Quasi-Monte Carlo sampling method for simulation-based dynamic probabilistic risk assessment of nuclear power plants

ABSTRACT

Dynamic probabilistic risk assessment (PRA), which handles epistemic and aleatory uncertainties by coupling the thermal-hydraulics simulation and probabilistic sampling, enables a more realistic and detailed analysis than conventional PRA. However, enormous calculation costs are incurred by these improvements. One solution is to select an appropriate sampling method. In this paper, we applied the Monte Carlo, Latin hypercube, grid-point, and quasi-Monte Carlo sampling methods to the dynamic PRA of a station blackout sequence in a boiling water reactor and compared each method. The result indicated that quasi-Monte Carlo sampling method handles the uncertainties most effectively in the assumed scenario.

KEYWORDS:

• Risk analysis
• uncertainty
• simulation
• boiling water reactor
• probabilistic risk assessment
• dynamic PRA
• Monte Carlo method
• Latin hypercube sampling
• quasi-Monte Carlo method

Acknowledgments

The first author would like to thank Mr. Satoshi Fujita of NRA, J for pointing out the idea of using advanced sampling methods. This research was conducted using the supercomputer SGI ICE X and HPE SGI8600 in JAEA. The development of RAPID is supported financially by NRA, J, but the implementation of new features in this study, including the QMC method, is the exception.

Disclosure statement

No potential conflict of interest was reported by the author(s).