Effect of limited controller and sensor datasets on the performance of data-driven model predictive control for residential buildings

Data-driven model predictive control (MPC) has gained popularity as a proactive control strategy due to the development of the Internet of Things (IoT) and data storage technologies. To apply the data-driven MPC to residential buildings, it is necessary to consider what data to measure, how to control a target system, and which prediction models are inserted. In this paper, we investigate the effect of different datasets due to sensor and controller set selection on data-driven model and data-driven MPC performance. We confirmed that the control variable according to the controller set affects MPC performance considerably.