There are many disadvantages in the fault detection of lowvoltage electrical apparatus in shields, such as lack of convenient and intelligent detection methods and judgment basis, complicated process, low detection efficiency, low accuracy and reliability. As a result, a method combining wavelet transform and deep learning technology is proposed to detect the performance index of lowvoltage electrical apparatus in shield. Based on several typical lowvoltage electrical faults in shield, the wavelet transform method is employed to extract the fault feature vector, which is taken as the input of the network, establishing a BP neural network for fault diagnosis network model. Finally, a detection method that can simulate the actual working conditions is proposed. The results show that: (1) The algorithm can quickly and accurately determine the fault problems and causes of the lowvoltage electrical apparatus in shield, and solve the problems of the existing detection device, such as single function, low efficiency, and no advance prediction ability of the electrical system remanufacture, full disassembly, and full inspection process when the shield stops. (2) The detection method is practical and stable, which can realize the simultaneous detection of multiple electrical components, thus greatly improve the work efficiency.