Scholarlink

<p>Energy consumption forecasting in green buildings remains challenging due to complex climate-building interactions and temporal dependencies in energy usage patterns. Existing prediction models often fail to capture long-term dependencies and adapt to diverse climatic conditions, limiting their practical applicability. This study presents an integrated forecasting framework that combines sequence-to-sequence (Seq2Seq) architecture with reinforcement learning and transfer learning techniques. The framework employs long short-term memory (LSTM) networks enhanced with attention mechanisms to model temporal dependencies and climate variability in energy consumption data. The attention mechanism enables the model to focus on relevant temporal features while transfer learning facilitates adaptation across different climate zones. Experimental validation on two publicly available green building datasets demonstrates superior performance, achieving 96.2% accuracy, mean square error of 0.2635, and coefficient of determination (<InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41598_2025_16953_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="21" /> </InlineMediaObject> <EquationSource Format="TEX">\(R^2\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>R</mi> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation>) of 0.98. The proposed framework exhibits strong generalization capabilities across diverse climate conditions and building types. However, the framework requires substantial training data (6-12 months of high-quality sensor data) and shows reduced performance during extreme weather events, with RMSE increases of 15-20% under such conditions. These results suggest significant potential for improving energy management strategies in green buildings, contributing to enhanced energy efficiency and reduced carbon emissions in the construction sector. The framework is applicable to green buildings with reliable sensor infrastructure and adequate historical data, with performance optimized for standard operational conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Climate-adaptive energy forecasting in green buildings via attention-enhanced Seq2Seq transfer learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月29日

查看全文
  • Fang Peng,
  • Tao Su,
  • Qing Zeng,
  • Xiaojuan Han

摘要

Energy consumption forecasting in green buildings remains challenging due to complex climate-building interactions and temporal dependencies in energy usage patterns. Existing prediction models often fail to capture long-term dependencies and adapt to diverse climatic conditions, limiting their practical applicability. This study presents an integrated forecasting framework that combines sequence-to-sequence (Seq2Seq) architecture with reinforcement learning and transfer learning techniques. The framework employs long short-term memory (LSTM) networks enhanced with attention mechanisms to model temporal dependencies and climate variability in energy consumption data. The attention mechanism enables the model to focus on relevant temporal features while transfer learning facilitates adaptation across different climate zones. Experimental validation on two publicly available green building datasets demonstrates superior performance, achieving 96.2% accuracy, mean square error of 0.2635, and coefficient of determination ( \(R^2\) R 2 ) of 0.98. The proposed framework exhibits strong generalization capabilities across diverse climate conditions and building types. However, the framework requires substantial training data (6-12 months of high-quality sensor data) and shows reduced performance during extreme weather events, with RMSE increases of 15-20% under such conditions. These results suggest significant potential for improving energy management strategies in green buildings, contributing to enhanced energy efficiency and reduced carbon emissions in the construction sector. The framework is applicable to green buildings with reliable sensor infrastructure and adequate historical data, with performance optimized for standard operational conditions.