抄録
The Fukushima Daiichi Nuclear Power Plant accident raised a concern for site risk involving multiple reactors. As a result, there has been a considerable interest in Multi-Unit PRA (MUPRA), which evaluates the risk of multiple units in a site. However, there is few international experiences of MUPRA, and standardized methodologies for MUPRA are required to estimate realistic site risk. This paper addresses a methodology to conduct MUPRA by extending the Single-Unit (SU) PRA models based on the conventional ET/FT methods. The proposed methodology focuses on interaction between units, consistency with the SUPRA models, and application to real nuclear power stations. To systematically conduct MUPRA, Initiating Events (IEs) are classified into three categories: independent Multi-Unit Initiating Events (MU-IEs), common cause MU-IEs, and dependent (cascading) MU-IEs. For independent MU-IEs, simplified evaluations are employed. For common cause and cascading MU-IEs, detailed evaluations are employed to capture multi-unit (MU) interactions such as common cause failures (CCF) and influence of adjacent units’ accident progress that affect availabilities and success criteria of shared system, human reliability and resource, and so on. Human reliability and CCF are major contributors to SUPRA and considerably important in MUPRA. Hence, Multi-unit Human Reliability Analysis (MU-HRA) and interunit CCF Analysis were also considered. The proposed methodology uses a simple MU-HRA method, in which human error probabilities (HEPs) are products of coefficients determined by decision trees and HEPs from single unit HRA. For inter-unit CCF, the screening process to select inter-unit CCF has been clearly defined, and we defined new parameters that reflect the commonality of components between units, such as similarity of maintenance procedures in units, on the parametric CCF models. Thus, it is possible to create an interunit CCF having a different viewpoint from the intra-unit CCF to be newly added to MUPRA.
