Research Projects
Maleg 4.0

MALEG 4.0

Project:
Applied machine learning to improve geothermal energy production by digitalisation

Duration
01.09.2026 – 31.05.2029

Website
in progress

Project partner

• Fraunhofer Institut für Solare Energiesysteme (ISE)
• Karlsruher Institut für Technologie (KIT) – Institut für Kern- und
  Energietechnik – Arbeitsgruppe Energie und Verfahrenstechnik (EVT)
• BWG Geochemische Beratung GmbH
• Stadtwerke München GmbH
• Stadtwerke Neuruppin GmbH
• Badenova Wärmeplus GmbH & Co KG

Short description:
To ensure the low-risk and cost-effective operation of geothermal plants, the chemical reactions that occur as a result of changes in pressure and temperature must be predictable and controllable.

Lowering the reinjection temperature and the system pressure must be regarded as particularly ambivalent, as whilst this increases the plant’s energy efficiency on the one hand, it also gives rise to considerable risks caused by undesirable reactions within the chemical system on the other. To optimise plant performance and ensure the most efficient possible multiple use of the geothermal resource, the critical limit conditions of the site-specific chemical system must therefore be known in detail. As these limits depend on numerous, often complexly interrelated parameters (e.g. T, p, X, pH, redox, gas content, etc.) and thermodynamic data are scarce, their quantification is associated with considerable uncertainties.

On the other hand, hydrochemical systems – with their complex interactions and multidimensional interdependencies – represent an ideal field of application for data-driven AI models, enabling optimisation and the deployment of predictive control systems such as digital twins. However, the use of AI has so far been virtually non-existent because, amongst other things, there is currently a widespread lack of high-resolution, site-specific thermodynamic-hydrochemical data and large systematic datasets that quantitatively capture the geochemical system’s responses to forced variations in key parameters, which are required to develop powerful AI models.

Supported by BMWE, support code: 03EE4083C