Vous êtes sur la page 1sur 8

A non-iterative method of decline curve analysis

Hamid Hosseinzade Khanamiri


Petroleum University of Technology, Khosro Jonoobi St., Sattar Khan Ave., Tehran 1453953153, Iran
a b s t r a c t a r t i c l e i n f o
Article history:
Received 3 January 2008
Accepted 10 May 2010
Keywords:
decline curve
ratetime equation
boundary-dominated ow
non-declining period
declining period
Decline curve analysis is one of the reserve estimation methods. It is widely used by companies and
engineers. The oldest and most popular method is Arp's empirical equation. There is no theoretical basis for
Arp's equations. Exponential equation almost always underestimates the ow rate and cumulative
production and on the other hand harmonic decline overestimates. Also the parameters of hyperbolic
equation cannot be calculated directly. In that case, using non-linear type curves or iteration is necessary.
In this article, rst, two semi-theoretical ratetime equations were derived of which one is for oil reservoir
and the other for gas reservoir. Flow equation in case of a boundary-dominated ow was modied to obtain
an equation that consists of ow rate, time and some constants. In the second step, various approaches of
calculating coefcients were discussed. In fact, an integrated study of the stated problem was done by
evaluating a number of physical concepts and applying dimensional analysis. In this way, two correlations
were proposed with which the coefcients were related to the inverse of time. Finally the proposed method,
exponential, harmonic and hyperbolic equations were applied to actual eld data and the results were
analyzed and compared. In case of oil reservoir, the comparison between output of the proposed method and
that of hyperbolic equation led to the same average error and in case of two gas wells, the proposed method
was as reliable as the Arp's method. In the aspect of applicability, there is a signicant distinction between
the two methods. In the case of Arp's hyperbolic equation, iteration was used to nd the best match;
however, applying the new method does not need any iteration and it is enough to do a simple calculation to
get the result.
2010 Elsevier B.V. All rights reserved.
1. Introduction
Estimating reserves and predicting production in reservoirs has
been a challenge for a long time. Many methods have been developed
in the last several decades. One frequently used technique is decline
curve analysis approach (Li and Horne, 2003).
Decline curve analysis (DCA) is a method to t observed production
rates of individual wells, group of wells, or reservoirs by a mathematical
functioninorder to predict the performance of the future production by
extrapolating the tted decline function. Arps (1945) introduced the
decline curve analysis method using mathematical equations. The
method is a mathematical equation with no physical basis other than
that the equation shows a declining trend. The function introduced by
Arp is characterized by three parameters; initial ow rate (q
i
), initial
decline rate (a
i
), and decline exponent (n). When n=0, the decline is
exponential (Eq. (1)). When n=1, the decline is harmonic (Eq. (2)).
When0bnb1, the decline is saidtobe hyperbolic (Eq. (3)). Gaskari et al.
(2006) emphasize that the Arp's methodis still being usedbecause of its
simplicity and since it is an empirical method, it does not need any
reservoir or well parameters. The mentioned ratetime equations are as
follow:
Exponential q = q
i
exp a
i
t 1
Harmonicq =
q
i
1 + a
i
t
2
Hyperbolicq =
q
i
1 + na
i
t
1=n
3
It is difcult to foresee which equation the reservoir will follow. On
the other hand, each approach has some disadvantages (Li and Horne,
2003). For example, the exponential decline curve tends to underes-
timate reserves and production rates; the harmonic decline curve has
a tendency to overpredict the reservoir performance (Agbi and Ng,
1987). In some cases, production decline data do not followany model
but cross over the entire set of curves (Camacho and Raghavan, 1989).
