Application of surrogate modelling to improve the thermal performance of single-family homes through archetype development

This paper investigates the application of surrogate modelling to create bottom-up archetype models, to improve the energy performance of specific residential building archetypes. The objective of this study is to quantify the application range(s) of surrogate models to predict and improve the annual energy performance of single-family home archetypes. Six distinct surrogate models were developed using multivariate regression and artificial neural networks, based on 6000 simulated samples extracted from four different century home archetypes in Toronto (Canada) as a result of random field study within a target neighbourhood. The results show that a single surrogate model can predict the annual energy requirements for four distinct archetypes with 7.87% average error with a multivariate regression model and 4.13% average error using an artificial neural network in 0.00047 and 0.013 ms respectively. These results indicate that the bottom-up surrogate model can describe the relationship between retrofit strategies and annual energy demand with substantial computational time savings. The findings of this study support the future application of surrogate modelling to develop bottom-up archetype models to continue the development of the Toronto Archetype Project and to support a national retrofit strategy in Canada.