You are on page 1of 65

Applied Reservoir Simulation

Course
Workshop Problems
Reservoir Simulation Application Training

Schlumberger Public
Course
and (Eclipse) Workshop
GeoQuest Training and Development,
And NExT
Denver and Houston

March 06 Applied Reseervoir Simulation Day 1 1


Outline of Workshop Problems

Problem 1:
A. IMPES and Implicit Comparison
B. Time Truncation Tests

Schlumberger Public
C. Numerical Dispersion

March 06 Applied Reseervoir Simulation Day 1 2


Outline of Workshop Problems

2. Problem 2:
A. Water Coning Critical Coning Rate
B. Water Influx History Matching Kh

Schlumberger Public
C. Water Coning
I. History Matching Kv
II. Creation of Pseudo Krw in Coarse Grid to
Match Coning
D. Vertical Equilibrium Comparison

March 06 Applied Reseervoir Simulation Day 1 3


Outline of Workshop Problems

3. Problem 3: History Matching


Primary Production
A. Sensitivity Simulations

Schlumberger Public
B. History Match
C. Predictions

March 06 Applied Reseervoir Simulation Day 1 4


Schlumberger Public
Problem 1

March 06 Applied Reseervoir Simulation Day 1 5


Problem 1 Part A: IMPES Fully

Schlumberger Public
Implicit Comparison Coarse Grid
IMPES Sub-directory

March 06 Applied Reseervoir Simulation Day 1 6


Reservoir Details

We have a linear reservoir 5000 x 100


x 30 feet. Grid is 50 x 1 x 1
Water injector at one end and

Schlumberger Public
producer at other end
Two phase (oil-water) system
Oil viscosity = 2 cp and water
viscosity = 0.5 cp
Simulate 2000 days.

March 06 Applied Reseervoir Simulation Day 1 7


FloViz View

Schlumberger Public
March 06 Applied Reseervoir Simulation Day 1 8
Choice in Solution of the
Equations
IMPES Implicit Pressure Explicit
Saturation
Linear problem smaller

Schlumberger Public
Through put limitations: 5 10% PV
Stability problems
Timestep size - small

March 06 Applied Reseervoir Simulation Day 1 9


Choice in Solution of the
Equations
Fully Implicit
Unconditionally Stable large timesteps
Timesteps controlled by time truncation

Schlumberger Public
error
Few long timesteps

March 06 Applied Reseervoir Simulation Day 1 10


Comparison IMPES and Fully
Implicit Solution
Amount of Numerical Dispersion
(Error) in the results
Work to obtain solution

Schlumberger Public
Number of linear iterations
Number of non-linear (Newton)
iterations
CPU time per time step
Total CPU time

March 06 Applied Reseervoir Simulation Day 1 11


Iteration Process in Reservoir
Simulation

Schlumberger Public
Example of linear
and non-linear
iteration process:
4 non-linear
iterations Usually a non-linear iteration
requires 10 to 30 linears to
converge pressure and saturations

March 06 Applied Reseervoir Simulation Day 1 12


Data Set Names

IMPES.DATA uses IMPES solution

FULLIMP.DATA uses fully implicit

Schlumberger Public
solution
Location sub-directory:
Problem 1/IMPES

March 06 Applied Reseervoir Simulation Day 1 13


View Results: Comparison

Oil Production Rate


Production Water Cut
Map of Water Saturations

Schlumberger Public
At 500 Days
At 1000 Days
Number of Linear Iterations in Each
Timestep
Sum of Linear Iterations

March 06 Applied Reseervoir Simulation Day 1 14


View Results: Comparison

Newton (non-linear) Iterations in


each timestep
Sum of Newton Iteration

Schlumberger Public
Timestep Length
Note: Fully Implicit takes 31 timesteps
IMPES takes 207 timesteps
CPU Time Per Timestep
Total CPU Time

March 06 Applied Reseervoir Simulation Day 1 15


Your Task

View the .data sets


Difference IMPES (in Schedule
Section) Fully Implicit (E100 Default)

Schlumberger Public
Run the data sets with ECLIPSE 100
Plot the results
Study the results and answer the
following questions

March 06 Applied Reseervoir Simulation Day 1 16


Questions

Which technique takes more linear


iterations per timestep?
Which technique takes more Newton

Schlumberger Public
iterations per timestep?
Which technique requires more work
to solve the simulation?
Why?
We will discuss the results in class.

