| Keywords |
Energy consumption pattern ; Machine learning ; Clustering analysis ; Building usage |
| Abstract |
The analysis of energy use patterns is considered to be one of the important building energy performance evaluation, because it can improve the understanding of energy consumption characteristics of building. Pattern analysis can also be used for benchmarking with other buildings. In this study, we analyzed the building energy consumption patterns based on the energy consumption data from building energy management system(BEMS) using machine learning techniques, especially k-means clustering. Energy consumption data were collected from 8 buildings with different building usage for one year. As a result, the office type buildings (A, B, D, E) showed different characteristics in seasons, weekday and holiday, etc according to the number k. The residential building (C) showed no significant difference in the weekday and weekday, but was more sensitive to seasonal changes. The buildings that operate 24 hours, such as F(Police station), G(Fire station), showed similar energy use patterns on weekdays and weekends. The school building (H) was divided into 11 clusters according to weekdays and holidays. Clustering based on building energy consumption could reveal different energy patterns and characteristics according to building operation, and the results were not always match to the nominal building usage. |