Вы находитесь на странице: 1из 2

Analytical Representation of the van

Everdingen-Hurst Aquifer Influence


Functions for Reservoir Simulation
J .R. Fanchi, * SPE, Keplinger and Assoc. Inc.

Abstract
Linear regression analysis has been used to develop some information about dimensionless pressure P tD and its first
simple analytical expressions for the van Everdingen- derivative P~D as functions of dimensionless time t D.
Hurst aquifer influx influence functions. Regression Usually, the relationship between t D and P tD is available
results are presented for a variety of aquifer radius/reser- in the reservoir simulator in tabular form for the infinite-
voir radius ratios. The regression equations are designed acting constant terminal rate case only. The program
for use in reservoir engineering applications, especially determines PtD and P~D for a given tD by using a
reservoir simulation. numerical interpolation scheme. An alternative approach
that requires less computer work while providing
Introduction equivalent or greater accuracy than the table look-up
A reservoir-aquifer system can be modeled by using a method is presented here.
reservoir simulator in which small gridblocks define the
reservoir and increasingly larger gridblocks define the Description of Method
aquifer. This approach has the disadvantage of increased A linear regression analysis 3 has been used to develop
computer storage and computing time requirements analytical representations of the Carter-Tracy influence
because additional gridblocks are needed to model the functions. The regression equations, the regression coef-
aquifer. A widely used and more cost-effective means of ficients, correlation range limits, and measures of the
representing an aquifer is to compute aquifer influx with linear regression validity are presented in Table 1 for a
an analytical model. Among the more popular analytical number of commonly encountered r e/r w cases. Plots of
aquifer models in use today is the Carter-Tracy these expressions are shown in Figs. la and lb. Dif-
modification 1 of the van Everdingen-Hurst 2 unsteady- ferences between the van Everdingen-Hurst tabular values
state aquifer influx calculation. and the calculated values based on linear regression
, The Carter-Tracy aquifer influx rate calculation requires analysis are insignificant within the correlation range

TABLE 1-CARTER·TRACY INFLUENCE FUNCTION REGRESSION COEFFICIENTS FOR THE CONSTANT TERMINAL RATE CASE

Regression Equation: PtO =80 +8 1 t O +8 2 In to +8 3 (ln t o)2 Average


Number
Multiple of Deviation
Correlation Objective Correlation Input Standard From
Case Regression Coefficients Range of Function Coefficient Values Error of Actual
r.lr w _8_0_ 8 1 8 2 83 to S' R2' N' Estimate' ~
1.5 0.10371 1.66657 -0.04579 -0.01023 0.06 to 0.6 1.7x10- 6 0.99999 19 3.2x 10- 4 0.06
2.0 0.30210 0.68178 -0.01599 -0.01356 0.22 to 5.0 3.9x10- 5 0.99999 23 1.4x10- 3 0.16
3.0 0.51243 0.29317 0.01534 -0.06732 0.52 to 5.0 1.4x10- 5 0.99999 19 9.1 x10- 4 0.09
4.0 0.63656 0.16101 0.15812 -0.09104 1.5 to 10.0 6.7x 10- 6 0.99999 22 5.8x10- 4 0.03
5.0 0.65106 0.10414 0.30953 -0.11258 3.0 to 15.0 4.5x10- 6 0.99999 29 4.1x10- 4 0.02
6.0 0.63367 0.06940 0.41750 -0.11137 4.0 to 30.0 1.5x10- 5 0.99999 25 8.1 x10- 4 0.04
8.0 0.40132 0,04104 0.69592 -0.14350 8.0 to 45.0 4.8x10- 6 0.99999 24 4.7x10- 4 0.02
10.0 0.14386 0.02649 0.89646 -0.15502 12,0 to 70.0 4.6x10- 6 0.99999 26 4.3x 10- 4 0.01
00 0.82092 -3.68x10- 4 0.28908 0.02882 0.01 to 1,000,0 8.2x10- 3 0.99978 30 1.71 x10- 2 1.50

