Technology Program. US DOE Geothermal. Holland, Ole Kaven. Art McGarr, Austin. And

Transcription

1 Update to NRC Com mmittee on Induced Seismicity Associated with Energy Applic cations Transferring Science to Regulation: Needs and Issues Ernest L. Majer Lawrence Berkeley National Laboratory Nov 4, 2013

2 Acknowledgements US DOE Geothermal Technology Program And James Nelson; Ann Robertson-Tate; Jean Savy; Ivan Wong, Bill Foxall, Eric Son nnenthal, Tom Daley, Steve Hickman, Nick Davatzes, Ezra Zermat, Bill Ellsworth, Art McGarr, Austin Holland, Ole Kaven Plus many more who have participated in many Induced seismicity meetings and workshops and provided d results and data

3 Importance of Unde erstanding Induced Seismicity Technical ( reservoir management) One of few means to understand volumetric permeability enhancement/fluid paths Proper uses could optimize reservoir performance Policy/Regulatory ( hazard management)* Potential to side track projects important energy supply Technology must be put on a solid scientific basis to get public acceptance Accurate risk assessment must be done to advance energy projects

4 Recent Dev velopments *EGS CO2 sequestration Waste water injectionn Hydraulic Fracturing

5 Fig. 2 Cumulative count of earthquakes with M 3 in the central and eastern United States, The dashed line corresponds to the long-term rate of 21.2 earthquakes/year. Published by AAAS W L Ellsworth Science 2013;341:

6 Fig. 1 Earthquake and geothermal facility locations and activity.(a) Regional map with faults and location of the Salton Sea Geothermal Field. E E Brodsky, and L J Lajoie Science 2013;341: Published by AAAS

7 Fig. 1 Remote triggering in the midwestern United States, from the composite ANSS catalog.(a) Cataloged earthquakes above 3.0 M between 2003 and 2013 (ANSS). N J van der Elst et al. Science 2013;341: ublished by AAAS

8 Draft Best Practices for Addressing Induced Seismicity Associated With Enhanced Geothermal Systems (EGS) seismicity/egs/ By Ernie Majer, Lawrence Berkeley National Laboratory, Berkeley, CA James Nelson, Wilson Ihrig & Associates, Emeryville, CA Ann Robertson-Tait, GeothermEx, Inc., Richmond, CA Jean Savy, Savy Risk Consulting, Oakland, CA Ivan Wong, URS Corporation, Oakland, CA May 23, 2013

9 The 7 Main Steps 1) Perform a preliminary screening evaluation 2) Implement an outreach and communication program 3) Identify criteria for ground vibration and noise 4) *Seismic monitoring 5) **Quantify the hazard a from natural a and induced seismic events 6) **Characterize the ris sk from induced seismic events 7) Develop risk-based mitigation plans

10 Risk in this context can Earthqua ake Risk be thought of as: R = AF(a eq)*(pr(f a)* *C($;LL f) Where R= risk, AF= annual frequency of ground motion a, given occurrence of an earthquake(s), Pr(f a) ) =probability of failure of something of interest given ground motion a, and C=consequences (dollars, or any metric of interest). AF developed using Prob abilistic Seismic Hazard Analysis (PSHA)

11 There efore: Three main issues to address How does one assess risk? How o does one minimize risk? How does one determine acceptable risk?

12 Uncer rtainty (on the characteriz zation of inherent random mness) Assuming probability distributions that describe physically random proce esses are fully known The Hazard curve characterizes the randomness in the physical world of earthquake initiation, and propagation The vulnerability functions characterize the randomness in the failure process of buildings but they usually are not fully known, and insufficient data often leads to many viable alternative interpretation ns that represent our knowledge uncertainty about our probability estimates. Use experts interpretations 11/25/

13 EGS Site 1 Seismicity bef ore (2 years) Injection and during injection (11,3 363 M cubed) 12 days Before After

14 EGS Site 2: Seismicity be efore (3 years) Stimulation (11,320 M cubed ) and du ring stimulation ( 3 days) Historical seismicity Injection seismicity

15 EGS Site 3, Seismicity during injection 41,600 M cubed ( None before)

16 Newberry Volcano EGS Stimulation ( 41,666 M cubed) MEQs and Modeled Pressure and Thermal Tracer from 3-D Thermal-Hydro ological-chemical Model Pressure differential (left; P total -P hydrostatic > 0.05 MPa) and thermally-degrading tracer (right) plotted at 50 days (close to the end of stimulation). Yellow circle outlines a region having a 1 km radius from the wellhead (black circle at top). Cased interval is in black. Red symbols are relocated MEQs (Cladouhos et al., 2013).

17 Risk Ass sessment Traditional PSHA Semi empirical Physics based Adaptive

18 Elements of PSHA Earthquake size & location Frequency of occurrence Ground motion calculation faul t M max N(>m)/y yr n (g) Acceleratio Uncertainty Aleatory: Stochastic variability inherent in physical processes Epistemic: Lack of knowledge, e.g. in choice of alternative fault and crustal structure models or parameters Multiple realizations to sample uncertainty distributions Monte Carlo Magnitude Hazard calculation Distance Hazard curve

19 Difference betw een Natural and Induced Seismicity F-m distribution Tectonic historical eq. cat alog assumed Poissonian M min ~4.5 Depth >5 km Distance >5 km GM frequency Hz GM estimation generic empirical Induced no catalog prior to injection non-uniform in time and space km 1 5 km 1-50 Hz local, site-specific Standard ground motion prediction relations are very poorly constrained at short distances and small magnitudes

