Version 2 2021-10-26, 17:37Version 2 2021-10-26, 17:37
Version 1 2021-10-15, 15:08Version 1 2021-10-15, 15:08
thesis
posted on 2021-10-26, 17:37authored byCameron Rochon Lawrence
This research investigates the application of surrogate modelling to improve the energy performance of single-family homes. EnergyPlus was used to simulate 6000 energy models for four different semi-detached and detached century archetypes in Toronto, ON. Multivariate regression and a novel forward stepwise selection methodology were used to develop the surrogate models for each archetype. These models predicted energy use between 7.02%- 7.54% error. A combined model that contained all four archetypes was developed to determine if a single model can replace multiple models. This model predicted annual energy use with 7.03% error and the number of samples required per archetype was reduced by a factor of 3-4. Elastic net regression was tested and found to be equally as effective as the proposed stepwise selection methodology. The findings of this research support the future application of surrogate modelling as a powerful tool to develop bottom-up archetype models for century homes in Toronto, ON.