Development of regression models by peer group for energy performance evaluation of office buildings using national big data

Abstract

With energy consumption in the building sector being a key area within which to reduce national greenhouse gas emissions, analytical techniques that can systematically evaluate building energy performance are increasingly needed. Therefore, this study aims to develop regression models by peer groups to evaluate building energy performance using national big data, including the energy consumption data of all office buildings in Korea, to separate energy consumption into cooling, heating, and baseload energy. For this, Simplified Weather-related Building Energy Disaggregation (SED) and Change Point Model (CPM) were used with data from 2018 to 2019, and correlation analysis was used to derive major variables that affect energy consumption. Nine peer groups were set based on gross floor area and permit year, and energy consumption characteristics were analyzed. Finally, regression models were developed for each group and energy use, and the forecast accuracy of the models was evaluated through the adjusted coefficient of determination (adj. R2). The analysis results showed that the larger the building size, the higher the forecast accuracy (adj. R2 up to 0.737), and that weather variables, such as cooling and heating slopes, Heating Degree Days (HDD), and Cooling Degree Days (CDD), were selected as major variables.