Estimating Hyperbolic Decline Curves Parameters

Sri Wahyuningsih1,4), Sutawanir Darwis1), Agus Yodi Gunawan2), and Asep Kurnia Permadi3)
1)
Statistics Research Division, Faculty of Mathematics and Natural Sciences,
Bandung Institute of Technology, Bandung
2)
Industrial and Financial Mathematics Research Division,Faculty of Mathematics and Natural Sciences,
Bandung Institute of Technology,Bandung
3)
Reservoir Engineering Research Division, Faculty of Mining and Petroleum Engineering,
Bandung Institute of Technology, Bandung
4)
Statistics Study Program, Faculty of Mathematics and Natural Sciences,Universitas Mulawarman
[email protected], {sdarwis, ayodi}@math.itb.ac.id

Diterima 20 Agustus 2008, disetujui untuk dipublikasi 3 Desember 2008

Abstract
The decline curve analysis refers to the estimation of some measure of well performance. The linearization of Arps
hyperbolic decline curves leads to the estimation of initial value of decline exponent. This work proposes a graphical
method to estimate the initial value of decline exponent and a binomial expansion to estimate the decline rate and the
initial production value. The graphical approach is used to determine the initial value of the decline exponent for
binomial expansion of hyperbolic Arps equations. Experimental study from some small data sets shows that our
proposed method is capable in reducing the number of iterations needed in binomial linearization of hyperbolic decline
curves. The main contribution of this work is a development of initial decline exponent estimation using graphical
approach where the traditional approach is based on trial and error. Bias in estimating the initial values can produce
large number of iterations in estimating the decline exponent. The proposed method is successfully reduce the number
of iterations in estimating the decline exponent. Experimentally, this study indicates the existence of lower bound of
decline exponent b* for each iteration using hyperbolic binomial expansion.
Keywords: Hyperbolic DCA, Graphical method, Binomial expansion
Abstrak
Analisis kurva penurunan mengacu pada estimasi beberapa ukuran kinerja sumur minyak. Pelinieran kurva penurunan
hiperbolik Arps berdasarkan estimasi nilai awal eksponen penurunan. Kajian ini mengusulkan metode grafik untuk
mengestimasi nilai awal eksponen penurunan dan menggunakan uraian binomial untuk mengestimasi laju penurunan
dan nilai awal produksi. Pendekatan grafik digunakan untuk menentukan nilai awal eksponen penurunan untuk uraian
binomial dari persamaan hiperbolik Arps. Hasil studi eksperimen dari beberapa data menunjukkan bahwa metode yang
diusulkan berhasil mereduksi jumlah iterasi pada pendekatan uraian binomial hiperbolik. Kontribusi utama dari
pekerjaan ini adalah pengembangan eksponen penurunan awal menggunakan metode grafik, dimana pendekatan
tradisional berdasarkan coba dan salah. Bias pada estimasi nilai awal dapat menghasilkan jumlah iterasi yang besar
dalam estimasi eksponen penurunan. Metode yang diusulkan berhasil mereduksi jumlah iterasi dalam mengestimasi
eksponen penurunan. Secara eksperimen, studi ini mengindikasikan eksistensi batas bawah eksponen penurunan b*
untuk setiap iterasi menggunakan uraian binomial hiperbolik.
Kata kunci: Penurunan hiperbolik, Metode grafik, Pendekatan binomial
1. Introduction
q0
,0 < b <1
(1.2)
qt =
1
Decline curves analysis (DCA) is the oldest
b
(1 + bDt t )
and the most commonly used tools in analysing
This
solution represents the relationship
petroleum and gas production. Most of the DCA are
between
production
and time during pseudosteady-state
based on the empirical Arps equation (Arps, 1945), and
period.
Here,
q
is
the initial production rate, that can
0
is described by the differential equation
be obtained from the production history of the well.
dqt
(1.1)
= - Dt qtb +1 ,
Arps
(1945)
suggested
a
tentative
dt
classification of decline curves, based on loss ratios.
where qt is the production rate at time t, b is the
The loss ratio is defined as the production rate per unit
of time divided by the first derivative of the rate-time
decline exponent, and Dt is the decline rate. A solution
of such differential equation is called Arps hyperbolic,
and is written as
125

