Invited talk | Hendrik Hamann: Quo vadis - Grid Foundation Models

The growing complexity of electric power systems, driven by renewable energy integration, distributed energy resources, and increasingly dynamic operating conditions, calls for new approaches to grid modeling, analysis, and control. While artificial intelligence and machine learning have demonstrated significant promise across individual power system applications, most existing approaches remain task-specific, system-dependent, and limited in their ability to generalize across different networks and operating conditions.

This talk explores the emerging concept of Grid Foundation Models: large-scale, general-purpose AI models that learn transferable representations of power systems from heterogeneous data, network topology, physical models, and operational experience. We will discuss how advances in foundation models, graph-based learning, and physics-informed AI could enable a shift from narrowly optimized solutions toward adaptable models capable of supporting diverse tasks, including power flow estimation, contingency analysis, stability assessment, forecasting, and operational decision support.

Particular attention will be given to the challenges that distinguish grid foundation models from foundation models in other domains, including the need to respect physical laws, capture network dependencies, generalize across grid topologies and scales, and provide trustworthy predictions under changing operating conditions. We will conclude with a perspective on the research opportunities and foundational advances needed to develop scalable, physically grounded, and reliable AI models for the next generation of power systems.