optimization of fluid injection and extraction for ... · max buildup along a fault . no co 2 flow...
TRANSCRIPT
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
September 27, 2016 ARB’s Carbon Capture and Sequestration (CCS) Program
CCS Technical Discussion Series: Health and Environmental Risks, and Environmental Justice
Abdullah Cihan and Jens Birkholzer
Energy Geosciences Division, Lawrence Berkeley National Laboratory
Optimization of Fluid Injection and Extraction for Reservoir Management and Risk Mitigation during CO2 Sequestration
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Reduced Storage Capacity
Interference Between Storage Sites
Pressure Buildup and Brine Displacement
Caprock Damage
Permitting and AoR
Effect on Other Georesources
Reservoir Management Via Brine Extraction
Induced Seismicity
Reservoir Integrity Issues Related to Industrial-Scale CO2 Injection
Beneficial Use of Extracted Brine
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
• Objective − Manage reservoir pressure with optimum extraction volumes and infrastructure needs
• Approach − Define specific reservoir performance criteria (e.g., maximum pressure near fault
zone, maximum caprock pressure, minimum AoR size, maximum CO2 trapping) − Via smart optimization algorithms, automatically find most suitable CO2 injection and
brine extraction solutions that meet performance criteria and constraints − Optimization searches for optimum number of wells, well locations, and rates
• Optimization − Optimization is based on a suite of gradient-based and global algorithms. − Global algorithms are preferred for coupled optimal well placement and control
problems − Must be conducted both during the planning stage and during the operation stage for
‘real-time’ management
Optimized Reservoir Management
BIRKHOLZER, J.T., CIHAN, A., ZHOU, Q. (2012): Impact-Driven Pressure Management Via Targeted Brine Extraction – Conceptual Studies of CO2 Storage in Saline Formations, International Journal of Greenhouse Gas Control, 7, pp. 168-180
CIHAN, A., BIRKHOLZER, J.T., BIANCHI, M. (2015): Optimal well placement and brine extraction for pressure management during CO2 sequestration, International Journal of Greenhouse Gas Control, 42, pp. 175-187
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Example Description of Constrained Global Optimization Problem
4
Minimize
Subject to
A general constrained optimization problem for pressure control using brine extraction can be formally expressed as
Extraction Ratio (if no brine extraction, f is chosen as 1/Vinj)
Pressure buildup constraints, e.g. max buildup along a fault
No CO2 flow into extraction wells (for brine extraction schemes)
Range of Cartesian coordinates for vertical wells (for partially penetrating and horizontal wells, additional parameters can be included)
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Example Description of Constrained Global Optimization Problem
5
Minimize
Subject to
A general constrained optimization problem for pressure control using brine extraction can be formally expressed as
Extraction Ratio (if no brine extraction, f is chosen as 1/Vinj)
Pressure buildup constraints, e.g. max buildup along a fault
No CO2 flow into extraction wells (for brine extraction schemes)
Range of Cartesian coordinates for vertical wells (for partially penetrating and horizontal wells, additional parameters can be included)
Other or Additional Objective Functions: • Minimize well number • Minimize size of AoR • Maximize CO2 trapping • Maximize EOR efficiency
Other or Additional Constraints: • Bottom hole pressure • Leakage rate in abandoned well field • TDS of extracted brine
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Design and Optimization of Injection, Extraction, and Monitoring Using a Suite of Forward Models
• Recently developed a constrained differential evolution (CDE) algorithm was coupled with a suite of forward models for solving global optimization problems involving well placement, injection and brine extraction.
• In addition, other local and global optimization algorithms built in Itough2 are available to use, depending on the nature of the optimization problems.