Fetkovich (1980) introduced decline curve analysis by type curves.
Type curve matching is essentially a graphical technique for visual
matching of production data using preplotted curves on a loglog
paper. Fetkovich used Arp's decline curves along with type curves for
Journal of Petroleum Science and Engineering 73 (2010) 5966
Corresponding author. Tel.: +98 21 44 23 2246; fax: +98 21 44 21 4222.
E-mail address: hamid.hosen@gmail.com.
0920-4105/$ see front matter 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.petrol.2010.05.007
Contents lists available at ScienceDirect
Journal of Petroleum Science and Engineering
j our nal homepage: www. el sevi er. com/ l ocat e/ pet r ol
transient radial symmetric ow of low-compressibility liquids at
constant bottom-hole pressures. He related Arp's decline parameters
to parameters like kh/B and r
e
/r
w
for production against constant
bottom-hole pressures. Fetkovich recommended harmonic decline
curves to be used for gas wells.
Carter (1985) presented gas production rate results in type curve
form for nite radial and linear ow system producing at a constant
bottom-hole pressure. He used a variable identifying the magnitude
of the pressure drawdown in gas wells. A curve with a value of 1
corresponds to n=0 in Fetkovich (1980) liquid decline curves and
represents a liquid system curve with an exponential decline. Curves
with =0.5 and 0.75 are for gas wells with an increasing magnitude
of pressure drawdown.
Fraim and Wattenbarger (1987) introduced a normalized time
function that linearizes the rate decline against normalized time for
gas reservoirs producing at constant bottom-hole pressures during
boundary-dominated ow. The calculation of the normalized times
involves an iterative process.
Palacio and Blasingame (1993) addressed the issue of variable,
non-constant bottom-hole pressures in gas wells. They introduced
new methods, which use a modied time function for analyzing the
performance of gas wells. They have also presented a new algorithm
along with the modied time function to compute gas in place, which
are capable of modeling the behavior of production data for variable
rate and/or variable pressure drop conditions. Like normalized time,
the calculation of pseudo-equivalent time is an iterative process.
Agarwal et al. (1999) introduced newtype curves, which represent
advancement over the Palacio and Blasingame (1993) type curves
because a clearer distinction can be made between transient and
boundary-dominated ow periods.
In this work, two ratetime equations are proposed, one for oil
reservoir and the other for gas reservoir. Type curve matching and
iteration is not needed in applying the equations, also the error of the
method is reasonable comparing with Arp's exponential, harmonic
and hyperbolic equations in the studied case.
2. Methodology
2.1. Gas reservoir formulization
Flow equation of a boundary-dominated gas reservoir is
q = 0:703kh
P
2
av
P
2
wf
z ln r
e
= r
w
0:75 + s
: 4
Assuming constant properties during depressurization, Eq. (4) can
be written as below:
q = c P
2
av
P
2
wf
_ _
5
In which P
av
and P
wf
are average reservoir pressure and well owing
pressure respectively, and c is a constant. Slider (1983) emphasized that
Eq. (5) should be modied to account for turbulency and the correction
can be carried out empirically by the addition of a fractional exponent to
the (P
av
2
P
wf
2
) term:
q = c P
2
av
P
2
wf
_ _
m
: 6
Eq. (6) is applied for inow performance tests with 0.5bmb1.0. In
such ow tests, P
av
is assumed to be constant. It is not a reasonable
assumption for long time, especially in decline curve analysis (DCA). It
means that m would not be between 0.5 and 1.0, if Eq. (6) used for
DCA.
Slider (1983) showed that the well pressure (P
wf
) changes linearly
vs. time in a radial reservoir under pseudosteady state owregime. He
also stated it can be shown that the average reservoir pressure alters
linearly vs. time, too. Flow regime in a boundary-dominated reservoir
is pseudosteady state; so that we can use Eqs. (7) and (8) for P
av
and
P
wf
.
P
av
= + t 7
P
wf
= +

