Topology-Aware Dynamic Power Flow Calculation Method for Power Grid Based on Graphormer-GE

Abstract

To balance prediction accuracy and computational efficiency in grid modeling, this paper proposes Graphormer-GE, a topology-aware dynamic power flow calculation method. It not only reflects the steady-state power and voltage distribution at a certain moment but also characterizes the continuous evolution behavior of the system under the influence of time-varying factors such as topology disturbances, load fluctuations, and new energy integration. The proposed Graphormer-GE method addresses the dynamic power flow challenge by constructing a high-dimensional graph integrating electrical characteristics and real-time communication status from IoT devices. It embeds electrical distance and physical topology through a geometric encoding mechanism to enhance sensitivity to topology changes. By establishing a unified attention computation space and jointly modeling temporal evolution and spatial relationships among nodes, the expressive capability of the model is significantly improved. Furthermore, a dynamic graph update strategy incorporating communication link quality ensures stable performance in network-constrained environments. The proposed framework provides an effective computational paradigm for intelligent power systems and electromagnetic energy transmission networks requiring accurate state perception and propagation analysis. Experiments on the IEEE 300-node system demonstrate superior accuracy (MAE 0.037 p.u.) and response speed (Postfault Error 0.07 p.u.), while maintaining stable performance during topological disturbances (error within 0.05 p.u.) across diverse communication environments, confirming its practical value for distributed power system monitoring and reliable electromagnetic infrastructure operation.