Decline Curve Analysis

Learning Objectives of Lecture 8:

Importance of decline curves
Decline curve models
Decline curve plots
Applications

Decline Curve Analysis

Preliminaries:

MBE analysis without further manipulations allows
estimation of only N and m for HCs.
Estimation of production rate specially as function of
time is also of great importance
Under natural depletion rate normally declines with
recovery
Majority of oil and gas reservoirs show natural
production rate decline according to standard trends
Unless natural trend is interrupted (water injection, well
shut in) the natural decline trend is expected to
continue until abandonment


Decline Curve Analysis


Natural decline trend is dictated by natural drive, rock and
fluid properties well completion, and so on. Thus, a major
advantage of this decline trend analysis is implicit inclusion
of all production and operating conditions that would
influence the performance.

The standard declines ( observed in field cases and whose
mathematical forms are derived empirically) are
Exponential decline
Harmonic decline
Hyperbolic decline
Decline Curve Analysis

When the average reservoir pressure decreases with time
due to oil and gas production, this in turn causes the well
and field production rates to decrease yielding a rate time
relation similar to that in the following figure.
Definition of normalized production rate decline, D:


0
/ /
lim
t
dq dt q t
D
q q A
A A
= =
Decline Curve Analysis
D = continuous production decline rate at time t
(1/time)
If t = years:
D
a
= annual continuous production decline rate (1/year)
If t = months:
D
m
= monthly continuous production decline rate
(1/month)

Unit of q is not important

Decline curve models

The general decline curve models is defined according
to their relation with q as follows:



where n is called as the decline exponent

The three standard decline models (usually observed in
field) are defined as follows.
n
i
i
q
D D
q
| |
=
|
\ .
Decline curve models

1. Exponential decline (n=0):


2. Harmonic decline (n=1):


3. Hyperbolic decline


where D
i
is the initial decline rate
i
i
q
D D
q
| |
=
|
\ .
tan
i
D D cons t = =
n
i
i
q
D D
q
| |
=
|
\ .
Decline curve models

Producing rate during decline period: substituting
the decline curve model equation into the normalized
production rate decline equation yields:




Separating the variables and integrating:
/
n
i
i
q dq dt
D
q q
| |
=
|
\ .
1
0
t q
i
n
t q
i
i
D dq
dt
q
q
+
=
=
} }
Decline curve models
Integrating and solving for q
yields the general
hyperbolic rate decline

substituting n=1 results in
the harmonic rate decline


Setting n=0 and integrating
and solving for q yields
exponential rate decline
( )
1/
( ) .1
1
i
n
i
q
q t Eq
nD t
=
+
( )
( ) .2
1
i
i
q
q t Eq
D t
=
+
( ) exp( ) .3
i i
q t q D t Eq =
Decline curve models
Cumulative production as a function of q for each model
are determined as:
ln( / ) .5
i
p i
i
q
G q q Eq
D
=
.4
i
p
i
q q
G Eq
D

=
1 1
1 1
.6
(1 )
n
i
p
n n
i
i
q
G Eq
D n
q q

| |
=
|
|

\ .

exponential decline


harmonic decline


hyperbolic decline
Decline curve models
Time at abandonment:
If we define the economic limit when the production rate is
q
a
then the exponential, harmonic and hyperbolic declines
would have the following abandonment times respectively:


1
ln .7
i
a
i a
q
t Eq
D q
| |
=
|
\ .
1
1 .8
i
a
i a
q
t Eq
D q
| |
=
|
\ .
1
1 .9
n
i
a
i a
q
t Eq
nD q
| |
| |
|
=
|
|
\ .
\ .
Decline curve models
Cumulative production during decline period:




Since q(dt/dq) from decline curve models is



0
( )
t q
p
q
i
dt
N q t dt q dq
dq
= =
} }
( / )
n
i
i
q
q dt dq D
q
| |
=
|
\ .
Decline curve models
Cumulative production as a function of q
becomes:



which integrates to ( for n0 or 1) to yield the
cumulative production expression for
hyperbolic
decline




1
n
q
i
p
q
i
i
q
N dq
D q
| |
=
|
\ .
}
1 1
1 1
.4
(1 )
n
i
p
n n
i
i
q
N Eq
D n
q q

| |
=
|
|

\ .
Graphical Features of Models
Cartesian plots yields


Graphical Features of Models
Seilog plots yield


Graphical Features of Models
Cartesian q vs Gp plots yield


Graphical Features of Models
Semilog q vs Gp plots yield


Decline curve models
For harmonic decline n=1, and hence, cumulative
production as a function of q:




which integrates to



1
q
i
p
q
i
i
q
N dq
D q
| |
=
|
\ .
}
ln( / ) .5
i
p i
i
q
N q q Eq
D
=
Decline curve models
For exponential decline n=0, Cumulative production as a
function of q:




which integrates to



1
q
p
q
i
i
N dq
D
=
}
.6
i
p
i
q q
N Eq
D

=
Decline curve models
Time at abandonment:
If we define the economic limit when the production rate is
q
a
then the exponential, harmonic and hyperbolic declines
would have the following abandonment times respectively:


