Table 4.

A summary of MPC-based frameworks, their characteristics and contributions in saving energy and optimizing thermal comfort

Work	ML model	Building type	System	Best performance	Control objective
Gao et al. (2020)	FNN + DDPGs	Public buildings	HVAC	4.31%HVAC energy saving	Save energy and improve thermal comfort.
Yang et al. (2020)	ANN-NARX	Office and lecture theater	ACMV	58.5% cooling thermal saving	Energy saving and thermal comfort optimization.
Yang et al. (2021)	RNN-NARX	Office and a lecture theater	ACMV	52% reduction of cooling energy	Save energy and optimize thermal comfort. in experimental testbeds
Chen et al. (2020)	MLP-based transfer learning	Residential buildings	HVAC	MSE=0.16	Optimize energy efficiency and thermal comfort.
Bünning et al. (2020)	RF	Residential buildings	HVAC	24.9% of cooling energy saving	Optimize energy consumption withoutcompromising thermal comfort
Yang and Wan (2022)	RNN-NARX	Office in a hospital	ACMV	26–31.6% cooling energy savings	Save energy and optimize thermal comfort
Li and Tong (2021)	Encoder-decoder RNN	Residential/public buildings	HVAC	4–7% energy saving	Energy saving and smart control of thermal environment
Mtibaa et al. (2021)	CAM- LSTM	Multi-zone buildings	HVAC	MAPE = 0.0872%	Save energy, predict peak power and improve thermal comfort