不同电极材料条件下改进Catboost算法的直流故障电弧检测算法

In order to study the interference of different electrode materials on arc fault detection in DC system, and to solve the malfunction problems of arc fault detection with avarious electrode materials, a DC arc fault detection algorithm based on Catboost algorithm is proposed. A DC arc fault experiment platform is built and the DC arc fault data from 6 electrode materials are obtained by the pulling-apart method. Through constructing the evaluation index of the arc fault detection characteristic, the wavelet transform method is used to extract time-frequency characteristics of arc faults more effectively for different electrode materials. The influences of different electrode materials on DC arc fault detection characteristics and efficiency of the algorithms are analyzed, and it is pointed out that the detection characteristic index has a positive correlation with the melting point and resistivity of electrode material. Compared with the existing threshold comparison method and the Adaboost algorithm, the proposed algorithm improves the range of arc fault detection and effectively solves the problem that it is difficult to accurately detect arc faults using pure aluminum electrode material. This algorithm can realize rapid detection of arc faults within 1.5 s for the 6 different electrode materials, and the detection accuracy reaches 100%, which meets the requirements of UL1699B standard.