Scholarlink

<p>This exploration investigates advanced machine learning (ML) tactics, specifically MLP, RBF, and XGBoost algorithms, for forecasting building energy usage, incorporating hyperparameter optimization tactics such as GBO and AO. The research shifts from traditional physics-based schemes to automated, scalable data-driven approaches. Four predictive schemes were developed and evaluated using an 80/20 training/testing data split. Outcomes have shown that the RBF model, with its R<sup>2</sup> value at 0.5770, runs contrary to the hybrid schemes, specifically XGBoost-GBO, performing at an R<sup>2</sup> value of 0.9210; some of the most influential variables are IV18, IV11, IV17, and IV12. These outcomes showed that enhanced performance and efficiency from the ML schemes, after optimization, served to enhance overall energy management practices, further improving building energy system forecasting. Anticipating peak demand periods allows organizations to adjust their usage accordingly, reducing the need to procure expensive peak-time energy and lowering overall energy costs.</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

Enhancing building energy forecasting with advanced ML techniques: An in-depth study of model performance, parameter optimization, and predictive accuracy

  • 期刊论文

  • 发表日期：

    2025年10月8日

查看全文
  • Gengqiang Huang,
  • Hailu Wan

摘要

This exploration investigates advanced machine learning (ML) tactics, specifically MLP, RBF, and XGBoost algorithms, for forecasting building energy usage, incorporating hyperparameter optimization tactics such as GBO and AO. The research shifts from traditional physics-based schemes to automated, scalable data-driven approaches. Four predictive schemes were developed and evaluated using an 80/20 training/testing data split. Outcomes have shown that the RBF model, with its R2 value at 0.5770, runs contrary to the hybrid schemes, specifically XGBoost-GBO, performing at an R2 value of 0.9210; some of the most influential variables are IV18, IV11, IV17, and IV12. These outcomes showed that enhanced performance and efficiency from the ML schemes, after optimization, served to enhance overall energy management practices, further improving building energy system forecasting. Anticipating peak demand periods allows organizations to adjust their usage accordingly, reducing the need to procure expensive peak-time energy and lowering overall energy costs.