| Title | Bayesian Data Assimilation for High-Dimensional Subsurface Parameter Inference in Modeling of Borehole Heat Exchanger Fields |
|---|---|
| Authors | Hesam Soltan Mohammadi, Lisa Maria Ringel, Christoph Bott, Selçuk Erol, Peter Bayer |
| Year | 2026 |
| Conference | World Geothermal Congress |
| Keywords | |
| Abstract | Accurate prediction of subsurface temperature evolution in borehole heat exchanger (BHE) systems relies on robust characterization of spatiotemporal model parameters. This study presents a Bayesian data assimilation framework to infer the statistical distribution of highly correlated parameters using a low-fidelity but computationally efficient modeling approach. The framework is tailored for complex, multi-layered subsurface environments with groundwater flow. By sequentially assimilating monthly temperature measurements, the method systematically explores and updates the model input parameter space—such as thermal conductivity, Darcy velocity, and heat extraction rates—capturing key parameter correlations in each layer and enhancing the reliability of temperature predictions.A synthetic case study simulating a five-year BHE field operation in COMSOL Multiphysics® is used to validate the framework. The setup includes four active and one inactive BHE, with temperature measurements collected monthly from the inactive BHE. These observations are used to update the model and identify parameter sets that best reproduce the thermal response of the system. Updated parameters are then propagated forward in time for improved predictions. After 32 months of sequential inference, the framework reduces uncertainty in predicted subsurface temperatures to 8%, while maintaining flexibility to accommodate evolving thermal conditions. This work underscores the potential of Bayesian data assimilation and data-driven approaches to enhance long-term thermal modeling of BHE systems by adaptively tuning model parameters in response to observed data, thereby avoiding the need for computationally intensive simulation tools. |