Fault Segment Recognition and Localization in Distribution Networks Based on Artificial Neural Networks
This article proposes a method for fault location in distribution networks.The proposed method utilizes an artificial neural network.To train the neural network,a series of specific features are extracted from the recorded relay fault signals.These characteristics are obtained by performing wavelet transform on three-phase currents and sequences and extracting high-frequency characteristics.Due to the occurrence of high-frequency faults,wavelet transform can be used to extract signal information.After wavelet transform,statistical methods can be used to obtain the small components of the sequence and the entropy of the three-phase signal,extract hidden features,and present them separately for training the neural network.In addition,since the input obtained for neural network training strongly depends on the fault angle,fault resistance,and fault location,training data should be selected so that these differences are appropriately presented,so that the neural network does not face any recognition problems.There-fore,selecting signal processing functions,data spectra,and subsequent statistical parameters and their combinations are very important.Finally,after implementing the neural network,the fault cross-section,fault location,and fault re-sistance can be estimated.The simulation results show that the neural network has good performance for faults with different angles,positions,and resistances.
万方数据
维普
