Stochastic nature of occupants' behavior toward air-conditioning operation in residential buildings

Occupants' behaviors (OBs) toward air conditioning (AC) usage in residential buildings have a strong influence on diverse time patterns of load profiles; thus, various stochastic OB modeling has been proposed. However, the validity of such models has not been fully assessed with metered data, particularly for modeling AC switching-off actions. In this study, the AC operation behavior of 20 dwellings in Japan was observed during the summer. The occurrence of AC-use events identified from the measured data was analyzed in relation to the time slot of an event, thermal conditions, and frequency of AC usage in each dwelling. The results exhibited a difference between switch-on and switch-off actions; the former is primarily dominated by indoor temperature because of thermal adaptation behavior, while the latter is more habitual. Based on this analysis, this study proposed a revised algorithm to synthesize AC operation schedules, which expresses the switch-off action as the probability of the duration of an AC-use event rather than a state transition probability function. The proposed algorithm can be incorporated into dynamic building energy simulations. The validity of the revised algorithm was also demonstrated through comparison with the measured data.