March 06 Applied Reseervoir Simulation Day 1 17


Problem 1 Part A: IMPES Fully

Schlumberger Public
Implicit Comparison Fine Grid
IMPES 2 Sub-directory

March 06 Applied Reseervoir Simulation Day 1 18


Reservoir Details

We have the same reservoir, fluids,


etc. except that now we will sub-
divide the grid into 2000 blocks in the

Schlumberger Public
x-direction instead of the previous
50.

The length of the grid blocks are now


2.5 feet instead of the previous
length of 100 feet.

March 06 Applied Reseervoir Simulation Day 1 19


Run and Compare Results

Run the IMPES.DATA and


FULLIMP.DATA in the IMPES2 sub-
directory and run Graf in the IMPES2

Schlumberger Public
sub-directory.

With these smaller grid blocks the


problem is numerically more difficult,
but the numerical dispersion is less.

March 06 Applied Reseervoir Simulation Day 1 20


Questions

Answer the same questions from the


previous part.
Note the differences in the IMPES

Schlumberger Public
and Fully Implicit runs.
Magnitude of the numerical dispersion in
the water cut, etc.
Instabilities in the IMPES run.
Number of linear and non-linear iterations
Sum of the linear and non-linear
Total CPU time for these cases.

March 06 Applied Reseervoir Simulation Day 1 21


Conclusion

For a blackoil simulation with


ECLIPSE we will prefer to use the
default solution Fully Implicit since

Schlumberger Public
the results are much more stable and
the CPU time is less.

March 06 Applied Reseervoir Simulation Day 1 22


Problem 1 Part B: Time Truncation

Schlumberger Public
Test: Quarter 5-Spot Water Flood

March 06 Applied Reseervoir Simulation Day 1 23


Time Truncation Error

We have the accumulation term in


our equations, for example
So

Schlumberger Public
t B o
When it is discretized we get terms
as follows S on+1 - S on
( x y z ) i


B w i t i

(1 - S oi ) P o in+1 - P o in
- ( x y z )i (1 )C f
B wi t

d (1/ B w )i P on+1 - P on P cow n+1 - P cow n


+ i (1 - S oi )
-
d p t t i

March 06 Applied Reseervoir Simulation Day 1 24


Time Truncation Error

This discretization process yields


and error that is

Schlumberger Public
O(t)
in the solution of the equations.
This error is called Time Truncation
Error

March 06 Applied Reseervoir Simulation Day 1 25


Simulation Situation

11x11x1 quarter 5-spot involving


water injection
Diagonal Grid

Schlumberger Public
1452 x 1452 x 50 meter flow field
K = 830 mD, = 0.17
Water Injector on water rate control
Producer on liquid rate control
Simulate 3660 days

March 06 Applied Reseervoir Simulation Day 1 26


Grid

Schlumberger Public
March 06 Applied Reseervoir Simulation Day 1 27
Time stepping
Three data sets are provided with
different time stepping regimes
maxts.data initial timestep = 10 days,
grows to max timestep = 183 days, 53

Schlumberger Public
total timesteps
smallts.data Initial timestep = 1 day,
max timestep = 10 days, 389 total
timesteps
mints.data Initial timestep = 1 day,
maximum timestep = 1 day, 3666 total
timesteps

March 06 Applied Reseervoir Simulation Day 1 28


Questions to Consider While
Viewing the Results
What is the effect of solving the flow
field with a few large timesteps?
What is the most accurate solution?

Schlumberger Public
Which solution requires the most
CPU time?
Given a balance between CPU time
and errors in the solution which
time stepping system would you
prefer?
March 06 Applied Reseervoir Simulation Day 1 29
Results to be Viewed with Graf

Timestep length
Water cut effect of time truncation
error

Schlumberger Public
Oil production rate effect of time
truncation error
Water saturation maps
CPU time per timestep
Total CPU time

March 06 Applied Reseervoir Simulation Day 1 30


Your Task

View the .data sets


Difference in the TUNING Keyword
timestep control

Schlumberger Public
Run the data sets with ECLIPSE 100
Plot the results
Study the results and answer the
following questions
We will discuss the results in class

