Experience from the comparison of two PSA-studies

Two probabilistic safety assessments (PSA) made for nearly identical reactors units (Forsmark 3 and Oskarshamn 3) have been compared. Two different analysis teams made the PSAs, and the analyses became quite different. The goal of the study is to identify, clarify and explain differences between PSA-studies. The purpose is to understand limitations and uncertainties in PSA, to explain reasons for differences between PSA-studies, and to give recommendations for comparison of PSA-studies and for improving the PSA-methodology. The reviews have been made by reading PSA-documentation, using the computer model and interviewing persons involved in the projects. The method and findings have been discussed within the project group. Both the PSA-projects and various parts in the PSA-model have been reviewed. A major finding was that the two projects had different purposes and thus had different resources, scope and even methods in their study. The study shows that comparison of PSA results from different plants is normally not meaningful. It takes a very deep knowledge of the PSA studies to make a comparison of the results and usually one has to ensure that the compared studies have the same scope and are based on the same analysis methods. Harmonisation of the PSA-methodology is recommended in the presentation of results, presentation of methods, scope, main limitations and assumptions, and definitions for end states, initiating events and common cause failures. This would facilitate the comparison of the studies. Methods for validation of PSA for different application areas should be developed. The developed PSA review standards can be applied for a general validation of a study. The most important way to evaluate the real feasibility of PSA can take place only with practical applications. The PSA-documentation and models can be developed to facilitate the communication between PSA-experts and users. In any application consultation with the PSA-expert is however needed. Many real uncertainties can be identified by comparing PSAs. Generally, comparisons are recommended as a method to review the quality of a PSA-study and as a method to analyse uncertainties of the study.