1
ln .9
i
a
i a
q
t Eq
D q
| |
=
|
\ .
1
1 .8
i
a
i a
q
t Eq
D q
| |
=
|
\ .
1
1 .7
n
i
a
i a
q
t Eq
nD q
| |
| |
|
=
|
|
\ .
\ .
Graphical Features of Models
For exponential decline
one can write:


Therefore a plot of
ln(q) vs t gives a
straight line with
slope equal to (-D
i
)
and intercept equal
to (ln(q
i
)).
ln( ) ( )
i i
q ln q D t =
( ) exp( )
i i
q t q D t =
Example. Exponential decline
q = 6049.1e
-0.0524 t
100
1000
10000
0 10 20 30 40 50 60
time (quarter year)
R
a
t
e
,

s
t
b
/
d





.
Graphical Features of Models
For hyperbolic decline no immediate straight form
is obtained, therefore a linear plot which allows us
to determine two parameters namely D
i
and q
i

simultaneously is not available.

In summary : The production plots allows us two
determine the nature of decline and then we can
obtain the decline model parameters.

Graphical Features of Models
For harmonic decline one can
write from cum. prod. eqn:

slope=(-D
i
/q
i
)
intercept=(ln(q
i
))
ln( ) ( ) /
i i p i
q ln q D N q =
log q = log (13694) - 0.0215 N
p
1000
10000
100000
0 5 10 15 20 25 30 35
Np (MMstb)
R
a
t
e

(
s
t
b
/
d
)







.
ln( / )
i
p i
i
q
N q q
D
=
Graphical Features of Models
For harmonic decline
one can also write from
decline rate. eqn:



slope = (D
i
/q
i
)
Intercept=(1/q
i
).



1/ 1/ ( / )
i i i
q q D q t = +
1/q = 4.98x10
-8
t + 7.336x10
-5
0.00000
0.00005
0.00010
0.00015
0.00020
0.00025
0.00030
0.00035
0.00040
0 1000 2000 3000 4000 5000 6000
Time (days)
1
/
q

(
d
/
s
t
b
)



.
Graphical Features of Models
For hyperbolic decline no immediate straight form
is obtained, therefore a linear plot which allows us
to determine two parameters namely D
i
and q
i

simultaneously is not available.

In summary : The production plots allows us two
determine the nature of decline and then we can
obtain the decline model parameters.

Production Plots
1. A plot of log(q) vs t is

Linear if decline is exponential
Concave upward if decline is hyperbolic (n>0) or harmonic

2. A plot of q vs Np is

Linear if decline is exponential
Concave upward if decline is hyperbolic(n>0) or harmonic

3. A plot of log(q) vs Np is

Linear if decline is harmonic
Concave downward if decline is hyperbolic (n<1) or exponential
Concave upward if decline is hyperbolic with n>1.

4. A plot of 1/q vs t is

Linear if decline is harmonic
Concave downward if decline is hyperbolic (n<1) or exponential
Concave upward if decline is hyperbolic with n>1.
Hyperbolic decline analysis
1. Since no wells have declines where n=0 or 1 exactly it is more
appropriate to use a regression technique to determine all
three parameters namley D
i
, q
i
and n simultaneously. Two
approaches are suggested by Towler:

An iterative linear regression
Nonlinear regression

Towler also pointed out that linear regression impose more weight
on smaller values of production rates as it involves logs of
variables. Furthermore, the two suggested procedures on linear
regression do not produce equivalent results.

Therefore, he suggests nonlinear regression as a method which
produces repeatable results, and weights the production rates
equally.
Hyperbolic decline analysis:
nonlinear regression steps in excel
1. From production data generate a spreadsheet
with a column of oil production q vs
production time t
2. Identify the part where finite acting period
starts this is the data to be curve fitted by
nonlinear regression
3. Rewrite the time and q columns for the
curvefit data only.
4. Set up cells for n, Di, and qi.
Hyperbolic decline analysis:
nonlinear regression steps in excel
5. Set up a cell to calculate average oil production
rate from q
6. Set up a column containing the hyperbolic decline
curve equation to fit the production q*
7. Set up a column to calculate errors squared from
(q-q*)
2
.
8. Set up a cell that calculates sum of errors
squared, S
SE
from step 7
9. Set up a column to calculate the total errors
squared from (q-q
ave
)
2
.

