Research on the Features of the Series Arc Fault Detection and Phase Selection Based on Singular Value Decomposition and Multi-scale Norm

Electrical fires caused by arc fault are very common. To effectively prevent the electrical fire accidents resulted by the arc fault and reduce the troubleshooting time, a feature extraction method of the series arc fault (SAF) detection and phase selection in the three-phase motor and frequency converter load circuit was proposed. First, the SAF experiments under different lines conditions were conducted by using the three-phase motor and frequency converter loads. Second, a reconstruction of attractor track matrix was constructed, and combined it with the singular value decomposition to enhance the fault features. Third, the features of the SAF detection and phase selection were extracted from the feature enhancement matrix by using the multi-scale norm. Finally, by employing particle swarm optimization algorithm to refine a support vector machine, a detection and line selection model for the SAF was developed. According to the test results, the detection and phase selection accuracy of the SAF in the three-phase motor and the frequency converter load circuit were 98.12% and 91.63% respectively. It can serve as a point of reference for the development of the industrial arc fault circuit breakers with the SAF phase selection function.