A novel selectivity technique for high impedance arcing fault detection in compensated MV networks

SUMMARY In this paper, the initial transients due to arc reignitions associated with high impedance faults caused by leaning trees are extracted using discrete wavelet transform (DWT). In this way, the fault occurrence is localized. The feature extraction is carried out for the phase quantities corresponding to a band frequency 12.5–6.25 kHz. The detection security is enhanced because the DWT corresponds to the periodicity of these transients. The selectivity term of the faulty feeder is based on a novel technique, in which the power polarity is examined. This power is mathematically processed by multiplying the DWT detail coefficients of the phase voltage and current for each feeder. Its polarity identifies the faulty feeder. In order to reduce the computational burden of the technique, the extraction of the fault features from the residual components is examined. The same methodology of computing the power is considered by taking into account the residual voltage and current detail coefficients where the proposed algorithm performs best. Test cases provide evidence of the efficacy of the proposed technique.