The distribution network has complex topology, large amount of branches, unbalanced power flow distribution, and imperfect communication network coverage. These disadvantages bring great difficulty to accurate fault diagnosis. In this paper, a fault diagnosis system based on 5G bearer network is constructed, the network delay and packet loss model are built, and the time delay of the actual network system. Then the Empirical Wavelet Transform (EWT) and Convolutional neural network (CNN) is introduced to the fault diagnosis. The different frequency components are obtained through the EWT transmission of the recorded electric quantity. The probability neural network model is constructed for each component to form the fault diagnosis method of EWT-CNN distribution network, and the fault judgment results are given. The experimental result demonstrates the effectiveness of the proposed EWT-CNN fault diagnosis method for distributed grid under 5G bearer network.