-NOTE: These symbols are defined in Ref. 3 as follows. The linear regreSSion objective function S is derived by
N

s. ~ (Y,-Y,)'.
i-1

where Y, is the van Everdingen-Hurst value of Pta. 'Yi is the regression equation value of P2'D' and N is the total nu~ber of van Everdingen-Hurst pairs of (P IO and tD)' A positive value of S near
zero indicates a good fit with S - 0 being a perfect fit. The "multiple correlation coefficient" R of Ref. 3 is defined by R = sum of squares using regresston equation values divided by sum of squares
using van Everdingen-Hurst values. or

N
~ (1',_1')'
i_1
R2 - - N - - - - - ' where Y is the average van Everdingen-Hurst value=.2.. ~
(N
Yi
)

N iz1
L; (y,_y)2
;_1

The range of R 2 is between 0 and' with R 2 = 1 being a perfect fit. The standard error of estimate is defined by .JS/(N - 2). where S is the objective function.

'Now with Marathon Oil Co. Copyright 1.985 Society of Petroleum Engineers

JUNE 1985 405


0.00.1l .10 .0 .0 100.0
at • f., . . Is- 'It

I
!
g

)
A

Dimensionless Time tD

Fig. 1-Aegression equation fit of the van Everdingen-Hurst influence functions. The smooth curve
is computed from the regression equation.

limits tabulated by van Everdingen and Hurst for all cases Nomenclature
except the infinite ratio case. P tD = dimensionless pressure
The average deviation of regression values from van P~D = derivative of PtD with respect to tD
Everdingen-Hurst tabular values for the infinite ratio case
is 1.5% (Table 1). Most of this error appears at early times
re = external aquifer radius, ft [m]
(0.01 < t D < O. 1) when aquifer influx is often small
rw = external reservoir radius, ft [m]
relative to influx for the total life of the reservoir. A cor- tD = dimensionless time
relation for the infinite ratio case with an average devia-
tion of 0.2 % and a standard error of the estimate of References
1.1 x 10 - 3 has been presented by Edwardson et al. 4 1. Carter, R.D. and Tracy, G.W.: "An Improved Method for
Their additional accuracy requires approximately twice Calculating Water Influx," J. Pet. Tech. (Dec. 1960) 58-60; Trans.,
the computational labor as the correlation presented in AIME,219.
2. van Everdingen, A.F. and Hurst, W.: "The Application of the
Table 1. They do not present correlations for other r efr w Laplace Transform to Row Problems in Reservoirs," Trans., AIME
values. 186 (1949) 305-24.
It is interesting to note that the form of the regression 3. Kuester, J.L. and Mize, J.H.: Optimization Techniques With For-
equation in Table 1 is the same for each r efr w case. This tran, McGraw-Hill Book Co. Inc., New York City (1973) 205.
4. Edwardson, M.J. et al.: "Calculation of Formation Temperature
simplifies the coding needed to incorporate these expres- Disturbances Caused by Mud Circulation," J. Pet. Tech. (April
sions into a reservoir simulator such as that described in 1962) 416-26.
Ref. 5. Besides simplifying the programming effort, the 5. Fanchi, J.R., Harpole, K.J., and Bujnowski, S.W.: "BOAST: A
computer work is lessened because it is no longer Three Dimensional, Three-Phase Black Oil Applied Simulation
necessary to perform a table look-up. Furthermore, the Tool," Vols. I and II, U.S. DOE Report DOE/BC-lOO33-3,
Bartlesville Energy Technology Center, Bartlesville, OK (1982).
derivative P~D is obtained directly from differentiation of
the regression equation for P tD by t D. This avoids the SPEJ
necessity of performing a numerical differentiation and
Original manuscript (SPE 12565) received in the Society of Petroleum Engineers of·
ensures that a mathematically smooth function is always fice Sept. 15. 1983. Paper accepted for publication June 4.1984. Revised manuscript
used. received July 9. 1984.

SOCIETY OF PETROLEUM ENGINEERS JOURNAL

Вам также может понравиться