POTENTIAL USE OF RISK INFORMATION IN THE FUKUSHIMA DAIICHI NUCLEAR POWER PLANT DECOMMISSIONING

Transcription

1 presented at Tokyo Institute of Technology Tokyo Japan 22 January 2016 POTENTIAL USE OF RISK INFORMATION IN THE FUKUSHIMA DAIICHI NUCLEAR POWER PLANT DECOMMISSIONING Michael W. Golay Professor of Nuclear Science and Engineering Massachusetts Institute of Technology Cambridge, MA

2 1 BIOGRAPHICAL SKETCH DR. MICHAEL W. GOLAY is a professor of nuclear science and engineering at the Massachusetts Institute of Technology, where he has worked since He is director of the Reactor Technology Course for Utility Executives and the Nuclear Operational Risk Management Course, both cosponsored by MIT and the National Academy for Nuclear Training. He has focused his work upon improving nuclear power performance both in the United States and internationally, particularly through use of probabilistic and dynamic methods of analysis. Recent activities include low carbon energy strategies, riskinforming nuclear enterprises and projects, and service on the National Academy of Sciences Fukushima Study Committee. He has also been an active advisor to governmental and industrial organizations, particularly concerning risk-informed regulation and decision-making under uncertainty. Professor Golay received his Ph.D. in nuclear engineering from Cornell University, and performed post-doctoral research at Rennselaer Polytechnic Institute. He was also a visiting researcher at Electricité de France. He has served on the INPO Advisory Council, the NRC s Research Review Committee, the DOE s TOPS Committee (on non-proliferation), and national laboratory and nuclear power plant oversight committees. He is a Fellow of the American Association for the Advancement of Science and of the American Nuclear Society.

3 2 PRESENTATION OUTLINE SAFETY IN FUKUSHIMA DECOMMISSIONING Creation of High Quality Systems and Processes Uses Classical Method n High design standards n Deterministic performance requirements Risk Information Uses Complementary Probabilistic Method for System Refinement n System failures and end-states n Deterministic and probabilistic risk analyses n Bayesian treatments of uncertain knowledge

4 3 CONTEXT FOR USING RISK INFORMATION 1. Deterministic Design, Maintenance and Processes of High Quality Systems n Reflecting high design standards n Governed by conservative, deterministic performance requirements 2. Use of Best-Estimate Risk Information for Improvement and Failure Prevention for These Systems n Identification of system failure events and outcome possibilities n Deterministic and probabilistic risk analyses for system performance evaluation and requirements n Treatments of uncertain knowledge u Influence and sensitivities of uncertainties u Bases for belief in alternative explanations

5 4 USES OF RISK ASSESSMENT RESULTS Quantification of Risks of Alternative Activities* For a Particular Activity, Identification of Most Important: n Risk Contributors n Risks Most Sensitive to Event Uncertainties n System Vulnerabilities (Unacceptable End-States) n Uncertainties in System Performance For a Particular Activity, Identification of Most Effective Means of Reducing: n System Vulnerabilities n System Risks *Least valuable use of results

6 5 STRUCTURE OF RISK ASSESSMENT What Can Happen How Likely is the Event What are the Consequences of the Event Risk = Expected Consequences of an Activity* = Events ( Prob. Event Consequence Event ) *e.g., Transfer of all radioactive material from Fukushima site to interim repository site

7 6 DEFINITION OF RISK Event Risk Expected Consequences From an Event R i = <C i > = (Probability of Event, i) * (Consequences of Event, i) = [(Frequency of Event, i) * (Time Interval of Interest)] * (Consequences of Event, i) CORE DAMAGE RISK DUE TO N DIFFERENT CORE DAMAGE EVENTS R total = $ Consequence ' N 1, i ) = p i ) i=1 % Consequence ) M, i ( N R i i=1

8 7 EXAMPLE OF CORE-DAMAGE ACCIDENT CONSEQUENCES! Consequences of # Core Damage, # # Due to Core "# Damage Event, i $ = %! Consequence 1,i $ # Consequence 2,i # e.g., = # "# Consequence M,i %! Core Damage $ # i ECCS Damage # i # Control Room Contanimation i # "# Containment Damage i % e.g., Event i Could Be a Loss-of-Coolant-Accident (LOCA) R i =! Risks Due to # # Core Damage, "# Event, i $ % = p i! Consequence 1,i $ # Consequence 2,i # # "# Consequence M,i %

9 RISK VECTOR CALCULATION #!!!! " Risk = C # i p i = # C 1 % C ( C = % 2 ( All Event % ( Sequences $ C n '! C i = Vector of consequences associated with the ith event sequence p i = Probability of the ith! event sequence C = Mean, or expected, consequence vector EXAMPLE! C i = " Offsite acute fatalities due to event i% $ Offsite latent fatalities due to event i' $ Onsite acute fatalities due to event i ' $ Onsite latent fatalities due to event i ' $ Offsite property loss due to event i ' $ Onsite property loss due to event i ' # $ Costs to other NPPs due to event i ' Department of Nuclear Science and Engineering 8

10 9 THREE MILE ISLAND (TMI)

11 10 INSIDE TMI The vessel head sitting in its support stand alongside of the reactor. The head was removed to gain access to the damaged reactor

12 11 index.htm

13 12 THE MACHINES THAT CLEANED UP TMI Rover Core Sampler

14 13 TMI CLEANUP

15 INDEPENDENT SPENT FUEL STORAGE INSTALLATION (ISFSI) AT SHUTDOWN CONNECTICUT YANKEE SITE Department of Nuclear Science and Engineering 14

16 15 IDAHO SPENT FUEL FACILITY AND THREE MILE ISLAND PLANT DEBRIS

17 16 POTENTIAL CONSEQUENCES OF CLEANUP FAILURE EVENT Pollution n Sea n Ground and surface waters n Air Radiation Exposure Classes n Occupational n Public

18 17 FUKUSHIMA CLEANUP TASKS Outside Power Plant n Identification of radiation-polluted spaces n Collection and sequestering of contaminated materials n Interruption of means of radiation spreading: u Air/water based Within Power Plant n Identification of radiation-contaminated spaces n Identification and remediation of transport pathways for radioactive material n Pathway preparation for removal of radioactive materials n Removal, packaging and transport away of radioactive materials* n Removal and sequestration of surface-contaminated materials n Removal of civil works and power conversion system materials n Restoration to brown-field power plant site conditions n Long-term surveillance of radioactive materials, as sequestered *Example subject for risk-informed design and regulatory treatments

19 18 WORKER RADIATION EXPOSURE EVENT TREE End States E 0 : Small exposure E 1 : Elevated exposure E 2 : High exposure

20 19 INTIATING EVENT FAULT TREE

21 20 EVENT, W, WORKER ENTERING INTO RADIATION FIELD Probability Event 0.20 Incorrect Worker Choices 0.1 Worker does not follow procedure 0.1 Worker chooses to enter radiation field Incorrect Worker Information 0.05 Erroneous communication Erroneous procedure Detector failure 0.01 Negligent Worker Behavior 0.01 Shielding Error 0.27

22 BOSTON USA, MBTA OPERATOR TIED Red DIDN T Line probe points to MBTA OFF THROTTLE, SET BRAKE ON RUNAWAY TRAIN, 12/10/15 mistakes The lever used to regulate speed on a series subway car Driverless train sped through four stations and more than 8 km with about 50 passengers on board Department of Nuclear Science and Engineering 21

23 22 EVENT, M, RADIOACTIVE MATERIAL TRANSFER INTO WORKER SPACE Probability Event 0.02 Material Transfer Error 0.01 Transfer operator error Transfer machine error Mechanical error Control system error 0.01 Detector Error Robotic Communications Error Cask Failures Cask drop and rupture Cask closure failure Cask fire Cask random rupture Material Spill 0.032

24 23 EXAMPLE PROBABILITY DENSITY DISTRIBUTION Normal Distribution, φ(x i ), Alternative Independent Variables, X i : Distribution of radiation consequences, C; given marginal accident radiation exposure, D A Event probability value; probability of spill of radioactive material of mass, M f ( X i ) = 1 e 1 2 2πσ i $ % X i µ i σ i ' ) ( 2 X = random variable µ i = mean value of X i σ i = standard deviation of X i

