AI With Memory Helps Cities Predict Energy Needs More Accurately

What is it about?

Cities across Europe are struggling to manage electricity demand as weather becomes more unpredictable and renewable energy use increases. Traditional forecasting tools often miss important patterns, especially when weather conditions change quickly. This study introduces a new type of artificial intelligence system that uses two kinds of memory — one for fast learning and one for long‑term knowledge — similar to how the human brain works. The system reads weather data and energy‑grid information in natural language, allowing it to understand subtle weather patterns that affect electricity use. When tested in major European cities, the model predicted energy demand far more accurately than standard methods. It also helped coordinate “sand battery” thermal storage, a new technology that stores heat efficiently. Overall, the system could help cities save millions of euros each year by reducing peak demand, improving renewable energy use, and making the grid more stable.

Featured Image

Photo by Anthony Indraus on Unsplash

Photo by Anthony Indraus on Unsplash

Why is it important?

This work is timely because energy systems are under pressure from climate change, rising demand, and the shift toward renewable power. Existing forecasting tools cannot keep up with rapidly changing weather patterns or seasonal variations. The dual‑memory LLM architecture offers a major step forward by combining fast adaptation with long‑term stability. It reduces forecasting errors by up to 85% and enables cost savings of more than €150 million per city annually. Its built‑in explainability also makes it suitable for high‑stakes infrastructure decisions. This approach has the potential to transform energy grids from reactive systems into predictive, resilient, and climate‑ready infrastructures.

Perspectives