Bourhis, PaulCousin, BenoîtRotta Loria, Alessandro F.Laloui, Lyesse2021-06-252021-06-252021-06-252021-05-1910.1016/j.geothermics.2021.102132https://infoscience.epfl.ch/handle/20.500.14299/179541This study explores and validates a machine learning approach for the practical, effective, and precise prediction of the thermo-physical characteristics that are essential for the analysis and design of shallow geothermal systems, including borehole heat exchangers: (i) undisturbed ground temperature, (ii) ground effective thermal conductivity, and (iii) borehole thermal resistance. Benefiting from 174 thermal response tests from central and western Switzerland, the algorithm is used to provide accurate site-specific as well as regional-scale predictions of the investigated thermo-physical characteristics, which in turn can serve preliminary yet representative evaluations of the geothermal potential of even very broad areas.Geothermal energyMachine learningThermal response testingBorehole heat exchangersData-driven predictionstext::journal::journal article::research article