Hyperbolic decline analysis:
nonlinear regression steps in excel
10. Set up a cell that calculates the sum of the total
errors squared, S
ST
, from step 9
11. Set up a cell to calculate the square of the
regression coefficient, R
2
, from R
2
=1-S
SE
/S
ST

12. Initialize the solver to optimize the contents of the
cells that contain n, D
i
and q
i
by maximizing the
cell that contains R
2

Note alternative to maximizing R
2
is to minimize the
sum the squares of the residuals S
SE
which gives
the same result.
Example. Exponential decline
Example. Well production history
0
1000
2000
3000
4000
5000
6000
7000
0 2 4 6 8 10 12 14
time (quarter year)
R
a
t
e
,
S
T
B
/
D
0
1000
2000
3000
4000
5000
6000
7000
C
u
m
.

P
r
o
d
u
c
t
i
o
n

(
M
S
T
B
)
Caution for applicability
The emprical decline curve equations assume that the
well/field analyzed is produced at constant BHP. If the
BHP changes, the character of the well's decline
changes.
They also assume that the well analyzed is producing
from an unchanging drainage area (i.e., fixed size) with
no-flow boundaries, If the size of the drainage area
changes (e.g., from relative changes in reservoir rates),
the character of the well's decline changes. If, for
example, water is entering the well's drainage area, the
character of the well's decline may change suddenly,
abruptly, and negatively.
Caution for applicability
The equation assumes that the well analyzed
has constant permeability and skin factor. If
permeability decreases as pore pressure
decreases, or if skin factor changes because of
changing damage or deliberate stimulation, the
character of the well's decline changes.
It must be applied only to boundary-dominated
(stabilized) flow data if we want to predict future
performance of even limited duration.
Example. Exponential decline
Example. Exponential decline
q = 6049.1e
-0.0524 t
100
1000
10000
0 10 20 30 40 50 60
time (quarter year)
R
a
t
e
,

s
t
b
/
d





.
Slope=-D 1/quarter year
Example. Exponential decline
Example. Rate decline with production
q = -0.4301Np + 5768.7
0
1000
2000
3000
4000
5000
6000
7000
0 2000 4000 6000 8000 10000 12000 14000
Cum. prod, MSTB
q

s
t
b
/
d
q abondonment
Reserves
Example. Harmonic decline
0
2000
4000
6000
8000
10000
12000
0 2 4 6 8 10 12 14 16
Time (years)
R
a
t
e

(
s
t
b
/
d
)
0
5
10
15
20
25
30
35
40
C
u
m
.

P
r
o
d
u
c
t
i
o
n

(
M
M
s
t
b
)
Example. Harmonic decline
1/q = 4.98x10
-8
t + 7.336x10
-5
0.00000
0.00005
0.00010
0.00015
0.00020
0.00025
0.00030
0.00035
0.00040
0 1000 2000 3000 4000 5000 6000
Time (days)
1
/
q

(
d
/
s
t
b
)



.
Example. Harmonic decline
01/7.336e-5 = 13631
log q = log (13694) - 0.0215 N
p
1000
10000
100000
0 5 10 15 20 25 30 35
Np (MMstb)
R
a
t
e

(
s
t
b
/
d
)







.
Example. Hyperbolic decline
n Di qi R2 qave SSE SST
0.62194 0.04522 23450.00 0.986350 2498.00 814938 59701960
Month q STB/month t q q*
(q-q*)2 (q-qave)2
q
-0.622
Dec-90 213
Jan-91 1033 time rate
Feb-91 1520 30 8840 8768 5126 40220964 0.00351
Mar-91 1453 60 4320 4784 215335 3319684 0.00548
Apr-91 1334 90 3800 3084 512477 1695204 0.00594
May-91 1284 120 2000 2185 34351 248004 0.00885
Jun-91 1077 150 1550 1646 9175 898704 0.01037
Jul-91 1081 180 1430 1293 18666 1140624 0.0109
Aug-91 1057 210 1000 1049 2391 2244004 0.01362
Sep-91 973 240 760 871 12422 3020644 0.01616
Oct-91 962 270 710 738 786 3196944 0.01685
Nov-91 924 300 570 635 4208 3717184 0.01932
Decline curve analysis with nonlinear regression
Example. Hyperbolic decline
Hyperbolic Decline curve
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
0 50 100 150 200 250 300 350
days
q

S
T
B
/
D
Example. Hyperbolic decline
n=0.622
q
i
-n
=1.51e-3

nD
i
q
i
-n
=5.81e-5
q
-0.622
= 5.81E-05 t + 1.51E-03
R
2
= 9.88E-01
0
0.005
0.01
0.015
0.02
0.025
0 50 100 150 200 250 300 350
time, days
q
^
(
-
0
.
6
2
2
)






.
n n n
i i i
q q q nD t

= +
Compare with the previously obtained values from nonlinear
regression