Tag: AI

As flexibility needs and energy security concerns grow, artificial intelligence (AI) is playing an increasingly central role in managing, optimizing, and securing energy systems. At the 2025 Energy Talk: AI and the Future of Energy, organized by the Stockholm Institute for Transition Economics (SITE) in collaboration with Energiforsk, several key experts and innovators showcased how AI is shaping the energy system, from household-level optimization to national infrastructure forecasting and regulation. The discussions highlighted AI’s potential to enhance efficiency, resilience, and user responsiveness, while also raising critical issues around data governance, cybersecurity, and value distribution. This policy brief summarizes the main takeaways from the event.

AI as an Actor in Energy Networks

AI is now embedded in everything from electricity generation forecasts to district heating systems and real-time price optimization. As Chloé Le Coq, Research Fellow at SITE and Professor at Paris Panthéon-Assas University, noted in her opening remarks, this marks not just a technological upgrade but a systemic shift in how energy systems operate. Where systems were once reactive, AI opens the door to adaptive, self-learning networks that can respond dynamically to demand and supply. Examples shared at the 2025 Energy Talk showed how this transformation is already underway across Europe.

In the Baltic region, AI is contributing to a broader transformation of the energy system. Dzintars Jaunziems, Advisor for Energy and Climate Policy at Latvijas Banka and Assistant Professor at Riga Technical University, explained that the region has undergone several major transitions over the past three decades by building liberalized market economies and opening energy markets. More recently, the Baltics have halted energy imports from Russia, fully disconnected from the BRELL grid, and completed synchronization with the EU electricity network.

Against this backdrop, AI is now supporting the Baltics’ transition from fossil fuels to renewables. In grid operations, AI is used to assess overhead lines and remotely monitor systems in real-time. It also optimizes transmission capacity and supports renewable energy forecasting, particularly for solar generation. In district heating, digital twin technologies are being introduced, while in the mobility sector, AI helps manage electric vehicle (EV) charging and route planning. The region has one of the highest smart meter penetration rates in Europe, although full-scale utilization is still pending.

In Ukraine, AI plays a crucial role in managing the energy system—both in daily operations and in maintaining resilience during wartime. Andrii Starzhynskyi, co-founder and CEO of a-Gnostics, presented several examples of how AI is already deeply embedded in the sector. Since 2018, machine learning has been used to forecast electricity consumption and generation with over 98 percent accuracy. This fully automated system also enables predictive maintenance by detecting failures in critical equipment such as transformers. One example is an app that analyses the sound of machines to detect faults early and prevent breakdowns.

AI also supports automated decision-making around electricity flows—for instance, whether to buy, sell, or store solar-generated electricity. At a-Gnostics, multiple AI models—primarily based on time series data—are used to manage and coordinate forecasts across different applications. According to Starzhynskyi, these solutions are already used daily by customers in sectors such as mining, agritech, energy production, and energy trading.

AI is also being used at the household level to enhance energy system efficiency. Björn Berg, CEO of Ngenic, presented how their system integrates AI in real time to control and optimize heat pumps, using live data rather than historical averages. Reported benefits include over 20 percent energy savings, fewer boiler starts, and reduced system losses. Berg noted that if optimization were scaled to one million heat pumps, the aggregate impact could exceed the output of Sweden’s three nuclear reactors—highlighting the potential of household-level AI integration at scale. At the same time, he pointed to current forecasting limitations, referencing a recent two-gigawatt prediction error as a reminder that learning models still need improvement.

Infrastructure, Governance, and Cybersecurity

The shift in how energy systems operate today also adds complexity. As energy systems become more decentralized, with growing integration of intermittent sources, and data volumes expand rapidly, new governance challenges emerge. Key questions include: Who owns the data and the algorithms? How can we ensure fairness, accountability, and cybersecurity?

Filip Kjellgren, Strategic Initiative Developer Energy at AI Sweden, shared how interactive visualization tools are making future energy needs more accessible to individuals. Traditional methods, such as static bar charts, often fail to engage. In contrast, tools like the so-called Behovskartan allow users to explore different demand scenarios and, visualize, and test assumptions such as reduced fossil fuel use. Kjellgren emphasized that while solar and wind installations are expanding rapidly due to falling costs, public resistance to local infrastructure remains strong—often stronger than for other infrastructure projects. In this context, AI-driven visualizations can help bridge the gap between energy system planning and public acceptance, improving both actual and perceived fairness and facilitating green transition.

Figure 1. Behovskartan

Source: Printscreen from behovskartan.se

Similarly, Michael Karlsson, Programme Coordinator Heat & Power at Energiforsk, introduced the organization’s newly launched AI cluster—an initiative designed to disseminate research and applied insights about AI in the energy sector through webinars, seminars, and other outreach activities. He also highlighted the limited involvement of energy economists in AI projects and called for greater interdisciplinary collaboration to close that gap and broaden the field.

These highlighted initiatives set the stage for a panel discussion focused on the broader policy and structural questions facing AI in energy systems. As AI becomes embedded in critical infrastructure, concerns have been raised about the controls over data and algorithms that drive energy decisions. Speakers warned against relying on proprietary “black-box” models, calling instead for open-source alternatives and domestic oversight. The discussion also highlighted the importance of building national capabilities to avoid overdependence on international actors with limited public accountability and at times questionable agendas. Legal frameworks were seen as lagging technological development—particularly regarding new forms of data, such as sound recordings from equipment, which are not clearly covered in existing regulations.

Cybersecurity and system resilience emerged as recurring themes. AI can help detect anomalies, anticipate grid stress, and support decentralized energy configurations. One example illustrates how AI can detect abnormal behavior in connected devices—so-called Internet of Things (IoT) components—by analyzing how equipment behaves in real-time, rather than relying solely on code-level protections. Several participants stressed the need to build resilience into infrastructure design. In the case of cyber-attacks or physical disruption—like those experienced by Ukraine—systems should be capable of switching to “island mode”, operating autonomously during crisis. Others pointed to privacy-preserving data architectures, where AI models are deployed to the data, avoiding the need to centralize sensitive information—an approach already used in sectors like healthcare and finance.

The panel also raised the question of fairness: Who benefits from AI in the energy sector? While large industrial users are already reaping the rewards, such as a farm that significantly lowered its electricity costs using AI-based forecasting, it remains unclear whether smaller consumers are seeing comparable gains. In regulated systems, efficiency improvements may translate into lower tariffs; however, several speakers noted that public acceptance of AI will depend on whether consumers can clearly perceive and share the benefits. Ultimately, the long-term legitimacy of AI will depend on how these gains are distributed in practice.

Concluding Remarks

The 2025 Energy Talk AI and the Future of Energy made clear that AI is no longer a future consideration—it is already transforming how energy is produced, distributed, and consumed. From national-level forecasting to household-level optimization and strategic planning, AI is increasingly present in every part of the energy system. Yet, as participants emphasized, its rapid deployment has outpaced both regulation and public awareness. Successfully integrating AI into the energy system requires a broader policy dialogue—one that goes beyond the technical regulation to address economic and social matters. The Energy Talk brought these intersecting areas into focus and highlighted the need for broader conversations on AI in energy.

Disclaimer: Opinions expressed in policy briefs and other publications are those of the authors; they do not necessarily reflect those of the FREE Network and its research institutes.