25 24 ALTERNATIVE SYSTEM PERFORMANCE REQUIREMENTS Deterministic requirements n Are limited to small set of specified situations (DBAs) n Can be definitive, clear, unambiguous n Uncertainties remain unstated are treated via conservative assumptions, required redundancy Probabilistic requirements n Are formulated for all end-state situations of interest (event tree branch outcomes) n Can be elaborate, complex n Use best-estimate analyses u Event combinations u Probabilities u Event importance and sensitivity evaluations n Portray uncertainty estimates quantitatively Both apporoaches utilize models, data, expert judgments

26 25 EXAMPLE DESIGN BASIS ACCIDENT: MATERIAL IS SPILLED FROM TRANSFER/STORAGE CONTAINER DURING TRANSPORT BY RAIL TO IFSCE AT ROKKASHIO Deterministic Treatment n All material in a cask, M, spills into environment at location n Occurs at worst place, worst population, worst time n Occurs with worst weather n No protective or evasive population protection action is allowed n Population dose = d! r MAg F all space % ' '!!! ( ) f i J i r! ( r! r ) Species Released, i " $ # $ % Transport function Uptake Mechanisms, j!! r ( K ij Effectiveness * i )* n Cumulative Dose Consequence, C = i, species released ( ) d! r g(! r) J i ( " " MAf i All Space r " r) E i j, Uptake Pathway k ij

27

28

29 28 EXAMPLE LIMITED SET OF BOUNDING DESIGN BASIS ACCIDENTS (DBAs), SPECIFIED AT A SET OF TIMES AND PLACES OF OCCURRENCE, cont DBA Case 4 n All waste material in transfer container is released spontaneously in fine form n Weather is calm, with stable stratification n Released material is neutrally buoyant n No precipitation occurs over greatest population center DBA Case 5 n Transfer container is exposed to sustained kerosene fire, causing material to be released in form of volatile vapors, and residual non-volatile species, escaping from transport container n Weather is high turbulent, at high velocity, directed toward greatest population n Released material is neutrally buoyant n Sustained rain occurs over greatest population center DBA Case 6 n Transfer container is exposed to sustained kerosene fire, causing material to be released in form of volatile vapors, and residual non-volatile species, escaping from transport container n n n Weather is calm, with stable stratification Released material is positively buoyant Moisture precipitation does not occur

30 29 RISK-INFORMED TREATMENT OF RADIATION RELEASE EVENTS Expected Consequence (Best Estimate-Based) of Radiation Release during Material Transfer from Fukushima to Interim Storage Site Consequence = Expected Consequence = L L C l = C(D) l = C(D) m R n J n l=1 l n Container Travel Segments l=1 People exposed at site,!!r r to dose, D Total Dose, D Total, accumulated at site! r due to release at site,! r #, and weather, J Prob. (Release of type K at site! r #) J reflects all sets of categories of transfer and radiation pathway mechanisms ( ) Prob n released material magnitudes types, transport + uptake categories Releases of type K at site! r # K reflects all sets of categories of material types, magnitudes, forms! released at site r!

31 30 FACTORS IN RADIATION RELEASE ACCIDENT DURING RADIOACTIVE MATERIAL TRANSPORT Symbol C D l!r! r! J i (!! R M Prob. n i r " r)!! Definition Consequence (e.g., radiation-induced disease in an individual)! Radiation dose delivered to an individual at site, r Segment of trajectory traveled by radioactive material being shipped from Fukushima to interim repository Site of exposure of individual Site of release of radioactive material Transport function for unit material of type i released at! site, r, and transported to site, r Release fraction of radioactive material of type i! being transported by event of type n occurring at site r! Mass of radioactive material being transported! Probability of event of type n occurring at site r! ith category of radioactive! material released in accident event, n, at site r!

