1. History Matching Flowmeter Data in the Ghawar Field
A Combined Facies and
Discrete Fracture
Reservoir Model
J. Voelker

2. Acknowledgement:
Jim Liu, Saudi Aramco

3. Principal points: the necessity of a super-k characterization
generation of combined facies/discrete fracture models
use of the well model to simulate discrete fracture flow
history matching flowmeter data with probability perturbation

4. Ghawar Field (Saudi Aramco)
largest onshore field in the world
50 years of production
mixed clastic/carbonate deposition
study area:
8 km x 3 km x 250 ft
Meyer, 2000

5. Study area simulation grid reservoir grid super-k bed grainstone
background
reservoir grid

6. The necessity of a super-k characterization
abnormally high flow capacity measured at the well, not predicted by inflow performance (kh, skin)
flow not correlated with permeability
Meyer, 2000

7. The necessity of a super-k characterization
problem: abnormal flow capacity extends beyond the well
super-k causes early injection water breakthrough at adjacent producing wells
possible producing well abandonment due to inability to shut off water
solution: predict the location of super-k
final history matched model must have the correct geology, including the essential flow elements of super-k
method of solution:
maintain the correct geology throughout the history matching process: probability perturbation

8. Generation of combined facies/discrete fracture models
essential super-k elements:
grainstone
discrete fracture
super-k bed
Meyer, 2000

9. Generation of combined facies/discrete fracture models
existence of faulting
region
study area

10. Generation of combined facies/discrete fracture models
the facies model
N/G=0.54
N/G=0.38
N/G=0.29
super-k bed model N/G = 0.18
grainstone model
super-k bed
grainstone
background
combined model

11. + = Generation of combined facies/discrete fracture models
stochastic: super-k beds + =
super-k beds imbedded into facies model

12. Generation of combined facies/discrete fracture models
Facies alone cannot predict super-k flow
Grainstone model
Grainstone + super-k bed model
15 equiprobable realizations
model incomplete
additional flow component required: fractures

13. Use of the well model to simulate discrete fracture flow
fracture flow simulation candidates:
pro: good fracture geometry description, no limit on number of fractures
con: every new model requires a new flow simulation grid
discretization
pro: flow simulation grid remains the same
con: poor fracture geometry description, limited number of fractures
dual porosity
pro: easily implemented, grid update not required, no limit on number of fractures
con: fracture geometry description almost as good as discretization (but better than DP), new
well model

14. Use of the well model to simulate discrete fracture flow
discretization
dual porosity
well model

15. Use of the well model to simulate discrete fracture flow
most effective placement of one fracture end: at the well blocks
implementation
west injection well
implication: the discrete fracture is near the Peaceman radius of the well,
discrete fracture does not intersect the well : no well stimulation
b/d/ft

16. Use of the well model to simulate discrete fracture flow
most effective placement of other fracture end: at a super-k bed located in a different pressure regime
implementation 1 4 2 5 y y 3
z = 51 x
west injection well 1 2 3 4 5

17. Use of the well model to simulate discrete fracture flow
super-k bed / fracture realizations:
perturbation of super-k beds results in perturbation of fractures
well blocks: conditioning
super-k bed blocks:search

18. History matching flowmeter data with probability perturbation?
objective of history match:
super-k predictions

19. History match: 2 tests, separated by 3 years
facies
flowmeter
fracture
History match: 2 tests, separated by 3 years
(total 18 tests in 25 years)

20. History matching flowmeter data with probability perturbation
equiprobable realizations
Facies ony
Facies ony

21. History matching flowmeter data with probability perturbation
probability perturbation results
(5 iterations) 4 2 1

22. History matching flowmeter data with probability perturbation
probability perturbation results
(5 iterations) 4 5 2 3 1

23. History matching flowmeter data with probability perturbation
probability perturbation results
(5 iterations) 4 3 2 1

24. i=0

25. i=1

26. i=2

27. i=3

28. i=4

29. History matching flowmeter data with probability perturbation
simulation behavior (2-2.8 GHz, 2 GB)
cpu cost per flow simulation:
3 year run, no fractures: ~10 minutes
3 year run, 2 fractures: ~12 minutes
25 year run, no fractures: ~30 minutes
25 year run, 5 fractures: ~100 minutes
cpu cost increases significantly with number of fractures, number of connections, and connection transmissibility

30. Conclusions Future work
super-k flow behavior may be described with a combined facies / discrete fracture model
discrete fracture flow may be modeled effectively with conventional well models
probability perturbation is an effective history matching tool in this case study
Future work
extend the technique to a larger study area
improve the convergence properties of simulations
develop well placement strategies

31. Extra slides

32. History matching flowmeter data with probability perturbation
essential characteristics of the model:
discrete fractures
connections to depletion/injection regions
super-k beds only
short fractures

33. History matching flowmeter data with probability perturbation
probability perturbation results 4 3 2 1 4 1 2 3

34. Analytical characterization of super-k
Given a facies reservoir model:
facies proportions
facies geometry
facies well conditioning
facies permeability
reliable flowmeter data
known well damage or stimulation
Then:
flowmeter data is governed by zone permeability

35. Analytical characterization of super-k
Flowmeter plot
Binary facies model
Well data
layered model

36. Analytical characterization of super-k
Layered reservoir
Flowmeter-west injection well
Correct facies geology
Super-k

37. The discrete fracture / super-k bed requirement
Added elements must:
be constituents of the real geology
increase well flow capacity
exist in constrained vertical intervals
Simplest conventional candidate:
discrete fractures

38. Use of the well model to simulate discrete fracture flow
maximum fracture connection transmissibility,