126 JURNAL MATEMATIKA DAN SAINS, DESEMBER 2008, VOL. 13 NO. 4

qt
. If the loss ratio is not constant, but the
dqt / dt
first derivative of this loss ratio with respect to time t
is constant or nearly constant, then the curve can be
assumed to be of hyperbolic type. The hyperbolic
decline equation has been used for many years to
predict future well performance from production
records. Several investigators (Gentry, 1972, Spivey,
1986, Li and Horne, 2005, 2007) have reported that the
actual oil production usually follows a hyperbolic
decline. Gentry (1972) published two figures for
solving all kinds of decline curve problems. All of the
equations used to describe decline are solutions of the
differential Equation (1.1). The decline rate Dt can be

curve

found given the q0 , t, qt , and cumulative production

Qt . Spivey (1986) developed algorithm that minimizes
the sum of squares of the producing rate from the
predicted value using binomial expansion.
In this paper, we develop a combination of
graphical (Gentry, 1972), and binomial expansion
methods (Spivey, 1986) to estimate hyperbolic DCA
parameters, and also propose an algorithm based on
graphical method. The graphical method is used to
estimate the initial value of decline exponent. The
resulting initial decline exponent is used in updating
the value of decline exponent in linearizing the
hyperbolic decline curves. We shall compare the results
of the methods. The article is organized as follows.
Review of graphical method is given in Section 2.
Review of the hyperbolic binomial expansion is
described in Section 3. In Section 4 we propose an
algorithm based on graphical method. We illustrate the
application of the methods and compare the results in
Section 5. Results are discussed in Section 6. The final
section of the paper consists of some concluding
remarks.
2. Graphical Method
Cumulative production for hyperbolic decline

q0
1
q
Dt = t
.
(2.3)
bt
Using two Equations (2.2) and (2.3), the parameter b
and Dt can be estimated.

3. Hyperbolic Binomial Expansion

Let us consider the hyperbolic decline
equation. Let b0 be an initial estimate for b . This
value can be estimated using Arps or graphical
methods. Rearranging the hyperbolic equation, we have
qt b0 = q0b0 (1 + bDt t )(b0 / b ) .

We assume

(3.1)

bDt t < 1 , then we can expand the right

hand side of Equation (3.1) as a binomial series, to

obtain

D 2t 2
qt b0 = q0b0 [1 + Dt tb0 + t b0 (b0 b)] .
(3.2)
2!
It is evident that the second and higher order
terms of the binomial expansion are identically zero if
b0 = b . Spivey (1986) assumed the convergence of
Equation (3.2) is rapid enough, then he ignored the
third and higher order terms. In this paper, we also
approximate Equation (3.2) including up to the second
order, since q t has a convex function. The right hand
side of the resulting Equation (3.2) is quadratic in t .

Thus, we may apply linear regression q0 b0 vs t and

t 2 to Equation (3.2) and fit a quadratic equation to the

resulting plot :

Qt = q d

is written as

b 1

q0
1
Q t b qt
=
.
(2.2)

q0 t b 1 q b
0

1
qt
The rate-time relationship is manipulated to solve for
q
the value of Dt in terms of ratio 0 . Thus the Arps
qt
equation can be written as

1
(1 + bDt t ) b 1 . (2.1)
1

bDt (1 )
b

q0

Gentry (1972) proposed the relationship between

Q
q
ratios t and 0 given by
q0 t
qt

qt b0 = 0 + 1t + 2t 2 + t .

(3.3)

Here 0 , 1 , and 2 are coefficients of regression

(3.3).
Now, we determine parameters of the decline
equation from the coefficients of regression. An
iterative scheme is set up to a certain degree of
accuracy of parameter updating. Following Spivey
(1986), and using binomial expansion the data are

Sri Wahyuningsih et al., Estimating Hyperbolic Decline Curves Parameters 127

fitted iteratively by a linear regression. The updating

scheme for decline exponent parameter b is given by

2 (i ) (i )
b(i +1) = b(i ) 1 0 2 ,

12(i )

(3.4)

1(i )
b (i ) 0 (i )

(3.5)

4. Proposed Algorithm
The proposed algorithm is based on the
graphical method. The steps of the proposed algorithm
are:
Step 1. Estimate q0 , t , and qt .

