SafePowerGraph-HIL: Real-Time HIL Validation of Heterogeneous GNNs for Bridging Sim-to-Real Gap in Power Grids
Aoxiang Ma, Salah Ghamizi, Jun Cao, Pedro Rodriguez·January 21, 2025
Summary
The SafePowerGraph-HIL framework validates heterogeneous GNNs for power grid state estimation and analysis using real-time HIL simulations on the IEEE 9-bus system. This dataset, generated by Hypersim, enhances HGNN training and accuracy, demonstrating potential for intelligent, adaptive control strategies in evolving power grids. The framework uses a heterogeneous GNN model for state estimation and dynamic analysis, integrated with AWS for scalable storage and processing. It fine-tunes a GNN with 500 sets of data, predicting system parameters like active and reactive powers, bus voltages, and angles, with an MSE loss for training. The model's effectiveness is validated through fine-tuning with real-time data, reducing validation loss for bus voltages and external grid by 13% and 75%, respectively. Future work aims to extend the framework to more complex networks with renewable energy sources.
Advanced features