This paper presents a predictive control method for power variation reduction and maximum power limitation in building heating systems. A case study with heating supply pipeline of a skyscraper floor with 13 zones is chosen. Simulation environment is developed to simulate floor thermal dynamics with fan coils as heating elements. Fan coils operation in different zones is orchestrated via model predictive control in order to achieve as smooth as possible cumulative heating power consumption profile while respecting individual comfort requirements in zones. The approach inherently limits the maximum power measured on a calorimeter. Compared to conventional decentralized hysteresis-based temperature control in zones, simulation results show that the heating power variance is successfully reduced and fan coils smooth operation is ensured while the overall energy efficiency increased. By replicating the presented control system through all the segments of a building, this approach yields important benefits of smoothing the building power consumption curve and significant savings in contracted heating power.