Generation Risk Assessment: System Design, Economic Impacts and Importance Measures

Probabilistic risk assessment (PRA) is the systematic process of constructing a model representing risks from hazards. A complete PRA identifies the scenarios that may lead to an undesired outcome, the likelihood of said scenarios, and the magnitude of the consequences. The scope of PRA can address a variety of undesired situations that may occur at a facility. One specific situation of interest is for cases where electrical power generation is lost. This case, loss of electricity generation, is evaluated using a Generation Risk Assessment (GRA). In a PRA, it is first determined "what can go wrong" by identifying hazards that may facilitate off-normal conditions or initiating events. Then, a scenario model for each initiating event that describes how the nuclear facility will respond to the off-normal condition is created. Next, the model is populated by assigning likelihoods for the initiating event and failures of applicable systems, components, software, or operators. The plant response modeling includes structures, systems, component (SSC) failures, and human error conditions. A GRA uses a similar approach to modeling. First, the conditions that initiate a loss of electrical generation are determined. Then these conditions are evaluated in a scenario model to represent the off-normal condition. One element of GRA models, though, is they are time-dependent for a variety of factors including seasonal elements, demand conditions, and repair/restoration processes following a loss of electrical production. Given the specificity of GRA, we need a specialised methodology to assess the importance of SSCs and gather insights through sensitivity analysis. We address the extension of traditional importance measures within this framework first. We then study the application of novel methods to manage the simultaneous presence of multiple and dynamic outputs.

Keywords: Generational Risk, GRA, Importance Measures, Reliability Analysis, Complex Systems Simulation.