Extreme precipitation events for nuclear power plants have traditionally been evaluated using the concept of a Probable Maximum Precipitation (PMP). Contrary to "PMP" including the word "probable" in its definition, the PMP concept does not assign a probability to the PMP rainfall depth. Thus, it is difficult to quantify the risk associated with a PMP event. For example, how can a PMP event be included in a Probable Risk Assessment (PRA) for a nuclear power plant? A probabilistic analysis of precipitation must be conducted to evaluate probabilities associated with extreme rainfall events. A methodology is presented for a regional extreme rainfall analysis using the method of L-Moments. The method involves applying one or more probability density function types. Each probability density function type is evaluated individually for its goodness-of-fit for the regional precipitation data. Multiple distributions can be retained in the analysis with probability weights being assigned to each distribution type based on appropriate criteria (e.g., goodness-of-fit or expert opinion). Consideration is given to the various sources of uncertainty, and methods for quantifying and reducing uncertainty are discussed. The probabilistic precipitation analysis produces a family of hazard curves that relate precipitation depths to annual exceedance probabilities. The hazard curves are used as input to a run-off model that produces hazard curves at individual structures (e.g., hydrodynamic forcing versus annual exceedance probability). These structure-specific hazard curves could be used as input for a PRA analysis, including fragility analysis for individual structures. This paper includes an example study for illustration purposes.
展开▼
