Detecting Arcing Current Faults in Medium-to-Low Voltage Transformers

Arcing current faults occurring in a medium-to-low voltage (MV-LV) transformer can pose a challenge for transformer protection. Such faults trigger currents with lower magnitudes than those triggered by conventional faults, thus making it difficult for transformer protection to accurately detect and respond to arcing current faults. This paper proposes a method for accurate detection and identification of arcing current faults in MV-LV transformers. The presented method is based on extracting the magnitudes and phases of low frequency harmonics from the differential currents. Unlike magnetizing inrush and conventional fault currents, arcing current faults trigger currents that have harmonic components with non-stationary phases. These non-stationary phases can provide signature information of arcing current faults. Non-stationary phases can be extracted using a multi-channel filter bank that is composed of digital bandpass finite impulse response filters, each of which has a linear phase over a wide frequency range. The non-stationary phase method is tested during various fault and non-fault events. Test results demonstrate accurate and reliable detection, identification, and response to arcing current faults with minor sensitivity to loading level and/or arcing current fault type (series or parallel).