1 07 April 2010 Denver, CO OPTIMIZING DEVELOPMENT STRATEGIES TO INCREASE RESERVES IN UNCONVENTIONAL GAS RESERVOIRS Presented to: RPSEA Unconventional Gas Program Prepared by: Duane McVay, Gulcan Turkarslan, and Rubiel Ortiz, Texas A&M University J. Eric Bickel and Luis Montiel, The University of Texas at Austin
40
Embed
07 April 2010 Denver, CO OPTIMIZING DEVELOPMENT STRATEGIES TO INCREASE RESERVES IN UNCONVENTIONAL GAS RESERVOIRS Presented to: RPSEA Unconventional Gas.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
07 April 2010
Denver, CO
OPTIMIZING DEVELOPMENT STRATEGIES TO INCREASE RESERVES IN
UNCONVENTIONAL GAS RESERVOIRS
Presented to:RPSEA
Unconventional Gas Program
Prepared by:Duane McVay, Gulcan Turkarslan, and Rubiel Ortiz, Texas A&M University
J. Eric Bickel and Luis Montiel, The University of Texas at Austin
2
Agenda
Background and Motivation
Reservoir Model
Decline Curve Model
Decision Model
3
Project Objectives
Objectives-- Develop new technologies for determining optimal dynamic development strategies and testing programs in gas shale and tight sand reservoirs.
-- Understand the tradeoff between Stage 1 spacing and duration.
Application
University and industry partners to determine optimal well spacing and completion methods in the Barnett Shale, Parker County, Texas, and the Gething tight gas formation in Alberta, Canada.
Impact: Technology incorporation into operators’ development processes will enable reaching optimal spacing as quickly as possible, accelerating production and increasing reserves.
4
Our goal is to integrate a reservoir and a decision model.
+640
320
160
80
InitialSpacing
ProductionResults
640
320
160
80
FirstDownspacing
ProductionResults
…
640
320
160
80
FinalDownspacing
ProductionResults
CompletionMethod
CompletionMethod
CompletionMethod…
Decision Uncertainty
640
320
160
80
640
320
160
80
InitialSpacing
ProductionResults
640
320
160
80
FirstDownspacing
ProductionResults
…
640
320
160
80
FinalDownspacing
ProductionResults
CompletionMethod
CompletionMethod
CompletionMethod…
Decision Uncertainty
The reservoir model will be based on statistical or reservoir simulation techniques.
The decision model will employ decision tress or dynamic programming to determine the optimal development program.
The models will incorporate uncertainty in reservoir parameters.
5
Agenda
Background and Motivation
Reservoir Model
Decline Curve Model
Decision Model
6
Our initial work is based on the Berland River Formation.
The study area encompasses 650 km2 and 46 existing wells.
7
We have constructed a stochastic reservoir modeling tool.
Single-well, 1-layer, single-phase reservoir model used
A stochastic modeling tool, @Risk, coupled to CMG IMEX
Generated VBA code in Excel to create the simulation data files and run the simulator
Perform Monte Carlo combined with reservoir simulation
8
We consider drilling either 1, 2, 4 and 8 wells in a 640 acre section.
Drill up to 8 wells in a section
640 (1 well/section)
320 (2 wells/section)
160 (4 wells/section)
80 acres (8 wells/section)
9
We consider uncertainty in net pay, porosity, formation depth, and drainage area.
Uncertain Parameters & Distributions
Other input parameters
10
We correlate permeability and porosity.
0.01
0.1
1
10
0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12
Gas Porosity, %
Per
mea
bil
ity,
md
Probabilistic model Correlation model (Deterministic) Expon. (Probabilistic model)
This analysis provides several insights regarding learning.
Correlation in qi increases with:
Decreasing stage 1 spacing
Decreasing stage 2 spacing
Increasing stage 1 length
Our ability to learn about the initial decline rate for Package 2 is limited.
Our ability to learn about b is muted until we get down to about 160 ac for stage 1.
19
Agenda
Background and Motivation
Reservoir Model
Decline Curve Model
Decision Model
20
Our initial decision model allows for one stage of downspacing.
Key
Decision
Uncertainty
Calculation
Evocative
Value
Influence
Package 1(Stage 1)
Package 2(Stage 2)
qi (Pack 1)
D (Pack 1)
b (Pack 1)
qi (Pack 2)
b (Pack 2)
D (Pack 2)
NPV ($)
This is an “option” analysis. We make the Stage 2 decision after learning about Stage 1.