Q
and t
.
q0t actual
=0
Step 3. Generate parameter b , 0 < b < 1 , and calculate
Q
using the formula (2.2).
the ratio t
q0t predicted
t

q ,

Step 2. Calculate Qt =

Step

4.

Calculate

the

absolute

difference

Qt
Q
t
and D t using Equation

q0t predicted q0 t actual

(2.3).
Step

5.

We apply the proposed method using two

data: the production history taken from the data of
Arbuckle Lime in Kansas reported by Arps (1945)
(Figure 1), and the data reported by Permadi (2004)
(Figure 2). Before estimating the hyperbolic DCA
parameters, we identify the data by Arps method.
Figure 3 and Figure 4 show that the loss ratio is not
constant. Figure 5 shows that the double exponential
smoothing of the first derivative of loss ratio from
Arps data is nearly constant, and Figure 6 shows that
the single exponential smoothing of the first derivative
of loss ratio from Permadis data is also nearly
constant. These lead is to a conclusion that the
production rate is of hyperbolic type. For both data,
the decline exponent of the mean of the first derivative
of loss ratio is 0.508.
An S-PLUS script was developed to
implement the algorithm of graphical method, and the
hyperbolic binomial expansion. For the first data, an
application of the graphical method, with q 0 = 28,200

qt = 1,027 bbl, t =

bbl,

(12x6)+1=73 months,

q0
= 27.5, Qt = 80,552, and
qt

Qt
= 0.219 , yields
q0 t

b = 0.34 (see Figure 7). Figure 8 confirms that the

Qt

q0t predicted

Plot

this value determine Dt .

5. Application

where i is the iteration indices, i = 0,1,....

Once b has been determined, the other parameters, q0
and
be
determined
from
D can
(i )
(i )
q0(i +1) = 0 (-1/ b ) and D(i +1) =

Q
Q
b = Arg min t
t
, and with
b
q0t predicted q0t actual

b,

vs

Qt
Q
t
vs b , and Dt vs b .

q0 t predicted q0 t actual

minimum of

Qt
Q
= 0.0003 is
t

q
t
0 predicted q0t actual

attained at b = 0.34. Figure 6 shows that the values of

D = 0.08 /month, for b = 0.34.
t

Step 6. Determine b as the solution of

2000
30000

1800
1600

25000

1400
Production

Production q(t)

20000

15000

1200
1000
800
600

10000

400
5000

200
0
Jan-43

Jul-42

Oct-42

Apr-42

Jan-42

Jul-41

Oct-41

Apr-41

Jul-40

Oct-40

Jan-41

Apr-40

Jan-40

Jul-39

Oct-39

Apr-39

Jan-39

Jul-38

Oct-38

Apr-38

Jan-38

Jul-37

Oct-37

Apr-37

Jan-37

Time

Figure 1. Plot of Production Rate vs Time of Lease

Producing from Arbuckle Lime in Kansas (Arps,
1945).

Jan
1st
year

Jan
Jul
1st 2nd
year year

Jan
Jul
2nd 3rd
year year

Jan
Jul
4th
3rd
year year

Jan
Jul
5th
4th
year year

Jan
Jul
6th
5th
year year

Jan
Jul
7th
6th
year year

Jul
7th
year

Time

Figure 2. Plot of Production Rate vs Time (Permadi,

2004).

128 JURNAL MATEMATIKA DAN SAINS, DESEMBER 2008, VOL. 13 NO. 4

0
0

-1
-2

-3

Loss Ratio/6months

Loss ratio/6months

-2

-4
-5

-4

-6

-8

-6
-10

-7

Jan-43

Jul-42

Oct-42

Apr-42

Jan-42

Jul-41

Oct-41

Apr-41

Jan-41

Jul-40

Oct-40

Apr-40

Jan-40

Jul-39

Oct-39

Apr-39

Jan-39

Jul-38

Oct-38

Apr-38

Jan-38

Jul-37

Oct-37

Apr-37

Jan-37

-8

-12
Jan
1st
year

Jan
Jul
2nd
1st
year year

Jan
Jul
2nd 3rd
year year

Jan
Jul
4th
3rd
year year

Time

Jan
Jul
5th
4th
year year

Jan
Jul
6th
5th
year year

Jan
Jul
7th
6th
year year

Jul
7th
year

Time

Figure 3. Loss Ratio for Lease Producing from

Arbuckle Lime in Kansas.

