The MALEG Project

The machine learning for enhancing geothermal energy production (MALEG) project is a collaboration of multiple partners to develop an artificial intelligence (AI) to optimize power plant efficiency on three geothermal sites. To investigate enhanced heat extraction, an increased cooling of the thermal water is necessary. These changes in the operating parameters could lead to scaling and corrosion.

To train the AI, an on-site demonstrator ‘hardware-twin’ as well as a ‘digital twin’ are developed. The ‘hardware-twin’ emulate power plant process, brine treatments and mineral extraction to investigate the occurrence of scaling and corrosion. The ‘digital-twin’ simulates the geochemical processes of these potential hazards.

The AI-based tool (MALEG) will be trained through extensive geochemical sampling campaigns, data collection and analysis of plant operating parameters at the associated power plants. In the end, MALEG is able to predict the geochemical composition and the scaling potential based on live measurements of the geothermal brine optimizing the individual power plant.

The Project Timeline

The implementation of artificial intelligence (AI) can be divided into several development stages. In the beginning there is always data, usually the bigger the better. To build up a large geochemical database, several sampling campaigns are conducted during the project. Therefore, a hardware twin is build to investigate specific geochemical scenarios at each geothermal power plant. This generated data is the foundation for the prospective training of the AI. There are different algorithms for data handling because not every data set is suitable for every machine learning algorithm. Thus, the MALEG database is developed first. Afterwards, suitable algorithms are evaluated and implemented for each geothermal power plant. The MALEG-AI is trained and validated with the gathered geochemical data. After verification, the AI will be implemented into the power plants. MALEG is able to predict the geochemical composition from the in-situ data feed and improve the efficiency of the heat generation and provide actions for the power plant operator.

Establishing machine learning algorithm to the demonstrator

Development of
digital twin

Creating hydrogeochemical database

Constructing demonstration plant

Initial sampling