Estimation of SCRAM Rate Trends in Nuclear Power Plants Using Hierarchical Bayes Models

Abstract

Nuclear reactors are equipped with reactor scram systems to ensure rapid shutdown of the system in the event of leaks, failure of power conversion systems, or other operational abnormalities. The U.S. Nuclear Regulatory Commission (NRC) collects data on scram rates from various nuclear power plants over time to estimate the trend of proper functioning of the plants which in turn is used to regulate them. The annual scram data obtained from 66 commercial nuclear power plants indicate an increase in the number of plants having no scrams from 1.5% in 1986 to 33% in 1993. To analyze correlated count data with excess zeros (e.g., no scrams), a zero-inflated model that accounts for both temporal and plant-to-plant variation is being developed in this article. A wide class of possibly non-nested models was fitted using Markov Chain Monte Carlo (MCMC) methods and compared using a predictive criterion. Out-of-sample tests were also performed to study the performance of the models in predicting the scram rates of the plants. For the NRC data on scram rates, the stochastic time trend models that account for zero-inflation were found to provide a much better fit compared to the deterministic trend models.