Figure 4. Plot of Loss Ratio Rate vs Time of

Production Rate.
Single Exponential Smoothing

Double Exponential Smoothing

0.4
The 1st derivative of the loss ratio

Variable
Actual
Fits

3
2
1
0
-1
-2
-3
-4

0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

6
Time

10

11

Q(t)/q(0)t

Figure 5. The First Derivative of Loss Ratio for Lease

Producing from Arbuckle Lime in Kansas.

Q
vs Decline Exponent
Figure 7. Plot of t
q0t predicted

b Using Graphical Method.

Variable
Actual
Fits

10

11

12

Time

Figure 6. The First Derivative of Loss Ratio for

Production Rate.
Difference of Q9t)/q(0)t prediction and actual

The 1st derivative of the loss ratio

Q
Q
t
vs
Figure 8. Plot of t
q
t
0 predicted q0t actual
Decline Exponent b Using Graphical Method.
Now, we consider the application of
hyperbolic binomial expansion method. We take as an
initial value data resulted from the graphical method

Sri Wahyuningsih et al., Estimating Hyperbolic Decline Curves Parameters 129

i.e.

b0 = 0.34. By use of this value, we obtain

b = 0.522 , q0 = 74, 011 , and Dt = 0.209/month. For

the second data, application of the graphical method,
q0 = 1,897 bbl,
qt = 406 bbl,
with
t = 14
q0
Qt =
12,567 bbl, and
months,
= 4.672,
qt
Qt
= 0.473 , yields b = 0.38 (see Figure 10). Figure 11
q0 t

confirms that the minimum of

method to the second data. We take as an initial value

data resulted from the graphical method i.e. b0 = 0.38.
By the use of this value, we obtain b = 0.517 ,
q0 = 2, 275 , and Dt = 1.94 /month.
Figure 9 shows that the first derivative of the
loss ratio is nearly constant for t > 5 . Exploring only
the part of the second data, we obtain the decline
exponent by Arps method to be 0.508. For this data,
application of the graphical method, with q0 = 903 bbl,

qt = 406 bbl, t = 9 months,

Qt
Q
t
= 0.0004

q0t predicted q0t actual

is

attained

bbl, and

Dt =

Qt
= 0.677 , yields b = 0.32 (see Figure 13).
q0t

Q(t)/q(0)t
P di i

b = 0.38. Figure 12 shows the value of

0.161/month, for b = 0.38.

at

q0
= 2.224, Qt = 5,502
qt

Figure 10. Plot of

Difference of Q9t)/q(0)t prediction

Figure 9. Plot of D t vs Decline Exponent b Using

Graphical Method.

Q
Q
t
Figure 11. t
vs Decline
q
t
0 predicted q0t actual
Exponent b Using Graphical Method.
As applied in the first data, we also consider
the application of hyperbolic binomial expansion

Qt

q 0 t predicted

vs Decline

Exponent b Using Graphical Method.

Figure 12. Plot of Dt vs Decline Exponent b Using

Graphical Method.
Figure

14

confirms

that

the

minimum

of

Qt
Q
t
= 0.00003 is attained at

q0t predicted q0t actual

b = 0.32. Figure 15 shows for b = 0.32 the value of

Dt = 0.101/month. By use of this decline exponent
from the graphical method, we obtain b = 0.517 ,

130 JURNAL MATEMATIKA DAN SAINS, DESEMBER 2008, VOL. 13 NO. 4

q0 = 1, 014 , and Dt = 1.33 /month from the hyperbolic

binomial expansion method (see Table 3).
Application of the hyperbolic binomial
expansion method to the both data shows that the
minimum iteration occurs at b 0 = 0.5 (see Table 1,
Table 2, Table 3). The last column of Table 1, Table 2,
and Table 3 show that the terms

2 0(i ) 2(i )

12(i )

of Equation

(3.4) are negative for 0.1 b(0) 0.5 . However, for

0.5 < b (0) 1 these terms are positive. Therefore,
b (i +1) > b(i ) for 0.1 b(0) 0.5 , and b (i +1) < b (i ) for

Figure 14. Plot of

Qt
Q
t

q0t predicted q0 t actual

vs

Decline Exponent b Using Graphical Method.

