安全、健康和环境2025，Vol.25Issue(9)：26-33,8.DOI:10.3969/j.issn.1672-7932.2025.09.004
基于MIC-LSTM的乙烯裂解炉氮氧化物浓度预测模型
Prediction of the NOX Concentration at Ethylene Cracking Furnace Based on MIC-LSTM Model
摘要
Abstract
The generation mechanism of NOX in the flue gas emissions of ethylene cracking furnaces is com-plex and difficult to control,making it one of the most challenging pollutants to treat in the flue gas of cracking furnaces.To improve the accuracy of the NOX emission prediction model,a prediction model screening method that combines the maximum mutual information(MIC)criterion with the structural equation model is proposed.A screening method for selecting input variables for the prediction model by combining maximum mutual infor-mation with structural equation modeling(SEM)is proposed.Firstly,MIC is used to screen out the variables with strong correlation as potential model inputs.Further,the SEM model with higher adaptability is optimized.Twelve modeling parameters,such as dilution vapor volume,fan speed,furnace cross-section temperature,fur-nace flue gas temperature,furnace negative pressure,and furnace oxygen content,are selected.These parame-ters provide a reasonable parameter set for the prediction model,effectively improving the model's predictive ac-curacy.The single-layer LSTM neural network structure was adopted,and the NOX prediction model for the nor-mal production stage of the ethylene cracking furnace was optimized based on MIC-SEM.The average R2 value between the predicted results and actual values was 0.908.The prediction accuracy of the LSTM neural network model was superior to that of the support vector machine(SVM)model,and the model accuracy was improved.关键词
NOX浓度预测/乙烯裂解炉/变量选择/最大互信息数/长短期记忆神经网络Key words
NOX concentration prediction/ethylene cracking furnace/variable selection/maximal informa-tion coefficient/long-short term memory network分类
资源环境引用本文复制引用
张子玥,杨文玉,张树才..基于MIC-LSTM的乙烯裂解炉氮氧化物浓度预测模型[J].安全、健康和环境,2025,25(9):26-33,8.基金项目
中国石化化工事业部课题(321142),数字孪生的智能乙烯工厂. （321142）
