期刊
SUSTAINABLE CITIES AND SOCIETY
卷 60, 期 -, 页码 -出版社
ELSEVIER
DOI: 10.1016/j.scs.2020.102194
关键词
Residential building; Air conditioner; Energy use prediction; Ensemble learning; Feature selection; Energy management
资金
- National Key RAMP
- D Program of China [2016YFC0700300]
- graduate research and innovation foundation of Chongqing, China [CYS18027, CYB17006]
The energy use prediction of residential buildings has an increasingly important role in urban energy management. This study proposed a prediction model for the cooling energy use of air conditioners in residential buildings. Large-scale monitoring data of the operation of 1325 air conditioners in Chongqing were collected from the networking cloud platform of an air conditioner manufacturer, including setting parameters by occupants, indoor environmental parameters, time parameters and energy use parameters. The historical monitoring data of previous week before the forecast day, the meteorological data of the forecast day and the apparatus parameters of AC were employed as the original data set in this study. Feature selection engineering, including correlation analysis, importance analysis and collinearity analysis, were performed in sequence to select the most correlated and important input features for energy use prediction. Afterwards, prediction models that use four ensemble learning methods and two single learning methods were developed and compared by evaluation metrics. The best model for prediction was proposed. The results show that eleven input features have a great relationship to the daily cooling energy use and were considered the inputs to the prediction model. The XGBoost model was chosen as the best model in this study. The proposed prediction model can help researchers understand which historical features are important for the future daily cooling energy use prediction of AC. This prediction model can provide some references for different groups to implement energy management for residential buildings.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据