Best Practices: Numerical Modeling & Risk Assessment

Rajesh Pawar

Los Alamos National Laboratory

Grant Bromhal

National Energy Technology Laboratory

Best Practices Symposium for GCS

Taipei, Taiwan

October 31, 2014

Modeling Goals

• Predict CO2 injection and migration as well as reservoir response in support of:– Permit Application

– Site Operations (short-term)

– Risk assessment & management (short & long-term)

• Help design & deploy monitoring techniques and approaches

• Identify impact of uncertainties on project goals

Models

• Models and modeling approaches are dependent on questions to be answered as well as type and amount of available data

– Can range from simple to complex

– Analytical/semi-analytical/numerical

– Deterministic or probabilistic

• CO2 Sequestration applications require multi-scale, multi-physics models

– Type of questions to be answered

– Hydrologic, thermal, chemical (reactions), mechanical processes

– Focused on specific parts of a sequestration site or entire system

Models and Data

• Pre-characterization Phase:

– Limited data: site specific or analog

– More analytical and/or simplistic (homogeneous) reservoir simulations

– Simulations with single/multiple wells

– Area of review, Plume boundary, preliminary assessment of storage capacity, injectivity

– Risk screening (FEPs analysis)

– Higher uncertainty

Models and Data

• Characterization Phase:

– Increased data availability

– Site-specific data

– Complex models including reservoir simulations

– Define operational parameters

– Risk assessment and management

– Inform effective monitoring deployment

• Operational Phase:

– Site-specific observations for model validation/update

– Improved monitoring design

– Improve predictions for post-injection monitoring deployment and long-term risk management

Lab Reservoir Modeling Codes• Codes were mostly developed out of

nuclear program, “hot dry rock”

• Meant to handle single, two and three-phase flow in porous media, with heat transfer

• All can handle CO2 generally

• Each has certain things that they do differently/better than others, but essentially comparable capabilities

• All are available for no or limited fee, may require license agreement

• Some have recently added capabilities to handle oil and gas situations…

• NETL codes came out of oil and gas program; BOAST front ends developed; MASTER suitable for CO2 but no longer supported

Code Lab

TOUGH2 LBL

STOMP PNNL

FEHM LANL

NUFT LLNL

BOAST, MASTER

NETL

Industry (Service Company) Oil and Gas Codes

• Developed to handle most oil and gas situations

• Majors have their own codes

• In general, can handle CO2, as long as you purchase the right modules

• In general, much simpler user interface than any of the lab codes

• Very expensive, but often free to university users

Code Company

Eclipse Schlumberger

GEM (GHG) CMG

Landmark Halliburton

Industry (Service Company) Oil and Gas Codes

• Developed to handle most oil and gas situations

• Majors have their own codes

• In general, can handle CO2, as long as you purchase the right modules

• In general, much simpler user interface than any of the lab codes

• Very expensive, but often free to university users

Code Company

Eclipse Schlumberger

GEM (GHG) CMG

Landmark Halliburton

Specialty Codes: CBM, Shale

• Dual porosity codes that handle CBM and gas recovery from highly fractured media

• Built for coal, augmented recently for shale gas

Code Source

PSU-Coalcomp

Penn State, NETL

COMET3 ARI

Specialty Codes: Fractured Reservoirs

• Handle large numbers of fractures discretely (not dual porosity)

• FRACMAN built for hard rock environments (geothermal)

• NFFLOW suitable for seal leakage

Code Source

FRACGEN-NFFLOW

NETL

FRACMAN Golder

Specialty Codes: Geomechanics• Most couple flow and mechanics loosely

(no flow in GMI?), relatively new

• Generally much slower than just flow models

• Good at handling near wellbore mechanical issues

• Being used for specific purposes (e.g., to assess impact on seal integrity)

Code Source

GMI-SFIB Geomech Intl

ABAQUS SIMULIA

TOUGH-FLAC

LBNL (+FLAC)

FEHM LANL

COMSOL? COMSOL

Specialty Codes: Geochemistry

• Couple flow and chemistry, mostly single phase, but some handle multiphase

• Generally for smaller sections, but some are used on field scale, though are fairly slow

• The reactions that they handle are only as good as the databases that they use

• Kinetic data often sparse

Code Source

Geochemist’s Workbench

U Illinois

CrunchFlow LLNL

NUFT-C LLNL

PFLOTRAN, FEHM

LANL

TOUGHREACT LBNL

Reservoir simulation requires an appropriate geologic model

• Develop a geologic model• Reservoir tops and bottoms

• Depths

• Boundaries – no flow? Constant pressure? Include other layers?

