基于XGBoost算法的夏热冬冷地区办公建筑围护结构的负荷预测

打开文本图片集

中图分类号：TU832 文献标志码：A

Abstract： Accurate air conditioning load prediction enhances building energy management and optimization， demonstrating significant potential for energy savings. To enable rapid hourly cooling load prediction for diverse buildings in hot summer and cold winter zones，this study decouples building cooling loads，focusing specifically on envelope load prediction. First， a baseline XGBoost-based model for hourly envelope cooling load prediction was developed， with comparative analysis offour feature combinations revealing feature set D as optimal. Subsequently， a generalized prediction model adaptable to various oice buildings was created by applying differential corrections to the baseline model. Validated against EnergyPlus simulations using test buildings in Shanghai and Hangzhou， the XGBoost-predicted hourly envelope cooling loads showed strong agreement， confirming the model's generalizability and accuracy in predicting envelope thermal performance across different buildings.

Keywords： load prediction; machine learning; EnergyPlus software; XGBoost algorithm

随着人们对舒适生活的不断追求，社会对于能源的需求与日俱增，其中建筑能耗占社会总能耗的比重也持续递增。（剩余7613字）