During the 2022 energy crisis, French electricity prices jumped more than almost anywhere in Europe — even though France barely uses gas for power. Our new study uses causal graphs & Shapley Flow to trace this back to high gas prices and low nuclear availability. www.nature.com/articles/s41...

Understanding the European energy crisis through structural causal models - Nature Communications

Despite a low share of gas-fired generation, French electricity prices showed stronger relative increases during the energy crisis. In the present paper, authors demonstrate how nuclear unavailability...

nature.com

AI makes energy systems more resilient. However, energy systems require explanations, i.e. explainable AI. We developed SHAPformer, an explainable forecasting tool, demonstrated on load and price data. Article: www.nature.com/articles/s41... and press release (German): www.kit.edu/kit/pi_2026_...

Explainable time-series forecasting with sampling-free SHAP for Transformers - Nature Communications

Time-series forecasting is crucial for decision making in many domains. The authors propose SHAPformer, a method for accurate and explainable time-series forecasting, delivering explanations in less t...

nature.com

Determining optimal heat pump usage is non-trivial. We explore how Deep Reinforcement Learning (DRL) can compete with Mode Predictive Control (MPC), interestingly a simple DQN shows promising results: link.springer.com/article/10.1...

Deep Reinforcement Learning for Price-Aware Building Heating Control - KI - Künstliche Intelligenz

KI - Künstliche Intelligenz - Heating systems account for a significant share of residential energy consumption, and rising energy prices call for intelligent, cost-aware control strategies....

link.springer.com

Energy systems are often highly correlated. With Alexandra Nikoltchovska, Sebastian Pütz and Markus Götz, we investigated how explanations of machine learning models should be derived for such correlated data: link.springer.com/article/10.1...

The statistics of power grids can be quite complex. My PhD student Xinyi Wen analyzed data from multiple continents and found heavy tails, bimodal distributions and more. Check out our latest article in Scientific Reports: www.nature.com/articles/s41...

Nonstandard power grid frequency statistics across continents - Scientific Reports

Scientific Reports - Nonstandard power grid frequency statistics across continents

nature.com

Excited to kick off #HAICON25 at Messe Karlsruhe for the next three days, presenting research results on "AI for energy" @helmholtzai.bsky.social

Power grids are complex systems with both deterministic and stochastic aspects. We extract the stochastic properties (drift and diffusion) from European and Australian data and train explainable AI (XAI) models, identifying key factors, see our article in Chaos: pubs.aip.org/aip/cha/arti...

Analyzing deterministic and stochastic influences on the power grid frequency dynamics with explainable artificial intelligence

Power grids are essential for our society, connecting consumers and generators. Their frequency stability is impacted by supply and demand changes, including de

pubs.aip.org