Difference of Q9t)/q(0)t prediction

Q(t)/q(0)t

0.5 < b (0) 1 . These results indicate that the decline

exponent using graphical method is a lower bound for
the decline exponent using hyperbolic binomial
expansion approach.
The results of fitting the production data
using Arps, graphical and binomial expansion methods
are shown in Table 4, Table 5, and Table 6. The
production prediction using combination of graphical
and hyperbolic binomial expansion methods is more
accurate compared to the traditional methods.

Figure 15. Plot of Dt vs Decline Exponent b Using

Graphical Method.
Q
Figure 13. Plot of t
vs Decline
q0 t predicted

Exponent b Using Graphical Method (9 data).

Sri Wahyuningsih et al., Estimating Hyperbolic Decline Curves Parameters 131

Table 1. Estimating Hyperbolic Decline Curve Parameter Using Binomial Expansion

b0

Number of
Iteration until
convergence

q0

Dt
(per 6 months)

0.1

40

0.522

74,011

1.256

0.2

38

0.522

73,971

1.256

0.3

36

0.522

74,128

1.256

0.34

34

0.522

74,011

1.256

0.4

32

0.522

74,259

1.256

0.5

18

0.522

74,203

1.256

0.6

28

0.529

78,391

1.263

0.7

38

0.529

78,137

1.263

0.8

42

0.529

78,388

1.263

0.9

46

0.529

78,157

1.263

44

0.529

78,134

1.263

2 0(i ) 2(i )

12(i )

(i
-1.076
(1)
-0.393
(1)
-0.172
(1)
-0.122
(1)
-0.068
(1)
-0.010
(1)
0.023
(1)
0.042
(1)
0.055
(1)
0.083
(1)
0.226
(1)

-0.082
(10)
-0.060
(10)
-0.039
(10)
-0.032
(10)
-0.021
(10)
-0.006
(5)
0.011
(10)
0.024
(10)
0.035
(10)
0.042
(10)
0.037
(10)

= iteration)
-0.019
(20)
-0.015
(20)
-0.011
(20)
-0.009
(20)
-0.006
(20)
-0.004
(10)
0.004
(20)
0.010
(20)
0.016
(20)
0.021
(20)
0.017
(20)

-0.006
(30)
-0.005
(30)
-0.003
(30)
-0.003
(30)
-0.002
(30)
-0.003
(15)
0.003
(24)
0.004
(30)
0.006
(30)
0.008
(30)
0.007
(30)

Table 2. Estimating Hyperbolic Decline Curve Parameter Using Binomial Expansion

q0

Dt

0.1

Number of
Iteration until
Convergence
16

0.517

2,269

0.202

0.2

16

0.517

2,279

0.201

0.3

14

0.517

2,269

0.202

0.38

14

0.517

2,282

0.201

0.4

14

0.517

2,285

0.201

0.5

0.517

2,275

0.201

0.6

12

0.518

2,321

0.200

0.7

14

0.517

2,317

0.200

0.8

14

0.518

2,320

0.200

0.9

14

0.517

2,319

0.200

12

0.517

2,319

0.200

b0

2 0(i ) 2(i )

12(i )
-2.028
(1)
-0.794
(1)
-0.376
(1)
-0.194
(1)
-0.157
(1)
-0.020
(1)
0.084
(1)
0.171
(1)
0.255
(1)
0.345
(1)
0.455
(1)

-0.136
(5)
-0.093
(5)
-0.057
(5)
-0.033
(5)
-0.027
(5)
-0.008
(3)
0.015
(5)
0.027
(5)
0.033
(5)
0.031
(5)
0.013
(5)

( i = iteration)