Forward Predictors
(1) Analytical Solutions
(2) Vertically Integrated Numerical Two-Phase Flow Models – Sharp-Interface Models – Vertically Integrated
Multi-Phase Models
(3) Multi-phase flow in full 3D systems
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Application to Hypothetical CO2 Injection in the Vedder Formation (Southern San Joaquin Valley,
California)
Sier
ra N
evad
a
Southern San Joaquin Basin
g(
)
260 270 280 290 300 310 320 330 340
3890
3900
3910
3920
3930
3940
3950
3960
3970
3980
3990
4000
GreeleyPond
New
Hope
1 New
Hop
e2
Poso Creek
Kern Front
Kern Gorge
km
km
Top of Vedder Elevation
Major faults not to experience pressure buildup above 1 MPa
Injection Target: 5 Mt CO2 per Year Over 50 Years
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Injection of 5 Mt CO2 per Year Over 50 Years Without Pressure Management
Pressure-buildup contour lines (black) in MPa and CO2 plume extent (red) CO2 Saturation
Results for Realization 1
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Optimal Well Placement and Pressure-Buildup
Maximum pressure buildup w/ and w/o brine extraction Results for Realization 1
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
R
Optimal Injection Rates without Brine Extraction for ΔPmax=1MPa along
the Faults
Optimal Extraction Ratios for 5Mt/year injection and ΔPmax=1MPa along the
Faults 1 1.968 Mt/yr 0.672 2 2.739 Mt/yr 0.393 3 3.229 Mt/yr 0.380 4 2.061 Mt/yr 0.654 5 2.451 Mt/yr 0.538
Optimized Extraction Rates for all Realizations
Realization-1 Lower permeability zones between the faults result in stronger pressure buildups and thus higher volume of brine extraction
Realization-3 Higher permeability zones exist between the faults. One extraction well achieves the performance goal
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Heterogeneity plays a significant role for optimal well placement and injection/extraction rate control, which points to:
• Importance of site characterization
• Use of adaptive management strategies combining monitoring+inversion+optimization in an integrated framework
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Optimization Framework and Integrated Management
Objectives • Develop and apply an adaptive optimization framework for CO2 storage projects which utilizes
advanced automated optimization algorithms and suitable process models for coupled multi-phase behavior,
• Use this new framework to investigate process behavior and then develop optimal schemes for key technical challenges in CO2 sequestration
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Aquifer 1
Aquitard
Q1(t)
Aquifer 2
Injection well
Obs. well
An Example Application of Adaptive Approach to Project Control and Optimization
Parameters assumed to be Unknown
Actual Values
K1 (m/d) 1.015E-01 S1 (Specific Storage Coef, 1/m) 1.693E-06
Horizontal Anisotropy Ratio 2.000E+00
K2 (m/d) 1.037E-01 S2 (1/m) 1.704E-06 K’ (aquitard) 1.028E-06 S’ (aquitard) 1.988E-06
Objective: • Critical pressure buildup is assumed
to be 0.2 MPa along the fault • Fluid injection rate is 1Mt/yr over 30
yrs • A adaptive optimization algorithm
involving optimization +monitoring + model update is applied to manage the system, minimizing extraction rates while attempting to reduce risk of fault slippage.
Extraction Well
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
1. Optimization With Existing Model(s)
2. Monitoring/Model
Testing
3. Model Calibration
Input (Planning
Stage)
High pressure in violation of constraint indicates: • Importance of site
characterization at the planning stage
• Monitoring and model update need to be done at a higher temporal frequency at earlier times to detect higher pressures and change extraction rates accordingly
For Observed vs. Predicted at Step 2
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
1. Optimization With Existing Model(s)
2. Monitoring/Model
Testing
3. Model Calibration
Input (Planning
Stage)
Constraint violation eliminated by higher-frequency monitoring and model update at earlier times
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
• Developed an optimization method that minimizes brine extraction volumes while meeting defined reservoir management targets
• Applied optimization method to realistic complex problems involving pressure control near faults and other management objectives
• Optimization allows for significant reduction in brine extraction volumes (extraction ratios < 1)
• Heterogeneity plays a significant role for optimal well placement and injection/extraction rate control for minimizing risks
Summary
• Investigation and development of adaptive management strategies for pressure control and mitigation
• Optimization and risk management under uncertainty conditions
Ongoing Work
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
BACKUP SLIDES
17
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
There is a need for robust and efficient optimization strategies for large-scale pressure
management problems
• Selections of well locations and injection/extraction rates are critical for maximizing storage and minimizing brine extraction.
• Objective functions can have multiple local optima in the solution space.
• Global optimization methods are needed.