March 06 Applied Reseervoir Simulation Day 1 31


My New Computer and ECLIPSE
2004a runs so fast has trouble

Schlumberger Public
storing timestep lengths

Following are results from old


Pentium III computer

March 06 Applied Reseervoir Simulation Day 1 32


CPU Time per Timestep in
Seconds

Schlumberger Public
March 06 Applied Reseervoir Simulation Day 1 33
Total CPU Time

Schlumberger Public
March 06 Applied Reseervoir Simulation Day 1 34
Problem 1 Part C: Numerical

Schlumberger Public
Dispersion: Radial Model with Gas
Coning

March 06 Applied Reseervoir Simulation Day 1 35


Numerical Dispersion

Definition: The error that occurs as


we discritize the partial differential
equations on a grid to create the

Schlumberger Public
finite difference equations.

March 06 Applied Reseervoir Simulation Day 1 36


Finite Difference Approximations
to First Derivatives
forward difference
f f (x + x) f (x)
=
x x

Schlumberger Public
Error term:
x f x f 2 2 3
.....
2! x 2
3! x 3
forward error

March 06 Applied Reseervoir Simulation Day 1 37


Objective

To see the effect of number of grid


block on a simulation model, a 2-D
radial (r, z) model was set up

Schlumberger Public
The reservoir has constant
properties; see data set RAD4.DATA
(rad4.data)

March 06 Applied Reseervoir Simulation Day 1 38


Description

Metric Units
Oil, Water, Gas, Dissolved Gas
The reservoir is 200 meters thick with

Schlumberger Public
the top 50 meters in the gas cap
Flat Top at 2950 meters
The width (radius) of the model is
500 meters

March 06 Applied Reseervoir Simulation Day 1 39


Description

The well is completed in the bottom


50 meters of the model

Schlumberger Public
The oil production rate is set at 6000
Sm3 / day

March 06 Applied Reseervoir Simulation Day 1 40


Description
The flow field was sub-divided into grids
that were 4x1x4, 8x1x8, 16x1x16, 32x1x32
and 64x1x64 blocks

Schlumberger Public
The well completions are located in the
proper layers so that the simulation
problems are identical except for the nr
and nz (number of grid blocks in the r and
z directions) values

March 06 Applied Reseervoir Simulation Day 1 41


Data Sets Provided
Five data sets are provided for you:

RAD4.DATA 4x1x4 grid

Schlumberger Public
RAD8.DATA 8x1x8 grid
RAD16.DATA 16x1x16 grid
RAD32.DATA 32x1x32 grid
RAD64.DATA 64x1x64 grid
128 x 1 x 128 grid takes half a day to
run so not provided
March 06 Applied Reseervoir Simulation Day 1 42
Creation of the Radial Grid
Spacing
In ECLIPSE radial grid

Have the option of letting simulator


calculate the radial grid spacing

Schlumberger Public
Required input: INRAD internal (well)
radius and OUTRAD outside radius and
NR number of grid blocks in the radial
direction

ECLIPSE will then calculate the radial grid


spacing with exponential growth
March 06 Applied Reseervoir Simulation Day 1 43
Equations for Calculating
Exponential Growth Grid Spacing
The outer cell radius for cells
logarithmically spaced can be
calculated by: ni

Schlumberger Public
ni re r nc
ri = rw exp ln ri = rw
e
n
c r
w OR
rw

ri = outer radius of cell i


rw = well radius
re = external radius
nc = number of cells
ni = number of cell i
March 06 Applied Reseervoir Simulation Day 1 44
Creation of the Radial Grid
Spacing
For our case:
INRAD = 0.22 meters
OUTRAD = 500 meters

