Novel Approach to Arc Fault Identification with Both Transient and Steady State Based Time-Frequency Analysis

The occurrence of electric arcs will pose a huge threat to personal safety and equipment safety. The identification and prevention of electric arcs is an indispensable part of ensuring electrical safety. Due to the dynamic nature of arcing behaviors and the various factors related complex characteristics of fault representation, the fast and reliable identification of arc fault face many of challenges. This paper proposed a novel method of arc fault identification with both transient and steady-state based time-frequency analysis. The Complete EEMD with Adaptive Noise with Hilbert Transform has adopted to analyze the actual arc faults and the extraction method of arcing initiation feature in the time-frequency domain is described. In addition, the correlation and sensitivity analysis to screen out the key-signatures of arc faults, to fit the calculation formula of steady-state current. At the same time, to fast and reliable detect the potential arc fault, a LSTM based arc fault identification strategy is proposed. With the series of actual arc fault cases under different configurations, the effectiveness has been validated thought multi-scenarios based calculation and comparisons.