To study an optimized extraction method of partial discharge（PD） fingerprint parameters under DC voltage

XLPE cable was used to construct 3 kinds of insulation defects

including inner semi-conductive layer breakage

internal air cavity defect

and insulation surface scratch defect. The PD characteristics of the 3 defects were tested under DC voltage with different amplitudes and polarities. Moreover

the traditional fingerprint extraction method was compared with the GA-BP algorithm-based characteristic parameters extraction method to find their effects on defect type identification. The research results indicate that the PD repetition rate and average discharge amount under positive voltage are higher than those under negative voltage with inner semi-conductive layer breakage. With internal air cavity defect

PD repetition rate under negative voltage is higher than that under positive voltage

while average discharge amounts are basically equal. With insulation surface scratch defect

the PD repetition rate and average discharge amount vary within a large range

and the PD discharge repetition rate is significantly higher than that of other two kinds of defects. The fingerprint parameter of an insulation defect obtained by the traditional fingerprint extraction method is susceptible to the influence of voltage amplitude and polarity. The GA-BP algorithm can exclude this effect after extracting fingerprint parameters. We use traditional fingerprint and optimized fingerprint to identify fault types

with the same defect under identical voltage amplitude and polarity

the recognition rates of them are higher than 97%; with the same defect under various voltage amplitudes and polarities

the recognition rate of traditional fingerprint drops to 87%

the recognition rate of optimized fingerprint is still above 97%.