Scholarlink

<p>Heating Load (HL) management is crucial in optimizing energy use and ensuring indoor comfort in residential buildings. Accurate HL prediction is vital for designing efficient HVAC systems, reducing energy consumption, and decreasing environmental impact. Machine learning (ML) schemes, i.e., Gaussian Process Regression (GPR), Random Forest Regression (RFR), and Adaptive Boosting Regression (ADAR) were employed in this investigation to predict HL from a database containing features like relative compactness, surface area, and wall area. Stochastic Paint Optimizer (SPO) and Motion-Encoded Particle Swarm Optimization (MPSO) enhancement frameworks were employed to improve model performance by optimizing the hyperparameters and obtaining greater predictive power. The findings indicate that hybrid models, which integrate machine learning with enhancement frameworks, significantly enhance forecasting accuracy. Among the schemes experimented on, RFR_SPO was the most accurate in the test phase with an R² value of 0.989 and RMSE of 1.024, attesting to the success of SPO in optimizing model performance. GPR_SPO also performed extremely well with R² of 0.976 and RMSE of 1.513, followed by RFR_MPSO with R² of 0.982 and RMSE of 1.319. Conversely, the ADAR scheme performed the worst with R² of 0.850 and RMSE of 3.740, indicating its poorer prediction accuracy compared to the optimized models. The hybrid models used enhancement tactics to derive more precise predictions, which can foster energy efficiency in buildings, thereby enhancing environmental sustainability via minimizing energy waste. The findings accentuate the overriding impact of enhancement tactics, particularly SPO, on enhancing HL forecasting models, demonstrating a viable framework for sustainable and energy-efficient residential HL management.</p> Graphical Abstract <p></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

Harnessing machine learning for predictive heating load management in residential buildings

  • 期刊论文

  • 发表日期：

    2025年10月3日

查看全文
  • Xiangdong Yin,
  • Tangsen Huang

摘要

Heating Load (HL) management is crucial in optimizing energy use and ensuring indoor comfort in residential buildings. Accurate HL prediction is vital for designing efficient HVAC systems, reducing energy consumption, and decreasing environmental impact. Machine learning (ML) schemes, i.e., Gaussian Process Regression (GPR), Random Forest Regression (RFR), and Adaptive Boosting Regression (ADAR) were employed in this investigation to predict HL from a database containing features like relative compactness, surface area, and wall area. Stochastic Paint Optimizer (SPO) and Motion-Encoded Particle Swarm Optimization (MPSO) enhancement frameworks were employed to improve model performance by optimizing the hyperparameters and obtaining greater predictive power. The findings indicate that hybrid models, which integrate machine learning with enhancement frameworks, significantly enhance forecasting accuracy. Among the schemes experimented on, RFR_SPO was the most accurate in the test phase with an R² value of 0.989 and RMSE of 1.024, attesting to the success of SPO in optimizing model performance. GPR_SPO also performed extremely well with R² of 0.976 and RMSE of 1.513, followed by RFR_MPSO with R² of 0.982 and RMSE of 1.319. Conversely, the ADAR scheme performed the worst with R² of 0.850 and RMSE of 3.740, indicating its poorer prediction accuracy compared to the optimized models. The hybrid models used enhancement tactics to derive more precise predictions, which can foster energy efficiency in buildings, thereby enhancing environmental sustainability via minimizing energy waste. The findings accentuate the overriding impact of enhancement tactics, particularly SPO, on enhancing HL forecasting models, demonstrating a viable framework for sustainable and energy-efficient residential HL management.

Graphical Abstract