基于DNN-SVM的无线专网设备故障识别与定位系统研究

Research on Fault Detection and Localization System for Wireless Private Network Devices Based on DNN⁃SVM

针对电力无线专网系统中故障识别与定位问题,提出了一种基于DNN-SVM的解决方案。电力无线专网系统在面临故障时往往难以及时、准确地进行故障识别和定位,给运维和维修工作带来了困难。本文通过专用设备对电力无线专网系统的实时数据进行采集,包括信号强度、信号质量、 CPE运行状态等信息,构建一个DNN-SVM算法,以同时实现无线专网故障识别与故障定位,通过DNN判别故障状态,多层二分类SVM判别故障类型。通过在实际电力无线专网数据集上进行实验和验证,对于单组数据的判别时间在毫秒级,而综合平均准确率为80%。

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 clas⁃sification.Experimental validation is conducted on an actual electric power wireless mesh network data⁃set.The decision time for a single data sample is in the millisecond range,and the overall average accu⁃racy rate is 80%.