Feature Extraction of High Impedance Arcing Faults in Compensated MV Networks. Part I: DWT-Based Analysis of Phase Quantities

In this paper, the initial transients in the electrical network due to arc reignitions associated with high impedance faults caused by leaning trees are extracted from phase voltages and currents. The discrete wavelet transform (DWT) is used for the fault feature extraction corresponding to band frequency 12.5- 6.25 kHz and therefore localizing this fault occurrence. The detection security is enhanced because the DWT is responded to a periodicity of the initial transients. A novel technique for the selectivity term of the faulty feeder is introduced. It is based on the power polarity, in which, this power is computed by multiplying the DWT detail coefficient of the current and voltage for each feeder. The fault due to a leaning tree occurring at different locations in an unearthed 20 kV network is simulated by ATP/EMTP and the arc model is implemented using the universal arc representation.