Six AI initiatives for a smarter and more resilient power grid

The transition to fossil-free energy is progressing too slowly, and the challenges exist across multiple levels. At the municipal level, there is a need to forecast future energy demands in order to make informed decisions about where to locate large energy-intensive businesses, add local power generation, and upgrade the grid to reflect changing consumption patterns. Similar forecasts need to exist at both regional and national levels.

"Sweden doesn’t have one. Instead, customers have to send inquiries to potential grid companies. And because they don’t know where capacity is available, they usually contact several companies at once. The result is a major issue with 'phantom bookings' – projects that never materialize but still consume time and resources from both grid companies and developers. This is one of the challenges we are tackling in a project together with Svenska Kraftnät and a few other grid companies," says Filip Kjellgren.

One key focus of that project is to investigate how synthetic data — data generated by computers based on real data — can be used to create so-called dynamic typical load profiles with the help of federated learning.These load profiles typically show how electricity consumption varies over time within an electrical grid, often broken down by customer segment, and they form the basis for analyzing grid capacity.

Several other projects are focused on the day-to-day operation of the power grid – specifically on how to  maintain a balanced electricity grid, even when there are large energy consumers in the close vicinity. One example is E-Charge 2, led by Lindholmen Science Park, with participation from Volvo Group and Scania, that investigates fast charging systems for electric heavy-duty trucks. Other similar initiatives are underway, all based on the same principle: using machine learning to analyze historical data to create real-time updates on energy consumption, production schedules, and grid availability.

Two of AI Sweden's energy projects have already been completed. One, carried out in collaboration with the Västra Götaland Region, resulted in a visualization tool to explore the possibilities for local energy production. The other, developed with Region Östergötland, focused on a similar need of visualization, to help municipalities understand how land use planning can support energy resilience.

"A challenge we’re seeing is that the energy sector's use of AI is limited by a shortage of expertise. That’s why collaboration is so incredibly important, not only to share knowledge, but also to scale ready-to-use solutions. We're seeing a growing interest in open-source development as a way of reaching a wider circle of users," says Filip Kjellgren.

Anna Svensson continues:

"The fact that we have projects underway in the energy sector is a good start. But our ambition is to expand into more energy-related projects that involve actors outside the energy industry as well. This involves, among other things, building shared knowledge and learning from analysis methods that have already proven successful in other sectors. This fall, we plan to put together a team to support a national collective effort, where the energy sector and the real estate sector meet in the application of AI."

AI Sweden's Energy Projects

AI Sweden has several energy projects underway, all part of a broader sector initiative.

Ongoing

• Federated machine learning for creating typical load profiles: This project uses federated machine learning to create dynamic electricity usage profiles across customer segments. These profiles can serve as a basis for national grid capacity maps that can guide decisions on suitable locations for new industries and/or power generation.
   
• Simulation and visualization of Sweden's future electricity system: Building on the existing "Behovskartan" (Needs Map) and collaborating with the Swedish Energy Agency, this project aims to create an improved digital decision-support tool for examining Sweden's future electricity needs.
   
• The future power grid created with federated machine learning: In partnership with several grid companies, this project investigates how to create models that identify different types of electricity grid customers to support future grid expansion.
   
• E-Charge2: AI solutions to optimize charging infrastructure for heavy transport: This project will analyze and optimize the design of charging infrastructure for electric heavy-duty trucks. The goal is to ensure efficient grid use while also minimizing charging infrastructure costs.

Completed

• Generation Toolkit: VGR (Västra Götaland Region): This project developed a decision-support tool to analyze the possibilities for new power production in municipalities within the Västra Götaland Region and to identify the most cost-effective energy sources to build.
   
• Capacity maps for municipal planning: This project analyzed the design and function of a potential digital decision-support tool. The aim was to help municipalities align land use with the development of a resilient and competitive municipal power grid.

For more information, please contact: