Energy Anomaly Identification & Diagnosis using HPC Trained AI model

The ENLIGHT-AI project aims to leverage High-Performance Computing (HPC) to develop an advanced energy management solution for optimising energy consumption and operational efficiency. The core objective is to create a hybrid Physics-Informed Neural Network (PINN) architecture that combines data-driven learning with established physical principles. The model is designed to capture temporal patterns, correlations between sensor data, and underlying system behaviour to deliver accurate energy consumption forecasts.

In addition, the project will develop a real-time anomaly detection framework to support rapid identification of abnormal energy usage patterns and facilitate informed operational decision-making. The solution is being optimised for scalability and performance on the Leonardo Booster EuroHPC supercomputer, enabling the efficient training and deployment of complex AI models.

Continuous feedback from pilot customers will guide the development process to ensure that the final solution addresses real market needs and can be effectively adopted in operational environments. The project’s success will be measured through indicators such as reduced energy costs, market growth potential, and the demonstration of how HPC can accelerate innovation and create business value in the energy sector.

Organisations involved: 

End-User:  iMTED

Technology Expert: Forschungszentrum Jülich GmbH