Application of machine learning for predicting shallow geothermal temperatures

Research output: ThesisMaster's Thesis

9 Downloads (Pure)


Nowadays, the energy sector is experiencing the transition towards clean energy sources that reduce carbon emissions. Geothermal energy can cover the demand for energy grid stability and contribute to decarbonisation. The heat extracted from underground can be used directly as heat for space cooling and heating without any conversion. The extraction of heat from shallow underground depths (up to 200 m) is the basis of this research. To provide long-term sustainability of projects involving geothermal deployment and heat extraction, it is necessary to carefully monitor and manage the geothermal resources. Groundwater temperature (GWT) is a critical environmental parameter influencing the utilisation of geothermal systems. Accurate prediction of GWT is essential for assessing the efficiency and optimising the performance of geothermal installations. This thesis investigates the predictive modelling of GWT using machine learning (ML) techniques. To accurately predict GWT with the help of ML algorithms, a comprehensive dataset with selected features was assembled, incorporating measurements from the eHYD database, along with weather data from Visual Crossing. The data from 1996 to 2016 of shallow geothermal wells in Vienna, district 22, was used for this purpose. Steps for dataset preprocessing are described, and ML algorithms are applied and evaluated to predict the temperature. The analysis revealed the K-Nearest Neighbors model to be the model providing the highest accuracy in predicting GWT, as well as showed the importance of additional accounting for urban factors when investigating geothermal applications on the city scale.
Translated title of the contributionAnwendung von maschinellem Lernen zur Vorhersage flacher geothermischer Temperaturen
Original languageEnglish
Awarding Institution
  • Montanuniversität
  • Yoshioka, Keita, Supervisor (internal)
Award date28 Jun 2024
Publication statusPublished - 2024

Bibliographical note

no embargo


  • Machine learning
  • temperature prediction
  • shallow geothermal energy
  • decarbonisation

Cite this