(Frankfurt am Main, July 30, 2025) The power grid is a dynamic system with millions of players in which feed-in and consumption must be well coordinated. This is why grid control technology is so important, as it is part of the critical infrastructure that ensures this coordination runs smoothly. In its new background paper "Artificial Intelligence in Grid Control Technology," the Power Engineering Society within VDE (VDE ETG) highlights the key role AI systems can play in this task. They can collect and analyze vast amounts of data and provide decision-making support – whether it's forecasting the output of photovoltaic systems, electricity consumption at a specific time of day, or identifying potentially critical conditions. "Artificial intelligence expands the toolbox for decision-makers," says Dr. Ralf Petri, Managing Director of VDE ETG. "It's not about losing control, but about achieving better decision-making quality and thus ensuring that grid operation remains responsible in the future."
Artificial intelligence is not new in grid operation
The use of AI in critical infrastructures is often questioned. However, VDE ETG makes it clear that there are already various applications in the power grid in which AI systems work reliably or have proven themselves in field tests. For example, load and feed-in forecasts are determined on the basis of weather and camera data. Another example are field trials conducted as part of the GridAnalysis joint project (German Federal Ministry for Economic Affairs and Energy), which show that AI-based condition assessments in low-voltage grids can be carried out with a high degree of accuracy. Programs such as ChatGPT can be used as assistance systems in grid management. By integrating them into so-called Retrieval Augmented Generation (RAG) systems, they are able to provide relevant information from internal documents, technical regulations, or external sources on request. Such solutions can also highlight relevant information, thereby accelerating decision-making processes.
Four-stage implementation model for the use of AI in network control technology
"The introduction of artificial intelligence in network control technology is a sensitive issue," says Petri. "It requires clear requirement profiles, verifiable development, robust testing, and controlled operation." The VDE ETG has therefore developed a four-stage implementation model based on established processes such as Technical Safety Management (TSM), the Information Security Management System (ISMS, ISO/IEC 27001) and software development in accordance with the V-model (ISO 26262-6). The regulatory framework for the use of AI is the EU AI Act, whose requirements are also taken into account by VDE ETG in the proposed model.
However, the authors of the background paper point out that the term "AI systems" is very broadly defined in the AI Act. They also note that the definition of technical documentation leaves room for interpretation, making it difficult for manufacturers to comply with the documentation requirements. "In our view, regulators, manufacturers, and operators need to engage in dialogue to clarify the situation. We need a technical definition of the term AI, and the classification of high-risk use cases needs to be tightened up," Petri states. "The clearer the requirements are, the easier it will be to enforce and monitor compliance."
The background paper "Artificial Intelligence in Network Control Technology" (German) is now available for download free of charge.
