基于声纹信号的变压器故障在线监测装置设计

Design of Online Transformer Fault Monitoring Device Based on Voiceprint Signal

针对目前不停电检测技术中,对变压器绕组变形、直流偏磁等直接或间接引发的机械缺陷不能及时作出响应的问题,研发了一套基于变压器的振动声纹信号的全天候在线故障监测装置。声纹传感器采集变压器的声纹信号,通过与现场分布式采集模块相连接的存储、分析控制箱的现场分析和疑似上传的模式进行信号上传。最后由存储分析软件将一级简单分析后的疑似故障信号通过网络上传至数据中心,再结合深度学习对数据进行深入分析,判定故障具体原因和解决方案。装置不但对提高电网主设备状态监测水平,减少巡检人力、物力和完善故障诊断技术具有重要意义,而且还可提高整个电网的稳定运行水平。

Aiming at the current uninterrupted power detection technology that can not respond in time to mechanical defects directly or indirectly caused by transformer winding deformation, DC bias and etc., a set of all-weather online fault monitoring devices based on transformer vibration and voiceprint signals was developed. The voiceprint sensor collects the voiceprint signal of the transformer, and uploads the signal through the on-site analysis and suspected uploading mode of the storage and analysis control box connected to the on-site distributed acquisition module. Finally, the storage analysis software uploads the suspected fault signals after the first level simple analysis to the data center through the network, and then conducts in-depth analysis of the data in combination with deep learning to determine the specific causes and the just solution of the fault. The device not only improves the status monitoring level of the main equipment of the power grid, reduces inspection manpower and material resources and makes perfect the fault diagnosis technology, which is of great significance, but also enhances the stable operation level of the entire power grid.