文献类型：期刊文献

英文题名：A gated multimode transformer with feature-adaptive partitioning and cross-mode fusion for industrial flue gas emission prediction

作者：Long, Jian[1];Li, Xu[1];Wang, Luyao[1];Hu, Guihua[1]

机构：[1]East China Univ Sci & Technol, Key Lab Smart Mfg Energy Chem Proc, Shanghai 200237, Peoples R China

年份：2026

卷号：332

外文期刊名：CHEMICAL ENGINEERING SCIENCE

收录：;EI(收录号：20261820630065);WOS:【SCI-EXPANDED(收录号:WOS:001761593400001)】；

基金：This work was supported by National Natural Science Foundation of China (62373155, 62273149) .

语种：英文

外文关键词：Desulfurization and denitrification; Patternclassification; Multimode modeling; GMFTrans

摘要：Given the pressing demand for cleaner production in the refining industry, developing accurate and robust models for predicting NOX and SO2 concentrations is of great practical significance. However, the actual desulfurization and denitrification processes involve multiple operating modes, strong nonlinearity, and complex variable coupling, posing substantial challenges to predictive modeling. Therefore, this work presents a multimode modeling approach grounded in a gated multidimensional information fusion transformer (GMFTrans). First, the feature adaptive multimode partitioning method is employed to partition industrial data into different modes. The GMFTrans encoder is then employed to extract critical mode-specific features. To further leverage inter-mode information, a multi-layer perceptron captures cross-mode dependencies, while a gating mechanism dynamically integrates multi-dimensional information, facilitating deeper exploration of intra-mode and intermode correlations. Experimental results from three practical petrochemical processes substantiate the efficacy of the proposed framework. The method achieved notable predictive accuracy, with R2 values of 0.897, 0.991, and 0.998 for NOX concentration, SO2 concentration, and C4 content, respectively.