Modeling Natural Fracture Networks Using Coupled Multi-Point Geostatistics and Flow Simulation

You might also like

Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 22

Modeling Natural Fracture Networks

Using Coupled Multi-Point


Geostatistics and Flow Simulation
Petroleum Technology Alliance of Canada

Drs. Dale Wong & Sanjay Srinivasan


The University of Calgary
Prof. Dale Wong Bio
DR. WONG has 20 years of experience in reservoir engineering, software
development and R&D with a full range of companies from various
industry sectors, such as exploration and production, service, consulting
and software. He is also a founder of a U.S.-based reservoir simulation
software company.
Dr. Wong was the co-chair of the committee to create the new B. A. Sc. Oil
& Gas Engineering program at the Department of Chemical and
Petroleum Engineering.
One of the pioneers of the pressure derivative method now commonly used
in pressure transient analysis methods for well tests, his areas of primary
interest are advanced well test analysis methods, transient simulation,
reservoir simulation and reservoir engineering.
Prof. Sanjay Srinivasan Bio
Prof. Sanjay Srinivasan received his doctorate in Geostatistics from
Stanford University. His research focus is on multiple point
geostatistics and calibration of information from geological models,
seismic and production data.
Dr. Srinivasan has over 7 years work experience at Bechtel Corporation
as a senior petroleum engineer, working on both upstream and
downstream aspects of petroleum field development.
Dr. Srinivasan supervises a research team of 11 graduate students
working on characterization of fractured reservoirs and complex
sandstone depositional environments, automatic history matching, data
sufficiency in reservoir modeling etc…
Mission
• To fast-track integrated research into
modeling naturally fractured reservoirs –
“university-initiated and industry-focused”
• Creation of a center of learning for these
methods for CAGE (Center for the
Advancement of Geostatistics in
Engineering)
Technology Delivery Point

You Are Here

Time
Development of 2-D, Steady-
State Methods
Target  August, 2001
• Develop mathematical and numerical prototype for future
development – “prove the technology, prove the people”
• Develop a training-based algorithm for recognizing
fracture patterns from analogues – “pattern-based
geostatistics”
Development of Methods to Model
Large Fracture Networks
Target  April, 2002
• Investigate strategies for solving BEM equation (ex.
iterative solution methods)
• Approximate “far away” fractures
• Integrated software for pattern recognition from analogues,
simulation conditional to reservoir-specific fracture data
Automatic History Matching

Target  April, 2003


• Developing iterative numerical procedures for creating fracture
networks that match field behavior
• Calibration procedure for identifying fracture density +
orientation related information from production data
Integration with Simulation
Technologies
Target  August, 2003
• Coupling of Numerical Procedures (FDM, FEM or other BEM codes)
• Modular software:
pattern recognition from analogues
+ calibration of info. from historic data
+ creating history matched reservoir models
Technology Development
Timeline
2-D, Steady-State, Stochastic & BEM Prototype

3-D, Steady-State, Stochastic & BEM Prototype


Methods for Modeling Many Fracture Problems

History Matching

Integration into Other Simulation Technologies

JAN 2002 JAN 2003 JAN 2004


CAGE - CENTRE FOR ADVANCEMENT
OF GEOSTATISTICS IN ENGINEERING
THE CONFIGURATION OF THE MIXED UNIX/PC BASED
PLATFORM :
P-270
CENTRAL SERVER
HDD – 128GB
RAM – 4GB

P-170 P-170 P-170 P-170 PIII


HDD-10GB HDD-10GB HDD-10GB HDD-10GB HDD-10GB
RAM-1GB RAM-1GB RAM-1GB RAM-1GB RAM-1GB
Requested Support

• Funding, to be matched by government


agencies, for capital and operating costs
• Company human and information resources
• $10,000 per participating company per annum
Traditional Problem – Fracture
Placement in Numerical Simulation
Vert Well
Hz Well

Vert Well
Problems
• The location and scale of natural fractures
• The “true” representation is probabilistic and not deterministic 
Methodology?
• How can information from various sources be integrated into
stochastic fracture representation?
• How does one model the fractures or, alternatively, history match
their behavior?
Geostatistical modeling : Data Integration
Unknown true reservoir Observed soft response
Earth physics

response
Seismic
TF
Flow/pressure
response
Data acquisition

Hard variable(s) = time


variable(s) to be simulated Soft data
Rock properties Physical response

Traditional geostatistical approach


Model and reproduce two-point correlations
Reservoir modeling – modern paradigm

True reservoir Training reservoir


Training image
Data event
True reservoir
150.000 150.000

Sand Sand

North
North

Shale Shale

0.0 0.0
0.0 100.000 0.0 East 100.000
East

Set of data events = training data set


Integrated Reservoir Modeling
True reservoir Soft data Reservoir model
True reservoir Seismic data with soft data / conditional simulation
150.000 150.000 150.000

1.000

0.900

0.8000
Sand
0.7000

0.6000
North

North
North
0.5000

0.4000

0.3000
Shale
0.2000

0.1000

0.0

0.0 0.0 0.0


0.0 East 100.000 0.0 100.000 0.0 East 100.000
East

Pr( fracture | template data ) x 150.000


forward simulation

Pr(soft data | fracture )


= Pr ( model | hard + soft data)

North
forward simulated soft data
0.0
0.0 East 100.000
Goal: Match Well Pressure and Rate
History in Naturally Fractured Reservoirs
600,000.0

500,000.0

400,000.0

Rate
300,000.0

200,000.0

40,000
100,000.0

35,000

0.0
30,000 Feb-99 May-99 Aug-99 Nov-99 Feb-00
Time
25,000
Wellhead Pressure (kpa)

20,000

15,000

10,000

5,000

0
Feb-99 May-99 Aug-99 Nov-99 Feb-00
Time
BEM For Natural Fractures

Boundary Element
  on Fracture

 Lough, Lee and Kamath, paper SPE 36730


k f  7.776 10 9 h 2
How Fractures are Modeled
Boundary Element Method
   x   f  x   steady state, potential  
2  

  x      x g  x , x dx   f  x g  x , x dx


      
 
 
ln x  x
g  2  D, free space Green' s function 
2
 g  x , x   g  x , x 
 
q  x      x  dx    f  x  

dx 
 n  n

   x 
q  x 
n
Stochastic - BEM Method to
Model Natural Fractures
 A      B.C.' s 
ij i i

 B   q    B.C.' s 
ij i i

where Aij and Bij are determinis tic dimensional terms


Now,
 A      B.C.' s 
ij i i

 B   q    B.C.' s 
ij i i

where Aij and Bi are stochastic dimensional terms


Summary
• New, breakthrough technology to model
natural fractures in forward or inverse
simulations
• In the forward case, for a particular
fracture network density and orientation
determine reservoir response
• In the inverse case, given a reservoir
response, determine the fracture density
and orientation

You might also like