Evaluating Generating Unit Unavailability Using Bayesian Power Priors

Generating unit unavailability assessment is an important task in the power system generation expansion planning aimed at managing an acceptable degree of security of supply. In the case of the newly installed or planned units, the sample size of the data for unavailability assessment is limited and inadequate to provide the desired accuracy in the unavailability estimation. A new concept based on the Bayesian power prior approach has been developed to utilize the data from similar generating units. The original contribution of the present work is a model that incorporates the data of unavailability from other generating units into the statistical analysis of unavailability of the analysed generating unit to improve the accuracy of the estimation. The empirical results show that for unavailability estimation, the power prior Bayesian approach exhibits better than the classical statistical approach in both the standard error of estimate and confidence interval as the measures of accuracy.