t 8
t is time and , , and are constants.
Note: t =0 is commencement of stabilized ow in the reservoir.
Substituting Eqs. (7) and (8) into Eq. (6) and rearranging new equa-
tion results an equation of ow rate (q) vs. time (t) as below:
q = c + t
2

t
2
_ _
m
9 1
q = c
2

2
_ _
t
2
+ 2

t +
2

2
_ _ _ _
m
9 2
Assume q
i
as the initial owrate at t =0 (start of declining period).
Eq. (9-2) with t =0 is:
q
i
= c
2

2
_ _
m
: 10
Dividing both sides of Eq. (9-2) by Eq. (10) yields:
q = q
i

2
_ _
t
2
+ 2

2
_ _
+ 1
_ _
m
11
b
1
and b
2
are dened as:
b
1
=
2
2

2
_ _

2
12 1
b
2
=

2

2
: 12 2
Eq. (11) is simplied by Eqs. (12-1) and (12-2):
q = q
i
1 + b
1
t + b
2
t
2
_ _
m
: 13
Eq. (13) is the general form of Arp's decline function. If b
1
=na
i
,
b
2
=0 and m is converted to 1/n, Arp's hyperbolic function will be
obtained:
q =
q
i
1 + na
i
t
1=n
: 3
2.2. Oil reservoir formulization
Flow equation for a boundary-dominated ow regime in an oil
reservoir is:
q = c P
av
P
wf
_ _
14
c is a constant. This equation represents laminar ow. The owcan be
turbulent in real cases. Therefore, a correction is needed to generalize
Eq. (14) for turbulent ow. Slider (1983) added an exponent to ow
60 H.H. Khanamiri / Journal of Petroleum Science and Engineering 73 (2010) 5966
equation to account for turbulency in gas reservoirs. For simplicity, his
assumption is applied for ow equation of the oil reservoir, too.
q = c P
av
P
wf
_ _
m
15
Using Eqs. (7) and (8) as used for gas reservoir, the decline
equation is derived for oil reservoir as below:
q = q
i
1 + bt
m
16
b is a constant. The only difference between gas reservoir decline
function (Eq. (13)) and that of oil reservoir (Eq. (16)) is b
2
, which is
zero in Eq. (13).
Eq. (16) is again general form of Arp's equation. Assuming b=na
i
and changing m to 1/n, Arp's hyperbolic equation, Eq. (3), will be
obtained.
2.3. Discussion on determination of coefcients b, b
1
and b
2
There are two decline Eqs. (13) and(16) andthe mainchallenge is to
determine parameters b
1
and b
2
for gas and b for oil reservoirs decline
model. If these coefcients were known, the exponent (m) would be
found simply by a loglog plot of q vs. (1+b
1
t +b
2
t
2
) and (1+bt) for
gas and oil reservoirs, respectively.
One approach is nding the coefcients theoretically with reservoir
parameters and maybe rock and/or uid properties. Many researchers
have attemptedto apply this approachinthe case of parameters of Arp's
equation, but no clear and general solution was attained.
Another alternative approach is using type curves as used for Arp's
hyperbolic equation. Fetkovich (1980) and Slider (1983) have applied
this method for Arp's model. However, their methods are time-
consuming and erroneous because of non-linearity.
A brief explanation of this approach (introducing empirical type
curves), in the case of proposed equations, is as follows:
Eq. (13) can be converted to dimensionless form with q
D
=q/ q
i
as
dimensionless ow rate, t
D1
=b
1
t and t
D2
= t

b
2
_
as rst and second
dimensionless times:
q
D
= 1 + t
D1
+ t
2
D2
_ _
m
17
Eq. (17) has three parameters for which introducing an empirical
type curve is difcult.
Similarly, with q
D
=q/q
i
as dimensionless ow rate and t
D
=bt as
dimensionless time, Eq. (16) is written in dimensionless formas below:
q
D
= 1 + t
D

