The automatic IoT equipment of the new power system involves many types. Different equipment has specific performance parameters and working state indicators. The types of continuous attributes and qualitative attributes will increase due to the diversification of equipment types, which in turn increases the difficulty of fault diagnosis. Based on this, a fault diagnosis method for automated IoT devices based on rough set (RS) and dual-stream network (DSN) is proposed. Firstly, the fault judgment function of the IoT equipment of the new power system is constructed, and the three-phase current on the high-voltage side is calculated to determine the rejection or misoperation state. Then, for continuous attributes and qualitative attributes, the discretization algorithm is used to deal with them. The rough set method is used to set the equipment fault diagnosis decision table, and the positive region of the conditional attribute set is obtained, and the fault feature set is quickly output. Finally, the DSN composed of Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) is used to train the historical fault information samples to realize fault diagnosis. Experiments show that the proposed method can accurately diagnose the faults of automatic IoT equipment in the new power system and has good application performance.