The electrical fault in ship integrated electric propulsion system is the most important fault type that affects the stable operation of the system. When an electrical fault occurs on board, it may interrupt the continuity of the power service, and if the fault is not detected and isolated in time, it will lead to serious damage to the electrical equipment. In this study, fault waveforms of different electrical equipment were analyzed from the perspective of failure mechanism according to the characteristics of each electrical equipment in the ship integrated electric propulsion system, and a hybrid network model of convolutional neural network (CNN) superimposed bidirectional long short-term memory (BiLSTM) network based on deep learning is proposed. This model can effectively capture the classification features of input data at different time scales. Combined with the time series signal of SE attention mechanism, the correlation of feature sequence is highlighted by weight, so as to improve the accuracy of fault diagnosis. To verify the effectiveness of the proposed method, comparison experiments was performed to evaluate the diagnostic performance of the network model.