m
18
Plotting q
D
vs. (1+t
D
) of Eq. (18) in a loglog graph is a line with
slope m. In fact, a line could be plotted for every m. The resulted
empirical type curve will be linear but will be based on iteration.
Assuming a value for b, m can be calculated iteratively without using
this type curve. So it is not a useful method and it is better not to use
this type curve.
The third approach (the proposed one) for coefcient determina-
tion is to guess suitable values for coefcients using dimensional
analysis and analyzing physical concept of the rate decline. It is
reasonable to use this method in the absence of theoretical solutions
for coefcients.
In this work, it is tried to suggest formulas for b
1
, b
2
and b.
Eqs. (13) and (16) state that dimension of b
2
is 1/(time)
2
and the
dimension of b
1
and b is 1/(time). Using this characteristic could be
helpful.
At rst, it is necessary to determine factors that may affect the value
of coefcients. An overall review of a reservoir life shows that it takes
some time for a reservoir to reach a stabilized declining production. This
time interval is different from one reservoir to other and may depend
on:
(1) Reservoir geometry e.g. drainage area and thickness.
(2) Rock properties e.g. permeability and tortuosity.
(3) Fluid properties e.g. viscosity and compressibility.
(4) Operational conditions of production e.g. perforations, chock
size, drilling new wells, development of the eld and so on.
Assume t is the time interval between start of production and
beginning of stabilized declining production. t has the effect of four
mentioned factors that affect the decline trend of production. This
trend is shown by decline Eqs. (13) and (17) which contain the
coefcients b
1
, b
2
and b. So, the effects of four mentioned parameters
(which can be represented by t) have to be introduced in the
equations by means of coefcients. Regarding the dimensions of
coefcients, Eqs. (19) and (20) are proposed for coefcients b
1
, b
2
and
b:
b
2
= b
2
1
=
1
t
2
19
b =
1
t
20
Eqs. (19) and (20) could be reliable because they simultaneously
relate production decline to reservoir geometry, rock and uid
properties and production conditions.
Another advantage of the proposed method is that, t can be
simply determined. Initially owrate (q) vs. time (t) should be plotted
from the beginning of production. This curve demonstrates two
periods:
1. Non-declining period (t).
2. Declining period.
The non-declining period involves both developing and plateau
periods. Production starts with an increasing, and in some cases
uctuated, rate (developing) and is continued by a roughly constant
rate (plateau). Then the declining period is commenced and is
continued with a fairly regular declining production. Therefore, t can
be determined qualitatively using q vs. t plot during early time of
production. Fig. 1 illustrates t and declining period of the Gloyd
Mitchel zone of the Rodessa eld, which is explained later.
After prediction of production ow rate, cumulative production
can be predicted:
N
p
=
t
t = 0
qdt 21
N
p
(G
p
) is cumulative oil (gas) production from t =0 (start of
declining period) till any time t. Integral of Eq. (21) can be solved for
ratetime equations. The answers are as below:
(1) Exponential
N
p
=
q
i
q
a
i
22
(2) Harmonic
N
p
=
q
i
a
i
ln
q
i
q
_ _
23
(3) Hyperbolic
N
p
=
q
i
1n a
i
1
q
q
i
_ _
1n
_ _
24
61 H.H. Khanamiri / Journal of Petroleum Science and Engineering 73 (2010) 5966
(4) Proposed method (oil reservoir)
N
p
=
q
i
b m + 1
q
q
i
_ _
1 + 1=m
1
_ _
25
(5) Proposed method (gas reservoir)
G
p
=
q
i
b
1
1 + m
2
1 + m
3
m=2
i +

3
p
2ib
1
t
_ _
m
1i

3
p
+ 2b
1
t
_ _
1 + b
1
t + b
2
1
t
2
_ _
m

2
F
1
1 + m; m; 2 + m;
i +

3
p
+ 2ib
1
t
2

3
p
_ _

26
Eq. (25) can be converted to Eq. (24) using the conversion factors
whichwere introducedat theendof Section2.2(b=na
i
andm=1/n).
In Eq. (26), it is assumed that b
2
=b
1
2
, as stated in Eq. (19). The
integral of cumulative gas production, Eq. (26), has been solved by
means of Mathematica software (ver. 5.1).
2
F
1
is hypergeometric
function and has a series expansion of:
2
F
2
z
1
=

k=1
a
k
b
k
c
k
z
k
= k! 27
In which (a)
k
is dened as:
a
k
=
a + k
k
28
is gamma function. In the Eq. (26), i is a complex number
i =

