Fault Type Identification in Distribution System Based on HHT and Neural Network

In order to accurately and reliably identify the type of distribution network fault, the fault type was divided into two types with the characteristics of the zero-sequence current. The two fault types are asymmetric ground fault and other faults. Then, by using the Hilbert-huang transform, the energy distribution of the fault current in different frequency bands was extracted. The fault type was encoded with binary code. The energy distribution of the fault current and the codes of the fault type were used as the samples of the two neural networks. The two neural networks were trained. Finally, when the fault occurred, the energy distribution of the current was extracted and imported into the corresponding BP network. And the outputs of the BP network are the codes of the fault type. The specific type of fault was identified. A number of simulations prove that this identification method accurately and reliably identify the type of distribution network fault. It improves the accuracy of the identification of various fault types to a large extent. Moreover, the method is free from the fault distance, fault time, the grounding resistance and the system operation mode.