The energy saving air-conditioning systems can decrease power consumption under low cooling load condition. The variable refrigerant flow (VRF) air-conditioning systems provide variable refrigerant flow by changing the speed of compressor for energy saving. The compressors droved by a Brush less DC (BLDC) driver will improve the high heat generation problem in low load operation conditions. This study applies the direct neural control (DNC) with specified reference model to control the room temperature for a VRF air-conditioning system. The simulation results show the time response of room temperature will follow the responses of the reference model with better adaptability than the conventional PID control.