1
p
_ _
and if the cumulative production is accurately calculated
by Eq. (26), the answer will be a real number.
3. Case studies
Here, both the proposed method and Arp's equations were applied
to real eld data and results were analyzed.
3.1. Case 1
Craft and Hawkins (1991) say that the production of the Gloyd
Mitchel zone of the Rodessa led, Louisiana, is a good example of a
reservoir which was produced during the major portion of its life by
the dissolved gas-drive mechanism. It also produced by liquid
expansion and negligible gravitational segregation. The wells were
produced at high rates and had a rapid decline in production.
In this example, average monthly production of the whole Gloyd
Mitchel zone of the Rodessa eld was analyzed and production of
individual wells was not of interest.
Craft and Hawkins (1991) have published the production perfor-
mance data. There was a total 42 months of production data including
Fig. 1. Non-declining period (t =16 months) and declining period of the GloydMitchel zone of the Rodessa eld. The two periods are separated by vertical dashed line.
Fig. 2. Comparison between predicted ow rates by Arps' equations, the proposed method (Eq. (16)) and actual ow rate of the GloydMitchel zone of the Rodessa led.
62 H.H. Khanamiri / Journal of Petroleum Science and Engineering 73 (2010) 5966
both the non-declining and declining periods. Of this time interval,
16 months belongs to the non-declining period (t =16 months) as
illustrated in Fig. 1. Therefore b=1/(16 months).
The proposedapproachwas appliedtotheremainingdata (declining
period) and after history matching and nding exponent m, the ow
rates were predicted (Fig. 2).
The percentage of absolute value of relative error of each data
point was calculated by below equation:

= 100

actual ratepredicted rate

actual rate
: 29
The sign was ignored because there were both underprediction
and overprediction which causes sign of the error to be changed
alternatively. Average value of all errors was 5.68% (Table 1).
Similarly, Arp's exponential, harmonic and hyperbolic equations
were applied to predict the oil production rate. For that purpose,
16 months after start of production was assumed as t =0. For
simplicity hyperbolic equation was used as below:
q=q
i

n
= 1 + na
i
t: 30
Hyperbolic equation was tried with n=0.1, 0.2, 0.3 0.9. Plot of
(q/q
i
)
n
vs. t is linear. The slope is na
i
, and a
i
could be found. All n of step
0.1 starting from 0.1 was tried to nd the best match with minimum
error. The most accurate result was attained with n=0.5. Also
hyperbolic equation with all 9 ns predicted the ow rate better than
exponential and harmonic.
Comparison between Arp's different ratetime equations and the
proposedmethodwithactual owrate is showninFig. 2. Hyperbolic of
n=0.5 is almost the same as the proposed method with high accuracy.
Flow rate was underpredicted by exponential equation while over-
predicted by harmonic equation both not as accurate as hyperbolic
(n=0.5) and proposed method. Calculated errors of exponential,
harmonic, hyperbolic (n=0.5) and proposed method (Eq. (16)) are pre-
sented in Table 1.
m is 1.9234 in Fig. 2. As it was mentioned at the end of
Section 2.2 using m=1/n, we can obtain the hyperbolic equation
from the proposed one for oil reservoir. It means that the equivalent
exponent in hyperbolic equation will be 1/1.9234 which is about 0.52.
This value is very close to 0.5, exponent of the most accurate
hyperbolic equation in oil rate prediction step. Steps of 0.1 are used to
nd the best exponent of hyperbolic equation. If we had used steps of
0.01, we would get 0.52 as the exponent for the most accurate
hyperbolic equation. Therefore, in this case, the proposed equation
estimated the best exponent of hyperbolic equation without iteration.
Averages of absolute values of relative errors for Arp's exponential,
hyperbolic with n=0.5 and harmonic are 21.01%, 5.78% and 17.62%
respectively. Only hyperbolic equation with n=0.5 predicted as well
as the proposed method, as shown in Fig. 2 and Table 1. Also, second
column of Table 1 shows coefcient of determination (correlation
coefcient) which is shown by (r
2
) and is dened in Eq. (31). The
highest r
2
shows the best t, but not necessarily the minimumerror of
prediction phase.
r
2
= 1

