Modelling and simulating occupant behaviour on air conditioning in residential buildings

Occupant behaviour on air conditioning directly influences building energy consumption. Current building energy simulation typically assumes that occupants operate air conditioners ideally ignoring the stochastic characteristic of occupant behaviour, which leads to a large deviation of predicted energy performance from real consumption. This paper proposed a methodology framework for modelling occupants' stochastic behaviour on air conditioning. This method combined measured data, statistical analysis and logistic regression to derive stochastic behaviour models that give the probability of when and above which temperature occupants may turn on air conditioning, at what temperature they may set and when they may turn off air conditioning. The behaviour models developed in this paper were further incorporated in widely used building simulation engine-EnergyPlus. The accuracy of the developed simulation model was validated against field measurement. Using this model, the cooling energy performance was simulated and compared with results from building energy standard based settings. The result shows that simulation using settings according to building energy standard over-predicts cooling load and total cooling energy requirements by 113% and 5.6 times respectively, which are very large discrepancies and may lead to a misunderstanding of energy savings potential of energy efficiency measures. The methodology developed in this paper has generality and can be extended and applied in other buildings for constructing behaviour models of air conditioning. The limitations of the current research as well as future research works were also discussed.