The evolution of the electricity grid towards the smart grid paradigm is fostering the integration of distributed renewable energy sources in Smart Buildings: a combination of local power generation, battery storage and controllable loads can greatly increase the energetic self-sufficiency of a Smart Building, enabling it to operate in islander mode or to participate in an Automatic Demand Response framework, thus taking advantage of time-variable tariffs to achieve economical savings. This paper proposes an energy management system specifically tailored for a Smart Office building, which relies on actual data and on forecasting algorithms to predict the future patterns of both local energy generation and power loads. Performance is compared to the optimal energy usage scheduling, which would be obtained assuming the exact knowledge of the future energy production and consumption trends, showing gaps below 10% w.r.t. the optimum.