k
j
= 1 q
real ; j
q
estimated; j
_ _
2

k
j
= 1 q
real; j
q
real;average

2
:
_ 31
After prediction of ow rates by proposed method and Arp's
equations, cumulative oil production (N
p
) was predicted. Eqs. (22)
Table 1
Relative errors and average relative errors of exponential, harmonic, hyperbolic
(n=0.5) and proposed equation (Eq. (16)) in prediction of ow rates of the Gloyd
Mitchel zone of the Rodessa led.
Equation r
2
(|| %)
Exponential 0.9905 21.01
Harmonic 0.9786 17.62
Hyperbolic (n=0.1) 0.9903 15.49
Hyperbolic (n=0.9) 0.9807 14.85
Hyperbolic (n=0.2) 0.9898 12.06
Hyperbolic (n=0.8) 0.9827 11.88
Hyperbolic (n=0.3) 0.9892 9.21
Hyperbolic (n=0.7) 0.9844 8.78
Hyperbolic (n=0.4) 0.9883 6.89
Hyperbolic (n=0.6) 0.9859 6.85
Hyperbolic (n=0.5) 0.9872 5.78
Proposed method 0.9848 5.68
Table 2
Relative errors of Arps' equations and the proposed equation in prediction of
cumulative production of the GloydMitchel zone of the Rodessa eld.
Equation (|| %) Condition
Exponential 12.86 Underprediction
Harmonic 4.00 Overprediction
Hyperbolic (n=0.5) 7.41 Underprediction
Proposed method 7.88 Underprediction
Fig. 3. Non-declining (t =7 months) and declining periods of well A.
63 H.H. Khanamiri / Journal of Petroleum Science and Engineering 73 (2010) 5966
(25) were used to predict N
p
in the last month of available data. N
p
is the cumulative production of the declining period. Relative errors of
each method are shown in Table 2.
Again error of the proposed method and hyperbolic equation with
n=0.5 are close to each other and they are smaller than that of
exponential but greater than the error of harmonic equation.
3.2. Case 2
Production rate of two wells from Oklahoma were studied. The
data is publicly available. The wells mainly produce gas with small
amount of liquid. Well A started to produce with produced GasOil
ratio (GOR) of about 9180 SCF/STB which was increased to over
90 000 SCF/STB during production. And well B had an initial produced
GOR of about 18660 SCF/STB which was increased to about
70 000 SCF/STB. GOR of both wells rose during production.
Gas ow rate of wells were matched by three Arp's equations
(Eqs. (1)(3)) andsecondequation of proposed method (Eq. (13)). t of
wells A and B were 7 and 10 months, respectively (Figs. 3 and 4).
29 months were usedto matchthe data withequations for eachwell and
production of 35 and 43 months were predicted respectively for each
well. The value of the exponent mis 0.2835 and 0.5113 respectively
for well A(Fig. 5) andwell B (Fig. 6). Predicted gas owrates by different
equations are shown in Fig. 5 for well A and in Fig. 6 for well B.
In the case of well A, harmonic equation had the smallest deviation
fromactual gas rate. Hyperbolic (n=0.9) and proposed equation were
close to actual rate, too (Fig. 5). The average relative errors of three
mentioned equations were 7.43, 9 and 9.95%, respectively. In the case
of well B, proposed equation had the minimum error (4.14%), and the
errors of harmonic and hyperbolic (n=0.9) were 6.71 and 8.51%
respectively. Absolute values of average relative errors (with descend-
ing order) and the regression parameter (r
2
) for both wells are shown
in Table 3.
Also cumulative production of wells in the declining period was
predicted using Eqs. (22)(24) and the proposed equation for gas
(Eq. (26)). Relative errors are shown in Table 4.
High amount of relative error in proposed method, especially in
the case of well A, may be due to hypergeometric function (
