Smart Wi-Fi physics-informed thermostat enabled estimation of residential passive solar heat gain for any residence

In the climate emergency we now face, developing capabilities to find and address opportunities for carbon mitigation at scale are desperately needed. Here, we address the opportunity to find buildings with high solar heat gain and thus cooling requirements, in order to identify candidate buildings for shading technologies. Prior studies have theoretically documented derived estimates of solar heat gain to buildings, both through heat transport through the envelope and through fenestration. Additionally, various sensors have been added to residences to measure solar heat gain contributions. Such approaches are not replicable at scale. Here, acquired residential smart Wi-Fi thermostat data is combined with solar radiation and weather data. From this data, physics-informed Long Short-term Memory (LSTM), Gated Recurrent Unit (GRU), and LSTM/GRU combined machine learning models to predict indoor temperature and temporal cooling are trained and validated. Solar heat gain contributions can be assessed by applying the developed model to inputs with solar heating conditions effectively turned off. The difference in cooling requirements, both real time and in total over time, between the actual cooling required and the cooling predicted in the absence of solar heating, reveals the passive solar heat gain for a building. Test residences included in the study offered different energy effectiveness and shading. The passive solar estimations developed uniquely for each residence matched expectations. The results shows that the percentage of cooling required to accommodate solar heat gain for the targeted residences ranged from 48% to 72%, dependent upon both the energy effectiveness and degree of solar shading. Most exciting is how the approach developed here can be implemented in any residence in which smart Wi-Fi thermostats exist.(c) 2022 Elsevier B.V. All rights reserved.