Verification of DWT-based detection of high impedance faults in MV networks

The faults due to leaning trees are common in Nordic Distribution networks. The detection of this fault type has previously been discussed and the faulty section has been estimated using a novel selectivity technique. This detection was based on an absolute sum of Discrete Wavelet Transform (DWT) detail coefficients of the residual voltages. The detection selectivity function depended on the transient directionality of the residual voltage and currents which are extracted using DWT. In this presentation, this technique is evaluated using 17 staged earth fault cases at a sampling frequency 10 kHz. These faults are accomplished in a 20 kV network using known fault resistances and natural trees. However, theses fault cases are not associated with arcs and the transients are only generated at the starting fault instant without repetition because the arc reignition impact on the network does not exist. Therefore, the transient features are localized using DWT at the fault instant only. The test cases can provide evidence of the detection efficacy.