IN Brief:
- The Novo Nordisk Foundation is providing up to DKK100m for a four-year Nordic electricity-grid research programme.
- DTU, Lund University and NTNU will investigate AI methods combined with power-system models and controls.
- Research is scheduled to start in the first quarter of 2027; proposed speed and resilience gains are not operational results.
The Novo Nordisk Foundation is awarding up to DKK100m, approximately €13.3m, over four years for a Nordic research programme investigating artificial intelligence in electricity-system operation. AI4GRIDS will be led by the Technical University of Denmark, working with Lund University in Sweden and the Norwegian University of Science and Technology. The programme is expected to begin in the first quarter of 2027.
Wind, solar, batteries, electric vehicles, heat pumps and large industrial loads are increasing the number and variability of inputs that electricity-network operators must manage. AI4GRIDS will investigate whether existing network capacity can be used more effectively while respecting electrical limits and maintaining resilience. Optimising dispatch or control is different from building additional lines or substations; the programme does not claim to replace physical reinforcement where it is required.
Researchers intend to combine artificial intelligence with mathematical models of electrical networks, forecasting, optimisation and control theory. Physical models can represent constraints such as thermal loading, voltage limits and the balance of active and reactive power, while data-driven methods may help search for feasible operating choices. A proposed action is only useful if it remains valid for the actual network configuration and can be executed within the operating and protection arrangements already in place.
Interconnected electricity systems must accommodate fast-changing weather-dependent output and demand whose timing is only partly predictable. A forecast rise in solar generation may coincide with charging demand or a network element being out of service, changing the options available to operators. The programme therefore brings together different specialisms rather than treating a numerical prediction as an instruction to operate switchgear or dispatch generation.
The foundation describes a longer-term ambition for some AI-supported calculations to be performed up to 100 times faster. This is a target for research, not a benchmark demonstrated by AI4GRIDS in an operational control room. Neither the programme announcement nor its funding decision establishes a measured improvement in network capacity, reduced curtailment or fewer outages. Those outcomes would require defined test cases and validation against practical operating constraints.
Professor Spyros Chatzivasileiadis of DTU Wind and Energy Systems is a named leader in the initiative. DTU will provide expertise in electrical power systems, trustworthy AI, energy data and computing. Lund University will contribute mathematical statistics, forecasting and control, while NTNU will focus on markets, optimisation and decision-making when inputs and outcomes are uncertain.
Power networks impose a particular challenge on automated decisions because an apparent optimisation in one part of an interconnected system may affect flows elsewhere. Transmission corridors, voltage conditions and the ability of generators or batteries to respond create linked constraints. A useful operating algorithm must therefore account for the network representation and operating rules, rather than simply matching predicted supply and demand in each location.
The researchers envisage AI-based tools assisting with safety-critical decisions while human operators retain supervisory control. In practice, proposed control actions would need to be assessed against incomplete measurements, changing network conditions and credible contingencies, with a route for intervention where an automated recommendation cannot be justified. The release describes research intentions rather than a certified or deployed autonomous grid operating system.
Nordic electricity systems offer a cross-border setting for this work. Denmark, Sweden and Norway have interconnected power markets and networks with different generation and demand characteristics, including substantial renewable and hydroelectric resources. Coordination across national boundaries and different network areas raises both forecasting and control questions. Research findings may ultimately be applicable elsewhere in Europe, but deployment will depend on the relevant grid codes, operating procedures and data access.
The project will also study demand-side flexibility. When an electricity user can change its consumption within an agreed operating envelope, the response may help relieve a short-lived local constraint or balance supply and demand. Such flexibility cannot raise the physical rating of an overloaded cable or transformer, and participation depends on communication, commercial arrangements and the availability of controllable demand at the relevant time.
More than 30 doctoral researchers, postdoctoral researchers and other specialists are expected to receive training through the programme. The participating universities plan to work with system operators, equipment developers, regulators and other organisations, connecting academic research with the practical requirements of electricity-network control rooms.
The four-year award is intended to fund development and validation of new research methods, with work starting in the first quarter of 2027. Operators, technology companies and regulators are expected to contribute practical requirements alongside the university teams. Adoption on a live network would still depend on demonstrations of reliability, integration with existing controls and governance suited to critical infrastructure; no commissioning date has been announced for an AI4GRIDS operational product.