g(

)

260 270 280 290 300 310 320 330 340

3890

3900

3910

3920

3930

3940

3950

3960

3970

3980

3990

4000

Greeley

Pond

New

Hop

e1 New

Hop

e2

PosoC

reek

KernFront

Kern Gorge

The geocellular model is generated from geologic data

• Reservoir model starts with geocellular model

– Determine how many simulation layers to use

– Define a grid, often higher res around important features

Easting (km)

No

rth

ing

(km

)

260 280 300 320 340

3890

3895

3900

3905

3910

3915

3920

3925

3930

3935

3940

3945

3950

3955

3960

3965

3970

3975

3980

3985

3990

3995

4000

New Hope Fault

Pond-Poso Fault

Greeley Fault

Rock and fluid properties are generated for each cell

• Permeability (3 directions)

• Porosity

• Relative permeability (usually constant)

• Fluid properties (black oil, EOS)

• Others (chemistry, stress, sorption, …)

Perform field performance history match, if data is available

0

1000

2000

3000

4000

5000

6000

7000

0 1000 2000 3000 4000 5000 6000 7000 8000

Production Rate (MCF/day)

Time (days)

HOWELL D 351

Measured

SS Model

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

0 1000 2000 3000 4000 5000 6000 7000 8000

Production Rate (MCF/day)

Time (days)

EPNG COM A 300

Measured

SS Model

Use the models to perform simulations to answer various questions

• How much CO2 can be stored? (capacity)

• How quickly can CO2 be injected and for how long? (injectivity)

• Where will the CO2 go? (Area of Review, permitting)

• Where will it end up over the long-term? (trapping, risk)

CO2 Saturation after 50 years Injectivity over 50 years

Model Updating

• Geologic/Reservoir models are typically updated as field operational/monitoring data becomes available

• Iterative process

• Updated models are used to redo forward predictions

Risk Assessment: Definition

• Identify and quantify potential risks due to subsurface injection of CO2

Risk = Pevent x Cevent

Probability Consequence

• For example: risk of shallow aquifer groundwater quality change due to leakage through wellbores

• Risk assessment can be part of site application, site operation design and long-term site safety plan

Types of stakeholder concerns

• Will there be sufficient capacity & injectivityover time?

• What will the impacts be from introducing CO2

into the reservoir?

• Will CO2 injection lead to earthquakes?

• What is the chance that all of the CO2 will remain in the reservoir?

• What might the impacts be of CO2 release from the reservoir?

injectio

n p

hase

po

st injectio

n

ph

ase

Risk Assessment: Approaches• Dependent on requirements: regulatory, operational,

safety• Qualitative:

• Probabilities and consequences are categorized in classes: High, Medium, Low

• Limited or no time-dependence• Risk ranking

• Semi-Quantitative/Quantitative:• Probabilities and consequences are computed• Deterministic or Probabilistic• Time dependent risk profiles• Fault-tree-analysis, event-tree-analysis, Systems

modeling• Approaches for GCS are being developed

Risk Assessment: Steps

Identify what’s at risk:Project Values, Risk Metrics

Identify failures (hazards) & impacts (consequences)

FEPs analysis, characterization data, analogs, expert elicitation

Identify scenarios: develop models

Compute probabilities of failure and consequences: risk ranking

Uncertainty analysis

Features, Events, Processes (FEPs)

• Factors that may influence safety of a storage reservoirs• Features: Parameters such as reservoir permeability,

caprock thickness• Events: Processes such as wellbore failure, earthquake• Processes: Physical or chemical processes such as

multiphase flow, reactions, geomechanical stress changes• Quintessa: Extensive FEPs database

Scenarios

• Possible future states of storage facility and events that result in leakage or other risks

• Scenario: Assemblage of FEPs• Need for rationalization of scenario: likelihoods,

evidence for or against, characterization data, analogs

Qualitative Risk Assessment: Risk Ranking

• Qualitative risk analysis results can be used for development of risk rankings:• Which risks are important to address• Develop corrective actions/mitigation plans• Site ranking, Permit applications

Example Risk Scales for Qualitative Risk Analysis

Ranking Factor

Severity of Negative Impact (S)

5 Catastrophic Multiple fatalities. Damage comparable to total project budget. Project shutdown. Useless results.

4 Serious One fatality. Damage = big chunk of project budget. Delay or lost operating time 1 year. Evacuation.

3 Significant Injury with permanent disability. Damage $100’s k. Lost operating time 1 month.

2 Moderate Injury with temporary disability. Damage $50k. Lost operating time 1 week. Regulatory notice.

1 Light Minor injury or illness. Damage $10k. Project lost time less than 1 week.

Ranking Factor

Likelihood of Impact or Failure Occurring (L)

5 Very Likely Happens every year, or more often. Nearly certain to happen during the project.

4 Likely Happens every few years. Probably will happen during the project.

3 Unlikely Happens every few decades. If we're lucky, it won't happen during the project.

2 Very Unlikely Would happen less often than every century, in projects similar to this one.

1Incredible or Impossible

If this projects like this went on forever, wouldn't happen in a thousand years.