2
F
1
) in
Eq. (26) which is an innite series and has a chopping error in
calculations. Also accumulation of successive round-off errors in this
equation which includes large number of calculations with respect to
Arp's equations may cause the total error to increase.
Fig. 4. Non-declining (t =10 months) and declining periods of well B.
Fig. 5. Comparison between predicted ow rates by Arps' equations, the proposed method (Eq. (13)) and actual ow rate of well A.
64 H.H. Khanamiri / Journal of Petroleum Science and Engineering 73 (2010) 5966
4. Conclusion
First, a model was proposed based on owequation of a boundary-
dominated ow regime. In this regard, two ratetime equations were
introduced of which one is for oil reservoir and the other is for gas
reservoir. Second, correlations are proposed to correlate the coef-
cients of equations to the unstabilized non-declining period of
production. Finally, this model and Arp's equations were applied to
two real cases and the results are as follows:
(1) The two equations of proposed model are more applicable than
Arp's hyperbolic equation because iteration or type curve
matching is not needed.
(2) The model is closer to theory in comparison with Arp's model
because it integrates the ow equation and the denition of
pressure in boundary-dominated ow regime as well as
introducing the correlations for coefcients regarding reservoir
parameters.
(3) The coefcients of equations can be determined simply by a
plot of ow rate vs. time.
(4) In the case of studied oil reservoir, the proposed method had the
best prediction of ow rate comparing Arp's equations. Also,
cumulative production was predicted by this method as accurate
as Arp's hyperbolic equation. Harmonic equation was the most
accurate one in prediction of cumulative oil production.
(5) In the case of two gas wells, the error of proposed method in
prediction of ow rate was a little more than that of hyperbolic
equation for well A and it was the least for well B. However, the
error of the model is more than that of Arp's equations in
prediction of gas cumulative production. Perhaps, the reason is
the innite series of hypergeometric function (
2
F
1
) in Eq. (26)
which has a chopping error in numerical calculations. Such long
formula has also high round-off error comparing Arp's equations.
This problem could be solved by improving the methods of
numerical calculations in the computer-aided analysis.
Nomenclature
a Decline rate, 1/month
a
i
Initial decline rate, 1/month
b Coefcient of theproposedequationfor oil reservoir, 1/month
b
1
First coefcient of the proposed method for gas reservoir,
1/month
b
2
Second coefcient of the proposed method for gas reservoir,
(1/month)
2
c Coefcient of Darcy equation
h Formation thickness (ft)
k Absolute permeability (D)
m Exponent of proposed equations
n Exponent of the Arp's equation
P
av
Average reservoir pressure, psia
P
wf
Well owing pressure, psia
Table 3
Average relative error and correlation coefcient of the matching phase (r
2
) of
exponential, harmonic, hyperbolic and proposed method (Eq. 13) for wells A and B.
Equation r
2
(|| %)
A
Hyperbolic (n=0.1) 0.9107 21.13
Exponential 0.9068 21.09
Hyperbolic (n=0.2) 0.9143 18.34
Hyperbolic (n=0.3) 0.9147 16.93
Hyperbolic (n=0.4) 0.9202 15.20
Hyperbolic (n=0.5) 0.9227 13.68
Hyperbolic (n=0.6) 0.9247 12.39
Hyperbolic (n=0.7) 0.9264 11.11
Hyperbolic (n=0.8) 0.9277 10.01
Proposed Method 0.959 9.95
Hyperbolic (n=0.9) 0.9288 9.00
