Increasing the Efficiency of Geothermal Power Plants by Coupling with Artificial Intelligence

The most effective way to utilize a geothermal resource is through a cascade utilization approach. Applications are arranged in order of decreasing temperature, ultimately leading to reinjection temperature. This sequence begins with power generation, followed by heat supply, direct-use applications such as heating and greenhouse cultivation, and finally, raw material extraction. However, implementing such an extensive utilization strategy is often hindered by hydrochemical challenges. Highly mineralized geothermal fluids tend to precipitate minerals (scaling) uncontrollably due to temperature, pH and pressure variations between production and reinjection, which reduce the overall efficiency and sustainability of energy production.
In the Upper Rhine Graben, reinjection temperatures range between approximately 60°C and 80°C. Lowering these temperatures further could enhance efficiency, it would also intensify scaling and other hydrochemical challenges. In the future, raw material extraction could intensify these issues even more. Unlike conventional geothermal energy production, where chemical disturbances are mainly driven by temperature and pressure changes, raw material extraction directly alters fluid chemistry through interventions such as chemical additives, pH adjustments, and increased solute concentrations. ... mehrThese modifications significantly disrupt the system’s equilibrium. Traditional deterministic geochemical models, which oversimplify the complex and site-specific nature of geothermal water chemistry, struggle to accurately capture these intricate interactions. Consequently, predicting hydrochemical challenges—particularly mineral precipitation and degassing pressures—remains highly uncertain. It is imperative to address these uncertainties to ensure the optimal utilization of the full range of site-specific parameter boundaries in geothermal cascade applications. The MALEG project aims to address these challenges by integrating artificial intelligence (AI) into geochemical modeling to achieve a more comprehensive understanding of hydro-geochemical systems. The AI tool will be trained using experimental data obtained from power plant twins deployed at operational geothermal facilities. These hardware twins will replicate geothermal plant processes in their entirety and will be connected to the operating plants via a bypass. This field laboratory setup will facilitate controlled experiments on hydrochemical precipitation and degassing, with operational parameters deliberately varied to extreme conditions. A highly detailed monitoring system will continuously track fluid and solid parameters, including temperature, pH, pressure, and ion concentration, enabling the identification of conditions that trigger mineral phase formation. Experiments will be conducted across at least three geothermal systems with varying geological settings and reservoir characteristics, ensuring the generation of a diverse and extensive hydro-geochemical dataset. This dataset will form the basis for the AI-driven MALEG prediction tool. The tool’s accuracy and reliability will be assessed using a digital twin that incorporates a hydrochemical modeling environment. By enhancing predictive capabilities, MALEG will contribute to the optimization of site-specific geothermal plant operations, improving efficiency and sustainability.
The ultimate objective is to develop a digital chemical system capable of interpolating discrete, snapshot-like chemical analyses within the multidimensional parameter space of complex hydrochemical systems. By integrating a wide range of chemical, physical, and thermodynamic parameters, this digital model will generate a continuous representation of fluid chemistry. This approach will enable more precise predictions of geochemical behavior under varying operational conditions, thereby enhancing the understanding and management of geothermal processes.