Feature extraction method of series fault arc based on ST-SVD-PCA

In order to study the characteristics and extract methods of series fault arc in underground coal mine power supply system,a series of fault arc experiments were carried out in motor and inverter load respectively. The time-fre- quency domain transform for loop current signal was conducted by using S-transform (ST). The amplitude matrix of S- transform was used as time-frequency feature matrix. The matrix singular value was obtained by conducting singular value decomposition (SVD) of the feature matrix. To reduce dimensions of feature vector,the principal component analysis (PCA) was carried out. The feature vector consists of many groups of singular value. The main component whose cumulative contribution rate higher than 95% was selected as fault feature. The validity of the extracted fault arc features were tested by using genetic algorithm (GA) optimized support vector machine (SVM). The compatibility of the arc fault identification method based on those fault arc features was also tested under different loads and operating conditions. It showed that the method can effectively identify the series arc fault occurred in motor and inverter load circuit.