Harmonic 0.9294 7.43
B
Exponential 0.9340 31.91
Hyperbolic (n=0.1) 0.9406 29.55
Hyperbolic (n=0.2) 0.9466 25.07
Hyperbolic (n=0.3) 0.9522 22.51
Hyperbolic (n=0.4) 0.9573 19.74
Hyperbolic (n=0.5) 0.9620 17.12
Hyperbolic (n=0.6) 0.9661 14.70
Hyperbolic (n=0.7) 0.9698 12.62
Hyperbolic (n=0.8) 0.9729 10.50
Hyperbolic (n=0.9) 0.9756 8.51
Harmonic 0.9778 6.71
Proposed Method 0.9695 4.14
Table 4
Relative errors of Arps' equations and the proposed equation in prediction of
cumulative production of wells A and B.
Well A Well B
Equation (|| %) Condition (|| %) Condition
Exponential 4.40 Underprediction 2.37 Underprediction
Hyperbolic (n=0.9) 2.52 Underprediction 1.67 Underprediction
Harmonic 2.26 Underprediction 1.74 Underprediction
Proposed method 9.29 Overprediction 4.35 Overprediction
Fig. 6. Comparison between predicted ow rates by Arps' equations, the proposed method (Eq. (13)) and actual ow rate of well B.
65 H.H. Khanamiri / Journal of Petroleum Science and Engineering 73 (2010) 5966
q Production ow rate, BPD
q
i
Initial production ow rate, BPD
q
D
Dimensionless ow rate
t Time, month
t
D
Dimensionless time
t
D1
First dimensionless time
t
D2
Second dimensionless time
z Gas compressibility factor
Constant, psi
Constant, psi
Constant, psi/month
Constant, psi/month
Viscosity (cP)
G
p
Cumulative gas production, SCF
N
p
Cumulative oil production, bbl
t Non-declining period of production, month
Acknowledgements
The author would like to thank Mr. H. Rezapoor and Mr. A.
Hosseinzade for their help to solve the integral in Eq. (26) and Mr. R.
Askarinezhad for revising the manuscript.
References
Agarwal, R., et al., 1999. Analyzing well production data using combined-type-curve
and decline-curve-analysis concepts. SPEREE 478.
Agbi, B., Ng, M.C., 1987. A numerical solution to two-parameter representation of
production decline curve analysis. SPE (16505). Presented at the SPE Petroleum
Industry Applications of Microcomputers, Montgomery, Texas.
Arps, J.J., 1945. Analysis of decline curves. Trans. AIME 160, 228247.
Camacho, R., Raghavan, R., 1989. Boundary-dominated ow in solutiongas drive
reservoirs. SPE Reserv. Eng. 503512.
Carter, R.D., 1985. Type curves for nite radial and linear gas-ow systems: constant
terminal-pressure case. SPEJ 719.
Craft, B.C., Hawkins, M.F., 1991. Applied PetroleumReservoir Engineering. Prentice Hall,
New Jersey.
Fetkovich, M.J., 1980. Decline curve analysis using type curves. JPT 1065.
Fraim, M.L., Wattenbarger, R.A., 1987. Gas reservoir decline-curve analysis using type
curves with real gas pseudopressure and normalized time. SPEFE 671.
Gaskari, R., Mohaghegh, S.D., Jalali, J., 2006. An Integrated Technique for Production
Data Analysis with Application to Mature Fields. SPE (100562). Presented at the SPE
Gas Technology Symposium, Calgary, Alberta, Canada.
Li, K., Horne, R.N., 2003. A Decline Curve Analysis Model Based on Fluid Flow
Mechanisms. SPE (83470). Presented at the SPE/AAPG Western Regional Meeting,
Long Beach, California.
Palacio, J.C., Blasingame, T.A., 1993. Decline-curve analysis using type curves analysis of
gas well production data. SPE (25909). Presented at the SPE Rocky Mountain
Regional Meeting/Low Permeability Reservoirs Symposium, Denver.
Slider, H.C., 1983. Worldwide Practical Petroleum Reservoir Engineering Methods.
PennWell Publishing Co., Tulsa, Oklahoma.
66 H.H. Khanamiri / Journal of Petroleum Science and Engineering 73 (2010) 5966

Vous aimerez peut-être aussi