摘要
由于长时间运行、设备老化、外部环境等原因,变电站设备存在着故障的风险。因此,对变电站设备的故障进行有效的监测和预警,成为电力系统安全运行的关键。而视觉几何组(VGG)算法是一种基于卷积神经网络的深度分析模型,该算法具有多层卷积和池化层,能够提取设备声频信号中故障特征来源,并进行分类,最终实现对变电站设备的运行状态进行实时分析。为此,通过改进VGG网络变电站设备的声频信号特征提取,实现了对设备故障的准确识别和预测。
Substation equipment has the risk of failure due to long-term operation,equipment aging,external environment and other reasons.Therefore,effective monitoring and early warning of substation equipment faults has become the key to the safe operation of power system.The visual geometry group(VGG)algorithm is an in-depth analysis model based on convolutional neural network,which has multi-layer convolution and pooling layers,and can extract the fault feature sources from the audio signals of equipment,classify them and finally realize real-time analysis of the operation state of substation equipment.Therefore,this paper improves the audio signal feature extraction of VGG network substation equipment,so as to realize accurate identification and prediction of equipment faults.
作者
李志锦
谭劲章
LI Zhijin;TAN Jinzhang(Foshan Power Supply Bureau of Guangdong Power Grid Co.,Ltd.,Foshan 528000,Guangdong,China)
出处
《流体测量与控制》
2024年第4期28-31,共4页
Fluid Measurement & Control