-0.024
(9)
-0.016
(9)
-0.024
(8)
-0.014
(8)
-0.012
(8)
-0.004
(5)
0.006
(8)
0.012
(8)
0.014
(8)
0.013
(8)
0.005
(8)

-0.010
(12)
-0.007
(12)
-0.004
(11)
-0.002
(11)
-0.002
(10)
-0.001
(7)
0.003
(10)
0.002
(11)
0.003
(11)
0.002
(11)
0.002
(10)

-0.002
(16)
-0.001
(16)
-0.002
(14)
-0.001
(14)
-0.001
(14)
-0.001
(8)
0.001
(12)
0.001
(14)
0.001
(14)
0.001
(14)
0.001
(12)

-0.002
(42)
-0.002
(38)
-0.002
(36)
-0.002
(34)
-0.002
(32)
-0.002
(18)
0.002
(28)
0.002
(38)
0.002
(38)
0.002
(46)
0.002
(44)

132 JURNAL MATEMATIKA DAN SAINS, DESEMBER 2008, VOL. 13 NO. 4

Table 3. Estimating Hyperbolic Decline Curve Parameter Using Binomial Expansion

b0

Number of
Iteration until
Convergence

q0

Dt

0.1

12

0.518

1,019

0.133

0.2

12

0.518

1,021

0.132

0.3

10

0.517

1,014

0.133

0.32

10

0.517

1,016

0.133

0.4

10

0.518

1,020

0.132

0.5

0.517

1,014

0.133

0.6

0.518

1,037

0.132

0.7

10

0.518

1,031

0.132

0.8

10

0.518

1,032

0.132

0.9

10

0.518

1,031

0.132

0.520

1,033

0.132

2 0(i ) 2(i )

12(i )
-2.716
(1)
-1.064
(1)
-0.502
(1)
-0.430
(1)
-0.211
(1)
-0.026
(1)
0.107
(1)
0.213
(1)
0.304
(1)
0.388
(1)
0.471
(1)

-0.279
(4)
-0.183
(4)
-0.109
(4)
-0.096
(4)
-0.051
(4)
-0.007
(3)
0.026
(3)
0.046
(4)
0.054
(4)
0.054
(4)
0.016
(3)

( i = iteration)
-0.016
(7)
-0.011
(7)
-0.007
(7)
-0.006
(7)
-0.003
(7)
-0.007
(4)
0.006
(5)
0.003
(7)
0.003
(7)
0.003
(7)
0.004
(5)

-0.004
(10)
-0.003
(10)
-0.002
(9)
-0.001
(9)
-0.001
(9)
-0.002
(5)
0.002
(7)
0.001
(9)
0.001
(9)
0.001
(9)
0.001
(7)

-0.001
(12)
-0.001
(12)
-0.002
(10)
-0.001
(10)
-0.001
(10)
-0.002
(6)
0.002
(8)
0.001
(10)
0.001
(10)
0.001
(10)
0.001
(8)

Table 4. Lease Producing from Arbuckle Lime in Kansas

Obser
vation

Month

Year

Real data qt (bbl)

Arps
Method

1
2
3
4
5
6
7
8
9
10
11
12
13

January
July
January
July
January
July
January
July
January
July
January
July
January

1937
1937
1938
1938
1939
1939
1940
1940
1941
1941
1942
1942
1943

28,200
15,680
9,700
6,635
4,475
3,628
2,850
2,300
1,905
1,610
1,365
1,177
1,027

15,920
10,228
7,128
5,254
4,035
3,197
2,596
2,151
1,811
1,546
1,336
1,166
1,027
21,054,950

MSE

Fitting Production qt (bbl)

Hyperbolic
Graphical
binomial
method
expansion
b0 =0.34
18,714
28,244
13,059
14,906
9,478
9,238
7,099
6,302
5,456
4,582
4,285
3,486
3,428
2,746
2,786
2,220
2,295
1,834
1,914
1,541
1,613
1,314
1,372
1,133
1,177
989
7,649,797
76,164

Hyperbolic
binomial
expansion
b0 =0.5
28,238
14,902
9,295
6,300
4,581
3,486
2,745
2,219
1,833
1,541
1,313
1,133
989
76,953

