Anomaly Detection for Transmission Loss Data Using Regression Method

Abstract

The large amount of historical line loss data accumulated by the power grid company can provide a basis for the work of loss reduction. To explore the cause of transmission loss in the case of poor data quality, a novel abnormal transmission loss data detection method is proposed. Firstly, the type and reliability of the line loss data are analyzed based on the engineering practice, and the loss models are established from the aspects of line loss and transformer loss. Secondly, the nonlinear regression models are established according to the established loss models, to find out the abnormal data more accurately using the limited data. Further, the data detection process is designed to include data classifying, preprocessing, modeling, and anomaly detection. The case study shows that the proposed abnormal data detection method has better accuracy than the traditional method and can locate anomalies to specific lines and transformers, which provides a better basis for loss reduction work.