Value in Energy Data: Towards Interpretable AI Assistants for Power System Operations

Moving beyond purely data-driven models, these assistants integrate real-time data with structured domain knowledge and expert reasoning. By leveraging neurosymbolic learning approaches, they offer improved interpretability, robustness and generalisation—key requirements for deployment in safety-critical environments such as power system operations.

Drawing on the AI4REALNET project, the session will demonstrate how these AI assistants are embedded within a human–AI cooperative framework, where they are designed to augment rather than replace human expertise. Examples will illustrate how they support both predictive and real-time analytics, helping operators with tasks such as system monitoring, identifying remedial actions and informing operational planning.

Attendees will gain a clear understanding of how advanced AI architectures can be practically applied to deliver more explainable, reliable and scalable decision support in modern power systems, and what this means for the future role of human operators.