High Resistance Grounding Fault Detection Method for Distribution Network Based on Transfer Convolutional Neural Network
To address the issue of low operational safety in distribution networks caused by high resistance grounding faults,a transfer convolutional neural network based high resistance grounding fault detection method for distribution net-works is proposed.Firstly,the HHT method is used to extract feature quantities from the original signal,and the extracted results are input into the convolutional neural network structure.Through training and learning,the classification processing of feature quantities is achieved.Then,through transfer learning,the trained convolutional neural network model is put into a new task to detect again,so as to improve the detection ability of distribution network high resistance grounding fault.The experimental results show that after 160 iterations,the fault detection accuracy of this method is as high as 99.9％,and the network training errors are all below 1.5.In noisy environments,this method has strong noise resistance and is suitable for detecting faults in different types of operating conditions.The convolutional layer has a small impact on the detection accura-cy of the migrated CNN,and the stability of the migrated CNN in fault detection is good,which can improve the high resist-ance grounding fault detection ability of the distribution network.
transfer learningconvolutional neural networkdistribution networkhigh resistance groundingfault de-tection methodsfeature extraction
万方数据
维普
