Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Clustering analysis for building energy benchmarking based on weekly load profile
Buildings account for more than 30% of energy consumption and greenhouse gas emissions worldwide. Therefore, there are tremendous research to optimize energy usage in buildings. Nowadays, the energy rating system benchmark buildings, that is, to determine the peer groups of buildings, so as to evaluate the energy performance and uncover the energy saving potential of buildings. Current research usually uses daily load profile instead of primary space use (PSU) labels to benchmark buildings. But this method ignores information such as the difference in load shape profiles between weekdays and weekends. In this paper, the proposed framework focuses on weekly load profile for benchmarking. To avoid the curse of dimensionality, the smart meter data is firstly compressed by autoencoder. Then, by clustering the compressed results, a representative load profile of building is obtained, which corresponds to an electricity consumption pattern. The framework was tested on the Building Data Genome (BDG) Project. The results showed that compared to benchmarking through PSU or daily load profile, our methodology allows for better grouping of buildings, so that the building energy performance can be more accurately evaluated.
Clustering analysis for building energy benchmarking based on weekly load profile
Buildings account for more than 30% of energy consumption and greenhouse gas emissions worldwide. Therefore, there are tremendous research to optimize energy usage in buildings. Nowadays, the energy rating system benchmark buildings, that is, to determine the peer groups of buildings, so as to evaluate the energy performance and uncover the energy saving potential of buildings. Current research usually uses daily load profile instead of primary space use (PSU) labels to benchmark buildings. But this method ignores information such as the difference in load shape profiles between weekdays and weekends. In this paper, the proposed framework focuses on weekly load profile for benchmarking. To avoid the curse of dimensionality, the smart meter data is firstly compressed by autoencoder. Then, by clustering the compressed results, a representative load profile of building is obtained, which corresponds to an electricity consumption pattern. The framework was tested on the Building Data Genome (BDG) Project. The results showed that compared to benchmarking through PSU or daily load profile, our methodology allows for better grouping of buildings, so that the building energy performance can be more accurately evaluated.
Clustering analysis for building energy benchmarking based on weekly load profile
Ruan, Yiming (Autor:in) / Liu, Fuqiang (Autor:in)
01.09.2021
1521379 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Benchmarking Industrial Building Energy Performance
| British Library Conference Proceedings | 2005