TO-22-C065 - Virtual Energy Auditing: A Data-Driven Model for Predicting EnerGuide Rating System

The EnerGuide Rating System (ERS) is a program of the Government of Canada to rate the energy performance of low-rise housing. The process to rate a house requires a site visit by a certified energy advisor to measure energy affecting parameters such as building location, orientation, geometry, airtightness, and envelope characteristics, as well as mechanical equipment. These data are then used to simulate the energy performance of the house, from which the rating is determined. As this process is labour intensive and requires approval and attendance by homeowners, an alternative method for the calculation of the rating is sought. In this paper, we propose a black-box approach to derive an equivalent ERS rating based on a basic list of building characteristics to enable a virtual energy audit. Given the large number of parameters available in the original dataset, a random forest-based feature selection method is developed to find the most influential inputs. The selected parameters from more than 7,000 houses were used to estimate the EGH rating using black-box modeling. The results indicate that a reasonable accuracy for representation of the traditional methods is achieved given sufficient training data samples from previously rated buildings in the same region.