Chang Shu

Energy retrofits of existing residential buildings are essential for reducing energy consumption and greenhouse gas emissions. However, limited financial resources and practical constraints require effective prioritization of retrofit measures. Retrofit effectiveness varies across climate zones, end uses, building characteristics, and future climate conditions, making uniform retrofit strategies inefficient at national scales. This study develops a data-driven framework to identify climate-zone and end-use specific retrofit priorities for residential buildings in Canada under both historical and future climate scenarios. A synthetic database is generated through parametric simulations of residential building archetypes designed according to climate-zone-specific code requirements. Machine learning (ML) models are trained to predict heating and cooling energy use intensity (EUI) separately for each climate zone. Explainable artificial intelligence, based on SHapley Additive exPlanations (SHAP), is applied to interpret model predictions and quantify the relative importance and directional influence of individual parameters. Results show climate-zone-dependent differences in retrofit priorities. In warmer and transitional zones (CZ 4-6), cooling EUI is primarily driven by window solar control (SHGC and south-facing FDWR), cooling system efficiency, and ventilation, whereas in colder zones (CZ 7A-8), foundation type and ground heat loss have greater influence. Heating-related retrofit measures generally offer larger energy-saving potential than cooling-related measures, particularly in colder climates. Under future climate scenarios, mechanical cooling efficiency becomes increasingly important in CZ 4-7A, while heating performance in colder zones remains dominated by envelope, foundation, and ventilation system with heat recovery. Overall, the proposed ML-XAI framework provides transparent, climate-responsive insights to support regionally differentiated retrofit strategies and evidence-based energy policy development.

Climate heat waves occurring in urban centers are a serious threat to public health and wellbeing. Historically, most heat-related mortalities have arisen from excessive overheating of building interiors housing older occupants. This paper developed an approach that combines the results from building simulation and bioheat models to generate health-based limit criteria for overheating in long-term care homes (LTCHs) by which the body dehydration and core temperature of older residents are capped during overheating events. The models of the LTCHs were created for buildings representative of old and current construction practices for selected Canadian locations. The models were calibrated using measurements of indoor temperature and humidity acquired from monitoring the building interiors and the use of published building energy use intensity data. A general procedure to identify overheating events and quantify their attributes in terms of duration, intensity, and severity was developed and applied to LTCHs to generate the limit criteria. Comparing the limit criteria from the proposed and comfort-based methods showed evident differences. The proposed method predicted the overheating risk consistent with the overall thermal comfort during overheating events in contrast to the comfort-based methods. The new limit criteria are intended to be used in any study to evaluate overheating risk in similar buildings.