Model predictive control for indoor thermal comfort and energy optimization using occupant feedback

Abstract

This study developed two model predictive control (MPC) algorithms, a certainty-equivalence MPC and a chance-constrained MPC, for indoor thermal control to minimize energy consumption while maintaining occupant thermal comfort. It is assumed that occupant perceptions of thermal sensation can be continually collected and fed back to calibrate a dynamic thermal sensation model and to update the MPC. The performance of the proposed MPCs based on Actual Mean Vote (AMV) was compared to an MPC using Fanger's Predicted Mean Vote (PMV) as the thermal comfort index. Simulation results demonstrated that when the PMV gives an accurate prediction of occupants’ AMV, the proposed MPCs achieve a comparable level of energy consumption and thermal comfort, while it reduces the demand on continually sensing environmental and occupant parameters used by the PMV model. Simulation results also showed that when there is a discrepancy between the PMV and AMV, the proposed MPC controllers based on AMV expect to outperform the PMV based MPC by providing a better outcome in indoor thermal comfort and energy consumption. In addition, the proposed chance-constrained MPC offers an opportunity to adjust the probability of satisfying the thermal comfort constraint to achieve a balance between energy consumption and thermal comfort.