Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

8/30/2013

Production Forecasting for


Unconventional Resources
Lecture 4
Arps’ Decline Model

Instructional Objectives
 You will be able to
 Estimate future production using Arps’ model
 Estimate reserves using Arps’ decline model
 State theoretical basis for Arps’ model
 State assumptions inherent in Arps’ model
 State appropriate values of b for specific
situations
 Understand the risk of using b ≥ 1 in
extrapolating decline curves

1
8/30/2013

Production Decline — Cartesian Plot


25

20
Rate, MMscf/D

15

10

0
0 100 200 300 400 500 600 700 800
Time, Days

Production Decline – Semilog Plot


100

10
Rate, MMscf/D

0.1
0 100 200 300 400 500 600 700 800
Time, Days

2
8/30/2013

Some Definitions from Arps (SPE 945228)


 Decline rate, D (1/time)
𝑑𝑞
𝐷 = − 𝑑𝑡
𝑞
 Parameter b defined as change in 1/D (loss
ratio) with time
1
𝑑
𝑏= 𝐷
𝑑𝑡
 Empirical observation: for most wells Arps
analyzed, b = constant (including b = 0)

Arps’ Hyperbolic Decline Model


 Integration for constant b leads to Arps’
hyperbolic decline model
𝑞𝑖
q(t) = 1
1+𝑏𝐷𝑖 𝑡 𝑏
 Implication: Arps hyperbolic decline
equation valid only for constant b
 Hyperbolic model thoroughly validated
(decades of successful application) for
constant b, which requires BDF

3
8/30/2013

Variables, Parameters in Arps’ Decline Model


Stabilized rate Decline rate at
at time 0, time 0, 1/time
volume/time

q  qi 1  bDi t 
1
b

Time
Arps’ decline
Rate at time t,
constant,
volume/time
dimensionless

Types of Arps Decline Curves


 Three basic types (rate vs. time)
 Exponential (or constant percentage)
(b = 0)
q(t )  qi e  at

 Hyperbolic (0 <b < 1)


𝑞𝑖
q(t) = 1
1+𝑏𝐷𝑖 𝑡 𝑏

 Harmonic (b = 1)
qi
q(t ) 
1  Dit 

4
8/30/2013

Common Units for Variables in Arps’ Model


 q (rate at time t)
 STB/D, STB/mo, STB/yr
 Mscf/D, MMscf/D, ….
 Di (decline rate at time 0)
 1/D, 1/mo, 1/yr
 t (time)
 D, mo, yr

A Recollection: Fetkovich Solution for Constant


BHP Production, Boundary Dominated Flow (BDF)
 For slightly compressible liquid flow, constant ct, Bo
−𝑞𝑖 𝑡
𝑞 = 𝑞𝑖 𝑒𝑥𝑝 𝑝𝑤𝑓
1− 𝑁𝑝𝑖
𝑝𝑖
where
𝜋𝑟𝑒2 𝜑𝑐𝑡 ℎ𝑝𝑖
𝑁𝑝𝑖 = 5.615𝐵𝑜
 Note form same as Arps’ exponential decline model:
𝑞 = 𝑞𝑖 exp −𝑎𝑡

5
8/30/2013

Theoretical Basis for Arps’ Decline Equation


 Production at constant BHP
 Well or reservoir in boundary-dominated flow (BDF)
(sometimes inappropriately called “pseudosteady-
state flow”)
 No transient flow data

 Constant productivity index


 No change in damage or stimulation

 Skin factor constant

 Constant radius of drainage


 For stabilized flow (BDF) with no change in
productivity index, BHP, or radius of drainage, ‘b ’
should be constant for life of well

Exponential Decline (b = 0; Di =a = constant)


 Nominal decline rate, a (instantaneous
decline rate)
 at
q  qe i ln q1 
a 
q2 
t 2  t1
 Effective decline rate, d (decline rate over
one unit of time, usually a year)
qi  q
d  1  e a
qi a   ln 1  d 

6
8/30/2013

Exponential Decline
 Cumulative production
t t
N p or G p   q dt  qi  e at dt
o o
qi  q Watch units!
N p or G p  If q in STB/D, a in days-1
a If q in STB/mo, a in mo-1
If q in STB/yr, a in yr-1

Nominal/Effective Decline Compared


3
Diverging
at high
Similar
2 values
at low
Nominal values
Decline, a
1

0
0.0 0.2 0.4 0.6 0.8 1.0
Effective Decline, d

7
8/30/2013

Hyperbolic Decline
 When the rate/time line isn’t straight on
semilog coordinates
q  qi 1  bDi t 
1
b 0<b<1
t

