ENI-MITEI Annual Meeting, S. Donato M., 29 June 2010
Multiscale Reservoir Science for Enhanced Oil
Recovery: Technology Development and Field
Applications
Rob van der Hilst, Steve Brown, Dan Burns, Michael Fehler,
Brad Hager, Tom Herring, Ruben Juanes, Dennis McLaughlin
Earth Resources Laboratory
MIT
Overall Motivation:
To meet demand:



New fields (e.g., deep off-shore, near/beneath complex structures, arctic region)
Enhanced Oil Recovery (EOR) from existing fields (global average < 40%)
Unconventional oil/gas (heavy oils, tar sands, tight gas reservoirs, hydrates)
Challenge:
Increase production from reservoirs that are complex and
strongly heterogeneous (both for new and existing fields)
Reservoir management:
Predict reservoir performance to enable optimal operation:
– Maximize reservoir sweep
– Best well placement and completion design
Integration of geophysical description with reservoir
models  more reliable prediction of performance
For example: fractured reservoirs
deformation during passage
of a compressional wave
Carbonate cliffs
geology/geophysics ↔ flow modeling ↔ enhanced production
?
Oil Production
Oil
Seismic Data
Willis et al (2006)
Water Injection
Water Front
What do we want to know?
• Where are the fractures?
• What are the fracture orientations?
• What are the fluid-flow properties of fractures
(that is, how do fluids flow through them)?
Approach:
• Joint analysis of geophysical response (e.g.,
scattering from fractures and heterogeneity,
deformation) and flow
Using Geophysics to Constrain Flow Mode
Geophysics-constrained
reservoir description
Geophysics-constrained
permeability model
Kfrac
Reservoir description from geophysics
Response (e.g. well rate)
Qwell
model
Model updated
with new data
data
time
INTEGRATED RESERVOIR SCIENCE
1:
Reservoir Structure and
Response
–
–
–
–
Fracture Characterization (e.g., seismics)
Flow Simulation
Data assimilation & real-time control
Quantitative integration
INTEGRATED RESERVOIR SCIENCE
1:
Reservoir Structure and
Response
2:
Reservoir Evolution
and Performance
– Surface deformation (GPS & InSAR)
– Coupled geomechanical/reservoir modeling
INTEGRATED RESERVOIR SCIENCE
1:
Reservoir Structure and
Response
2:
Reservoir Evolution
and Performance
3:
Application of New Concepts
(Field Case Study)
Integration of Geophysics &
Reservoir performance modeling
Different Levels of Integration
Data
Model
• Surface deformation
• Production data
- tiltmeters
•Surface seismic
• Well logs
• Analogue reservoirs •Fracture characterization - InSAR, GPS
• Wellbore breakouts
• 3D seismic
• Induced seismicity
Flow models
Geophysical
interpretation
Geomechanical
modeling
coupled
Clearly insufficient
CTRW-RTT joint inversion methodology
Main outcomes:
3-way data assimilation methodology
• Better forecasts
• Optimal production to maximize recovery while controlling subsidence
Different Levels of Integration
Data
Model
• Surface deformation
• Production data
- tiltmeters
•Surface seismic
• Well logs
• Analogue reservoirs •Fracture characterization - InSAR, GPS
• Wellbore breakouts
• 3D seismic
• Induced seismicity
Flow models
Geophysical
Geomechanical
interpretation coupled
modeling
coupled
Main outcomes:
3-way data assimilation methodology
• Better forecasts
• Optimal production to maximize recovery while controlling subsidence
Different Levels of Integration
Data
Model
• Surface deformation
• Production data
- tiltmeters
•Surface seismic
• Well logs
• Analogue reservoirs •Fracture characterization - InSAR, GPS
• Wellbore breakouts
• 3D seismic
• Induced seismicity
Flow models
Geophysical
Geomechanical
interpretation coupled
modeling
coupled
CTRW-RTT joint inversion methodology
3-way data assimilation methodology
Main outcomes:
3-way data assimilation methodology
• Better forecasts
• Optimal production to maximize recovery while controlling subsidence
Different Levels of Integration
Data
Model
• Surface deformation
• Production data
- tiltmeters
•Surface seismic
• Well logs
• Analogue reservoirs •Fracture characterization - InSAR, GPS
• Wellbore breakouts
• 3D seismic
• Induced seismicity
Flow models
Geophysical
Geomechanical
interpretation coupled
modeling
coupled
CTRW-RTT joint inversion methodology
3-way data assimilation methodology
Main outcomes:
• Better forecasts
• Optimal production to maximize recovery while controlling subsidence
Numerical and Laboratory Modeling of
Scattering from Fractures
• Understand seismic response of fractures and
fracture systems
– Develop new field-data analysis approaches
