Abstract: A solution based on DNN-SVM is proposed for fault detection and localization in the electric power wireless mesh network system. The timely and accurate identification and localization of faults in the electric power wireless mesh network system pose challenges for maintenance and repair work. In this paper, real-time data from the electric power wireless mesh network system, including signal strength, signal quality, and PCE operating status, is collected using dedicated devices. A DNN-SVM algorithm is constructed to achieve simultaneous fault detection and localization in the wireless mesh network. The DNN is used to discriminate fault states, while the multilayer binary SVM is employed for fault-type classification. Experimental validation is conducted on an actual electric power wireless mesh network dataset. The decision time for a single data sample is in the millisecond range, and the overall average accuracy rate is 80%.