Research on Electric Energy Metering Anomaly Detection and Classification Algorithm Under Multi-Source Data Fusion

Abstract

The renewable energy grid is a distributed power grid, which needs to be parallel to the thermal power grid to ensure the stability of its own power generation. Therefore, renewable energy has the problem of multiple sources of power generation data and complex electric energy metering, which has always limited the development of renewable energy. In order to improve the grid-connected power supply quality of renewable energy and improve the accuracy of electric energy metering, this paper proposes a binary classification algorithm to collect, amplify and standardize multi-source data. Then, the Fourier function is used to normalize the multi-source data to shorten the differences between different devices. In particular, the attribute features are classified and the input data adjustment parameters are set. Finally, renewable energy's electric energy data value is measured, and the abnormal data characteristics are obtained through the second-order derivative to realize data mining and electric energy metering anomaly detection. The simulation results show that the binary classification can complete the metering and anomaly detection of electric energy in the distributed power grid, and the classification processing can simplify the complexity of multi-source data, improve the accuracy of metering anomaly detection, and make its dispersion more reasonable, and the results are between 90~95%, which is better than the online monitoring method. The time for anomaly result detection is 25s, which is shorter than the previous algorithm, and cross-domain data search is realized, with a cross-domain rate of 50~60%. Therefore, the classification algorithm proposed in this paper can meet the requirements of electric energy metering under multi-source data fusion and achieve rapid anomaly detection.