– Platform/data for testing & evaluation of new
methods
• Develop models to test relationships between
fracture compliance, roughness, permeability,
and seismic scattering
Numerical and Laboratory Modeling of
Scattering from Fractures
• Seismic response
– Numerical
•
•
•
•
•
Single and multiple fractures
2D and 3D
P-to-P and P-to-S scattering
Finite difference; semi-analytical; boundary element
Static models to estimate compliance
– Experimental
• Multiple fracture model
• Incorporate flowing fractures
wave length
fracture
seismic response
homogeneous
anisotropy zone
(1)
(2)
(3)
Focus Area
Linear-slip Fracture Model (Schoenberg, 1980)
Fracture Compliance
fracture
u  u2  u1  Z  T
u1
u2
displacement
compliance traction
length/stress [m/Pa]
“zero” thickness
Numerical Model 1
single fracture
(NB we can do this also in 3D)
P-wave
P scattered waves
2D P-to-P Fracture
Response Function (FRF)
Numerical Model
2
Numerical
Model
Multiple
Fractures
Multiple
(parallel)
Fractures)
Fracture Spacing 50 m
Aperture 5 m
Fracture Zone 50 m
thick
Ra
psNMO11
fracture10
Tr
900
az=0
800
9 az=10
700
8 az=20
Transverse
component
0
60 amplitude
shows strong
az=30
7
near 45 degrees
500
6 az=40
400
5 az=50
300
4 az=60
200
approach
3 az=70
New
to
100 scattering in
analyzing
2 az=80
field data?
00
1 az=90
0
2
0.5
Time (sec)
1
Vz
11
11
900
10 az=0
800
9 az=10
900
10 az=0
800
9 az=10
700
8 az=20
600
7 az=30
700
8 az=20
600
7 az=30
500
6 az=40
400
5 az=50
500
6 az=40
400
5 az=50
300
4 az=60
200
3 az=70
300
4 az=60
200
3 az=70
100
2 az=80
100
2 az=80
00
1 az=90
00
1 az=90
0
0.5
Time (sec)
1
0
0.5
Time (sec)
1
Laboratory Experiments: Current Status
30 cm
• Seismic acquisition geometries
– Iso-Offset acquisition at different azimuths
– Common source gathers at different azimuths
– CDP gathers at different azimuths
• Comparison with numerical models
• Move towards joint seismic-flow experiments
Laboratory Experiments: Acquisition Geometry
900
100
00
Offset = 6 cm
P Wave Source
P, S Receiver
PP 2nd interface
90
80
70
60
50
40
30
Transverse component
shows strong amplitude
near 45 degrees
(similar to numerical result)
20
10
0
PP Fracture Tip
P-S Converted
SS Fracture Tip
Conclusions Modeling
• The amplitude of scattered-waves scales with compliance (Z)
• Radiation patterns depend mostly on ratio of normal to tangential
compliance (ZN/ZT)
• On the transverse component, P-S Converted wave shows
maximum amplitude at about 40-500  possible new orientation
attribute
• On the inline component, P-S Converted wave shows systematic
increase in amplitude towards 900 (not shown)  possible new
orientation attribute
• Stacking enhances signal in a direction parallel to fracture
orientation (consistent with Scattering Index - Willis et al., 2006)
The insight thus obtained can be used to infer
fracture compliance from seismic field data
Compliance (e.g., from seismics)  Permeability
–
–
–
Elastic compliance is a key parameter
influencing seismic scattering in fractured rocks.
We want to know more about compliance
values, scaling, and relation to permeability
We are conducting numerical studies based on
realistic fracture roughness statistics
Fehler, Burns, Brown
Compliance (e.g., from seismics)  Permeability
Empirical Relationship (from fracture modeling)
1/compliance (relative)
Brown
Compliance (e.g., from seismics)  Permeability
–
–
–
–
Elastic compliance is a key parameter
influencing seismic scattering in fractured rocks.
We want to know more about compliance
values, scaling, and relation to permeability
We are conducting numerical studies based on
realistic fracture roughness statistics
We find:
•
•
Large fractures have much larger compliance
Clear relationships between permeability, compliance,
and stress
Brown
Fracture Response Function (FRF)
• Can be obtained directly from (multi-component) seismic data
• Methodology validated with numerical and laboratory data
• Provides information about fracture orientation, spacing, and
relative compliance (& permeability)
Now: Preliminary Application to data from Emilio Field
Fehler, Burns, Brown
Seismic profile across Emilio Field
Emilio Field
Geometry of the top of reservoir & wells
Vp~4km/s
Fehler, Burns, Brown
Fracture Orientation
Confidence
Confidence
Fehler, Burns, Brown
Fracture Spacing
Fehler, Burns, Brown
Fracture Response Function
Relative Compliance
scattering strength
~ fracture compliance x fracture density
With constraints from geodetic data
(below) and with (empirical) scaling
relationships from modeling  this can
be used to estimate permeability (and
flow)
?
