Abstract: Electric bicycles, balance bikes and other electrical devices have huge safety hazards in the charging process, which will lead to serious personal and property losses if they are not supervised. In this paper, we propose a feature fusion load identification method based on smart electric appliance network for the online supervision of electrical devices with safety risks. Firstly, we build a smart electricity network based on gateways, smart sockets and other IoT technologies, extract the features of risky electrical devices through sample data, and fuse the safety hazard features using principal component analysis. For the online identification of risky appliances, the SVDD algorithm is used to match the load features monitored by smart sockets to achieve accurate identification of the identity and hidden danger status of risky appliances. The algorithm shows that the proposed method has a high detection rate of high load devices, which verifies the effectiveness and accuracy of the proposed method and is of great practical significance for improving the supervision ability of public places and investigating the safety risks in libraries and dormitories.