Abstract: The growing importance of decarbonization and renewable energy sources to national power systems has brought about a need to implement large-scale energy storage technology. Adiabatic Compressed Air Energy Storage (A-CAES) systems comport with the environmental requirements of renewable energy storage better than traditional CAES systems because they eliminate the combustion of auxiliary fuel during discharge. This benefit is achieved with a Thermal Energy Storage (TES) tank that heats up during the air compression step, stores the thermal energy, and then releases it during discharge by heating the expanding air. This process increases the overall energy efficiency of the A-CAES system since a large portion of the heat of compression is used rather than lost. The subject of this paper is an experimental and numerical study of a slender basalt-filled TES tank designed for installation in decommissioned mine shafts. Storing energy in such otherwise unused, already depreciated infrastructure represents a unique application and economic opportunity for highly urbanized industrial areas. Situating A-CAES systems in mine shafts also significantly reduces their cost due to natural stress compensation from high air pressure. The slenderness of the reservoir compared to existing and previously studied systems has a positive effect on the efficiency of heat storage in the basalt due to the reduced heat conduction field, but results in a greater pressure drop of the flowing air. To investigate the performance of the TES storage tank, a laboratory bench was constructed with basalt grit as the accumulation material. The experimental results obtained were used to verify the numerical model used in ANSYS Fluent. The numerical model showed a theory–experiment discrepancy in the basalt temperature rise results of only 1.91% for the entire curves and 2.29% for the most intense temperature rise phase, and the difference between the pressure drop results did not exceed 4%. The flexibility of the numerical model was tested for various input parameters. A high level of model validation was achieved by selecting an accurate correlation for the Nusselt number, as well as implementation of temperature-dependent thermodynamic parameters for the storage material.