The continuous increase in electricity use for cooling of buildings is unsustainable. The combination of natural ventilation (NV) with traditional mechanical cooling systems can be part of the solution if an effective control system is able to combine the two systems while maintaining indoor air quality and thermal comfort. Currently used rule-based heuristic controls lack flexibility to tackle this control challenge, while advanced methods like reinforcement learning require substantial computational resources and extensive training data. In response to this challenge, this paper presents a real-world operation deployment of Model Predictive Control (MPC) MPC framework for natural ventilation control. The proposed MPC approach uses a resistance-capacitance thermal zone model coupled with a previously developed artificial neural network predictor (ANN) of NV airflow. Measured savings in electricity use for cooling were obtained in a university library building exposed to the warm climate of Lisbon, Portugal. Results show that the proposed MPC approach provided significant electricity use reductions. On mild summer days, using the MPC during the daytime reduced HVAC electricity energy use by 68%. When the MPC was used also for night cooling of the internal thermal mass the measured savings reached 97%. On hotter days, the savings ranged between 21% and 26%. Numerical simulations validated by these experimental results indicate seasonal electricity use for cooling reductions of 37% using daytime MPC and up to 56% when night cooling is added. • Proposes an easily deployable MPC that combines RC and ANN. • MPC performance was assessed with experimental measurements and numerical simulations. • NV use results in measured mechanical cooling energy savings of 68–97% on mild days and 21–26% on hot days. • Validated numerical simulations showed mechanical cooling energy savings of: 37%–56% in a cooling season.
Simões et al. (Sat,) studied this question.