Investigating Dwelling Overheating Risks in Cooler Climates Using Ensemble-Based Machine Learning Surrogates of Parametric Dynamic Simulation Models
摘要
Most research on housing overheating focuses on extreme heat events in countries with tropical climates during the summer. For countries with temperate climates, with typically mild summer periods, the focus, to date, has been on retrofitting existing dwellings to reduce heating energy consumption in the cold months. However, because of climate change, there is an expected rise in temperature, particularly during summer, which poses a significant risk to dwellings of cooler climates, especially in areas that do not have cooling systems. Hence, to investigate the risk of overheating in summer in cooler climates, the research aims to quantify the effect of climate change on overheating risk in the context of Ireland and appraise the significant contributory factors among the building features towards overheating using a mid-floor apartment typology. The research employs an ensemble-based machine learning surrogate methodology using the dataset generated from the parametric dynamic simulation models. The study indicates that dwellings that fall under more energy-efficient classes may be at greater risk of overheating, with window specification and thermal mass as key contributory factors. Researchers and policymakers can replicate and implement the methodology adopted in the research to investigate possible climate adaptation strategies in countries with similar colder climates, thus moving towards creating climate-resilient dwellings.