A smart meter based approach to power reliability index for enterprise-level power grid

We aim to develop a new, practical power reliability index (PRI) for enterprise-level power grids. Largely different from traditional evaluation methods that assume the exponential distribution of time between failures and use theoretical metrics such as mean time between failures (MTBF), our approach is data-driven and takes advantage of smart meters that are capable of recording various power quality indicators at selected monitoring points. The basic idea of our approach is to estimate the instantaneous network reliability, a concept describing current network reliability status, based on measurements from a limited number of smart meters in the power grid. We develop algorithms to solve the technical challenges in the accurate estimation of instantaneous network reliability, using Expectation Maximization (EM) and Monte Carlo Expectation Maximization (MCEM). The effectiveness of our algorithms is validated with empirical evaluation.