Assessing the Ensemble Smoother with Multiple Data Assimilation for Subsurface Fluvial Geothermal Systems

Abstract

Efficient geothermal resource development remains challenging due to inherent geological uncertainty and limited subsurface data. A proof-of-concept for a digital twin for a fluvial geothermal reservoir, similar to the Delft campus geothermal project, is presented. This digital twin has the aim to integrate geological scenario modeling, production simulation, uncertainty analysis, and data assimilation to mitigate operational risks, reduce maintenance costs, extend reservoir longevity, and enhance the overall sustainability of this project. In this contribution, we assess the efficiency of the ensemble smoother with multiple data assimilation (ESMDA) for subsurface property inversion of a fluvial geothermal system. First, we developed an efficient method that allows for the swift creation of multiple geological scenarios of channelized reservoir geometries, fully constrained to well information, using Rapid Reservoir Modeling (RRM). Next, we generated an ensemble containing multiple geological realizations for a given scenario representing the geothermal system using stochastic reservoir modelling. For a single scenario and its ensemble of stochastically generated property distributions, heat flow and production rates were simulated using the Delft Advanced Research Terra Simulator (DARTS). One of the ensemble members and its simulated production data were taken as the “truth” (or reference) case. ESMDA was then employed to invert the property distribution within the fluvial channels of all other ensemble members, using the “observed” temperature and pressure data along the injection and production well from the “truth” case. We also consider the presence of a monitoring borehole to analyze how additional monitoring data impacts the convergence of ESMDA. The simulation results of the posterior models demonstrated a significant reduction in root mean square error for temperature and pressure data which align more closely with the “observations” compared to the prior models. This outcome confirms the feasibility of applying ESMDA for data assimilation in fluvial geothermal systems, such as the Delft campus geothermal project.