Occupant control behavior is a key factor affecting the energy consumption of building air-conditioners (ACs). The operating behavior of ACs and their models in office buildings have been investigated extensively. However, although the thermal sensation of occupants is affected by their previous thermal experience, few researchers have attempted to incorporate this effect quantitatively in models of AC turning on behavior. Not considering the cumulative effect may result in inaccurate predictions. Therefore, in this study, a survival model is proposed to describe AC turning on behavior in office buildings under the cumulative dimension of time. Based on a dataset containing environmental parameters and occupant behavior information, as well as considering occupants entering a room as the starting event and turning on an air-conditioner as the end event, the endurance time before an AC is turned on is investigated, and a survival model is used to predict the probability of the AC turning on due to environmental factors. Based on a switch curve, confusion matrix, and tolerance–time curve, the prediction results of the survival model are analyzed and validated. The results show that a tolerance temperature of 29 ℃ and a tolerance duration setting of 1 h can effectively model the turning on behavior of the AC. In addition, based on comparison results of different models, the survival model presents a more stable switching curve, a higher F1 score, and a tolerance curve that is more similar to reality. Different tolerance durations, as well as static and dynamic tolerance temperature settings, are considered to optimize the model. Furthermore, the AC energy consumption is calculated under the survival model and the traditional Weibull model. Simulation results were compared with measurement, and the survival model verified the improvement effect of prediction accuracy by 8% than the Weibull model. By considering the time-transformed accumulation of physical environmental factors, the accuracy of AC turning on models can be improved, thus providing an effective reference for future building energy consumption simulations.

Reliable energy and performance prediction for building design and planning is important for newly-designed or retrofitted buildings. Window operating behavior has an important influence on the ventilation and energy consumption of these buildings under different realistic scenarios. Therefore, quantitatively describing this behavior and constructing a prediction model are important. In this work, an action-based Markov chain modeling approach for predicting window operating behavior in office spaces was proposed. Two summer measurement data (2016 and 2018) were used to verify the accuracy and validity of the modeling approach. The opening rate, outdoor temperature, time distribution, and on-off curve were proposed as four inspection standards. This study also compared the prediction performance between the action-based Markov chain modeling approach with the state-based Markov chain modeling approach, which is the most popular modeling approach to model occupant window operating behavior. This study proved that the yearly variation of occupants’ behavior performed a form of action that remained unchanged during a certain period. Meanwhile, the results also proved that the action-based Markov chain modeling approach can reflect the actual window operating behavior accurately within an open-plan office, which is a beneficial supplement for energy-consumption simulation software in a window-state prediction module. The state-based Markov chain modeling approach showed better stability and accuracy in terms of the opening rate, whereas the action-based Markov chain modeling approach showed good consistency with the measurement data in the on-off curves and in situations with little data. For the on-off curves, the accuracy of action-based modeling approach in the prediction of window open-state is 20% higher.

During technical evaluations of cooling systems in residential buildings, it is necessary to consider the influence of the household air-conditioning (AC) use modes. In other words, how the occupants control the AC, for instance, when it is turned on, what the temperature setting is, and how long it is used. Field measurements and spot interviews indicate that AC usage by residents should be environmental, event and random related. A reduced-order AC conditional probability (CP) model was developed in this study to describe AC usage. The AC CP model was integrated with a building energy modeling program (BEMP) to reflect the interaction of the AC operation and the indoor environment. With consideration of stochastic AC use modes, the uncertainty of user compositions was studied. Additionally, simulation results revealed that AC use modes and user compositions can cause up to a 4.5-fold difference in the system efficiency of district cooling systems. The Lorenz curve and Gini coefficient were applied in this study to describe the load distribution in a quantitative manner. Through a comparison with the constant schedule definition model, the study also identified inclusion of the stochastic feature of AC use modes and their compositions in simulations as being important to the technical evaluation of district cooling systems.