Research on semi-supervised anomaly detection in automatic verification system of intelligent watt-hour meter based on direct push support vector machine

With the continuous advancement of the construction of smart grid, the demand for smart watt-hour meters is growing day by day, and the importance of watt-hour meter verification is increasingly prominent. In this paper, the semi-supervised anomaly detection of intelligent watt-hour meter automatic verification system based on direct inference support vector machine (TSVM) model is studied. To solve the anomaly problem that may occur in the long run of automated verification system, a method combining unsupervised anomaly detection and manual inspection is proposed to obtain a small number of labeled samples and construct TSVM anomaly detection model. The model uses labeled samples and unlabeled samples for training, and can label unlabeled samples, and continue to expand and optimize through semi-supervised training in the future automated verification work. By analyzing the data of State Grid Shanghai Electric Power Company automatic verification system, the validity of this method is verified, the error rate of unsupervised anomaly detection algorithm is reduced, and the working efficiency of the automatic verification system is improved. The experimental results show that the TSVM model has a significant improvement in the accuracy of anomaly detection, which is of guiding significance for the intelligent operation and maintenance of the automatic verification system of intelligent energy meters.