Treating common-cause failures in multi-unit pras

The term "common-cause failures (CCFs)" refers to a failure of two or more components arising from a single shared cause during a short period of time. Methods for incorporating CCFs into Probabilistic Risk Assessments (PRAs) have evolved over the past several decades, including various models (for example Beta Factor, Alpha Factor and Multiple Greek Letter models) and clear guidelines (such as NUREG/CR-5485 and IAEA-TECDOC-1135). However, the PRAs were originally developed for a single reactor, i.e. the analysis is confined to one reactor at a time, rather than the integrated risk from all the reactors and other radioactive sources on the whole site. Regarding the treatment of CCFs in MUPRAs, several questions may arise: (1) should the common-cause component groups (CCCGs) be expanded to include inter-unit CCFs? (2) If the answer is "yes", how to handle the CCFGs of very high order (e.g. involving more than 5 or 6 components)? This paper conducts base case and sensitivity analyses using different CCF models in an MUPRA, trying to provide insights for the above questions. This paper is based on the MUPRA results of a real two-unit nuclear power station, through which the CCFs are recognized as a primary contributor to the site risk.