The development of methods of dynamic probabilistic risk assessment (PRA) is an ongoing topic of research at the Ohio State University. Recently, the ADAPT (Analysis of Dynamic Accident Progression Trees) software tool was developed to provide a flexible framework in which to perform a dynamic event tree analysis. Dynamic PRA methodologies have the advantage over conventional PRA methodologies in that a more realistic and mechanistically consistent analysis can be performed of a system in question. Dynamic PRA methodologies are designed to take the timing of events explicitly into account which can become very important especially when uncertainties in complex phenomena are considered. Despite the advantages that dynamic event methodologies offer, there is still considerable question in the community as to how dynamic methodologies can provide “better” results than classical methods, whether there is really a need the detailed modeling that dynamic methodologies provide within the context of a full PRA, and whether the implementation of dynamic methodologies on a real system is practical as dynamic methodologies can be computationally expensive.
The purpose of this work is to address those concerns just noted by performing a comparison of the results obtained for a particular scenario on a real system by using classical PRA analysis and a parallel analysis performed using a particular dynamic PRA method. In late 1980’s the NUREG-1150 study was commissioned to perform a full PRA using the best methods available at the time of five U.S. nuclear power plants. The power plants that were chosen were: Surry Unit 1 (PWR), Zion Unit 1 (PWR), Grand Gulf Unit 1 (BWR), Peach Bottom Unit 1 (BWR), and Sequoyah Unit 1 (PWR). For each of these plants, a detailed analysis of all systems and potential accident pathways was performed using conventional PRA methodology. For this study, the results obtained in NUREG-1150 for the Zion Unit 1 plant will be compared to the results obtained in a dynamic PRA analysis. Specifically, the results obtained for one of the initiating events examined in NUREG-1150, namely the Loss of Offsite Power (LOSP) initiating event with loss of all diesel generators (commonly known as a Station-Blackout (SBO) Event), will be compared. Data from all supporting documentation on the Zion Unit 1 NUREG-1150 analysis have been gathered and a comparable dynamic model will be built and executed using the MELCOR severe accident analysis code.
The end-goals of this research are to 1) evaluate the advantages and disadvantages of both conventional and dynamic PRA with respect to one another, 2) compare the numerical results of the conventional and dynamic analysis with respect to the chosen event’s contribution to core damage frequency and large early-release frequency, and 3) to make a comparison of the accident sequences generated by both analyses to determine if dynamic analysis shows additional risk-significant scenarios not discovered by classical methods.
Results from the dynamic analysis show consistency with the classical PRA results. However, the dynamic analysis was able to provide additional resolution and detail for some classical plant damage stages. In addition, new accident sequences which were not considered by the classical analysis were discovered. Dynamic event tree analysis proved to be a powerful tool in modeling plant response in a physically consistent manner given an input probabilistic model and provided additional insight into the potential accident progression.