Copyright © 2013, SAS Institute Inc. All rights reserved. RESERVOIR MANAGEMENT: WATER DRIVE OPTIMIZATION
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
RESERVOIR MANAGEMENT:
WATER DRIVE OPTIMIZATION
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
OPTIMIZE PRODUCTION WITH INTELLIGENT WELL MANAGEMENT
2. Explore
8. Development Strategy
9. Risk management
1. Sample 3. Uncertainty
5. Matching
6. Strategy Optimization
7. Probabilistic Strategy
4. Probabilistic Analysis
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
• Traditional DCA
• Probabilistic methodology
• Well Forecasting Solution • Bootstrapping module
• Clustering module
• Data mining workflow
FUNCTIONAL
DATA ANALYSIS
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
Paper 167428•Data Mining Methodologies enhance Probabilistic Well Forecasting•K.R.Holdaway
1. Cumulative liquid production
2. Cumulative oil or gas production
3. Water cut (Percentage determined
by water production/liquid
production)
4. B exponent (Decline type curve)
5. Initial rate of decline
6. Initial rate of production
7. Average liquid production
CLUSTER
ANALYSIS
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
BIG DATA BIG ANALYTICS APPROACH
• Water cut curves normalization based on Cumulative
Liquid produced.
• Analysis should take into account increments wise
regions.
(there are 3 increments)
• Analysis should be based on the different time phases
(time windows)
• Water distance (minimum FWL Distance) should be
considered as one of the inputs in the clustering
• Well bore type (horizontal or vertical) should also be
taken into account
Modeling Design Parameters: It has been decided that the analysis
should be designed according to the following criteria:
Normalization
Spatial
Distribution
Time Windows
Water Distance
Well Bore Type
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
BIG DATA BIG ANALYTICS APPROACH
NORMALIZATION OF THE WATER CUT CURVES
In order to make valid comparisons between well data sets it is necessary to perform a remediation
step that entails a robust quality control work process to identify outliers and impute for missing and
erroneous values.
Well 158_1
Well 1436_0
Normalizing WCT curves for
each well using Cumulative
Liquid produced in this well
Normalization
A Water Cut normalization step is implemented based on Cumulative Liquid production.
That approach helps eliminating temporal aspect of the data to some extent
We refer to the
normalized
scatter plots of
the Water Cuts
as the Water
Cut Marks of
the wells
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
BIG DATA BIG ANALYTICS APPROACH
TAKING INCREMENTS OF HRDH INTO ACCOUNT
Spatial
Distribution
Increment 1
Increment 2
Increment 3
Spatial location of the well will play an important role in the clustering analysis. We
assessed the location with respect to increments
The field is divided
into three
increments,
however increments
are correlated with
production time
windows.
We analyzed the
distributions of the
wells into the
increments in
different time
windows, and
consolidated into
one categorization
where the clustering
analysis should be
based on.
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
BIG DATA BIG ANALYTICS APPROACH
AVOID BIAS IN COMPARING DIFFERENT WELLS
• It has been suggested to use the above time windows for the clustering analysis.
• A separate segmentation model will be developed for each of the three time
windows.
• As note in the spatial distribution section, the well locations are depending on the
increments, and increments are depending on time. Therefore it turns out that by
considering only the time windows one will also incorporate the effect of the
increment information
Production Starting Time
Time window 1 Time window 2 Time window 3
∞ 1 Jan 96 1 Jan 96 1 Oct 2003 ∞
Time Windows
Production has started in different decades in different wells. There are wells that initiated oil
production since 1960s.Therefore production amounts are in different scales for the old and new
wells, and behaviors are different in the depletion, injection and post-injection phases
1 Oct 2003
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
BIG DATA BIG ANALYTICS APPROACH
DISTANCE TO WATER LEVEL IN A WELL
INFLUENCES WATER CUTS
Water Distance
The minimum distance to free water level is an important factor to understand water cut
behavior of a well. Considering the wells in the region the majority of the wells are distributed
within 4000-6600 interval (Mean: 5533 Median: 5388).
An interesting research
question would be to analyze
the different behaviors of the
wells having significant different
FWL Dist values
Distribution of the Minimum Distance to Free Water Level
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
BIG DATA BIG ANALYTICS APPROACH
HORIZONTAL AND VERTICAL WELLS BEHAVE DIFFERENTLY
Well Bore Type
Time Window Number of Horizontal
Wells
Number of Vertical Wells
Window1 (… - 1996)
- 42
Window2 (1996 – Oct.2003)
88 55
Window3 (Oct2003 - …)
138 9
Deviated Wells
1
5
2
Majority of the wells are horizontal and are analyzed in Time Window 3.
There are also a small number of wells having deviated configuration
Distribution of wellbore type for the wells
considered throughout the analysis
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
WELL CONFIGURATION (HORIZONTAL / VERTICAL)
Distribution of Well Configuration (Horizontal / Vertical) in the Clusters
As it can be seen the distribution is characteristic for some clusters.
Clusters are represented by their numbers
*Deviated Wells Excluded
All wells in this
cluster are vertical
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
RESULTS / GENERAL COMMENTS
PRESSURE TRENDS (TIME WINDOW 1)
In the early years of the
production both clusters are
following the overall trend
of the wells in the time
window.
After 2003 the annual SWP
average for the wells in
cluster 2 starts increasing
and surpasses that of
cluster 1 although the SWP
trend of both clusters is
positive (probably due to
injection)
Annual Average SWP for the two clusters of Time Window 1
0
500
1000
1500
2000
2500
3000
3500
19
57
19
60
19
63
19
66
19
69
19
72
19
75
19
78
19
81
19
84
19
87
19
90
19
93
19
96
19
99
20
02
20
05
20
08
20
11
SWP
Year
1
2
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
RESULTS / GENERAL COMMENTS
PRESSURE TRENDS (TIME WINDOW 2)
In the early years of the
production the clusters
are following the overall
trend of the wells in the
time window.
After 2003 and Injection
period the pressure is
measured differently for
the wells in cluster 4. That
cluster has the largest
pressure. It is almost
3500 on the average.
Annual Average SWP for the two clusters of Time Window 1
0
500
1000
1500
2000
2500
3000
3500
4000
1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008 2011
SWP
Year
1
2
3
4
Different Average
SWP trends based
on most recent
observations
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
RESULTS / GENERAL COMMENTS
PRESSURE TRENDS (TIME WINDOW 3)
Since the wells in this time
window are relatively new
and the production started
in 2003, the pressure is
following a steady trend
However, it can be noted
that first and second
clusters are in the declining
phase, whereas 3,4 and 5
are increasing.
Annual Average SWP for the two clusters of Time Window 1
0
500
1000
1500
2000
2500
3000
3500
2003 2004 2005 2006 2007 2008 2009 2010 2011
SWP
Year
1
2
3
4
5
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
OPTIMIZE PRODUCTION WITH INTELLIGENT WELL MANAGEMENT
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
OPTIMIZE PRODUCTION WITH INTELLIGENT WELL MANAGEMENT
Copyr i g ht © 2013, SAS Ins t i tu t e Inc . A l l r ights reser ve d .
Q&A
ENERGY ANALYTICS SUMMIT 2014