Analysis of Abnormal Freezing Data and Updating Algorithm for Electromechanical Energy Meter Terminals

Abstract

Due to the rapid development of science and information technology, electricity information acquisition system has been widely used in the electricity data collection of users. With the massive electricity data collection, it is difficult to adopt traditional data processing methods to meet the abnormal data processing. In order to effectively mine abnormal information in electricity consumption data, an anomaly detection model based on the Isolation Forest (iForest) algorithm is proposed. Firstly, the daily load curve with strong regularity is used as the characteristic index of anomaly monitoring, and the users with abnormal electricity consumption data are preliminarily screened. Secondly, on the basis of electrical variables, the suspected abnormal users are further analyzed, and the anomaly identification model of electricity data is established to automatically classify the voltage at the metering point. Moreover, combined with the current data, the abnormality of the electric energy metering device is identified, and then the validity of the model is determined through on-site verification. Finally, according to the participating voltage of the fault phase, the 96-point voltage data frozen during the failure period is analyzed and the correction coefficient is adjusted. The results reveal that the electricity data detection model based on the iForest algorithm has significant advantages in computational efficiency. Through the cumulative recall and Precision-Recall (P-R) curves of the model, it is found that the majority of abnormal users can be detected only by detecting a few users with high abnormal scores, which shows that the model has high efficiency. The decision tree algorithm combined with the current data can effectively identify the anomalies of the energy metering device, which verifies the validity of the anomaly identification model of the electricity consumption data.