A hybrid model predictive control scheme for energy and cost savings in commercial buildings: Simulation and experiment

Abstract

This paper presents a hybrid model predictive control (MPC) scheme for energy-saving control in commercial buildings. The proposed method combines a linear MPC with a neural network feedback linearisation (NNFL) method. The control model for the linear MPC is developed using a simplified physical model, while nonlinearities associated with the building system are handled by an affine recurrent neural network (ARNN) model through system feedback. The proposed MPC integrates several advanced air-conditioning control strategies, such as an economizer control, an optimal start-stop control, and a pre-cooling control. The developed MPC has been tested in the check-in hall of T-1 building, Adelaide Airport, through both simulation and field experiment. The result shows that the proposed control scheme can achieve a considerable amount of savings without violating occupants' thermal comfort.