Improving predicted mean vote with inversely determined metabolic rate

Inaccurate thermal comfort prediction would lead to thermal discomfort and energy wastage of overcooling/overheating. Predicted Mean Vote (PMV) is widely used for thermal comfort management in air-conditioned buildings. The metabolic rate is the most important input of the PMV. However, existing measurements of the metabolic rate are practically inconvenient or technically inaccurate. This study proposes a method to improve the PMV for the thermal sensation prediction by inversely determining the metabolic rate. The metabolic rate is expressed as a function of the room air temperature and velocity considering the effects of the physiological adaptation, and inversely determined using an optimizer (variable metric algorithm) to reduce the deviation between the PMV and thermal sensation vote. Experiments in environmental chambers configured as a stratum ventilated classroom and an aircraft cabin and field experiments in a real air-conditioned building from the ASHRAE database validate the proposed method. Results show that the proposed method improves the accuracy and robustness of the PMV in the thermal sensation prediction by more than 52.5% and 41.5% respectively. Essentially, the proposed method develops a grey-box model using model calibration, which outperforms the black-box model using machine learning algorithms.