface yield correlation, Eqs. 9 through 11, 851 data points
IPTC 10320 9
1.1
1.0
0.9
γgRr Calculated
0.8
0.7
0.6
0.5
0.4
0.4 0.5 0.6 0.7 0.8 0.9 1.0
γgRr Measured
Fig. 3 – Comparison of measured and calculated values of reservoir gas specific gravity indicates the accuracy and precision
of the new correlation, Eqs. 14 through 16, 851 data points
1000000
100000
Rd, scf/STB
10000
1000
0 1 10 100
Heptanes Plus in Reservoir Fluid, mole %
Fig. 4 – Data for Eq. 17 showing underlying power function, 862 data points
10 IPTC 10320
14
12
Calculated Composition C7+
10
0
0 2 4 6 8 10 12 14
Measured Composition C7+
Fig. 5 – Comparison of measured and calculated values of heptanes plus composition indicates the accuracy and precision of
the new correlation, Eq. 17, 862 data points