Search

Energy performance certificates (EPC) aim to provide transparency about building energy performance (BEP) and benchmark buildings. Despite having qualified auditors examining buildings through on-site visits, BEP accuracy in EPCs is frequently criticized. Qualified auditors are often bound to engineering-based energy quantification methods. However, recent studies have revealed data-driven methods to be more accurate regarding benchmarking. Unlike engineering methods, data-driven methods can learn from data that non-experts might collect. This raises the question of whether data-driven methods allow for simplified data collection while still achieving the same accuracy as prescribed engineering-based methods. This study presents a method for selecting building variables, which even occupants can reliably collect and which at the same time contribute most to a data-driven method's predictive power. The method is tested and validated in a case study on a real-world data set containing 25,000 German single-family houses. Having all data collected by non-experts, results show that the data-driven method achieves about 35% higher accuracy than the currently used engineering method by qualified auditors. Our study proposes a stepwise method to design data-driven EPCs, outlines design recommendations, and derives policy implications. [ABSTRACT FROM AUTHOR]

Energy performance certificates (EPC) aim to provide transparency about building energy performance (BEP) and benchmark buildings. Despite having qualified auditors examining buildings through on-site visits, BEP accuracy in EPCs is frequently criticized. Qualified auditors are often bound to engineering-based energy quantification methods. However, recent studies have revealed data-driven methods to be more accurate regarding benchmarking. Unlike engineering methods, data-driven methods can learn from data that non-experts might collect. This raises the question of whether data-driven methods allow for simplified data collection while still achieving the same accuracy as prescribed engineering-based methods. This study presents a method for selecting building variables, which even occupants can reliably collect and which at the same time contribute most to a data-driven method's predictive power. The method is tested and validated in a case study on a real-world data set containing 25,000 German single-family houses. Having all data collected by non-experts, results show that the data-driven method achieves about 35% higher accuracy than the currently used engineering method by qualified auditors. Our study proposes a stepwise method to design data-driven EPCs, outlines design recommendations, and derives policy implications. [ABSTRACT FROM AUTHOR]