Fault Location Method for Distribution Network with Distributed Generation Based on Deep Learning

Abstract

Aiming at the problems of insufficient feature extraction ability and low accuracy of current distribution network fault location methods, a distribution network fault location method based on deep learning is proposed. Firstly, the one hot coding rule is used to preprocess the fault branch. In order to prevent the order of magnitude difference in the data, the collected data are normalized. Secondly, in order to better mine the characteristics of data, an improved Bi-long short term memory (BiLSTM) fault location method is proposed, which combines the advantages of BiLSTM and batch normalization mechanism. Finally, the I-BiLSTM network and U-BiLSTM network are trained respectively, and the output of dual network neurons is selected at the Softmax output layer to formulate logic gate rules. By configuring measuring points at each node in the distribution system topology, the fast and accurate mining of section dual terminal timing characteristics is realized. The experimental results show that the accuracy of the proposed method is 0.958, which is much higher than the comparison method. Therefore, the proposed method has good fault location ability in distribution network.