Professional Documents
Culture Documents
Abstract
Computer models of oil reservoirs have become increasingly
more complex in order to represent geological reality and its
impact on fluid flow. Memory and CPU time limitations by
finite difference/volume simulators force a coarser resolution
of reservoirs models through upscaling.
Upscaling can lead to significant difficulties in reservoirs
studies: (1) while the fine-scale geological model is build from
petrophysical, log, and seismic data, its dynamic behaviour is
never checked. As a result, a coarse-scale reservoir study can
be linked to a fine-scale geological model but the two might
be inconsistent in their dynamic behaviour. (2) Conversely, the
upscaled model cannot be properly tested since the flow and
production behaviour at the fine-scale level is not available.
There is no reference solution for guiding important decisions
for building a consistent upscaled model. (3) A large number
of sector models are required in designing optimal well
patterns.
Streamlines simulation is now an attractive alternative to
overcome some of these drawbacks since it offers substantial
computational efficiency while minimising numerical
diffusion and grid orientation effects. It allows the integration
of fine-scale geological models into the reservoir engineering
workflow.
In this paper, we demonstrate the usefulness and efficiency of
a streamline simulator (3DSL5) in the reservoir engineering
workflow. We evaluate its speed, memory requirements and
scalability using tracer and black oil test data sets on an SGI
Origin 2000 (250 MHz MIPS). Our data are based on real
fields and range from 200000 to 7 millions cells with cells as
small as 30x30x0;5 meters. We examine problems with preand post-processing of large data sets and visualising such
simulations. Streamlines allowed us to check the validity of a
S j
v
v
ut
ut =
,
equation
SPE 66362
for
f j
v
1
Gj = 0,
SPE 66362
2.
3.
4.
5.
fine grid
Commercial open BC
Commercial sealed BC
FEM unconfined
Analytical
FEM confined
30000
25000
Oil rate
1.
35000
20000
15000
10000
0
200
400
600
800
1000
1200
1400
1600
1800
2000
1800
2000
Time (days)
0,3
0,25
0,2
Wcut
0,15
fine grid
Commercial -open BC
Commercial sealed
FEM unconfined
Cardwell&Parsons
FEM confined
0,1
0,05
0
0
200
400
600
800
1000
1200
1400
1600
Time (days)
comparisons
for
SPE 66362
Figure 5
Commercial -open BC
Commercial sealed
FEM unconfined
1950
Cardwell&Parsons
FEM confined
Oil rate
1750
1550
1350
1150
950
750
0
200
400
600
800
1000
1200
1400
1600
1800
2000
Time (days)
p =
p averagefin e (i ) p upscaled (i )
n _ coarse
2
P2S
P3S
P2C
P1S
i =1, ncoarse
ID2
Oil
sat.
norm
Pressure
norm
Confined
FEM
unconfd
FEM
Commer.
Open BC
Commer.
sealed BC
Analytic.
0.078
0.079
0.0837
0.0832
0.0941
30.56
33.97
35.19
32.78
155.85
I1C
I1
IS
ID1
P2S
84%
P3S
11%
P2C
4%
P1S
0.7%
P3C
ID2
72%
I1C
21%
I1
26%
IS
0.2%
Fig. 7- Allocations factors for injector well ID1 and producer well
P3C
SPE 66362
Full grid
Coarse grid
200 x 544 x 80
136 x 50 x 20
CPU time Memory
CPU time Memory
Eclipse
~ 240h
4,7 GB
6 mn
210 MB
3DSL
5h42
1 GB
4mn16
90 MB
CPU time per timestep CPU time per TS
Eclipse
36 mn
20s
3DSL
42 mn
24s
Table 2 - CPU time and memory for incompressible runs
The streamline simulator uses 3 to 4 time less memory than
Eclipse, and is estimated to be 20 times faster than Eclipse for
the full grid model (2,7 million active cells) but only 1,5 times
for the coarse grid (80 000 active cells). The CPU time for the
coarse grid (6 mn versus 4mn) is not significant because it
comprises much more I/O time than calculation time.
It is important to underline that the numbers mentioned here
are meant as a practical indication for the reservoir engineer to
estimate the time necessary to perform validation simulations
in an upscaling study, and are not meant as general
comparison of the efficiency of the finite volume scheme
versus streamline-based simulation. The streamline simulator
we used has a specific option for incompressible runs, while
the FV simulator is not suited for this type of simulations.
