An Edge Computing Method for Partial Discharge under the Power Internet of Things

Abstract

The recognition of impulse interference for the partial discharge data on the edge computing platform can effectively reduce the false alarm rate in the risk assessment of gas insulated switchgear (GIS). Therefore, an edge computing method for partial discharge is proposed in the paper, which uses the similarity computing among the aggregation of data for each sensor after clustering to match homologous signals. Finally, the algorithm will jointly locate the interference signal and recognize the impulse interference signal by comparing the received signal strength indicator (RSSI) of multiple UHF sensors in/out of GIS equipment. The training set, being consisted of partial discharge and interference signal from the test, is used to contrast the algorithms on the edge computing platform, the result of which indicates that the accuracy rate of the recognition of the impulse interference with the proposed algorithm is 81.4% on the computing platform, with the average processing time for each sample of 7.15s.