| Title | Integrating geological uncertainty into geothermal reservoir simulations with Bayesian optimization workflows |
|---|---|
| Authors | J. Rihet, M. Abd-allah, S. Pearson-Grant, J. Patterson |
| Year | 2025 |
| Conference | New Zealand Geothermal Workshop |
| Keywords | Modelling framework, Geothermal Reservoir Simulation, Waiwera, Automation, Aspen SKUA, Aspen Tempest ENABLE |
| Abstract | Geothermal reservoir predictions are typically done under a high degree of uncertainty. Hence, accurately understanding and factoring in all uncertainties is critical for robust resource development and management. This study presents an integrated workflow that combines geological uncertainty quantification, Bayesian optimization, and automated history matching for geothermal reservoir simulation. Using a fully controlled modelling environment—SKUA for geomodelling and Waiwera for simulation—we construct a synthetic geothermal reservoir exhibiting typical features such as a clay cap and heterogeneous lithologies. In addition to flow-governing reservoir uncertainties such as rock permeability and upflow rate, we account for spatial uncertainty in geologic features such as clay cap depth, clay cap hole size, permeable reservoir extent, and fault dip/strike/extent, which are progressively constrained through Bayesian Optimisation techniques. The approach leverages a workflow orchestrator to automate the Natural State Calibration and the dynamic History Matching phases. Multiple realisations of the reservoir model are generated and compared against synthetic production data. To address the high computational cost of ensemble methods, we use established assisted history matching tools that support uncertainty sampling, ensemble management, calibration point setup, and efficient matching with several optimisation algorithms available. The result is an ensemble of calibrated, geologically consistent reservoir models with quantified predictive uncertainty. This work demonstrates a proof of concept for a systematic and practical approach for geothermal reservoir forecasting. Because the workflow is built around a modular, software-agnostic orchestrator, a wide number of setups using different modelling approaches can be efficiently integrated within the larger history-matching framework. This flexibility offers a scalable path for researchers and operators wanting to integrate the impact of geological uncertainty into their reservoir predictions. |