32 PROBABILISTIC MODELS PERMIT TREATMENTS OF VALIDITY BELIEFS OF ALTENATIVE HYPOTHESES Department of Nuclear Science and Engineering 31 Levels of belief by experts of alternative hypotheses can be stated probabilitisically Belief assessments can be propagated within probabilistic risk analyses to show implications for performance evaluations Levels of belief can be changed with logical consistency as new evidence becomes available (Bayesian updating) Such uses avoid the need to select among substantially uncertain hypotheses in assessing system performance, acceptability

33 32 DOSE-CONSEQUENCE MODEL Probability Density Function, f(c µ, σ, D 0 ) Shows Relative Likelihood of Observing C, Given D 0

34 ALTERNATIVE INDIVIDUAL DOSE- EFFECT RELATIONSHIPS CONSIDERED BY NRC IN SOARCA* EXERCISE Department of Nuclear Science and Engineering 33 *State-of-the-Art Reactor Consequence Analyses

35 34 LEGEND Model Number Model 1 Linear, No Threshold (LNT) 2 Threshold, D T = 10 mrem/yr, 0.10 msv/yr, LNT for D > D T 3 Threshold, D T = 620 mrem/yr, 6.20 msv/yr, LNT for D > D T 4 Threshold, D T = 5,000 mrem/yr, 50.0 msv/yr, or 10,000 mrem/yr, 100 msv/yr, lifetime dose

36 35 BAYESIAN EVIDENCE Dose, D, Individual Consequence, C! $ P# je = P! E j $! $ # P# j For j-th hypothesis " % " % " %! $! $ P# E jp# j and Evidence, E = C(D 0 ) " % " % j D 0 = observed dose D 0 =D B +D A D B = background dose exposure D A = additional dose exposure Alternative Dose-Response Models Considered by NRC A CON total = Consequence due to dose exposure of accident = dd 0 * H 0 Consequence Model* ( C D 0 )g Dose Dist ( D 0 ) g DoseDist = distribution of individuals receiving dose, D 0 C = C(D) = Individual Consequence of Exposure, D 0

37 36 BAYESIAN UPDATING OF ALTERNARIVE EXPOSURE CONSEQUENCE MODEL LIKELIHOODS Normalization: p i ( H i ) =1 i Illustrative prior hypothesis: probability distribution, p i (H i ) describes probability that H i is the true hypothesis New Evidence: Exposure D Observed New Consequence: C Observed New Evidence: E = C Observed (D Observed ) Posterior hypothesis probability distribution,!.! p i ( H i E) = p ( E H i)p i ( H i ) p( E H i )p i H i p E H i i ( ) p i ( ) H i, describes revised probability that H i is the true hypothesis, where ( ) is probability of observing evidence, E; given that H i is the true hypothesis.

38 37 ALTERNATIVE INDIVIDUAL EXPOSURE CONSEQUENCE MODELS Evidence: E = C Observed = C(D 0 ); D Observed = D Background + D Additional or D 0 = D B + D A 1) Linear Model:Prob L ( C D 0 ) = φ* ( C [µ, σ, D 0 ]), 0 C < µ = αd 0, σ = << µ 2) Threshold Model: Prob T Alternative Results: ( C D) = $ % ' [ ] φ( C µ, σ, D 0 ), µ = σd 0, σ << µ, D 0 D T 0, D 0 < D T ( ) 0 << αd 0T 4 A) C D 0 D T3 < D 0 < D T4 B) C( D 0 ) αd 0T D 4, T4 >> D T3 >> D T2 D 0 > D T4 * Approximate relationship, ignore cases where C < 0, and re-normalize as needed over domain, C > 0

39 38 SUMMARY Risk information can be used for improvement of previously-created systems Risk analysis permits consideration of: n All system performance contributors, situations n Relative importance of individual system performance contributors n Effects of uncertainties PRA offers a complementary alternative to deterministic treatments of system performance evaluations, requirements Bayesian analyses permit implications of alternative hypotheses to be accommodated in decision support

40 39 Backup Slides

41 40 DEFINITION OF RISK Event Risk Expected Consequences From an Event R i = <C i > = (Probability of Event, i) * (Consequences of Event, i) = [(Frequency of Event, i) * (Time Interval of Interest)] * (Consequences of Event, i) CORE DAMAGE RISK DUE TO N DIFFERENT CORE DAMAGE EVENTS R total = $ Consequence ' N 1, i ) = p i ) i=1 % Consequence ) M, i ( N R i i=1

