A fault prediction method of distribution network based on the improved RelieF-Softmax algorithm

Abstract

A fault risk prediction method for distribution network based on improved RelieF-Softmax algorithm is proposed. Firstly, the fault data of distribution network is obtained and 14 fault characteristics are extracted through data investigation. The improved relief filtering feature extraction method is adopted to screen the initial fault feature set and select key fault features. Finally, a distribution network fault warning method based on the cross entropy loss function of focal loss is proposed, and the results of calculation examples demonstrate the effectiveness of the method proposed in this paper.