ML21277A085
| ML21277A085 | |
| Person / Time | |
|---|---|
| Issue date: | 08/18/2021 |
| From: | Nathanael Hudson, Tehrani N, Peter Yarsky NRC/RES/DSA |
| To: | |
| Dennis M | |
| References | |
| Download: ML21277A085 (13) | |
Text
Autonomous Control Algorithms to Simulate Boiling Water Reactor Cycle Depletion NRC Team:
Nate Hudson, Ph.D. Nathanael.Hudson@nrc.gov Nazila Tehrani, Ph.D. Nazila.Tehrani @nrc.gov Peter Yarsky, Ph.D. Peter.Yarsky@nrc.gov
Regulatory Purpose NRC then
- Reactive
- Plant specific
- Traditional methods NRC now
- Proactive
- Generic model
- Advanced methods 9/8/2021 Data Sci/AI Reg Applications Workshop 2
NRC is exploring innovative tools
- Perform independent transients/accidents analysis
- Increase staff efficiency
- Identify/focus safety significant issues
- Boost confidence in licensee results
- uncertainty high
- margin low
- Address emergent issues 9/8/2021 Data Sci/AI Reg Applications Workshop 3
Motivation
- Automate LWR core/cycle design
- Create models not associated with any specific licensing action
- Use autonomous control methods 9/8/2021 Data Sci/AI Reg Applications Workshop 4
RES develops generic LWR TRACE models for accident/transient analysis Boiling water reactor Peter Yarsky (Project Lead, Senior Reactor Systems Engineer, Code And Reactor Analysis Branch)
Nate Hudson (Reactor Systems Engineer, Code And Reactor Analysis Branch)
Nazila Tehrani (Reactor Systems Engineer, Accident Analysis Branch)
Pressurized water reactor Andy Bielen (Project Lead, Nuclear Engineer (Fuels/Neutronics), Fuel
& Source Term Code Development Branch)
Mike Rose (Reactor System Engineer (Neutronics Analyst, Fuel
& Source Term Code Development Branch)
Alice Chung (Reactor System Engineer (Fuel Analyst), Fuel &
Source Term Code Development Branch) 9/8/2021 Data Sci/AI Reg Applications Workshop 5
Control excess reactivity during cycle
- burnable poisons
- control blades
- flow control windows Competing goals
- meeting desired cycle energy
- maintaining safety margins
- minimizing duty related fuel failures High-dimensionality of BWR cores
- issues with symmetry
- very large number of types of fuel elements BWR cores are complex to design 9/8/2021 Data Sci/AI Reg Applications Workshop 6
TRACE/PARCS models of PWR/BWR transients/accidents analysis
- Coupled neutronics/TH transient
- normal
- anticipated transients
- accident conditions
- current reactors
- advanced reactors
- 12 major plant types (5 BWR)
- Standard models
- Efficient confirmatory calculations 9/8/2021 Data Sci/AI Reg Applications Workshop 7
Necessity Traditional NRC
- NRC used PARCS to simulate BWR cycle depletion
- calculate cycle power and burnup distributions
- transient analyses need burnup-dependent
- rod patterns
- flow rate
- EOC bundle shuffle sequence Innovative NRC
- theoretical concept
- operate a typical plant
- given fuel design
- over a long time 9/8/2021 Data Sci/AI Reg Applications Workshop 8
Autonomous control algorithms Literature review Proposed micro-reactor sense reactor conditions sense reactor coolant system judge qualification of signals evaluate current state of system make decisions about actions implement actions for operation NRC goals PARCS models dynamically adjust
- fuel loading between cycles
- control rod pattern
- flow rate during cycle yield all statepoint information over full cycle 9/8/2021 Data Sci/AI Reg Applications Workshop 9
Literature Review Bayesian networks for dynamic (PRA)
Kim, et al.
DNN to optimize core loading pattern for BWR Saleem, et al.
Annealing method Hays and Turinsky 9/8/2021 Data Sci/AI Reg Applications Workshop 10
Literature Review Bayesian networks for dynamic (PRA) studies evolution of risk during postulated events makes decisions during a transient for reducing risk artificial reasoning rely on surrogate models NRC wants to create surrogate models DNNs to optimize core loading pattern for BWR DNNs trained against core simulator artificial reasoning makes decisions about core loading patterns meeting power peaking limits cycle energy demand reasonable computational expense/accuracy used to find optimal designs
~10,000 simulations to train NRC wants to optimize loading patterns use core simulator use sampling of design choices similar to particles distribution at a temperature iteratively lowering temperature, algorithm finds optimal solution
~100,000 core simulator runs NRC wants to optimize loading patterns Annealing method 9/8/2021 Data Sci/AI Reg Applications Workshop 11
Work to be Performed
- Autonomous control for BWR core/cycle design feasible?
- Apply combination of existing decision-making methods
- Bayesian
- Neural Networks
- etc.
- Approximate core loading design/control rod sequence
- Contingency (Traditional Methods)
- Need feedback 9/8/2021 Data Sci/AI Reg Applications Workshop 12
Definitions
- BWR: Boiling water reactor
- DNN: Deep neural networks
- EOC: End of cycle
- LWR: Light water reactor
- PARCS: NRC Reactor Kinetics code
- PWR: Pressurized water reactor
- RES: Office of Nuclear Regulatory Research
- TH: Thermal-Hydraulics
- TRACE: NRC Thermal-Hydraulics code 9/8/2021 Data Sci/AI Reg Applications Workshop 13