[1M4-GS-10] AI application

[1M4-GS-10-04] Diagnosis of Ground Fault Cause on Distribution Line using SVM

〇Ryoma Matsubara¹, Takashi Onoda¹ (1. Aoyama Gakuin Univ.)

Keywords:AI, equipment diagnosis, ground fault cause diagnosis, pattern recognition

The cause of ground fault in distribution line for large customers is diagnosed by experts looking at the current waveform at the time of the ground fault, and it takes a lot of time to restore. Therefore, it is required to automate the diagnosis and reduce time until the restoration. In previous study, five causes of ground fault (cable, gap, insulator damage/fouling, complete, bird and beast contact) were able to be identified by performing 2-class classification of SVM in 4 steps using the current waveform at the time of ground fault. On the other hand, previous study wasn’t considered the load on distribution lines, and it is known that the current waveform becomes complicated in the distribution line with load. Therefore, in this study, we propose features that enables diagnosis of ground faults in distribution lines with load. Through experiments using actual data, it has been confirmed that five causes of ground fault can be diagnosed with an accuracy rate of 90% even by using the proposed feature.