Improving Location Estimates of Distribution Faults With Imperfect Grid Edge Measurements

This paper proposes a hybrid algorithm to improve location estimates of momentary and incipient faults in distribution operations. The proposed scheme is implemented and tested as a software program using actual field data collected at distribution substations and grid edge. The estimation is achieved by modeling the relationship graph between fault recorder waveforms at the substation and smart meter voltage exceptions that captures the spatial and temporal dependency. A windowed transformation of the waveforms is utilized to mine the temporal dependency for the location estimation and anomaly scoring. The performance of the proposed method is demonstrated on the failure of an underground residential distribution switch to a 50 kVA transformer. In the pilot testing, the estimated fault area of the affected structures proved to be 100% accurate whenever the spatiotemporal alignment was successful.