Research on Extracting Potential DC Arc Fault Features Based on Data Mining Methods

Due to the great difference in the properties of cable materials used in DC lines, the time-frequency features of DC arc fault that are easy to extract are difficult to take into account various cable materials. Therefore, it is necessary to mine potential arc fault features of different materials from the arc fault signals to meet the needs of distinguishing fault state from normal state. Firstly, the current waveforms of DC arc fault are obtained from the experiments of different electrode materials, four type of data mining methods are used to mine the potential DC arc fault information in the current signals to extract the DC arc fault detection features. Then, based on the principles of the proposed data mining methods, the mining results are compared to obtain the optimal mining features for various electrode materials. Finally, the construction of DC arc fault detection algorithm for different electrode materials is realized based on SVM model. The detection results indicate that the potential arc fault features can increase the accuracy of arc fault detection.