基于色度图与卷积神经网络的开关柜运行状态监测方法

Monitoring Method for Operation State of Switchgear Based on Chromaticity Diagram and Convolutional Neural Network

为更有效地监测开关柜设备的工作状态,以开关柜为研究对象,根据非接触式的声纹信号对设备工况进行评估,提出了一种基于色度图与卷积神经网络的开关柜运行状态监测方法。首先将采集得到的声纹信号变换为色度图谱,进而构造开关柜正常、异常状态的特征图谱集,再利用卷积神经网络进行特征的深度挖掘,最终实现开关柜不同状态的辨识。实验结果表明,所提方法能够有效表征开关柜不同的工作状态,辨识准确率可达99.2%。

In order to monitor the working state of switchgear more effectively,the switchgear was taken as the research object,and the working state of the equipment was evaluated according to the non-contact voiceprint signal.A monitoring method of working state of switchgear based on chromaticity diagram and convolutional neural network is proposed.Firstly,the collected voiceprint signal is converted into chromaticity diagram,and then the characteristic atlas of normal and abnormal state of switchgear are constructed.Then the convolution neural network is used to deeply mine the features,and finally the different states of the switchgear can be identified.The experimental results show that the proposed method can effectively characterize different working states of switchgear,and the identification accuracy can reach 99.2%.