Simplified probabilistic selectivity technique for earth fault detection in unearthed MV networks

A novel selectivity technique to identify the faulty feeder in unearthed Medium Voltage (MV) networks is introduced. The proposed technique is based on a simplified probabilistic method applied to transient features extracted from the residual currents only using the discrete wavelet transform (DWT). DWT will enhance the accuracy of localising fault events with different fault resistances and the probability will give the protection decision (detection/selectivity). The standard deviation of a window of the DWT detail coefficient is used to detect the fault and an adaptive threshold is used as input to the probability function to estimate the faulty feeder. A faulty unearthed 20 kV network is simulated by ATP/EMTP imitating practical fault cases such as arcing and tree leaning faults. The time-domain test cases provide evidence of the efficacy of the proposed technique.