Moreover, all the runs were not performed on a dedicated
machine and some CPU times may have been overestimated
while the simulations were running in an overloaded
environment. The processors on our SGI Origin 2000 are 250
MHz old-fashioned MIPS processors. For example, the 5h40
streamline CPU time can be reduced to less than 1h using a
more recent 866 MHz Pentium III dedicated PC.
Production curves were viewed using Eclipses postprocessor GRAF (load user option), while visualization of
TOF, pressures, and saturations maps was done using Gocad
user interface files. While post-processing of grid properties
such as pressure and saturations maps is becoming easier for
large grids, post-processing of streamlines results (time of
flight TOF, drainage zones for producers PW or injectors IW)
is not considered standard output and therefore lacks an
efficient output standard. For now, 3DSL interfaces with
Gocads p-line ASCII format. Storing the streamlines plus
properties such as TOF and well pore volumes produces large
ASCII files (1,4 GB per output step for the fine grid) causing
difficulties with our Gocad version limited to 1,5 GB of
memory. It is important to note that in many cases attempting
to view all streamlines is actually beyond the resolution of the
screen. The solution is to reduce strongly the number of
outputted streamlines (for example 1 every 200). For now, this
reduction needs to be performed inside the output section of
the streamline simulator and not inside the 3D modeler tool.
3D modelers must account for post-processing streamlines
simulations results coming from large models. Indeed, treating
simulation results in 3 dimensions with a very high resolution
grid can lead to the following problems encountered in most
commercial 3D modelers and post-processing softwares:
20000
water - coarse
10000
0
0
1000
Time in days
2000
Eclipse
3DSL
Eclipse
3DSL
SPE 66362
Fine grid
200 x 544 x 80
CPU time Total
Memory
~ 650h
10 GB
Upscaled grid
50 x 136 x 20
CPU time Total
Memory
24mn
- 280 MB
46 TS
12h28
1 GB
25mn
90 MB
11 TS
CPU Time per timestep CPU time per TS
55 mn
26s
93 mn
70s
SPE 66362
72h
2 GB
CPU time per timestep
230 mn
Upscaled grid
50 x 136 x 20
CPU time Memory
46 mn - 400 MB
55Ts
33 mn
90 MB
CPU time per TS
50s
3mn
while the white zones are not. For the considered test case,
the injectors seem to maintain the pressure on the flanks but
will not sweep much oil to the central area towards the
producer. An injector pattern with some injectors in the
central area is likely to be more satisfactory to increase oil
recovery.
Superimposing the time of flight (Fig. 10) with the mobile
oil zone (Fig. 9) and the positions of producers and injectors
wells also allows to detect potential injectors that would
cause early water breakthrough.
2. Another typical results from a streamline simulator are the
pore volumes associated with each well as indicated in Figure
11, which shows reservoir volumes associated with the
different injection wells. The regions are much larger than the
swept area shown in Figure 10, because the display of the pore
volumes associated with each well is a purely geometrical
construction. It says nothing about how long it would take to
sweep the entire area. This type of figure helps to find the
geometrical influence zone of the injectors.
For the considered test case with 6 injectors, we can notice
that the injectors have little influence on the central area and
that increasing the injected rate or enabling a longer period of
injection will have little effect. For incompressible system,
such detection might more difficult since taken together all
injectors must exactly sweep the entire pore volume, although
it might take a long time. Figure 11 also indicates that one
injector (dark grey color - upper left) only has a very small
pore volume associated with it, suggesting that this injector is
not efficient and should be placed elsewhere. As the previous
one, this second type of results is difficult to obtain using a
conventional FV simulator.
3. A third type of typical result is the geometrical producer
influence volume as indicated in Fig. 12. Each color indicates
the volume of the reservoir tied to a given producer. Figure 12
depicts the reservoir volume associated with each producer for
6 vertical wells while Figure 13 for 6 horizontal wells having
one end at the same XY position. In this case where all the
faults are communicating, the horizontal well scheme has the
potentiality to drain almost all the oil field but it doesn't
indicate the time necessary to reach this objective. The
theoretical drainage zone for horizontal well P4 (white color
Fig. 13) is enormous but since the streamlines are very
elongated, the time indicated by the simulation for the oil in
the northern part of the reservoir to reach the producer is too
long and other economical constraints do intervene. The
volumes associated with the 2 northern wells are smaller using
horizontal wells. This observation is in line with the results of
the simulation which demonstrates that implementing 2
vertical wells instead of 2 horizontal wells in the northern part
of the reservoir was more efficient.
