Scholarlink

This study investigates the potential of machine learning models to predict the Higher Heating Value (HHV) of biochar derived from agricultural crop waste, specifically focusing on food grain crops such as wheat, rice, and corn. Using three machine learning methods—random forest (RF), gradient boosting regressor (GBR), and extreme learning machine (ELM)—the aim was to provide consistent prognostics. For both training and testing sets, the models were scored using mean squared error (MSE), R2, and mean absolute percentage error (MAPE). With the lowest Train MSE (0.0053) and Test MSE (0.4253) together with a high Train R2(0.9997) and Test R2(0.99336), ELM showed the best performance, per results. ELM also has the lowest MAPE values; hence, it is the most accurate model for biochar HHV prediction. While GBR showed overfitting tendencies with an exaggerated Train R2 of 0.9991 but a much smaller Test R2 of 0.8857, RF also performed well with a Test R2 of 0.9576. These findings show ELM’s potential as a strong instrument for biochar energy output prediction, thereby supporting the optimization of biomass-to-biochar conversion techniques.

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

Agriculture Crop Waste to Biochar: Extreme Learning Machine-Based Prognostics for Higher Heating Values

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Duy Tan Nguyen,
  • Minh Tung Phung,
  • Duc Chuan Nguyen,
  • Huu Cuong Le

摘要

This study investigates the potential of machine learning models to predict the Higher Heating Value (HHV) of biochar derived from agricultural crop waste, specifically focusing on food grain crops such as wheat, rice, and corn. Using three machine learning methods—random forest (RF), gradient boosting regressor (GBR), and extreme learning machine (ELM)—the aim was to provide consistent prognostics. For both training and testing sets, the models were scored using mean squared error (MSE), R2, and mean absolute percentage error (MAPE). With the lowest Train MSE (0.0053) and Test MSE (0.4253) together with a high Train R2(0.9997) and Test R2(0.99336), ELM showed the best performance, per results. ELM also has the lowest MAPE values; hence, it is the most accurate model for biochar HHV prediction. While GBR showed overfitting tendencies with an exaggerated Train R2 of 0.9991 but a much smaller Test R2 of 0.8857, RF also performed well with a Test R2 of 0.9576. These findings show ELM’s potential as a strong instrument for biochar energy output prediction, thereby supporting the optimization of biomass-to-biochar conversion techniques.