Scholarlink

<p>Accurate estimation of latent heat storage in advanced solar thermal systems is essential for optimizing energy efficiency. This study develops and evaluates Extreme Gradient Boosting (XGBoost) models to predict the thermal performance in a customized conical-fin latent heat storage system enhanced with graphene-alumina hybrid-nano-additives. Unlike prior work relying on simulation or generalized databases, this research leverages a unique, high-fidelity experimental dataset reflecting the complex, nonlinear behavior of this specific system configuration. Three tailored XGBoost models were constructed using key input variables, including fin configuration, charging time, and heat-transfer fluid temperature, to estimate Melt Fraction and Stored Energy. The best-performing model achieved exceptional accuracy, demonstrated by a Mean Squared Error of 1.58E-03 and a Coefficient of Determination of 0.99892. Predictions were highly stable and unbiased, with the Margin of Deviation remaining below ± 0.6% and an Index of Agreement of 0.99. These findings validate the effectiveness of XGBoost in modeling complex thermophysical systems and establish a scalable framework for intelligent energy storage estimation in next-generation solar technologies.</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

Extreme gradient boosting (XGBoost) prediction of latent heat storage in Conical‑Fin solar thermal units with Graphene-Alumina Hybrid-Nano‑Additives

  • 期刊论文

  • 发表日期：

    2026年1月28日

查看全文
  • Andaç Batur Çolak

摘要

Accurate estimation of latent heat storage in advanced solar thermal systems is essential for optimizing energy efficiency. This study develops and evaluates Extreme Gradient Boosting (XGBoost) models to predict the thermal performance in a customized conical-fin latent heat storage system enhanced with graphene-alumina hybrid-nano-additives. Unlike prior work relying on simulation or generalized databases, this research leverages a unique, high-fidelity experimental dataset reflecting the complex, nonlinear behavior of this specific system configuration. Three tailored XGBoost models were constructed using key input variables, including fin configuration, charging time, and heat-transfer fluid temperature, to estimate Melt Fraction and Stored Energy. The best-performing model achieved exceptional accuracy, demonstrated by a Mean Squared Error of 1.58E-03 and a Coefficient of Determination of 0.99892. Predictions were highly stable and unbiased, with the Margin of Deviation remaining below ± 0.6% and an Index of Agreement of 0.99. These findings validate the effectiveness of XGBoost in modeling complex thermophysical systems and establish a scalable framework for intelligent energy storage estimation in next-generation solar technologies.