Integrating model-driven and data-driven methods for fast state estimation

With the high heterogeneousness of power loads and the large-scale penetration of renewable energy systems, power system states are changing more frequently. Accurately and efficiently achieving power system state estimation is becoming challenging. Data-driven methods shows strong application potential in state estimate for the high time efficiency. While ignoring the physical topology of power systems limits previous data-driven methods mining of complete information. In this paper, we propose an integrated model-driven and data driven model to monitor the dynamic change of system states accurately and rapidly. First, a model-driven method is used to extract the features with high entropy and reduce the computational complexity, which retains the strong coupling relationship between electrical parameters. Then, a graph neural network-based method is introduced to take full advantage of topology information. The edge and node features are selected as the input of data-driven method to fully adapt the internal laws of data and further improve accuracy. The simulation results show the superiority of the integrated method in terms of reliability, efficiency and accuracy.