Fault diagnosis of microgrid based on wavelet decomposition and convolution neural network

Abstract

The topological structure of the microgrid is different, and the internal power flow changes in both directions, which makes the fault diagnosis and protection of the microgrid difficult. In addition, the research on microgrid failures in the past years has mainly focused on internal failures, with few mentions of failures in network tie lines. This article proposes a fault diagnosis method combining Meyer wavelet decomposition and convolutional neural network. Through wavelet decomposition of the three-phase voltage and current signals at the time of the fault, approximate coefficients and detail coefficients are extracted, and the convolutional neural network input is formed to establish the fault. Diagnosis model to detect fault type and fault location. Experiments show that this method has high diagnostic accuracy, and it can efficiently diagnose faults in the network and the faults of the interconnection line between networks. Finally, compared with the traditional BP neural network and the original convolutional neural network, it also has better fault diagnosis result.