 Again, N p or G p   qdt
0

qi1b  q1b 
b
q
So N p or G p  i
1  bDi

Hyperbolic Decline
 How to fit (unknowns qi, Di, b)
 Regression analysis (computer; e.g.,
Excel ‘Solver’)
 risky, not guided by engineering
principles – but very common
 Fetkovich type curves
 Fetkovich recommendations using
expected values of b (Paper SPE
28628)

8
8/30/2013

Determination of Parameters in Arps Model


Using Procedure in SPE 161092
 Fit ratio q/Gp using MS Solver and equations
below (eliminate qi ) and find best fit b, Di

 Calculate qi using best fit of Gp data and


known b, Di and equation

 Strengths of method: preserves material


balance; solves for only two unknowns (at
most) at a time

Example Fit to Determine b, Di from 15 years of


Field Data (SPE 161092)—b = 1.82, Di = 0.109 day-1

9
8/30/2013

Example Fit to Determine qi from 15 years of


Field Data (SPE 161092)—qi = 1,850 Mscf/D

Fit to Rate Data After n, Di, qi Determined

10
8/30/2013

Example Fit to Determine b, Di from 52 Mo. of


Field Data (SPE 161092)—b = 1.46, Di = 0.121 days-1

Example Fit to Determine qi from 52 Mo. of Field Data


(SPE 161092)—qi = 8,300 MMscf/D (237 Wells)

11
8/30/2013

Fit to Rate Data After n, Di, qi Determined

Harmonic Decline
 Special case of hyperbolic decline for
b = 1 (rare in practice)
qi
q
1  Dit 
qi  qi 
N p or G p  ln  
Di  q 

12
8/30/2013

Decline Curve Shapes: Cartesian Rate-Time


8000

Exponential
7000
Hyperbolic
Harmonic
6000
Rate, STB/mo

5000

4000

3000

2000

1000

0
0 100 200 300 400 500 600
Time, months

Decline Curve Shapes: Semilog Rate-Time


10000

1000
Rate, STB/mo

100

Exponential
Hyperbolic
Harmonic

10
0 100 200 300 400 500 600
Time, months

13
8/30/2013

Decline Curve Shapes: Log-Log Rate-Time


(Remember Shape for Fetkovich Type Curve)
10000

1000
Rate, STB/mo

100

Exponential
Hyperbolic
Harmonic

10
1 10 100 1000
Time, months

Decline Curve Shapes: Cartesian Cum-Time


1600
Exponential
1400 Hyperbolic
Harmonic
1200
Cum Prod, MSTB

1000

800

600

400

200

0
0 100 200 300 400 500 600
Time, months

14
8/30/2013

Decline Curve Shapes: Cartesian Rate-Cum


8000

Exponential
7000
Hyperbolic
Harmonic
6000
Rate, STB/mo

5000

4000

3000

2000

1000

0
0 200 400 600 800 1000 1200 1400 1600
Cum Prod, MSTB

Cartesian Rate-Cum Plot Especially Useful


for Analyzing Field Data
 Producing time not explicit in plot
 Shut-in periods during month have no
effect on this plot, but must be taken into
account for improved accuracy in rate-time
plot
 For exponential decline, especially
simple to determine EUR: extrapolate
linear trend directly to economic limit
rate

15
8/30/2013

Decline Curve Shapes: Semilog Rate-Cum


10000

Exponential
Hyperbolic
Harmonic

1000
Rate, STB/mo

100

10
0 200 400 600 800 1000 1200 1400 1600
Cum Prod, MSTB

Observed Values of b

30

25

20
Frequency, %
15

10

5
0
0 .2 .4 .6 .8 1.0
From Cutler, Arps Values of b

16
8/30/2013

Decline Curve Equations


b0 0  b 1 b 1
Rate
q  qi e  at
Exponential
Hyperbolic
Harmonic
Cumulative
N p or G p q q
N p or G p  i
f qi , q 
 Nominal a

a  Di
a  f DEffective
i, t a   ln 1  d 
d  f a  
d  1  ea 
Life ln qi / q 
t t
a

b Values for Specific Situations


 b indeterminate
 Constant or increasing rate
 Flow all transient (infinite acting)

17
8/30/2013

Typical Arps ‘b ‘ Factors – Oil and Gas


Drive Mechanism b

Typical volumetric gas reservoir, BDF, Constant FBHP 0.4

Typical volumetric solution-gas drive oil reservoir, BDF, Constant


0.3
FBHP
Layered Volumetric Oil or Gas Reservoir, Constant FBHP 0.6 – 0.8
Low Permeability Oil or Gas Reservoir, Transient Formation Linear
2
Flow, Constant FBHP

Can b Be Greater Than 1?


qi  qt  qi
b = 0: lim N p t   lim 
t  q t 0 Di Di

b < 1: lim N p t   lim


t 
qi b
q 0 Di 1  b 
 qi

qi1b  q1b  D 1  b 
i

qi  qi 
b = 1: lim N p t   lim ln    
t  q 0 Di  q 

b > 1: lim N p t   lim


t 
qi b
q 0 Di 1  b 

qi1b  q1b   

18
8/30/2013

Can b Be Greater Than 1?


