Application of Bayesian Networks in Multi-Hazard Safety Assessment of Nuclear Power Plants
- 1Geo-Engineering Section, Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, The Netherlands (v.k.duvvurumohan@tudelft.nl)
- 2Geophysical Institute and Center for Disaster Management and Risk Reduction Technology (CEDIM), Karlsruhe Institute of Technology, Karlsruhe, Germany
- 3Bureau de Recherches Geologiques et Minieres (BRGM), Orleans, France
- 4Faculty of Technology, Policy and Management, Delft University of Technology, Delft, The Netherlands
- 5Nuclear Research and Consultancy Group (NRG), Petten, Netherlands
- 6Comissariat a L'Energie Atomique et aux Energies Alternatives (CEA), Paris, France
Low probability events occurring in sequence, within a limited operational time (damage and recovery window between events), are a key consideration in multi-hazard safety assessments of nuclear power plants (NPPs). Cascading effects from hazards and associated event sequences could potentially have a significant impact on risk estimates. The Bayesian network can act as a framework to consider aforementioned statistical dependencies between various hazards in multi-risk analyses of nuclear power plants.
Within the EU project NARSIS (New Approach to Reactor Safety Improvements), a Bayesian network-based risk assessment framework was developed to perform multi-hazard risk assessment of NPPs.
The Bayesian network method was applied for an external-event related station blackout (SBO) scenario at a NPP. Earthquake, flooding, and tornado were among the hazards considered at a decommissioned NPP site location in Europe. Both hazard dependency in time as well as a cascading scenario was also considered. The hazards, their interactions and the fragilities of selected systems, structures and components within the nuclear power plant were represented in the network and their probability distributions were obtained based on the multi-hazard and fragility approaches adopted within the NARSIS project.
Sensitivity analyses in the network were used to identify key hazards and interactions. Most influential hazard combinations and ranges of intensity measures were identified through diagnostic inference in the network. Discretisation of continuous variables (hazard curves in this case) is a key aspect of performing inference in Bayesian networks. The effect of various levels of discretisation of hazard probability distributions was assessed, to identify suitable discretisations of hazard data.
This application demonstrates the use and advantages of the Bayesian network methodology, developed in the NARSIS project, for probabilistic safety assessments of NPPs.
How to cite: D. Mohan, V. K., Vardon, P., Daniell, J., Gehl, P., Schafer, A., van Gelder, P., Natarajan, V., Molenaar, C., Foerster, E., and Ragon, F.: Application of Bayesian Networks in Multi-Hazard Safety Assessment of Nuclear Power Plants, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-21036, https://doi.org/10.5194/egusphere-egu2020-21036, 2020
