Hybrid Model Predictive Control Scheme for Controlling Temperature in Building Under Uncertainties

Abstract

This paper presents a study on temperature control with a heater in a building using Model Predictive Control (MPC) with a focus on addressing two uncertainties: in model and the weather forecast. In previous works, a grey-box model of the building system was developed, and the values of parameters were estimated by the estimation techniques. In this work, based on the model, simulations are conducted comparing four types of MPC controllers: Deterministic MPC, Multistage MPC, Chance-constrained MPC, and hybrid MPC. The hybrid framework integrates the strengths of the multistage and chance-constrained MPCs to achieve conservative performance and increased robustness in constraint satisfaction. The simulations demonstrate that while deterministic MPC may not always guarantee constraint satisfaction, the hybrid framework offers improved robustness by considering uncertainties in model mismatch and uncertain weather forecasts. The 95% confidence region of model uncertainty is used to assess the robustness of simulations. The results show that the hybrid MPC approach is effective in maintaining temperature in the desired range and ensuring constraint satisfaction in controlling the temperature in a building.

Keywords

• Chance-constrained MPC
• HVAC system
• model predictive control
• multistage MPC
• optimal control
• optimization