You are on page 1of 68

Reservoir Characterization

From Production and


Injection Fluctuations
Larry W. Lake
The University of Texas at Austin
Larry_Lake@mail.utexas.edu
Outline
Introduction
The Model
Applications of the Model
Synthetic Fields
(Synfields)
Field Applications
Uses of the Model
Validation
Prior and Current Work
Belkis Refunjol
Jorge SAntana Pizarro
(Petrobras)
Isolda Griffiths (Shell)
Alejandro Albertoni (Nexen)
Pablo Gentil (ENI)
Ali Al-Yousif (Aramco)
Danial Kaviani (TAMU)
Thang Bui (TAMU)
Xming Liang
Morteza Sayarpour (Chevron)
Sami Kaswas (Exxon)
Tom Edgar, ChE
Leon Lasdon, IROM
Jerry Jensen (U.Calgary)
Alireza Mollaei, PGE
Ahn Phoung Nguyen, ChE
Fei Cao, PGE
Jacob McGregor, PGE
Jong Suk Kim, ChE
Wenle Wang, PGE
Past
Present
What others say about modeling
Bratvold and BickelTwo types
Verisimilitude- the appearance of reality
Cogent- enables decisions
Haldorsen.the progress of ideas
Youth= simple, nave
Adolescence=complex, nave
Middle age=complex, sophisticated
Maturity= simple, sophisticated
Hypothesis
Characteristics of a reservoir can be
inferred from analyzing production
and injection data only
Boundary Conditions
Must be injection project
Rates are most abundant data type
Rates must vary
No geologic model required
Everything done in a spreadsheet
Outline
Introduction
The Model
Applications of the Model
Synthetic Fields
(Synfields)
Field Applications
Uses of the Model
Validation
q(t) = q(t
0
)e
(
tt
0

)
+ I(t) 1 e
(
tt
0

c
t
V
p
( )
p
wf,t
p
wf,0
t t
0

1 e
(
tt
0

CRM Continuity Equation


c
t
V
p
dp
dt
= i(t) q(t)
dq(t)
dt
+
1

q(t) =
1

i(t) J
dp
wf
dt
=
c
t
V
p
J
Ordinary Differential Equation:
Continuity:
Solution:
q(t)
i(t)
BHP
Injection
Primary
q(t) = J p p
wf
( )
Production Rate:
Signal Response
Production response to an injection signal
Connectivity
ij = 1 day
fij = 0%
Connectivity
ij = 1 day
fij = 100%
Connectivity
ij = 6 days
fij = 100%
Connectivity
ij = 6 days
fij = 65%
Capacitance-Resistance Model (CRMT)
( ) k
t t
k k
I e e q q

+ =


1
1

q(t) I(t)
J
V c
p t
=
Time constant
f
2j
f
6j
f
4j
f
3j
f
5j
j

f
1j
f
11
f
12
f
13
I
6
I
1
I
2
I
3
I
4
I
5
q
j
(t)
Capacitance-Resistance Model (CRMP)
( ) ik
n
i
ij
t t
k j jk
I f e e q q
i
j j

+ =
1
1
1

j
p t
j
J
V c

=
1
1

=
p
n
j
ij
f
Time constant
Inter-well connectivity or gain
Drainage volume
around a producer
Capacitance-Resistance Model (CRMIP)
I
i
(t)
q
j
(t)
f
ij

