Résumé : In this study, computational fluid dynamics (CFD) is employed to evaluate the influence of surrounding buildings on the performance of a roof-mounted, 2-bladed Savonius vertical-axis wind turbine (VAWT). The latter is planned to be located in the Bovisa Campus of Politecnico di Milano. In the present work a preliminary simulation campaign has been conducted, explicitly depicting the surrounding area and employing an advanced Reynolds-averaged Navier-Stokes (RANS) model. This closure is suitable for Atmospheric Boundary Layer (ABL) simulation, reliably reproducing the various ground roughness elements and employing a Building Influence Area (BIA) for a more accurate representation of the disturbed flowfield. After considering twelve main wind directions, the resulting velocity profiles are extracted and used as inlet conditions for a second session of simulations, related to the wind turbine. The final goal is to reproduce the effect of the surrounding buildings and to accurately forecast the energy production of the machine. This is a relevant aspect of the increasingly topical framework of smart city, implying the exploitation of wind energy. Outcomes indicate that the resulting energy production of the machine remarkably departs from ideal conditions and that accounting for the surrounding topography becomes an aspect of great relevance.