This paper proposes a supervisory demand-based temperature control for air-conditioning systems of large-scale rooms that aims to improve the control performance and energy performance. In a large-scale room, it is very likely that load may not be evenly distributed. To capture the load distribution, the large-scale room is divided to a number of zones, each of which is served by an independent Variable-Air-Volume (VAV) box. In each zone, wireless sensor nodes are installed at the breathing level to measure the temperature so as to guarantee that the controlled temperature is the temperature that occupants really need. A demand-based supervisory control is developed to realize temperature tracking only for zones with occupants but take account of thermal coupling between occupied and unoccupied zones. This paper will focus on the study of the energy saving potential of the proposed control. The potential is evaluated under different occupancy conditions by comparing with a conventional temperature control using computational fluid dynamics (CFD) simulations and lab-based experiments. (C) 2013 Elsevier B.V. All rights reserved.