ij
ij
p t
ij
J
V c

=
1
1

=
p
n
j
ij
f
Time constant
Inter-well connectivity or gain
( )
=

+ =
i
ij ij
n
i
ik ij
t t
k ij jk
I f e e q q
1
1
1

Steady-State Connectivity Map

0
0
0
0
0
0 20 40 60 80 100
Producer
Water Injector
Carbon Dioxide Injector
0 1,000 ft
Better CO2
Performance
Interwell Connectivity
Two Equally Viable Solutions
Transient Interwell Connectivity
After 10 days
Transient Interwell Connectivity
After 30 days
Transient Interwell Connectivity
After 90 days
Transient Interwell Connectivity
After 180 days
Transient Interwell Connectivity
After 365 days
Transient Interwell Connectivity
After 2 years
Transient Interwell Connectivity
After 4 years
Transient Interwell Connectivity
4 years <<
Gains >0.5
Mature West Texas Waterflood
Injector
Producer
Gains>0.5
Gains >0.4
Mature West Texas Waterflood
Injector
Producer
Gains >0.3
Mature West Texas Waterflood
Gains>0.3
Injector
Producer
Gains >0.2
Mature West Texas Waterflood
Gains>0.2
Injector
Producer
Mature West Texas Waterflood
R-squared
Producer Number
Time Constants
Reservoir A
Producer 184 Good Fit
R
2
= 0.961
err = 0.146
Bbl/
day
Month
Producer 127 Good Fit
R
2
= 0.696
err = 0.037
outliers
Bbl/
day
Month
Producer 74 Poor Fit
R
2
= -1.03
err = 0.143
Bbl/
day
Month
Producer 201 Poor Fit
R
2
= 0.793
err = 6.58
Bbl/
day
Month
CRM: Oil Fractional-Flow Model
f
o
(t) =
q
o
q
o
+ q
w
=
1
1+ WOR(t)
q
o
(t) = f
o
(t)q(t)
f
o
(t) =
1
1+ a CWI(t)
b
log
1
f
o
(t)
1

= loga + blog CWI(t)


Outline
Introduction
The Model
Applications of the Model
Synthetic Fields
(Synfields)
Field Applications
Uses of the Model
Validation
Future Injection
Historic Period 131 Active Injectors
Prediction Period 97 Active Injectors
Injection has been concentrated in fewer wells (37
injectors shut-in)
27.3% of historic field injection from injectors shut-
in throughout prediction period
Optimal Injection and Predicted Oil Production for
the Field
0 20 40 60 80 100 120 140 160 180 200
2
3
4
5
6
x 10
4
Month
b
b
l
/
d
a
y
Historic
Optimal
0 20 40 60 80 100 120 140 160 180 200
500
1000
1500
2000
2500
3000
Month
b
b
l
/
d
a
y
Historic Oil Production
Predicted Oil Production
Extrapolated Oil Production
Injection Shares
Injector Number
Percent of
Total
Production Shares
P112
P195
Producer Number
Percent of
Total
GardnerHypeCurve
TheGardnerGroup
40 J im Honefenger (P.E. Moseley & Associates, Inc.)
Outline
Introduction
The Model
Applications of the Model
Synthetic Fields
(Synfields)
Field Applications
Uses of the Model
Validation
Validation
Just how do we scientifically validate
geoscience hypotheses?
Remember:
Characteristics of a reservoir can be inferred
from analyzing production and injection data
only
Recognizing testable hypotheses can be subtle and
requires practice. To do it, ask how would one test this
hypothesis.
If the duck is lighter than this woman, then she is
a witch.
Synfield Cases
Heterogeneity
Large compressibility
Fractures
Barriers
Anisotropy
Partial completions
Large shut in times
Changing BHP
All agree with imposed geology
Validation
Field Injectant Independent Data Agree
With Data
Synfields Water Simulation Very well
Characteristics of a reservoir can be inferred from
analyzing production and injection data only
Retrodiction
Validation
Field Injectant Independent Data Agree
With Data
Synfields Water Simulation Very well
Synfields Water Retrodiction Very well
Characteristics of a reservoir can be inferred from
analyzing production and injection data only
Chihuido Field
Good correlation
Inferred faults are in yellow
Gainsand time constants
reproduce known geological features
Validation
Field Injectant Independent Data Agree
With Data
Synfields Water Simulation Very well
Synfields Water Retrodiction Very well
Chuido Water Faults from seismic Reasonably
Characteristics of a reservoir can be inferred from
analyzing production and injection data only
SWCF Flow Capacity
7516
7519
7523
7524
From Al-Yousef (2006)
Homogeneous
Validation
Field Injectant Independent Data Agree
With Data
Synfields Water Simulation Very well
Synfields Water Retrodiction Very well
Chuido Water Faults from seismic Reasonably
SWCFU Water Anecdotal fractures Reasonably
Characteristics of a reservoir can be inferred from
analyzing production and injection data only
North Sea Field II
Validation
Field Injectant Independent Data Agree
With Data
Synfields Water Simulation Very well
Synfields Water Retrodiction Very well
Chuido Water Faults from seismic Reasonably
SWCFU Water Anecdotal fractures Reasonably
NSF II Water Structure Well
Characteristics of a reservoir can be inferred from
analyzing production and injection data only
North Buck Draw Comparison
CM correlates with tracer breakthrough time
0
5
10
15
20
30 0 5 10 15 20 25 35
Tracer Breakthrough Time (months)
S
p
e
a
r
m
a
n