21
We make the following economic assumptions for illustrative purposes:
Gas Price $/MCF 5.50MC $/MCF 1.00FC MM $/yr/well 0.05Drilling Cost MM $/well 1.00Discout Rate 0.10
22
If Stage 1 is 1 yr in length then the optimal strategy is 640 ac to start and then downspace to 160 ac if high production is observed.
Stage 1 Prob qi 640 320 160 80 Opt D Value640 0.25 H $4.2 $3.4 $4.9 $0.1 160 $4.9
2.09$ 0.5 M $1.5 $0.7 $1.4 ($3.6) 640 $1.50.25 L $0.4 ($0.5) ($0.3) ($5.6) 640 $0.4
Stage 2
Stage 1 Prob qi 640 320 160 80 Opt D Value320 0.25 H $3.0 $4.5 ($0.4) 160 $4.5
1.72$ 0.5 M $0.6 $1.4 ($3.7) 160 $1.40.25 L ($0.5) ($0.3) ($5.6) 160 ($0.3)
Stage 2
Stage 1 Prob qi 640 320 160 80 Opt D Value160 0.25 H $3.7 ($1.1) 160 $3.7
1.43$ 0.5 M $1.2 ($3.9) 160 $1.20.25 L ($0.4) ($5.6) 160 ($0.4)
Stage 2
Stage 1 Prob qi 640 320 160 80 Opt D Value80 0.25 H ($0.3) 80 ($0.3)($3.27) 0.5 M ($3.6) 80 ($3.6)
0.25 L ($5.5) 80 ($5.5)
Stage 2
23
Notice that the dynamic strategy is about $660K better than a static 160 ac strategy.
Stage 1 Prob qi 640 320 160 80 Opt D Value640 0.25 H $4.2 $3.4 $4.9 $0.1 160 $4.9
2.09$ 0.5 M $1.5 $0.7 $1.4 ($3.6) 640 $1.50.25 L $0.4 ($0.5) ($0.3) ($5.6) 640 $0.4
Stage 2
Stage 1 Prob qi 640 320 160 80 Opt D Value160 0.25 H $3.7 ($1.1) 160 $3.7
1.43$ 0.5 M $1.2 ($3.9) 160 $1.20.25 L ($0.4) ($5.6) 160 ($0.4)
Stage 2
$2.09 MM - $1.43 MM = $0.66 MM
24
A three-year Stage 1 is worse than a 1-year Stage 1, but the dynamic strategy is still better.
Stage 1 Prob qi 640 320 160 80 Opt D Value640 0.25 H $1.4 $1.4 $3.4 ($1.0) 160 $3.4
$0.72 0.5 M $0.0 ($0.4) ($0.0) ($4.8) 640 $0.00.25 L ($0.6) ($1.4) ($2.1) ($7.2) 640 ($0.6)
Stage 2
Stage 1 Prob qi 640 320 160 80 Opt D Value320 0.25 H $1.4 $3.4 ($0.9) 160 $3.4
$0.48 0.5 M ($0.4) ($0.0) ($4.8) 160 ($0.0)0.25 L ($1.4) ($2.1) ($7.2) 320 ($1.4)
Stage 2
Stage 1 Prob qi 640 320 160 80 Opt D Value160 0.25 H $4.0 ($0.3) 160 $4.0
$0.63 0.5 M $0.2 ($4.6) 160 $0.20.25 L ($2.0) ($7.0) 160 ($2.0)
Stage 2
Stage 1 Prob qi 640 320 160 80 Opt D Value80 0.25 H ($0.6) 80 ($0.6)($4.26) 0.5 M ($4.7) 80 ($4.7)
0.25 L ($7.1) 80 ($7.1)
Stage 2
25
Let’s look at a higher price environment.
Gas Price $/MCF 9.00MC $/MCF 1.00FC MM $/yr/well 0.05Drilling Cost MM $/well 2.00Discout Rate 0.10
26
With a 1-year stage 1, 640s are best to start and we downspace to 160s unless we see low production.
