Arc fault detection on load-side based on sensitive features tracking

The detection of series arc faults using fault current is difficult to overcome the influence of load types, making it difficult to establish a unified fault detection criterion. In contrast, since the arc voltage waveform of fault point is less affected by the load types and is basically a square wave shape, which provide conditions for constructing a unified fault criterion. In terms of the fault information, the fault distortion point of voltage on load-side caused by the arc voltage transition edge provides the position information of the arc voltage transition edge, and its polarity, amplitude and rate of change make it possible to distinguish from the transition edge caused by normal harmonic voltage drop, which provide the theoretical basis for fault detection using the voltage on load-side. Based on the basic analysis of arc voltage waveform features, this paper proposes an arc fault detection method based on load-side voltage sensitive feature tracking for the purpose of identifying the existence of arc voltage transition edges. The method proposed in this paper highlights the transition edge by eliminating the fundamental wave component of the voltage on load-side, the phase areas where the fault distortion points may exist are used as the sensitive area for fault detection, and the identification and tracking of the transition edge is achieved based on the same direction of voltage change, finally, the presence of arc fault voltage is characterized through the polarity, amplitude and rate of transition edge by fusion. The detection method proposed in this paper has a clear physical meaning and has the advantage of being less affected by the load types. Compared with other similar methods, the method proposed in this paper has higher detection sensitivity and stronger ability to distinguish from voltage drop distortion. The experimental results show that the average detection accuracy of the proposed method for faults detection under various loads exceeds 96%, which verifies the effectiveness of the method.