24

Information from Ken Hnottavange-Telleen, Schlumberger Carbon Services

Sample Qualitative Risk Assessment ResultsRisk Matrix

0.5

1.5

2.5

3.5

4.5

5.5

0.5 1.5 2.5 3.5 4.5 5.5

Seve

rity

Likelihood

#01 #02

#03 #04

#05 #06

#07 #08

#09 #10

#11 #12

#13 #14

#15 #16

#17 #18

#19 #20

Information from Ken Hnottavange-Telleen, Schlumberger Carbon Services

Quantitative risk assessment requires simulation of CO2/brine movement through various parts of a sequestration site

Sequestration Reservoir

Wel

ls

Fau

lts

Groundwater Aquifer

Seals/Fractures

Injector

CO2/Brine Migration 3

CO2

Injection

Pressure Front1 CO2 Plume

2CO2/Brine MigrationIntermediate Zones

CO2/Brine migration through leakage pathways

2

Selecting modeling approach: Challenges

• Complexities:• 3-dimensional• Complex Geology: Heterogeneity, uncertainty • Complex processes for different parts (reservoir,

wells, faults, intermediate zones, shallow aquifer): different time-scales

• System parts are inter-dependent: Do we simulate the entire system together?

• Do we have computational abilities to do that?• How can we perform calculations in a probabilistic way?

to account for effects of uncertainties?

One way to simulate entire site is using an integrated assessment modeling approach

A.Divide system intodiscrete components

B.Develop detailed component models that are validated against lab/field data

D. Link ROMs via integrated assessment models (IAMs) to predict system performance & risk; calibrate using lab/field data from NRAP and other sources

C. Develop reduced-order models (ROMs) that rapidly reproduce component model predictions

E. Develop strategic monitoring protocols that allow verification of predicted system performance

Large number of calculations need computationally efficient models for predicting behavior of individual parts

ROMs are developed and tested through a comparison to validated simulators.

Detailed simulationsMultiple simulations of detailed component models (reservoir, wellbores, faults, aquifer)

Sensitivity analysis

Identify key variables that control component behavior

Develop ROMs• Look-up tables (LUTs)• Response surfaces (e.g. via PSUADE)• Artificial intelligence approaches • Simplified relationships (e.g., PCE)

Validate ROMs against simulations

ROMs are not simple models but models that capture complex processes and are computationally efficient

Risk calculations using the integrated system model

Utilize aquifer abstraction to estimate change in pH and TDS given chosen set of aquifer parameters and estimated CO2/brine leak rates

Utilize reservoir abstractions to estimate time-dependent saturation and pressure in sequestration reservoir given the chosen set of reservoir and injection parameters

Utilize leakage abstractions to estimate CO2 and brine leak rate given chosen set of parameters and estimated time/space-dependent pressure and CO2

saturation in sequestration reservoir

Multiple Realizations

Each Realization

Sample uncertain parameters (reservoir, wells, faults, seals, shallow/intermediate aquifers)

Each Realization: Dynamic simulation over 100s-1000s yr, run linked models simultaneously

Uncertainties

• Geologic media are inherently heterogeneous• Limited opportunities to collect data• Knowledge of parameters/processes controlling overall

behavior is incomplete: multiple sources of uncertainties• Need to understand/characterize which uncertain

parameters affect overall risks• Monte-Carlo simulations• Parameter sensitivity/importance

• Uncertainty analysis can be used to develop further data collection strategies to reduce uncertainties

Risk Assessment

Risk Management

• Risk management: manage/mitigate/reduce risks• Includes risk assessment, monitoring, risk mitigation

MonitoringMitigation

• Risk management is an inherent part of field projects• Qualitative risk assessment/risk rankings used to develop

monitoring & mitigation strategies

• Approaches to incorporate monitoring/mitigation in QRA need to be developed

• Mitigation strategies for CO2 storage operations need to be developed