For minimizing the possibility of misjudgment and false alarm of partial discharge (PD) signal identification in the operation of PD online monitoring system of cable as much as possible

PD’s interpretation mechanism and identification accuracy of partial discharge are improved and updated. The warning record data of tens of thousands suspected PD in cables collected by the field operation system is analyzed and classified

filtering process of inter-phase signal correlation is added to the automatic logic discrimination program of PD

the existed neural network is optimized in structure and the pre-processing of the learning data of the existing neural network is improved. Three kinds of partial discharge data

including field recording

analog signal generator

and artificial partial discharge model with voltage application experiment

are used to test the improved PD discrimination program. The results show that the enhanced PD discrimination program can not only reduce the warning rate of non-PD warning data to the expected level of 5%

but also

at the same time

improve the accuracy of identifying various types of PD signals generated at artificial simulation

reducing significantly the possibility of misjudgment and false reporting in the PD online monitoring.