The identification of aviation arc fault based on the BP neural network

Aviation arc fault belongs to low voltage arc fault. And it has the characteristics of short duration, low current intensity, etc. A method to identify aviation arc fault under DC power supply system based on BP neural network and wavelet packet is proposed in this paper. Firstly, the arc fault experiment platform is built to collect the current signal under normal circuit and arc condition. A base wavelet selection criteria maximum total energy to total Shannon entropy ration is used to select an appropriate wavelet for feature extraction. Then the Shannon entropy of characteristic frequency range is extracted as the characteristic quantity of arc fault. In addition, the effect of atmospheric pressure on the arc fault is also taken into account. Finally, the BP neural network optimized by Leverberg-Marquard algorithm is used to identify the arc fault. The final results show that the method proposed in this paper has a recognition rate of more than 98% for the arc fault.