@inproceedings{simbuild2024_2167,
	doi = {},
	url = {https://publications.ibpsa.org/conference/paper/?id=simbuild2024_2167},
	year = {2024},
	month = {May},
	publisher = {IBPSA-USA},
	author = {Kathryn Elaine Kaspar and  Mohamed M. Ouf and  Ursula Eicker},
	title  = {Data-Driven Occupant-Thermostat Override Models for Winter Heating in Quebec},
	booktitle = {Proceedings of SimBuild Conference 2024},
	volume  = {11},
	isbn = {},
	address  = {Denver, Colorado},
	series  = {IBPSA-USA Building Simulation Conference},
	pages = {725--734},
	abstract = {Understanding occupant thermostat preferences and thermostat setpoint override behavior is critical for exploiting thermostat setpoints or HVAC systems for energy flexibility and demand response potential. This study models occupant thermostat overrides in residential buildings during the heating season in Quebec, Canada. Two distinct occupant types, 'Average' and 'Tolerant', are identified based on their thermostat setpoint preferences. Discrete-time Markov logistic regression models are developed to predict the probability of thermostat setpoint overrides during 'Home', 'Sleep', and 'Work' hours for both occupant types. Random forest models are employed to classify the magnitude of the setpoint changes as either small (less than 0.5 °C) or large (greater than 0.5 °C). Using these models, we are thus able to estimate the probability of a setpoint override and the magnitude of the setpoint override for two distinct occupant types in Quebec. Results indicate good model performance with balanced accuracy, recall, and precision. However, limitations include data availability, model assumptions, and the need for more comprehensive occupancy data. These models offer potential applications in demand response and HVAC control strategies for residential buildings.},
	issn = {},
	Organisation = {IBPSA-USA},
	Editors = {}
}