8000
Exponential (b=0)
7000 Hyperbolic (b=0.5)
Harmonic (b=1)
Hyperbolic (b=2)
6000
Cum Prod, MSTB

5000

4000

3000

2000

1000

0
0 1000 2000 3000 4000 5000 6000
Time, months

Can b Be Greater Than 1?


 During certain (transient or unstabilized)
flow regimes, best-fit b can be greater
than 1
 Transient linear flow, b = 2
 Transient bilinear flow, b = 4
 Ultimately, there must be a flow regime
for which b < 1
 Never use b ≥ 1 for the last segment in any
forecast

19
8/30/2013

Is Arps’ Decline Model Valid for


Unconventional Oil or Gas Reservoirs?
 Short answer: no
 Flow transient for much of, possibly entire, life of
well in ultra-low permeability reservoirs
 Best-fit ‘b ’ values almost always > 1, and
decrease with time – no way to establish proper b
for extrapolation, since it will be less in future than
observed in the past
 Substantial risk of overstating reserves when Arps’
model used without modification (such as
minimum final decline rate)

Arps Model Applied to Simulated Data from


SPE 109625: Good Fit of History
Arps model b:1.5, D:0.007 1 days qi:4.823 mmcf d

2000

Arps model b:1.5, D:0.007 1 days qi:4.823 mmcf d


1500
Q, mmcf

10.0

1000
5.0
q, mmcf d

500
2.0

0 1.0
0 500 1000 1500
days 0.5

0 500 1000 1500


days

20
8/30/2013

Despite Good Match of History, Forecasting


Ability Poor, Especially with Limited Early Data
Years of Best Fit, Error in
History Arps “b” Remaining
Matched Reserves,
%

2 2.66 145
5 1.91 104
10 1.51 30.6
25 1.20 7.9
50 1.14 0

Dealing with Forecasts When b > 1


 After matching historical production and
determining the hyperbolic exponent, b, some
recommend that we define minimum final
Decline Rate, Dmin (exponential decline)
 Dmin should be used because hyperbolic curve
extrapolation with b >1 can seriously
overestimate future production and life of well
 Just as a reservoir in a given geological
environment has a typical “b” value, we
should identify and use typical “Dmin” value
 Analogy is best way to select Dmin

21
8/30/2013

Illustration of Minimum Final Decline Rate


qi 5000 STB/Day
ADR = 55% FIND MORE MATURE WELLS
OR OTHER ANALOGIES TO
ESTIMATE MINIMUM
HYPERBOLIC DECLINE RATE

qt 200 STB/Day
ADR = 15%

mADR = 10%

mADR = 7%

mADR = 5%

Some Questions …
 Shouldn’t we try to develop a decline model
specifically suited for transient flow, rather than force
the Arps model to fit transient data?
 Is the Dmin approach the inverse of what we should
be doing?
 Would it make more sense to use an appropriate
transient model for early (transient) data and
switch to the Arps model later when we reach
BDF? This is exactly what Fetkovich did in
constructing his well-known type curve for DCA.

22
8/30/2013

Final Thoughts
 We obtain more accurate results if we plot q/Δp (instead of q) vs. time
when values of BHP are readily available and then apply Arps’ models
 We also get better results if we eliminate data affected by fracture fluid
clean up
 Upside
 Accuracy greatly improved when BHP not constant (e.g., when
steadily decreasing) and when well is cleaning up
 Downside
 Conventional commercial decline analysis software may not allow
these alternative
 Alternatives also significantly more time consuming
 Does not transform transient flow to BDF – still require BDF data
for confident extrapolation

What We’ve Learned


 You are now able to
 Estimate future production using Arps’ model
 Estimate reserves using Arps’ decline model
 State theoretical basis for Arps’ model
 State assumptions inherent in Arps’ model
 State appropriate values of b for specific
situations
 Understand the risk of using b ≥ 1 in
extrapolating decline curves

23
8/30/2013

Production Forecasting for


Unconventional Resources
Lecture 4
Arps’ Decline Model
End

24

You might also like