Line Loss Anomaly Perception Method Based on MIC-IF Algorithm for Photovoltaic Low-Voltage Transformer Area

Abstract

The topology and line parameters of low-voltage transformer area are often difficult to obtain, and distributed photovoltaic (PV) access makes the distribution grid’s power flow characteristic changes, which leads to high difficulty in recognizing line loss anomalies in transformer areas. To address the above problems, a method for perceiving line loss anomaly in PV transformer area called MIC-IF is proposed. Based on this algorithm, the feature vectors of each transformer area are constructed by combining several operation indicators and line loss rate, and it is considered that the outlier vectors correspond to the transformer areas with abnormal operation states. After completing the judgment of the cause of anomalies, PV energy theft detection is carried out for areas in which the PV power generation is anomalous based on RUSBoost algorithm. Finally, the results of analysis are summarized and the conclusion on anomaly perception is obtained. The effectiveness of the proposed method is verified based on data from 20 simulation transformer areas, and the results show that the accuracy and F1-socre of MIC-IF reach 0.95 and 0.89, respectively, and are higher than the comparison algorithm. The detection framework takes into account the PV access and does not rely on line parameters, with high interpretability and accuracy, providing a certain reference for engineering applications.

Keywords

• Anomaly perception
• ensemble learning
• line loss rate
• maximal information coefficient
• photovoltaic low-voltage transformer area