Session Chairs: Pedro Vergara Barrios, (TU Delft) and Fei Ding, (NLR)

This session examines how generative AI can support distribution systems planning and operation. Distribution systems differ from transmission systems in several ways: large amounts of data are available from consumers, while most distributed energy resources (for example, electric vehicles, solar panels, electric heat pumps) are connected to the electricity infrastructure at medium and low voltage levels. This creates the perfect context for AI: large amounts of data are available, while generalization is needed to handle the rapid pace of context change. Topics include the use of synthetic data generation for distribution systems’ hosting capacity, foundation models using graph-based power flow for scenario-based planning, and risk quantification.

Designing AI models for Intelligent Distribution Grid Operation

Speaker: Zita Vale – ISEP/IPP – Instituto Superior de Engenharia do Instituto Politécnico do Porto / Engineering Institute – Polytechnic of Porto

Power and energy systems transition towards more efficient and sustainable operation requires distribution grids to adopt new decision-making models. With energy resources becoming increasingly distributed and new active roles seen for consumers, virtual power producers, and energy communities service procurement is becoming increasingly important for distribution grids making decision-making an important challenge. Artificial Intelligence-based models are needed to face this challenge, ensuring not only adequate approaches for sustainable and efficient energy use but also enabling the fair and efficient participation of all the involved actors.
This talk will discuss the requirements and options for the design of AI models for distribution grid operation. Data driven and knowledge-based approaches that can ensure the required intelligent decision-making in the frame of current and future power and energy systems will be presented and discussed. Artificial Intelligence traditional paradigms and the boom of the new Large Language Models (LLM) and agentic approaches will be covered and the perspectives of their use will be discussed.