o
r

C
M

T
i
m
e

(
m
o
n
t
h
s
)
Spearman
CM
Linear (CM)
Validation
Field Injectant Independent Data Agree
With Data
Synfields Water Simulation Very well
Snyfields Water Retrodiction Very well
Chuido Water Faults from seismic Reasonably
SWCFU Water Anecdotal fractures Reasonably
NSF II Water Structure Well
NBDU Gas Tracer data Fairly well
Characteristics of a reservoir can be inferred from
analyzing production and injection data only
Williston Basin Field
Validation
Field Injectant Independent Data Agree
With Data
Synfields Water Simulation Very well
Snyfields Water Retrodiction Very well
Chuido Water Faults from seismic Reasonably
SWCFU Water Anecdotal fractures Reasonably
NSF I Water Structure Well
NBDU Gas Tracer data Fairly well
Will. Basin Water Acoustic impedance Reasonably
Characteristics of a reservoir can be inferred from
analyzing production and injection data only
Future Work
Working spreadsheet
Couple to GAMS
Excel vs. MATLAB
Multiplotting (visualization)
Integrate with DA/VOI approaches
Propagating error/uncertainty
More validation (oil in tank)
Extend to primary recovery
Fluid allocation studies (conformance)
Optimize to produce more oil
Add EOR model(s)
Remove outliers
Maximize NPV of future oil recovery
Warm start
Gainfit
Remove
inactive wells
Remove gains
based on distance
Remove small
gains
Gainfit #2
Calculate residuals
and replace outliers
Gainfit #3
Gainfit #1
Fracfit #1
Calculate residuals
and remove outliers
Fracfit #2
Reservoir
model
Model Fit and Prediction Algorithm
~2.5 hrs
computation
time
Remove outliers
Maximize NPV of future oil recovery
Warm start
Gainfit
Remove
inactive wells
Remove gains
based on distance
Remove small
gains
Gainfit #2
Calculate residuals
and replace outliers
Gainfit #3
Gainfit #1
Fracfit #1
Calculate residuals
and remove outliers
Fracfit #2
Reservoir
model
Model Fit and Prediction Algorithm
<1 min
computation
time
Remove outliers
Maximize NPV of future oil recovery
Warm start
Gainfit
Remove
inactive wells
Remove gains
based on distance
Remove small
gains
Gainfit #2
Calculate residuals
and replace outliers
Gainfit #3
Gainfit #1
Fracfit #1
Calculate residuals
and remove outliers
Fracfit #2
Reservoir
model
Model Fit and Prediction Algorithm
<10 min
computation
time
Appraisal and
Conceptual
Analysis
GATE GATE
Evaluate
Alternatives
GATE
Define
Selected
Alternative
GATE Execute Operate
Inevitable
Dis-
appointment
Portfolio
Optimization
Uncertainty
Updating
Concept Selection & Development Optimization
Real Options
Portfolio Management and
Project Selection
Addressing Risks Throughout the E&P Asset Lifecycle
VOI; Impact
of Estimates
& Methods
Financial Risk
Management
Cost and Schedule Estimating; Execution Risk Management
HSE Risk Management
Real-Time Optimization
and Risk Management
Valuing Price
Forecasts
Capital
Allocation w/
Uncertain
Arrivals
FUTURE:
Life Cycle
Assessments
Contracting
Strategies
(lump sum v
cost plus?)
MPD &
Blowouts;
Drlg Safety;
Offshore
Spills
Simple Model
Development
Gain Map
Injector
Producer
P210
I 58
P103
Producer 210 (large distance)
093 . 0
882 . 0 R
2
=
=
err
Bbl/
day
Producer 103 (skipped over)
110 . 0
635 . 0 R
2
=
=
err
Bbl/
day
Injector Number
Lost Injection

1 f
ij
j=1
N
p

CRM Fit Total Field


R
2
= 0.956
Bbl/
day
Month

You might also like