Scholarlink

<p>Achieving robust control of nitrogen oxide (NO<sub><i>x</i></sub>) emissions in municipal solid waste incineration (MSWI) power generation is essential for the development of intelligent denitrification systems. Nevertheless, constructing a highly reliable prediction model remains challenging due to the dynamic non-linearity, time-variability, and strong variable coupling inherent within MSWI processes. To address these issues, this study introduces a novel soft-sensing framework that integrates a convolutional neural network (CNN), a convolutional block attention module (CBAM), and a Transformer architecture. Specifically, conditional mutual information maximization (CMIM) combined with mechanistic analysis is employed to optimize feature selection and determine appropriate input variables. Subsequently, the incorporation of CBAM into the CNN enhances deep feature extraction by adaptively emphasizing informative temporal and channel features. Finally, the Transformer component captures long-range dependencies and complements deep feature extraction, thereby improving both predictive accuracy and generalization capacity. Experimental evaluations demonstrate that the proposed model achieves superior performance, with the coefficient of determination (<i>R</i><sup>2</sup>), root mean squared error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) values of 0.96, 3.10&#xa0;mg/m<sup>3</sup>, 2.36&#xa0;mg/m<sup>3</sup>, and 4.17% on the training set and 0.91, 4.34&#xa0;mg/m<sup>3</sup>, 3.32&#xa0;mg/m<sup>3</sup>, and 5.89% on the test set, respectively. Overall, the model provides an effective reference for industrial soft-sensing applications and contributes valuable data support for intelligent process control.</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

Soft sensing of NOx emissions in municipal solid waste incineration using a CBAM–CNN–Transformer model

  • 期刊论文

  • 发表日期：

    2026年4月9日

查看全文
  • Wenjing Li,
  • Zhenghui Li,
  • Anli Zhou,
  • Wen Liu,
  • Jianghong Chen,
  • Wenbo Pi,
  • Zhimin Lu,
  • Shunchun Yao

摘要

Achieving robust control of nitrogen oxide (NOx) emissions in municipal solid waste incineration (MSWI) power generation is essential for the development of intelligent denitrification systems. Nevertheless, constructing a highly reliable prediction model remains challenging due to the dynamic non-linearity, time-variability, and strong variable coupling inherent within MSWI processes. To address these issues, this study introduces a novel soft-sensing framework that integrates a convolutional neural network (CNN), a convolutional block attention module (CBAM), and a Transformer architecture. Specifically, conditional mutual information maximization (CMIM) combined with mechanistic analysis is employed to optimize feature selection and determine appropriate input variables. Subsequently, the incorporation of CBAM into the CNN enhances deep feature extraction by adaptively emphasizing informative temporal and channel features. Finally, the Transformer component captures long-range dependencies and complements deep feature extraction, thereby improving both predictive accuracy and generalization capacity. Experimental evaluations demonstrate that the proposed model achieves superior performance, with the coefficient of determination (R2), root mean squared error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) values of 0.96, 3.10 mg/m3, 2.36 mg/m3, and 4.17% on the training set and 0.91, 4.34 mg/m3, 3.32 mg/m3, and 5.89% on the test set, respectively. Overall, the model provides an effective reference for industrial soft-sensing applications and contributes valuable data support for intelligent process control.

Graphical abstract