Global minimum
f
xi
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Brine extraction can be a significant factor in economic viability of a CCS project
Based on Harto et al., 10th Annual Conference on CCS, Pittsburgh, May 2011
• Construction of wells and treatment/disposal facilities • Operating cost (pumping, treatment, disposal, discharge) • Transportation • Permitting, monitoring, reporting
• Optimization can minimize extraction volumes and beneficial use can ease economic burden
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Constrained Differential Evolution Algorithm
Implementation of Constraints in CDE Objective function evaluations in the original DE is reformulated in the CDE based on the following criteria:
1. Any feasible (satisfying constraints) solution is preferred to any infeasible solution 2. Among feasible solutions, the one having the better objective function value is preferred 3. Among the infeasible solutions, the one having smaller constraint violation is preferred
Main Stages of Differential Evolution Algorithm (Storn and Price, 1997)
Initialization Selection Mutation Crossover
• A constrained differential evolution (CDE) algorithm was developed for solving optimization problems involving well placement, injection, and brine extraction.
• The DE algorithm is a very powerful evolutionary algorithm with good convergence properties, ease of use and suitability for parallelization.
• The original DE method is in general applicable to unconstrained optimization problems.
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Demonstration and Testing of Constrained Differential Evolution (CDE) Method
Impact Point (x=0, y=-20km) (Critical ΔP=0.2MPa)
Injection Well (x=0, y=0)
Extraction Well
Impact Point (x=20 km, y=0) (Critical ΔP=0.2MPa)
x
-y
Optimal Location
(10 km, -10 km)
A Simple Optimization Problem with Known Optimal Extraction Well Location
Norm
alized Extraction R
ate N
ormalized
Extraction Rate
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Comparison with Results from Gradient-Based Method for Extraction Rates
SQP: Sequential Quadratic Programming
• CDE finds optimal well location • CDE and SQP optimum extraction rates agree well • CDE needs much larger number of forward runs, but easily parallelizable
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Example application case - using different models for estimating optimal injection rates
Two-phase Flow in 3D Two-phase Vertical-Equilibrium (VE)
(Saturation distribution at the top layer, vertical
resolution~5m, average thickness~100m)
Model Type
Optimal Injection Rates without Brine Extraction for
ΔPmax=1MPa along the Faults
Two-phase 3D 0.051 Mt/yr Two-phase VE 0.048 Mt/yr
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
Effects of Critical Pressure and Fault Permeability on ‘Extraction Ratio’
Critical ∆P (bar)
Extr
actio
nR
atio
(m3 /m
3 )
2 4 6 8 100
0.2
0.4
0.6
0.8
1
kfault/kreserv=0kfault/kreserv=10-4
kfault/kreserv=10-2
Extraction Ratio: Total Volume of Extracted Brine Divided by Injected CO2
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
A Generic Optimization Problem for CO2 Injection and Brine Extraction
Model Easting (m)40000 60000 80000
Facies
4321
CO2 InjectionWellFault-1
Model Easting (m)
Mod
elN
orth
ing
(m)
40000 60000 80000
10000
20000
30000
40000
50000
54.543.532.521.510.5
CO2 InjectionWell
∆P (MPa)
• Injection target: 5 Mt CO2 per year over 30 years • Threshold pressure to prevent caprock fracturing: 8 MPa • Threshold pressure to prevent fault activation: 0.55 MPa • Optimization goal: Inject target amount of CO2 while honoring pressure constraints, with
minimum number of injection/extraction wells and minimum brine extraction volume
Pressure contours for single-well injection of 1.5 Mt CO2 at 30 years (both pressure constraints are violated)
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
5
5
55
5
5
5
10
10
10
10
10
20
20
2030
30
30
40
4050
70
Model Easting (m)
Mod
elNo
rthin
g(m
)
40000 60000 80000
10000
20000
30000
40000
500000.40.350.30.250.20.150.1
Ave SCO2
Injection Zone
Extraction Wells
A Generic Optimization Problem for CO2 Injection and Brine Extraction
Optimization Results Involving Four Injection Wells and Two Extraction Wells with Time-Dependent Extraction Rate
EARTH AND ENVIRONMENTAL SCIENCES • LAWRENCE BERKELEY NATIONAL LABORATORY
A Generic Optimization Problem for CO2 Injection and Brine Extraction
Time (yr)
Opt
imiz
edPu
mpi
ngR
ates
(m3 /s
)
0 10 20 30 400
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
Injection
Extraction
Time (yr)
Max
imum
Pres
sure
Bui
ldup
(bar
)
0 10 20 300
20
40
60
80
100
Injection ZoneFault Zone
Threshold for Fracking
Threshold for Fault Activation
Extraction Ratio: 22%