ENLIGHT-AI - Energy aNomaLy Identification & diaGnosis using HPC Trained AI model
Project duration
May 01, 2026 - July 31, 2027
Project partners
Funding
The project is funded by the European Union under the Fortissimo Plus porgram.
Project Description
The iMTED AI Energy Assistant is a German start-up that empowers energy/facility managers in identifying potential savings and malfunctioning equipment in buildings, and supports purchasing renewable energy from the spot market when prices dip. iMTED proposes to develop a high-performance computing (HPC)-driven, physics-informed artificial intelligence (AI) model, ENLIGHT-AI, specialized for energy management. This model enables short- and long-horizon energy demand forecasting of non-residential buildings. It detects and diagnoses in real time anomalous events, such as sensor defects, leakage, or valve hunting, and provides intuitive visualization of energy hot spots for actionable insights. The software uses physics-informed and data-driven deep learning (DL) technologies to study the anomalous behavior from large, complex, and aggregated data of existing smart meters of heating, cooling, electricity, and water sensors. This reduces the installation of new sensors for individual equipment and minimizes both capital and operational expenditure. Overall, it enables up to 20% reduction in energy and water consumption without major renovations. The supporting partner, the Simulation and Data Lab Fluids & Solids Engineering (SDL FSE) from the Jülich Supercomputing Centre (JSC) of Forschungszentrum Jülich (FZJ), with expertise in the field of development for highly-scalable engineering codes, code analysis, and optimization, supports the technical and engineering aspects of the ENLIGHT-AI. This business experiment validates the scalability, cost-effectiveness, and market potential of an HPC-driven AI model, paving the way for its deployment in smart buildings across Europe.