Stage 1 Prob qi 640 320 160 80 Opt D Value640 0.25 H $7.6 $6.4 $9.4 $1.3 160 $9.4
$4.14 0.5 M $2.8 $1.5 $3.2 ($5.2) 160 $3.20.25 L $0.8 ($0.6) $0.0 ($8.7) 640 $0.8
Stage 2
Stage 1 Prob qi 640 320 160 80 Opt D Value320 0.25 H $5.6 $8.6 $0.5 160 $8.6
$3.66 0.5 M $1.4 $3.0 ($5.4) 160 $3.00.25 L ($0.6) $0.0 ($8.7) 160 $0.0
Stage 2
Stage 1 Prob qi 640 320 160 80 Opt D Value160 0.25 H $7.3 ($0.8) 160 $7.3
$3.14 0.5 M $2.7 ($5.7) 160 $2.70.25 L ($0.1) ($8.7) 160 ($0.1)
Stage 2
Stage 1 Prob qi 640 320 160 80 Opt D Value80 0.25 H $0.6 80 $0.6($4.60) 0.5 M ($5.2) 80 ($5.2)
0.25 L ($8.6) 80 ($8.6)
Stage 2
27
However, under a 3-year Stage 1 we should start with 160s and not downspace.
Stage 1 Prob qi 640 320 160 80 Opt D Value640 0.25 H $2.6 $2.7 $6.6 ($0.5) 160 $6.6
$1.67 0.5 M $0.2 ($0.5) $0.5 ($7.4) 160 $0.50.25 L ($1.0) ($2.2) ($3.1) ($11.5) 640 ($1.0)
Stage 2
Stage 1 Prob qi 640 320 160 80 Opt D Value320 0.25 H $2.8 $6.7 ($0.4) 160 $6.7
$1.38 0.5 M ($0.5) $0.5 ($7.4) 160 $0.50.25 L ($2.2) ($3.1) ($11.5) 320 ($2.2)
Stage 2
Stage 1 Prob qi 640 320 160 80 Opt D Value160 0.25 H $7.8 $0.7 160 $7.8
$1.72 0.5 M $1.0 ($6.9) 160 $1.00.25 L ($2.9) ($11.3) 160 ($2.9)
Stage 2
Stage 1 Prob qi 640 320 160 80 Opt D Value80 0.25 H $0.2 80 $0.2($6.35) 0.5 M ($7.1) 80 ($7.1)
0.25 L ($11.4) 80 ($11.4)
Stage 2
28
We are able to integrate the reservoir model and the decision model via the use of decline curves.
The reservoir / decline curve studies show that we are able to learn about reservoir parameters.
Initial spacing and stage lengths are substitutes to some degree.
We believe we can learn about each decline curve parameter, but learning about initial production rate appears strongest.
Optimal dynamic development strategies may be worth millions more than static strategies.
In conclusion…
29
Future Plans
Refine reservoir model.
Expand decision model to allow for learning about all decline curve parameters.
Incorporate completion efficiency into the reservoir/decision modelling.
Construct an areal, multi-well model of the same reservoir to better model heterogeneity and interference between wells.
Incorporate pilot down spacing into the reservoir/decision modeling to determine the optimum number and length of pilots.
30
07 April 2010
Denver, CO
OPTIMIZING DEVELOPMENT STRATEGIES TO INCREASE RESERVES IN
UNCONVENTIONAL GAS RESERVOIRS
Presented to:RPSEA
Unconventional Gas Program
Prepared by:Duane McVay, Gulcan Turkarslan, and Rubiel Ortiz, Texas A&M University
J. Eric Bickel and Luis Montiel, The University of Texas at Austin
31
Deliverables
2/28/10 Preliminary integrated reservoir and decision making models based on synthetic data.
8/17/10 Extend model and apply to UGR test reservoir
5/17/11 Extend model to second test reservoir
8/17/11 Final report
32
Backup
33
Background
High risk associated with low-permeability gas sands Complex heterogeneities Variable reservoir properties Uncertain completion and stimulation efficiency
Sound development decisions needed Efficient completion practices Optimal well spacing
A trade-off between conserving capital and protecting the environment to avoid over drilling, but maximizing production by quickly achieving the optimum well spacing
34
Background
Determination of optimum development strategy in tight gas sands is technically challenging.
Large variability in rock quality
Wide range of depositional environments
Large number of wells
Limited reservoir information
Time & budget constraints
35
Background
Traditional methods for determining optimal well spacing
Statistical comparison of the performance of wells drilled at different spacings
Applicable only when sufficient production data from multiple infill programs are available
In emerging plays, historical infill programs are not available to evaluate optimal spacing with traditional methods.
36
Permeability Model
Permeability model based on a porosity-permeability correlation
A scalar factor of 1.92 added to calibrate the existing simulation model to the well performance data
Uncertainty incorporated into the permeability model by attributing a normal distribution to the porosity-permeability correlation.