苏州科技大学学报（自然科学版）2025，Vol.42Issue(3)：58-69,12.DOI:10.12084/j.issn.2096-3289.2025.03.008
基于滑动窗口迁移LSTM的建筑能耗预测方法研究
Research on building energy consumption prediction method based on sliding window transfer LSTM
摘要
Abstract
With the growth of global energy consumption and increased environmental awareness,the issue of building energy consumption has become a key challenge for sustainable development.Building energy consump-tion prediction is important for energy conservation,building design,energy management and green building de-velopment.Traditional engineering and statistical methods have limitations in dealing with complex nonlinear re-lationships,with insufficient accuracy and reliability.Although machine learning methods are excellent in analyz-ing nonlinear data,they are highly dependent on data features.Deep learning methods improve the performance and adaptability of models through multi-level feature extraction.However,most of the existing methods only tar-get a single building type,ignoring the impact of different time periods on energy consumption.To address these issues,a generalized framework for building energy consumption prediction which combines migration learning and sliding window LSTM models is proposed.The framework enables energy consumption prediction across building types,automatically focuses on important features,efficiently integrates cross-domain data,and solves the problem of heterogeneity of data sources,thus improving the accuracy of prediction.The proposed method is studied in real projects and compared with traditional bench-marking methods.The results show that the model maintains the prediction accuracy close to that when complete data is available despite the scarcity of building data,and that the model reduces the mean absolute and root-mean-square errors between the predicted and ac-tual values by 10.078 4 and 12.554 9 on average.关键词
迁移学习/LSTM/建筑能耗预测/深度学习/能源管理Key words
transfer learning/LSTM/building energy prediction/deep learning/energy management分类
信息技术与安全科学引用本文复制引用
尚婷婷,傅启明,崔志明,吴宏杰,王陆平..基于滑动窗口迁移LSTM的建筑能耗预测方法研究[J].苏州科技大学学报（自然科学版）,2025,42(3):58-69,12.基金项目
国家自然科学基金项目(62472301 （）
62372318 （）
62073231 （）
62176175) （）
苏州大学江苏省大数据智能工程实验室开放课题(SDGC2157) （SDGC2157）
