To increase the efficiency of corn drying an underground thermal energy storage (UTES) facility stores waste heat from a combined heat and power unit (CHPU) during the summer to use it for corn drying in autumn. The storage consists of a high permeable gravel layer, which is bordered by an impermeable clay layer at the top and the bottom and two artificial cylindrical walls. The storage has a diameter of 33m and a height of 11m. In this thesis, this real world system is replaced by a thermal reservoir simulator, in which a corn drying control system (CDCS) is implemented. The simulator is used to calculate transport of heat in the UTES during the charging and the discharging period and for the optimal determination of well locations and schedules to optimize the corn drying process. The created simulator uses the CSMP++ software library and is based on the hybrid finite-element finite-volume discretization method. This numerical scheme combines the robustness of FVM in solving advection - dominated problems, with the capability of FEM to handle diffusion-dominated problems. The simulator is verified by comparing the numerical solution with analytical ones, as well as benchmarking it against TOUGH ("Transport of Unsaturated Groundwater and Heat") simulator and the PHT ("Pressure Enthalpy Temperature") simulator. PHT is a compressible flow simulator, also generated from the CSMP++ libraries. Simulations prove that production wells during the controlled discharge period should be located near the top, and the injection wells near the bottom. Whenever possible, distance between wells should be maximized to prevent thermal breakthrough for as long as possible. The charging and discharging period should both be carried out at the highest possible rate to reduce diffusive heat loss. The highest energy output is thus achieved by charging the UTES with hot water at one well at the top section and discharging the storage from the same well. In such a scenario, composed by four months of charging followed by two months of discharging, fifty percent efficiency may be achieved. The simulator created and described in this thesis provides a useful and computationally efficient tool to design UTES systems.