Topology Identification Method for Distribution Network Based on Convolutional Neural Network

Switching operations cause changes in the topology of the distribution network, complicating topology identification. Additionally, anomalies in the measurement data further reduce the accuracy of topology identification. To address these issues, this paper proposes a topology identification method based on Convolutional Neural Networks (CNN). Firstly, a data repair method is proposed to enhance data quality and ensure the reliability of model training. Next, a CNN is used to extract the characteristics of voltage data under different topologies. By leveraging the potential correlation between voltage data and topology, identification can be achieved without the need for all-node or advanced measurements. Finally, simulations on a modified IEEE33 node distribution network validate the proposed method's accuracy, demonstrating its improvement over existing machine learning models.