Buildings belong to the large energy consumers class and account for 40% of worldwide primary energy consumption. Application of advanced control and diagnosis algorithms in buildings has been intensively investigated with the aim to improve their energy efficiency and bring the buildings sector into the smart city arena. In the thesis, a general methodology for minimizing thermal energy consumption using current energy sources and minimal retrofitting through the use of advanced identification and control techniques is presented and validated via simulation studies and on-line implementation on a real 248-zones skyscraper building. The developed framework paves the way to fast deployment of model-based energy management strategies in buildings and unleashes significant energy and cost-saving potentials of a smart building in a smart city. The following contributions of the developed approach are highlighted: i) a method for identification of an energy model of a system of fan coil units, which gives a direct relation between the thermal energy provided to a zone, actuation commands, heating/cooling medium conditions and the zone temperature, ii) a method for identification of a thermodynamic building model suitable for predictive control design and disturbance estimation, iii) predictive control strategy for zone temperature control and hierarchical coordination with the central heating/cooling medium conditioning system, building microgrid, smart grid or some other building subsystem sharing a common energy link with zones, iv) validation of the developed hierarchical predictive control system within a living-lab environment in terms of achieved thermal comfort and energy cost reduction, v) validation of the developed zone temperature control via on-line implementation on a full-scale skyscraper building.