Research on feature of series arc fault based on improved SVD

In order to study the feature and extraction methods of series arc fault, the series arc fault experiments under different current conditions were carried out with the motor load and inverter respectively. A method of feature extraction based on improved singular value decomposition was proposed, and arc faults were distinguished by support vector machine (SVM). SVM was optimized by genetic algorithm (GA). Current signals were used to structure the attractor track matrix, and the time- delay step of the matrix was reconstructed by autocorrelation analysis. By means of singular value decomposition of the trace matrix, singular values of the matrix were obtained, the feature of arc fault were obtained by screening these values. Finally, GA- SVM was used to test the feature of the arc fault. The results showed that the method could effectively extract the series arc fault feature in the motor and inverter load circuit.