20 Some Current Models for Fail ure (IS) Traditional (Mohr Coulomb) Rate and change mo dels Volume Change Pressure Diffusion Thermal Stress chan ges Scaling law/unified

21 Role of Fluid Pressure in Earthquake Generation Normal (clamp ping) Stress = n For an earthquake to occur one must exceed the critical shear stress on the fault: c + ( n p) In situ Shear Stre ess Water/fluid pressure in fault = p Elevated Fluid Pressure: C = Rock strength = coefficient of friction on slip plane Reduces effective normal stress on fault, lowering resistance to shearing. Implies that if pres ssure balance can be maintained seismicity can be controlled

22 Seismic Moment & Injection Volume McGarr (1976) M K.. V 0 Total Seismic Moment Fluid injected K ~ 0.5 Volume added to region in expansion in direction NW/SE(2) and NE-SW (3)

23 Example of a semi-empirical method Shapiro et al. (2009) is partly empiri ical. Uses a triggering i front concept (Kaiser effect) and assumptions on stochastic micro-crack distribution to predict quantity of cracks and earthquakes. Only accounts for induced earthqua kes in the EGS production field (not on existing faults) 11/25/

24 PSHA/RA computational framework Flow model: P(x,t) Eq. simulation Ground motion Eq Catalog Eq. source params Ground motion stress & fault params Hazard curve Fragility Green s fns. Risk calculation SIMRISK Risk Curve The framework is currently simulation-based but can accommodate other methods for estimating frequency-magnitude statistics and ground motions, e.g. empirical

25 Hazard Assessment Challenges Modeling Which model to use? Continuum Modified continuum Discrete element Move past Coulomb criteria to rate and state? Which model to use One should "first" check the validity of exiting predictive models; both commercial as well as research types beforee embarking on developing new 'black boxes (function of available data) Inverse modeling seems to be an interesting alternative

26 Generate seismicity catalogs using RSQSim* modified to incorporate an evolving pressure field * Dieterich (1995); Richards-Dinger & Dieterich (2012) Fault/frac cs embedded in elastic half-space Initial stochastic distributions of fault properties Use slip-rate and state friction law to simulate complete earthquake cycle under constant tectonic shear loading rate stress accumulation eq. nucleation rupture healing and distributions evolve with successive events - initial stress distribution rapidly forgotten during burn-in period P(x,t) from flow model modifies effective normal stresss on each element Coseismic slip on one cell changes stress/stressing rate on all other cells (3D BEM) cascading failures large events Three fault element states: (0) Locked, (1) nucleating slip, (2) earthquake rupture cell state transition times predicted analytically no predetermined time-stepping Outputs full rupture history for direct input to ground motion calculations moment, stress drop and slip velocity time funct tion at each cell Convolve with synthetic Green s functions: FKRPROG (Saikia, 1994) 1D layered half-space; DC 25 Hz; pre-computed library

27 Hazard Manage ment Questions How close is close? (multiple interacting wells?) Is the local and regional structure understood? Proximity to existing faults? Proximity of structures/people? Extent of natural seismicity? Total energy release ( short time versus long)? In situ stress? Pore pressure? Structure? K? Injected volume? Local/regional structure,? Largest event? In situ stress? Pore pressure? Structure? K? Injected volume? Local/regional structure,? Will small earthquakes lead to bigger ones? In situ stress? Pore pressure? Proximity to existing faults? Structure? K? Injected volume? Can induced seismicity trigger earthquakes on distant faults? Local/regional structure? In situ stress? Pore pressure? Proximity to existing faults? K? Injected volume?

28 Hazard Manage ment Questions Can induced seismicity be controlled? Injected volumes? Pumping rates? Viscosity? In situ Stress? What controls are (will be) in place to mitigate t future induced d seismicity? Monitoring system? Injec ction plans / protocols? What is the plan if a large earthquake occurs? Injection Plans/Protocols? How do you calculate risk? PSHA, Physics Based, etc? How do you communicate risk to the public? Outreach plan?

29 Mitigation O pportunities Tools ( variation of, rate of change) Location of events, Energy release ( Moment and Mw) Source mechanisms Stress Drop b- value (temporal an nd spatial) Velocity and amplitude changes (fracture/fault monitoring) Fluid manipulation (temp, viscosity, etc.)

30 Needs Deploy Improved/advanced monitoring systems Improved spatial, bandwidth and temporal coverage continuous data-stream as basis for operational control decisions during development and long-term operation Couple seismic with chemical,temp and hydrologic data Characterization ti needs Stress ( Initial and pore pressure data) Injection data Fault and fracture properties Geologic (mechanical. chemical, fluid,) Fluid properties es (Injectate and native) Modeling for risk-based decision adapt probabilistic seismic hazard/risk method coupled with physics-based approach incorporating uncertainty

31 Conclusions Economic drivers are advancing the knowledge base on induced seismicity relative to fluid injection However, there are few accepted best practices models and procedures for regulating induced seismicity (several in progress) To move forward the scientific community needs to interface with regulatory agencies and transfer knowledge in an understandable fashio on The regulatory side needs to recognize that induced seismicity knowledge is stilll evolving and technical expertise will be needed to meet the scientific community half way To achieve successful implementation of many current and new energy technolog ies, in a practical time frame, further scientific studies need to performed at dedicated field labs that are not constrained by commercial requirements.