Sri Wahyuningsih et al., Estimating Hyperbolic Decline Curves Parameters 133

Table 5. Lease of Producing

Month

Real data qt
(bbl)

Arps method

1,897
1,603
1,365
1,175
1,025
903
801
716
643
581
528
482
442
406
MSE

1,616
1,393
1,214
1,066
945
844
757
684
621
566
518
476
439
406
19,028

1
2
3
4
5
6
7
8
9
10
11
12
13
14

Fitting production qt (bbl)

Hyperbolic
Graphical
binomial
method
expansion
b0 =0.38
1,623
1,884
1,400
1,583
1,218
1,349
1,066
1,164
940
1,015
833
893
742
792
665
707
598
636
540
575
490
522
446
476
407
436
372
401
12,650
105

Hyperbolic
binomial
expansion b0 =0.5
1,878
1,578
1,345
1,160
1,011
890
789
705
633
572
520
474
434
400
188

Table 6. Lease of Producing

Mont
h
1
2
3
4
5
6
7
8
9

Real
data qt
(bbl)
903
801
716
643
581
528
482
442
406
MSE

Fitting of Production qt (bbl)

Hyperboli Hyperbolic
Arps
Graphical c binomial
binomial
method
method
expansion
expansion
b0 =0.32
b0 =0.5
812
818
894
892
734
742
792
790
666
676
708
706
608
618
636
634
557
565
575
573
512
519
522
521
473
477
476
475
437
440
435
435
406
407
401
400
1,947
1,478
51
72

6. Discussion
The graphical method depends on the initial
production rate q0 and the last observation at time t ,

qt . Using these information, the decline exponent b

and the decline rate are obtained from equation (2.2)
and (2.3). In binomial expansion, the decline exponent
is updated based on convergence criterion. This work
proposed using graphical method to estimate the initial
value of decline exponent. The convergence of the
hyperbolic binomial expansion depends on condition
bDt t < 1 , the initial value of b0 , and Equation (3.2).
The algorithm based on binomial expansion to estimate

the hyperbolic decline curves parameters uses trial and

error approach to estimate the initial value of decline
exponent. This work proposes a graphical approach to
estimate the initial value of decline exponent.

7. Conclusions
This paper developed the method of
estimating hyperbolic decline curve parameters by
combining of graphical and binomial expansion
approach, and proposed an algorithm based on the
graphical method. The graphical method is simple and
fast to estimate the initial value of decline exponent.
The experimental results showed that the combination
of the graphical and the hyperbolic binomial expansion

134 JURNAL MATEMATIKA DAN SAINS, DESEMBER 2008, VOL. 13 NO. 4

approaches is more accurate compared to traditional

method. The proposed method successfully reduces the
number of iterations in estimating the decline exponent.
Experimentally, this study indicates the existence of
lower bound b* of decline exponent using hyperbolic
binomial expansion for each iteration.

Acknowledgement
The authors thank to BPPS DIKTI and Hibah
Bersaing 2009 for support this paper. We are grateful
to anonymous referees whose comment led to a
significant improvement of the article.

References
Arps, J. J., 1945, Analysis of Decline Curves. Trans.
AIME,.160-247.
Gentry, R. W., 1972, Decline Curve Analysis, J. Pet.
Tech. 38-41.

Li, K. and R. N. Horne, 2005, Verification of Decline

Curve Analysis Models for Production
Prediction. SPE 93878 presented at the 2005
SPE Western Regional Meeting held in in
Irvine, CA, USA. 1-8.
Li, K., and R. N. Horne, 2007, Comparison and
Verification of Production Prediction Models.
Journal
of
Petroleum
Science
and
Engineering Vol. 55 Issues 3-4. 213-220.
Permadi, A. K., 2004, Teknik Reservoir I, KK Teknik
Reservoir, Fakultas Pertambangan dan Teknik
Perminyakan ITB, Bab VIII.
Spivey, J. P., 1986, A New Algorithm for Hyperbolic
Decline Curve Fitting. SPE 15293 presented
at the Symposium on Petroleum Industry
Application of Microcomputers, Silver Creek.
89-100.