Fault Identification and Classification Method of Distribution Network with Distributed Generation Based on IHBF-CNN

When faults occur in distribution networks with distributed generation, the fault characteristics are different from those of traditional distribution networks, so high accuracy fault classification is essential fault analysis. Based on Improved Hilbert Bandpass Filter (IHBF) and Convolutional Neural Networks (CNN), a fault classification model for distribution networks with distributed generation is proposed. IHBF is used to convert the fault signal into an energy matrix of time-frequency, and then the CNN neural network is constructed to classify the faults. Considering various parameters such as fault resistance, fault section, system frequency, as well as changes in network topology and neutral grounding mode, the proposed method is compared with four existing distribution network fault classification methods. It shows that the method outperforms other methods in terms of epoch consumption and fault identification and classification accuracy, and is not affected by fault parameters, network structure and neutral grounding mode, which fully demonstrates its robustness and great accuracy in the identification and classification of fault, which is applicable for distribution networks containing distributed power sources.