At Jülich Systems Analysis (ICE-2), we are contributing to the RenoVisor project by leading the building modelling activities and helping to establish a strong, data-driven foundation for the project.
With a research focus on the unbiased scientific investigation of technologies, value chains, and market developments, our work considers material requirements, sector coupling, and the broader policy and societal context of future energy systems.
Over the past months, our interdisciplinary team has grown with three new PhD researchers from computer science and physics — strengthening our expertise at the interface of AI, energy systems, and building modelling.
One important milestone so far:
We successfully created a first training dataset for an AI surrogate model for building energy simulations.
Why is this important?
Detailed bottom-up simulations can take several minutes per building configuration. To enable scalable and interactive renovation advisory tools, computation times need to be significantly reduced.
The AI surrogate model developed within RenoVisor is designed to accelerate these calculations from minutes to milliseconds by predicting simulation results for different building configurations and renovation pathways.
The first version of the model has already been developed and trained. Next steps will focus on expanding the dataset and further improving robustness and prediction accuracy.
An initial training dataset was successfully created and the first version of the AI surrogate model was developed and trained. Further work will focus on expanding the training dataset and improving the robustness and accuracy of the AI surrogate model.
What role do you see for AI surrogate models in accelerating the energy transition in buildings?