Stochastic modelling for identification of potential geothermal resources

Characterising the thermal state for geothermal assessment is important to highlight potentially interesting areas. We tested an approach based on stochastic modeling in the western sector of the Po basin. A new thermal database was created by collecting data from hydrocarbon wells, bottom hole temperatures (BHT) and temperatures from drill stem tests (DST). To identify areas with potential geothermal resources, we interpolated data using an original algorithm based on Gaussian simulations producing a 3D temperature field model. This led to generated temperature contour maps at different depths and along selected geological cross-sections. The stochastic modeling identified the area west of Milan as having the highest geothermal potential (temperatures about 180 °C at about 6 km depth). The results of the stochastic modelling were validated with 1D geothermal modelling of the deeper boreholes along the cross-sections. 1D models relied on thermophysical properties (thermal conductivity, volume heat capacity, density and porosity) measured in the laboratory on core samples extracted from the wells, and radiogenic heat production values inferred from gamma-ray logs. Thermal conductivity was inferred using an indirect approach that considers the temperature dependence of the matrix, the pore-fluid conductivity, and the porosity variation with depth. 1D thermal modelling assumes a steady-state purely conductive thermal regime. Geotherms and surface heat-flow estimations for each well were produced by minimising the root mean square error (RMSE) between the calculated temperature and the observed temperature corrected for the drilling mud circulation. The 1D thermal calculations and the temperatures inferred from the stochastic model are in good agreement, but the presence of outliers can lead to important deviations for the stochastic model. The average differences between the temperature profiles of the two models range from 0 to 10 °C, but in particular case reaches 15 °C. In general, in the case of a relatively simple structural setting, as it occurs for the selected cross-sections mainly characterised by horizontal strata, the stochastic model can provide a reliable picture of subsurface temperature distribution since thermal refraction effects are likely negligible.