Spectral Clustering for Fast Outage Detection and Visualization in Power Distribution Systems

Power outages has huge economic impact on the society and hence it is crucial to quickly detect outages and plan service restoration actions. Existing outage detection method either requires deployment of expensive sensors or are computationally expensive. In this work, we proposed the spectral clustering based outage detection algorithm that is easy to implement using available sensors at distribution-level and is computationally efficient in accurately detecting outages using load forecast data and smart meter outage notifications. We tested the performance of the proposed algorithm in three-phase unbalanced IEEE-123 node test feeder. The proposed algorithm can accurately detect the outages and is robust to the measurement noise.