Anomaly Data Mining Method of Electric Power Metering Automation System Based on Improved Threshold Algorithm

When analyzing abnormal data of electric power metering automation system, because the data itself exists in a dynamic form, the accuracy of the identification results of abnormal data is poor. Therefore, this paper proposes a research method for abnormal data mining of electric power metering automation system based on improved threshold algorithm. Considering that, when the least square method is directly used for analysis, the data of the power metering automation system is processed in a unified way, which is difficult to ensure the adaptation of the mining results to the data state of the real-time power metering automation system. When the threshold algorithm is improved, the weighted mechanism is introduced, and the weighted least square method is used to improve the threshold algorithm. In the abnormal data mining stage, the direct reconstruction method is used to analyze the relationship between the actual data and the norm, and the improved threshold function is used to analyze the difference of the reconstructed data to judge the abnormal data. In the test results, the accuracy of the design method for abnormal data identification reached 77.78%, with high accuracy.