Prioritising residential buildings retrofits using an

AI-data-driven tool to enhance EPC rating

Hamidreza Seraj

School of Computing and Engineering

Supervisors:

Professor Ali Bahadori-Jahromi

School of Computing and Engineering

More than 20 percent of all UK homes, around 6 million, were built before 1920. So, building energy retrofit plays a significant role in achieving national energy efficiency and greenhouse gas emissions reduction goals. In this context, energy performance certificate (EPC) rating scheme is one of the policies that suggests benchmarking the building stocks to identify their energy performance and energy retrofit requirement. This study aims to develop an AI data-driven tool which evaluates the effect of various retrofit strategies on buildings’ EPC rating. So, three machine learning models including XGBoost, support vector machine (SVM), and K-nearest neighbor (KNN) were developed based on the EPC dataset for residential buildings and in accordance with standard assessment procedure (SAP) guideline in the UK. Furthermore, a user interface was designed which enables users to observe the effect of different retrofit strategies on buildings’ EPC rating. The findings of this study indicated that the XGBoost model outperformed the other models, achieving an accuracy score of more than 0.79. Therefore, this model can be utilised as a supporting tool for prioritising retrofit strategies in different types of residential buildings.