Schlumberger Public
The r results for the 5 grids follow

March 06 Applied Reseervoir Simulation Day 1 45


r Values for NR = 4 Increasing
Exponentially

Next to the well

1.299 8.969 61.928 427.584

Schlumberger Public
Outer most grid block

March 06 Applied Reseervoir Simulation Day 1 46


r Values for NR = 8 Increasing
Exponentially
Next to the well

0.358 0.941 2.472 6.497 17.071

Schlumberger Public
44.857 117.868 309.716

Outer most grid block

March 06 Applied Reseervoir Simulation Day 1 47


r Values for NR = 16 Increasing
Exponentially
Next to the well

0.137 0.221 0.359 0.582

0.943 1.529 2.479 4.018

Schlumberger Public
6.513 10.558 17.114 27.742

44.971 72.897 118.167 191.549

Outer most grid block

March 06 Applied Reseervoir Simulation Day 1 48


r Values for NR = 32 Increasing
Exponentially
Next to the well

0.060 0.077 0.097 0.124 0.158 0.201 0.256 0.326

Schlumberger Public
0.415 0.528 0.673 0.856 1.090 1.388 1.768 2.250

2.865 3.648 4.645 5.913 7.529 9.586 12.204

15.538 19.783 25.187 32.068

40.829 51.983 66.184 84.265 107.285

Outer most grid block


March 06 Applied Reseervoir Simulation Day 1 49
r Values for NR = 64 Increasing
Next to the well Exponentially
0.0282 0.0319 0.0360 0.0406 0.0458 0.0516 0.0583 0.0658

0.0742 0.0837 0.0945 0.1066 0.1203 0.1357 0.1531

Schlumberger Public
0.1728 0.1950 0.2200 0.2482 0.2801 0.3161 0.3566 0.4024

0.4540 0.5123 0.5781 0.6523 0.7360 0.8305 0.9371

1.0574 1.1931 1.3462 1.5190 1.7140 1.9340 2.1822 2.4623

2.7784 3.1350 3.5374 3.9914 4.5037 5.0818 5.7341

6.4701 7.3006 8.2376 9.2950 10.4880 11.8342 13.3532 15.067

17.0012 19.1834 21.6457 24.4240 27.5590 31.0963 35.0877

39.5914 44.6732 50.4073 56.8773


Outer most grid block
March 06 Applied Reseervoir Simulation Day 1 50
z Values for the 4 Grids

nz = 4 z = 50 meters
nz = 8 z = 25 meters
nz = 16 z = 12.5 meters

Schlumberger Public
nz = 32 z = 6.25 meters
nz = 64 z = 3.125 meters

March 06 Applied Reseervoir Simulation Day 1 51


Grids for Radial Coning Problem

Schlumberger Public
March 06 Applied Reseervoir Simulation Day 1 52
64 x 1 x 64 Grid

Schlumberger Public
March 06 Applied Reseervoir Simulation Day 1 53
RAD4.DATA from FloViz

Schlumberger Public
March 06 Applied Reseervoir Simulation Day 1 54
RAD8.DATA from FloViz

Schlumberger Public
March 06 Applied Reseervoir Simulation Day 1 55
RAD16.DATA from FloViz

Schlumberger Public
March 06 Applied Reseervoir Simulation Day 1 56
RAD32.DATA from FloViz

Schlumberger Public
March 06 Applied Reseervoir Simulation Day 1 57
RAD64.DATA from FloViz

Schlumberger Public
March 06 Applied Reseervoir Simulation Day 1 58
Numerical Dispersion

Numerical Dispersion can occur in


both the r and z direction
By decreasing r and z the

Schlumberger Public
numerical dispersion from both
discretization errors will decrease

March 06 Applied Reseervoir Simulation Day 1 59


Your Task

View the 5 data sets with an editor


Run the 5 data sets with ECLIPSE
Plot the results

Schlumberger Public
Analyze the effect of numerical
dispersion on the Gas-Oil Ratio
We will discuss the results in class

March 06 Applied Reseervoir Simulation Day 1 60


Results with 128x1x128 Grid

Schlumberger Public
March 06 Applied Reseervoir Simulation Day 1 61
Question?

Which refinement is more important?


Refinement in the radial direction?
OR refinement in the vertical direction?

Schlumberger Public
For this gas coning situation?

March 06 Applied Reseervoir Simulation Day 1 62


To Answer the Question

We assume that the 64 x 1 x 64


numerical is close to the solution.

Schlumberger Public
We will run two cases:
radial-fine.data refined in the radial
direction only 64 x 1 x 8 grid
vertical-fine.data refined in the vertical
direction only 8 x 1 x 64 grid

March 06 Applied Reseervoir Simulation Day 1 63


Guess Which Refinement is More
Important
Before you run the 2 cases try to
estimate / guess which refinement
case will be closer to the 64 x 1 x 64

Schlumberger Public
very fine grid results.

Then run the cases, plot with fine.grf

Did you guess correctly?

March 06 Applied Reseervoir Simulation Day 1 64


Schlumberger Public
End of Workshop Problem 1

March 06 Applied Reseervoir Simulation Day 1 65

You might also like