Extension of probabilistic seismic hazard analysis to account for the spatial variability of ground motions at a multi-unit nuclear power plant hard-rock site

Abstract

Nuclear power plant (NPP) sites typically consist of multiple reactors (units). When earthquakes occur in the vicinity of these sites, effects will be experienced at all units. However, a seismic probabilistic risk assessment (PRA) of a multi-unit site typically only considers the possibility of adverse conditions at a single unit or, if multiple unit impacts are considered, typically assumes that all units at the site experience the same ground motion. This assumption of perfect correlation in ground motion is inconsistent with data from dense seismic arrays, which show that there is spatial variability in the ground motion between closely-spaced locations during the same earthquake. Moreover, such an assumption is not inherently conservative as is often assumed. To facilitate more realistic assessments, this paper proposes a method for capturing the effects of spatial variability of ground motion at an NPP site in the seismic PRA. The proposed method uses the results from an existing probabilistic seismic hazard analysis (PSHA) performed for a reference location at the site (e.g., the site seismic hazard curve and magnitude disaggregation) and then estimates the conditional probability distribution of the ground motion at a non-reference location. The proposed method accounts for the spatial variability in ground motion amplitude and includes mathematical formulations for NPP sites classified as “hard-rock.” The proposed method is developed using Bayesian networks because their graphical structure facilitates model transparency and communication. Moreover, Bayesian networks facilitate efficient probabilistic modeling and inference involving dependent random variables. However, the proposed method can be implemented more generally without using Bayesian networks as the calculation framework and commentary is provided regarding these alternative calculation options. This paper concludes with an example application using a PSHA for a hypothetical hard-rock site in the Central United States. Based on the results from the example application, we note that the perfect correlation assumption with respect to the ground motion at multiple locations/units at an NPP site may not necessarily be conservative.