Control sequence prioritising ceiling fan operation over air conditioners using machine learning for thermal comfort
Abstract
This paper proposes and tests the implementation of a sustainable cooling approach that uses a machine learning model to predict operative temperatures, and an automated control sequence that prioritises ceiling fans over air conditioners. The robustness of the machine learning model (MLM) is tested by comparing its prediction with that of a straight-line model (SLM) using the metrics of Mean Bias Error (MBE) and Root Mean Squared Error (RMSE). This comparison is done across several rooms to see how each prediction method performs when the conditions are different from those of the original room where the model was trained. A control sequence has been developed where the MLM’s prediction of Operative Temperature (OT) is used to adjust the adaptive thermal comfort band for increased air speed delivered by the ceiling fans to maintain acceptable OT. This control sequence is tested over a two-week period in two different buildings by comparing it with a constant air temperature setpoint (24oC). Analysis of the data showed that the MLM is more consistent with lower errors than the SLM across a variety of rooms. Compared to the constant air temperature setpoint control, the OT control sequence showed improved comfort reported by 70 occupants in the study and a cooling electrical energy savings of over 90% during the test conditions.
Keywords
Sustainable cooling, Ceiling fans, Adaptative comfort, Air speed, Machine learning, Ai, Control sequence