42 41 EXAMPLE OF CORE-DAMAGE ACCIDENT CONSEQUENCES! Consequences of # Core Damage, # # Due to Core "# Damage Event, i $ = %! Consequence 1,i $ # Consequence 2,i # e.g., = # "# Consequence M,i %! Core Damage $ # i ECCS Damage # i # Control Room Contanimation i # "# Containment Damage i % e.g., Event i Could Be Core Damage, due to a Loss-of-Coolant- Accident (LOCA) Initiating Event R i =! Risks Due to # # Core Damage, "# Event, i $ % = p i! Consequence 1,i $ # Consequence 2,i # # "# Consequence M,i %

43 42 CUT SETS AND MINIMAL CUT SETS CUT SET: A cut set is any set of failures of components and actions that will cause system failure. MINIMAL CUT SET: A minimal cut set is one where failure of every set element is necessary to cause system failure. It does not contain another cut set.

44 43 CUT SETS AND MINIMAL CUT SETS CUT SET: A cut set is any set of failures of components and actions that will cause system failure. MINIMAL CUT SET: A minimal cut set is one where failure of every set element is necessary to cause system failure. It does not contain another cut set.

45 44 RISK IMPORTANCE MEASURES Risk = R(q 1, q 2,, q n ), where r i = reliability of the i th plant component, action, or cut set q i = unreliability of the i th component = 1 - r i I Fussell-Veselyi = the fraction of total risk involving failure of element, i where R(q i ) R Nom m n I Fussell Veselyi = R ( q i) ( ) = R mcs i 1 + mcs i2 +! + mcs im R Nom R( mcs 1 +! + mcs n ) = risk arising from event sequences involving failure of component, action or cut set, i = nominal plant risk = number of minimal cut sets involving element (basic event) i = total number of minimal cut sets

46 45 RISK IMPORTANCE MEASURES Risk Achievement Worth (RAW i ) Maximum relative possible increase in total risk due to failure of element, i; the element is assumed always to fail (failure event probability, q i = 1). RAW i = R( q i = 1) R Nom where RAW i = the risk achievement worth of the i th component, action or cut set

47 46 RISK IMPORTANCE MEASURES Risk Reduction Worth (RRW i ) = Maximum possible relative reduction in risk due to perfection of event i reliability; the component is assumed always to succeed every time (failure event probability, q i = 1). RRW i = R Nom R( q i = 0) where RRW i = the relative risk decrease importance of the i th component, action or cut set

48 47 USES OF RISK IMPORTANCE MEASURES Fussell-Vesely n Measure a Component s or System s Participation in Risks n Can Be Used to Identify Which Components or Systems Contribute to Current Risks Risk Achievement Worth n Identifies Which Components or Systems Must Be Kept Reliable Risk Reduction Worth n Identifies Which Components or Systems Are Most Valuable for Improvement n Note I Fussell Vesely i =1 1 RRW i

49 48 LECTURE OBJECTIVES INTRODUCTION OF THE BASIC ELEMENTS OF PROBABILISTIC RISK (PRA) ANALYSES Risk PRA Structure PRA Results PRA Importance Measures

50 49 STRUCTURE OF RISK ASSESSMENT What Can Happen How Likely is the Event What are the Consequences of the Event Risk = Expected Consequences of an Activity* = ( Prob. Event Consequence Event ) Events *e.g., Transfer of all radioactive material from Fukushima site to interim repository site

51 50 USES OF RISK ASSESSMENT RESULTS Quantification of Risks of Alternative Activities* Identification of Most Important Contributors to Risks of a Particular Activity Identifications of Contributors to Risks of a Particular Activity that are Most Sensitive to Uncertainties Identification of Most Important System Vulnerabilities Identification of Most Effective Means of Reducing System Vulnerabilities Identification of Most Effective Means of Reducing System Risks Identification of Most Important Uncertainties in System Performance Identification of Most Effective Means of Reducing Sensitivity of Risks to Uncertainties *Least valuable use of results

52 51 IDAHO NATIONAL LABORATORY (INL) SPENT FUEL STORAGE FACILITY The Department of Energy stores three metric tons of spent fuel in pools like this at the Idaho National Laboratory, and 277 metric tons stored in dry casks