Scholarlink

<p>Prediction of higher heating value (HHV) of biochar is critical to optimize its application as a renewable energy source. A total of 13 biochar samples were produced by hydrochar method guided by a Response Surface Methodology design to ensure wide experimental coverage. It presents the development of robust models for HHV prediction using a total of 211 char data with the help of multilayer perceptron artificial neural networks (MLP-ANN). The combined use of machine learning and RSM is important both for experimental design and for producing high-accuracy outputs. Input datasets were derived from proximate analysis, ultimate analysis and a combined set of both. MLP-ANN models were trained and validated using a supervised learning approach using mean absolute error (MAE), root mean square error and coefficient of determination (R²) as key performance indicators to evaluate model accuracy. Among the MLP-ANN models produced, the combined proximate-ultimate analysis-based model produced the best results in terms of correlation coefficient (R<sup>2</sup>: 0.9785), root mean square error (RMSE: 0.5120), and mean absolute percentage error (MAPE: 0.0385). These findings demonstrate the superior ability of MLP-ANN to capture the complex, nonlinear relationships between biochar composition and energy content and highlight the added value of integrating proximate and ultimate analyses. The study provides a reliable tool for rapid estimation of biochar HHV, which supports the advancement of biomass energy applications.</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

Evaluation of HHV and Ash of Biochars Obtained from cow dung and Prediction of HHV using Machine Learning

  • 期刊论文

  • 发表日期：

    2026年5月23日

查看全文
  • Vedat Adıgüzel,
  • Sevilay Demirci,
  • Muhammet Ali Karabulut,
  • Erman Öztürk,
  • Fikret Akdeniz

摘要

Prediction of higher heating value (HHV) of biochar is critical to optimize its application as a renewable energy source. A total of 13 biochar samples were produced by hydrochar method guided by a Response Surface Methodology design to ensure wide experimental coverage. It presents the development of robust models for HHV prediction using a total of 211 char data with the help of multilayer perceptron artificial neural networks (MLP-ANN). The combined use of machine learning and RSM is important both for experimental design and for producing high-accuracy outputs. Input datasets were derived from proximate analysis, ultimate analysis and a combined set of both. MLP-ANN models were trained and validated using a supervised learning approach using mean absolute error (MAE), root mean square error and coefficient of determination (R²) as key performance indicators to evaluate model accuracy. Among the MLP-ANN models produced, the combined proximate-ultimate analysis-based model produced the best results in terms of correlation coefficient (R2: 0.9785), root mean square error (RMSE: 0.5120), and mean absolute percentage error (MAPE: 0.0385). These findings demonstrate the superior ability of MLP-ANN to capture the complex, nonlinear relationships between biochar composition and energy content and highlight the added value of integrating proximate and ultimate analyses. The study provides a reliable tool for rapid estimation of biochar HHV, which supports the advancement of biomass energy applications.

Graphical Abstract