Abstract:

The power distribution network is directly connected with the user.Its stability is closely related to the power system’s ability to deliver power to the user side.The operation environment of distribution network is complex and extensive,and high impedance fault (HIF) occurs when the operation line falls into contact with trees,grass and other conditions.At this point,the impedance of the fault point reaches several hundred or even several thousand ohms,and the amplitude of voltage and current changes weakly,so the fault is difficult to be detected.If the fault can’t be eliminated in time,the intermittent arc at fault point will cause immeasurable damage.In this paper,the Hilbert-Huang Transform (HHT) band-pass filter is used to extract the feature quantity,to form the time-frequency energy matrix,one-dimensional convolutional neural network (1D-CNN) is used to construct classifier for fault classification.The simulation model is verified and the adaptability is analyzed.The results show that the algorithm has high accuracy and good adaptability.

Key words: distribution networks, high impedance fault, one-dimensional convolutional neural network (1D-CNN), fault classificatio