The Laboratory of Soil Mechanics (LMS) at EPFL has launched a new research initiative exploring how artificial intelligence can improve the long-term performance and sustainability of geothermal heat pump systems. Conducted by Professor Lyesse Laloui and Dr Elena Ravera, the project aims to develop predictive tools capable of optimising geothermal operations in real time while helping preserve underground thermal resources.
As Switzerland accelerates its energy transition, geothermal energy is expected to play an increasingly important role in reducing dependence on fossil fuels. According to figures presented in the project documentation, geothermal systems already provide around 5% of Switzerland’s heating demand, with national objectives targeting a fivefold increase by 2050.
However, the rapid deployment of geothermal systems also raises new challenges. Many installations are designed primarily for heating demand and are operated with limited monitoring once in service. Over time, this can lead to declining efficiency, excessive thermal extraction, and in some cases degradation of the surrounding ground conditions.
The LMS project seeks to address these issues through continuous monitoring and machine learning. The research combines non-intrusive temperature sensing, long-term operational data collection, and AI-based predictive modelling to better understand how geothermal systems behave under real operating conditions.
By integrating operational data such as temperature and flow conditions, the team aims to create a predictive interface capable of helping operators anticipate system behaviour, optimise performance, and improve energy efficiency over time.
Beyond technical optimisation, the project also supports broader goals linked to climate resilience and sustainable urban energy infrastructure. The work aligns with growing national and cantonal ambitions for low-carbon heating solutions and reflects LMS’s continued leadership in advancing energy geostructures and sustainable subsurface engineering.
The project, titled Optimisation intelligente des systèmes géothermiques par apprentissage automatique, is funded by the Fonds pour l’Efficacité Energétique (FEE) and conducted in collaboration with the Services industriels de Lausanne (SiL).