Fault Status Monitoring of Power Metering and Detection Equipment Based on BiLSTM

Abstract

The power metering and detection equipment is a device used to measure the power consumption of users and energy loss of power lines. It is of great significance to timely identify and handle operational anomalies of electric power metering and detection equipment to ensure stable and reliable operation of the metering device. In response to the problems of low accuracy and poor robustness in existing fault diagnosis methods for electric power metering and detection equipment, this paper proposes an intelligent fault diagnosis method based on deep learning. The paper combines the characteristics of convolutional neural network (CNN) and bidirectional long short-term memory (BiLSTM) in extracting local and global features, and combines them with self-attention (SA) mechanism to propose a hybrid network model SA-CNN-BiLSTM for monitoring the status of electric power metering and detection equipment. This model first extracts local information through the CNN network. At the same time, BiLSTM can extract global features, and then the feature fusion layer combines these two different features. Then, the self-attention mechanism weight adjustment layer is used to calculate and adjust the weights of the fused feature vectors. Finally, the fused feature vectors are input into the fully connected layer, and fault classification predictions are obtained through a softmax classifier. System experiments have demonstrated the correctness and effectiveness of this method.