Heating demand and indoor air temperature prediction in a residential building using physical and statistical models: a comparative study

Abstract

Abstract In Canada, space heating accounts for the largest proportion of energy consumption in residential buildings. Therefore, accurately predicting the heating demand and interior temperature of a residential building plays a vital role in estimating the building’s total energy consumption with the consideration of thermal comfort. The prediction results obtained through different models could be used to develop predictive controllers to achieve peak shifting as well as to provide utility providers with valuable information for electric power distribution. Common methods to predict heating demand mainly include physical models and statistical methods. This study used two physical models (i.e. TRNSYS model and TRNSYS-CONTAM model) and one statistical model using supervised machine learning algorithm to predict the heating demand as well as the indoor temperature of a residential building, located in Quebec, Canada. Results show that TRNSYS-CONTAM model has higher accuracy than TRNSYS model no matter in terms of interior air temperature or heating demand prediction, while the statistical model shows better interior air temperature prediction result than physical models.