Smart choice or flawed approach? An exploration of connected thermostat data fidelity and use in data-driven modelling in high-rise residential buildings

Abstract

Connected thermostat data provide new opportunities to access heating, ventilation and air conditioning (HVAC) operation and indoor condition data in high-rise residential buildings. However, how well these thermostat data reflect actual conditions and operation is unclear and best-practices to leverage these data for energy use modelling are needed. Connected thermostat data from 54 suites in two high-rise residential buildings are used to investigate the accuracy of thermostat-reported suite condition and HVAC runtime data and the relationship between suite-HVAC runtime and thermal energy demand. Next, data-driven approaches for forecasting suite HVAC runtime are explored. Two key challenges when using these data for energy modelling were identified. First, while a linear relationship between HVAC runtime and thermal energy demand was observed, there was significant variation in this relationship between suites. Second, the simple, data-driven regression methods tested were largely ineffective in accurately predicting suite-level HVAC runtime for hourly intervals (average error: 30%).