A convolutional neural network combined with bidirectional long and short-term memory neural network (CNN-BiLSTM) fault diagnosis model is proposed for the open-circuit fault problem of three-phase inverters in permanent magnet synchronous motor drive systems. Firstly, different open-circuit fault types of the inverter are analyzed and labels are set, then the three-phase currents output from the corresponding permanent magnet synchronous motors are extracted as the fault feature signals, and the frequency-domain feature signals are extracted using the complementary ensemble empirical modal decomposition (CEEMD), and finally, the identification and classification of the open-circuit faults of the three-phase inverter are realized using the CNN-BiLSTM. Experimental results show that this method can effectively and accurately complete the identification and classification of three-phase inverter open-circuit faults.