Benchmark study of simulators for thermo-hydraulic modelling of low enthalpy geothermal processes

In order to assess the thermo-hydraulic modelling capabilities of various geothermal simulators, a comparative test suite was created, consisting of a set of cases designed with conditions relevant to the low-enthalpy range of geothermal operations within the European HEATSTORE research project. In an effort to increase confidence in the usage of each simulator, the suite was used as a benchmark by a set of 10 simulators of diverse origin, formulation, and licensing characteristics: COMSOL, MARTHE, ComPASS, Nexus-CSMP++, MOOSE, SEAWATv4, CODE_BRIGHT, Tough3, PFLOTRAN, and Eclipse 100. The synthetic test cases (TCs) consist of a transient pressure test verification (TC1), a well-test comparison (TC2), a thermal transport experiment validation (TC3), and a convection onset comparison (TC4), chosen to represent well-defined subsets of the coupled physical processes acting in subsurface geothermal operations. The results from the four test cases were compared among the participants, to known analytical solutions, and to experimental measurements where applicable, to establish them as reference expectations for future studies. A basic description, problem specification, and corresponding results are presented and discussed. Most participating simulators were able to perform most tests reliably at a level of accuracy that is considered sufficient for application to modelling tasks in real geothermal projects. Significant relative deviations from the reference solutions occurred where strong, sudden (e.g. initial) gradients affected the accuracy of the numerical discretization, but also due to sub-optimal model setup caused by simulator limitations (e.g. providing an equation of state for water properties).

Automated summary

A comparative test suite was created to assess the thermo-hydraulic modelling capabilities of various geothermal simulators, with 10 different simulators being used as benchmarks. Results from the tests compared among participants showed that most simulators were able to perform reliably, but significant relative deviations occurred in cases with strong gradients or due to sub-optimal model setup caused by simulator limitations.