A data-driven surrogate model was developed using Deep Neural Networks, trained from a set of Computational Fluid Dynamics (CFD)Computational Fluid Dynamics (CFD) simulations to predict rooftop velocities for a specific inlet velocity condition. The objective is to considerably reduce the number of CFDComputational Fluid Dynamics (CFD) simulations to assess the annual wind energyWind energy yield of rooftop urban wind turbinesWind turbine. Steady, incompressible RANS simulations were performed to obtain local wind velocities in the Northern Quarter region of Brussels (Belgium) using a modified k- SST turbulenceTurbulence model implemented with an improved Atmospheric Boundary Layer approach. Velocity fields from simulations served as the labels to train the surrogate model, and their corresponding inlet velocity conditions were the features. Rooftop velocity fields were then predicted for a new inlet velocity, and the performance of the trained model was assessed by comparing the results with the ones obtained from the CFDComputational Fluid Dynamics (CFD) simulations. The tested surrogate model was then used to predict rooftop velocities for additional inlet velocity conditions to finally estimate the rooftop Annual wind EnergyWind energy Production (AEP).