Metering voltage anomalies directly affect the distribution line loss,and manually identification of voltage a-nomalies is inferior in efficiency.To address this problem,a voltage anomaly detection method based on convolutional neu-ral network was proposed.First the voltage data were unified into three-phase data and the voltage differences among the three phases were calculated and standardized.Then the three-phase voltage was scaled to preserve the size information of the voltage.Subsequently the standardized voltage differences and the scaled voltage data were combined into 6-channel da-ta.Finally,using the convolutional neural network,a voltage anomaly classification model was designed and trained,which could distinguish voltage anomaly states.Experimental results showed that the overall accuracy and recall rate of this method can reach 97％,indicating high accuracy.
关键词
电能表/卷积神经网络/电压异常检测/台区线损
Key words
electric meter/convolutional neural network/voltage anomaly detection/distribution line loss
维普
万方数据