Vortrag von Christian Nauck zu "Dynamic Stability Assessment of Future Power Grids using Graph Neural Networks"
Abract
The integration of renewable energy is transforming power grids into decentralized systems with complex, multi-directional dynamics and reduced stability margins. This talk presents a machine-learning approach for fast and scalable dynamic stability assessment in future power grids. While several models are explored, graph neural networks prove especially effective, generalizing from small synthetic grids to real-sized systems. A key contribution is a novel graph layer, the Dirac–Bianconi Graph Neural Network, designed to capture long-range interactions in power grids. The work also introduces the prediction of full stability landscapes rather than a single scalar score, providing interpretable insights for grid operators. These methods can greatly reduce the computational cost of probabilistic stability analysis and support more efficient decision-making in modern power systems.
About the speaker
Christian Nauck is …