Scholarlink

<p>Biochar is a highly efficient and clean fuel. In recent years, significant progress has been made in machine learning technology to predict the higher heating value (HHV) of biochar. This study innovatively proposes a method to enhanced data for the HHV of biochar. The dataset was divided into three groups according to the characteristics of biomass, and the prediction model of HHV of biochar was established on the basis of three machine learning algorithms: LightGBM, CatBoost, and DNN. The effect of “enhanced data” on the prediction accuracy of the model is evaluated. Experiment results reveal that inclusion of “enhanced data” improves the model-fitting performance of the model, and the model of LightGBM is more suitable for biochar HHV prediction. The introduction of enhanced data improves the prediction accuracy of the model, with R<sup>2</sup> increasing by 0.068, MAE decreased by 0.421, and RMSE decreased by 0.180. The SHAP analysis demonstrated that inclusion of “enhanced data” changed the ranking of feature importance in that ash content of pyrolysis and temperature of pyrolysis stayed at the forefront of importance features. PDP and ICE analysis demonstrated that inclusion of “enhanced data” significantly changed the contribution of some of the features to HHV of biochar. This study provides significant reference and guidance for predicting other characteristics of biomass pyrolysis products.</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

Research on the Prediction of Higher Heating Value of Biochar Based on Machine Learning Trained with “Enhanced Data”

  • 期刊论文

  • 发表日期：

    2025年9月30日

查看全文
  • Chenxi Zhao,
  • Hang Yang,
  • Yiming Zhang,
  • Wenlong Yan,
  • Qiuxia Li,
  • Aihui Chen,
  • Xiaogang Liu

摘要

Biochar is a highly efficient and clean fuel. In recent years, significant progress has been made in machine learning technology to predict the higher heating value (HHV) of biochar. This study innovatively proposes a method to enhanced data for the HHV of biochar. The dataset was divided into three groups according to the characteristics of biomass, and the prediction model of HHV of biochar was established on the basis of three machine learning algorithms: LightGBM, CatBoost, and DNN. The effect of “enhanced data” on the prediction accuracy of the model is evaluated. Experiment results reveal that inclusion of “enhanced data” improves the model-fitting performance of the model, and the model of LightGBM is more suitable for biochar HHV prediction. The introduction of enhanced data improves the prediction accuracy of the model, with R2 increasing by 0.068, MAE decreased by 0.421, and RMSE decreased by 0.180. The SHAP analysis demonstrated that inclusion of “enhanced data” changed the ranking of feature importance in that ash content of pyrolysis and temperature of pyrolysis stayed at the forefront of importance features. PDP and ICE analysis demonstrated that inclusion of “enhanced data” significantly changed the contribution of some of the features to HHV of biochar. This study provides significant reference and guidance for predicting other characteristics of biomass pyrolysis products.