Superimposing Figure 13 with the oil in place map helps to
detect undrained areas, and locate positions for infill wells.
Misplacement of producers is easily detected when the
P6
80
P5
70
P4
P3
60
P2
50
P1
40
30
20
10
0
1
365
730
1095
1460
1865
SPE 66362
I7
I6
I5
I4
I3
I2
I1
80
volume percentage
70
60
50
40
30
20
10
0
1
365
730
1095
Time (days)
1460
1865
SPE 66362
Optimal Time Stepping: One key advantage of streamlinebased simulation is that large timestep can be taken. What the
optimal number is cannot be know apriori, but simple
numerical experiments can quickly give an indication. Figure
18 shows simulations using constant time step sizes starting
with 6 months and decreasing to 10days. The dots show nonuniform timesteps: 10x25 days followed by 5x50days,
5x100days and finally 5x200days for a total of 25 timesteps.
We use this timestepping for all subsequent runs.
Conclusions
The usage of a streamline simulator for two types of typical
reservoir simulation tasks has been presented: (1) validation of
the upscaling process for absolute permeabilities and for
pseudo relative permeabilities in water-oil system; and (2) aid
in optimizing well pattern design.
10
References
1. Batycky, R.P., Blunt, M.J., and Thiele, M.R.: A 3D Field
Scale Streamline-Based Reservoir Simulator, SPE
Reservoir Engineering (November 1997) 246-254.
2. Thiele, M.R., Batycky, R.P., Blunt, M.J., and Orr Jr.,
F.M.: Simulating Flow in Heterogeneous Systems Using
Streamtubes and Streamlines, SPE
Reservoir
Engineering (February 1996) 5-12.
3. Bratvedt, F., Bratvedt, K., Buchholz, C.F., Holden, L.,
Holden, H., and Risebro, N.H.: A New Front Tracking
Method for Reservoir Simualtion, SPE Reservoir
Engineering (February 1992), 107-116.
4. Bratvedt, F., Gimse, T., and Tegnander, C.: Streamline
Computations for Porous Media Flow Including Gravity,
Transport in Porous Media (October 1996) 63-78.
5. StreamSim Technologies: 3DSL v1.3 Reference Manual
(September 2000).
6. Thiele, M.R., Batycky, R.P., and Blunt, M.J.: A
Streamline-Based
3D
Field-Scale
Compositional
Reservoir Simulator, paper SPE #38889 presented at the
ATCE, San Antonio,TX (October 1997).
7. Crane, M., Bratvedt, F., Bratvedt, K., Childs, P., and
Olufsen, R.: A Fully Compositional Streamline
Simulator, paper SPE #63156 presented at the ATCE,
Dallas, TX (October 2000).
8. Christie, M.A. and Blunt, M.J.: Tenth SPE Comparative
Solution Project: A Comparison of Upscaling
Techniques, paper SPE #66599 presented at the
Reservoir Simulation Symposium, Houston, TX
(February 2001).
9. Pollock, D.W.: Semianalytical Computation of Path
Lines for Finite-Difference Models, Ground Water,
(November-December 1988) 26(6), 743-750.
10. King, M.J., Blunt, M.J., Mansfield, M., and Christie,
M.A.: Rapid Evaluation of the Impact of Heterogeneity
on Miscible Gas Injection, paper SPE #26079 presented
at the Western Regional Meeting, Anchorage, AK (1993).
11. Verdiere S, Quettier L, Samier P, Thompson A.:
Applications of a parallel simulator to industrial test
cases paper SPE #51887 presented at the SPE Reservoir
Simulation Symposium, Houston, TX (1999).
12. Baker, R., Chug, S., and Batycky, R.P.: Full-Field
Modeling Using Streamline-Based Simulation: 4 Case
Studies paper SPE presented at the Reservoir Simulation
Symposium, Houston, TX (February 2001).
13. Grinestaff, G.H., and Caffery D.J.: Waterflood
Management: A Case Study of the Northwest Fault Block
Area of Prudhoe Bay, Alaska, Using Streamline
Simulation and Traditional Waterflood Analysis, paper
SPE #63152 in proceedings of the 2000 ATCE, Dallas, Tx
(October).
14. Chakravarty, A., Liu, D., and Meddaugh, W.:
Application of 3D Streamline Methodology in the
Saladin Reservoir and Other Studies, paper SPE #63154
in proceedings of the 2000 ATCE, Dallas, Tx (October).
SPE 66362
SPE 66362
11
12
SPE 66362
SPE 66362
13