Physics-data hybrid driven based topology and line parameter identification for AC/DC distribution network

To achieve the accurate identification in AC/DC distribution network, A physics-data hybrid driven method is proposed to predict the real-time topology and line parameters of AC/DC distribution network. Firstly, a framework based on physics-data hybrid-driven approach is proposed, which enables the rapid online identification of real-time topology and line parameters. Secondly, a pseudo-measurement model of the AC/DC distribution network is proposed considering the control mode of voltage source converter (VSC). Then, the adaptive spectral clustering (ASC) algorithm is proposed to estimate the number of historical topology categories, and the label-free topology discrimination (LTD) model is trained by machine learning methods according to the clustered data. Then, a two-stage physics-driven model is proposed to deal with the topology and line parameters identification problem using only a small amount of historical data with the same topology label. By leveraging the relationship between the data with topology labels and the results of the physics-driven identification, a label-to-topology and line parameters mapping model is built using the graph convolutional neural network (GCN), enabling rapid prediction of the topology and line parameters. Finally, the effectiveness of the proposed method is verified by case study.