Relative Compliance
Fehler, Burns, Brown
Geophysical monitoring of sub-surface reservoirs (Hager, Herring)
 (sub)Surface deformation (GPS, InSAR)
 Fault (re-)activation
 Induced seismicity
seismic activity and subsidence
Surface subsidence due to
reservoir pumping
observed by GPS
monitoring
• effect on wells/production
• impact of fault activation
• potential seismic risk
Geodetic Characterization of Fractures:
fractures change surface deformation resulting from
pressure changes at depth
Isotropic porosity
NW-SE oriented vertical fracture
Vertical (color) and horizontal (vectors, max = 3) surface displacements for the same point
source volume change at unit depth. For the fracture, the maximum horizontal displacement is
greater than the vertical displacement.
Hager and Herring
Example of Observed Fracture Response:
In Salah CO2 Injection
Isotropic δv/v ~ 0.5%
Fracture opening ~ 7 cm
Observations (Onuma & Ohkawa, 2009)
Model (Vasco et al., 2010)
Sensitivity to fracture properties
• Geodesy
– Assume n cracks with width change δb
– Displacement ~ nδb
• Only the product is resolvable
• Assume δb ~ b
• Displacement is then proportional to nb
• Flow studies ( permeability k)
– k ~ nb3
 Joint inversion of displacement and flow
data can resolve n and b
Hager and Juanes
b+δb
b
Facies Identification in Petroleum Reservoirs
Objective:
Develop efficient and robust framework for the reconstruction of geologic facies
from reservoir data.
Reservoir : D
Problem Statement:
Given production data from wells, we are
interested in the following inverse problem: find
the region Ω (the facies) corresponding to the
high permeability of the reservoir.
McLaughlin group
D
  high permeability
(red region )
Synthetic Experiment:
Initial guess 1
Identification of Absolute Permeability given production data from wells
Data: Flow rates from 9 production wells and 4 injection wells.
Initial guess 1
Reference
McLaughlin group
(with known facies at the well
locations)
Synthetic Experiment:
Initial guess 1
Identification of Absolute Permeability given production data from wells
Reconstruction
Reference
McLaughlin group
Gradient-based
(180 iterations)
Flow Modeling – Research thrusts
Viscous fingering in a Hele-Shaw cell

 Coupled flow and geomechanics
 Computational aspects: discretization, staggered solution
 Reservoir modeling: response of fractures / faults
 Direct numerical simulation of flow in fractured reservoirs

 Continuous-time random walk (CTRW) modeling of flow in fractures
 Inversion / data assimilation
 Towards joint seismic-flow inversion: joint CTRW-RTT paradigm
 Towards 3-way inversion: flow, seismic, geomechanics
Juanes
Flow in fractured media – why a stochastic approach
 A deterministic multiscale approach is not attractive for inversion,
optimization, and control:
 Amount of data is insufficient to obtain a well-posed problem
 Resolution of data is insufficient to locate individual fractures
 Need a stochastic multiscale approach and, in particular:
 Parsimonious flow model (fewer parameters)
 Capture anomalous (non-Gaussian) behavior of transport
 Allows assessment
of predictability
Juanes
(Photograph by Jon Olson)
A simple fracture network – particle tracking
 Two sets of fractures
(constant orientation and density)
 Power-law distribution of velocities
(uncorrelated)
 Develop model of expected transport
(mean) and its confidence (variance)
Juanes
A simple fracture network – effective model
 The mean behavior is exactly
described by CTRW
Juanes
 The variance is exactly described
by a novel two-particle CTRW
“Continuous time random walk” and fractured reservoirs
 CTRW can model fast paths (fractures) and their directionality
along with slow paths (background matrix)
 Parameters for y(s,t) can be related to fracture orientation, spacing,
connectivity and transmissivity
Juanes, Fehler, Burns, Brown
Concluding Remarks
–
–
–
Progress in several areas
Fracture modeling and laboratory
experiments are catalysts for development of
new field data analysis methods
Seismic-to-permeability is helping to bridge
transition to reservoir modeling
–
Numerical simulation and laboratory experiments
Concluding Remarks
–
–
Inversion methodologies will be used to
combine geophysical and reservoir modeling
approaches
Reservoir analysis developing on many fronts
–
Attempt to find approach that makes best
overlap with geophysics