Geothermal energy extraction changes reservoir pressure and temperature due to colder fluid injection, which impact the stress field. In naturally fractured or faulted reservoirs, stress changes introduce risks of fault reactivation and induced seismicity.  In this study, the capabilities of the Delft Advanced Research Terra Simulator (DARTS) framework have been extended to account for the thermo-poroelastic response relevant to geothermal applications.  In the proposed modeling approach, a low-fidelity model is used for geomechanics where stresses are calculated from pressure and temperature changes. The fault slippage is calculated using Mohr-Coulomb friction criteria. The developed model has been applied to heterogeneous geothermal field to evaluate how energy production and thermal front can affect the geomechanical state of the field and trigger induced seismicity.

The Dinantian carbonate reservoir (DCR) is a proven geothermal play in Northwest Europe. In the UK, evidence points towards it being an important thermal spring source and a major aquifer in many basins. In the Midlands region, the carbonates have been buried to below 4000 m while they are below 2000 m in the south of England. Available data suggests a temperature range from 30 to in excess of 100 °C.
In the UK, DCR have very low primary porosity and permeability which may increase locally due to karstification, dolomitization and/or faulting. However, the influence on reservoir permeability of depositional environment, diagenesis, dolomitisation, connectivity between fault zones and high porosity zones is not yet fully understood. The current uncertainties about the physical properties of DCR have a direct impact on the sustainability of deep geothermal systems as they affect costs of exploration and drilling, accuracy and reliability of microseismic monitoring (e.g., microseismicity location) and of pore pressure distribution models. To reduce the level of uncertainties about reservoir properties, we integrated new microstructural and experimental (e.g., density, porosity, sonic velocity, permeability) datasets from pre-cut cylindrical samples (limestones, fractured limestones and karstified dolomites) collected from DCR sections of Roddlesworth-1 (1288-1298m) and Grove-3 (2308-2315m) boreholes in the UK.
Measured average bulk densities show high variation in the dolomites (2.39-2.75 g/cm3) and a much more constant value in limestones (2.58-2.62 g/cm3). The variability of bulk density within the dolomites can be attributed to the high variability of total porosity (averages ranging 3.2-15.9%) which also affects their sonic velocity. In contrast, the limestone samples showed uniform porosities of less than 5%. All limestone samples showed small permeability (>10-20 m2) as expected from their low porosities. Dolomitised samples show relatively low permeability compared to their high total porosity, ranging from 10-17 m2 to 4*10-19 m2. These results suggest that permeability of dolomitised samples are instead controlled by their effective (1.7-2.9%) rather than total porosity (3.2-15.9%). These interpretations are consistent with microstructural observations showing a rather complex pore network, consisting of intercrystalline pores and vugs.
The results improve our understanding of the porosity-permeability relationship within the DCR. The porous vuggy nature of dolomitised lithology means that an EGS hydrothermal type reservoir concept is possible, where reservoir stimulation would increase connectivity of the pores and improve permeability in order to circulate reservoir fluids saturating the pore network.

Human intervention in geothermal systems, such as fluid injection, can reactivate faults in the subsurface. This study investigates the Lower Carboniferous system and specifically the Dinantian formation. This specific formation is particularly interesting for deep geothermal energy (>100ºC) in the Netherlands, Belgium and Germany but can serve as a proxy for fractured carbonate deep geothermal reservoirs worldwide. This study experimentally investigated the stress conditions under which Dinantian carbonate faults, filled with gouge material, reactivate, heal and relax. Moreover, we tested different grain size distributions, which serve as a proxy of grain dissolution during fluid injection in geothermal systems. Finally, we investigated the effect of shearing history on the reactivation envelopes, healing and relaxation of carbonated faults. 
Experiments were performed in a rotary shear apparatus to shear a simulated gouge obtained from quarry material of the Dinantian carbonates. It was sieved to be in grain size clusters of 45 μm, 50% 45 μm - 50 125 μm, 15 μm, 5-125 μm and 125-212 μm. Experiments were performed at room temperature, dry or wet, using de-ionised water or, in 2 cases, bine as pore fluid, under drained conditions. The brine was obtained from an existing geothermal well and contained mainly different Na and Cl salts. In every single experiment and sample, we stepped the normal stress up from 1 to 10 MPa, and down from 10 to 1 MPa, using stress steps of 1 to 2 MPa. During each stress interval, we performed a slide-hold-slide procedure, where the hold times were 10-100-1000 sec, and the sliding interval lasted 10 seconds, with a velocity of 20 μm/sec. Using the peak shear stress after each hold period, we calculated the Mohr-Coulomb reactivation envelopes of each experiment, as well as the quantification of the fault healing and relaxation as a function of hold time. Furthermore, we quantified the fault healing rate and relaxation rate as a function of grain size and shearing history during stepping up and down the applied normal stress.
Our preliminary results showed that carbonate laboratory faults exhibited a friction coefficient of up to 0.74 for both dry and wet experiments. However, the friction coefficient of wet experiments showed a greater range of values as a function of hold time compared to dry. Dry and wet experiments displayed a clustered behaviour in healing and relaxation with the wet experiments preserving the highest values of both healing and relaxation throughout the shearing history of the samples. Wet samples display higher values of healing/relaxation rate compared to the dry ones. Grain size differentiation appears to have an effect on the healing/relaxation rate for wet samples (DI water and brine) and no effect on dry samples, but further investigation is required.
Future work includes the investigation statistical significance of the current regressed reactivation envelopes, microstructural analysis of sheared samples and further analysis of the impact of shearing history on the fault strength.