基于LSTM的多因素石灰窑煅烧带温度预测研究  

Study on Predicting Multi-factor Lime Kiln Calcining Zone Temperature Based on LSTM

作　　者：温后珍[1,2] 栾仪广 孟碧霞 卞庆舟[1] 陆建明 WEN Hou-zhen;LUAN Yi-guang;MENG Bi-xia;BIAN Qing-zhou;LU Jian-ming(School of Mechanical Science and Engineering,Northeast Petroleum University;Jiangxi Ruoshui New Material Technology Co.,Ltd.;Shandong Yulong Petrochemical Industrial Park Development Co.,Ltd.)

机构地区：[1]东北石油大学机械科学与工程学院 [2]江西若水新材料科技有限公司 [3]山东裕龙石化产业园发展有限公司

出　　处：《化工自动化及仪表》2024年第5期864-871,906,共9页Control and Instruments in Chemical Industry

基　　金：黑龙江省科学基金(批准号:QC2016003)资助的课题;东北石油大学青年科学基金(批准号:HBHZX202004)资助的课题。

摘　　要：针对石灰窑煅烧过程易出现燃烧不平衡的问题以及石灰窑煅烧系统的滞后性,提出了大数据分析+神经网络的解决方案。利用大数据分析对石灰窑多源历史数据进行数据融合插补,采用多元线性回归方程分析空间因素对温度的影响,通过时间滑窗提取特征,在此基础上利用长短期记忆神经网络(LSTM)算法构建多因素模型,并采用自适应运动估计算法进行优化。实验结果表明:较单因素LSTM模型,多因素LSTM模型有效提高了石灰窑温度预测精度,现场可根据预测值提前调整工艺参数,实现了石灰窑局部温度预测。Considering unbalanced combustion occurred in the lime kiln’s calcination process and the lag of lime kiln calcination system,a big data analysis+neural network solution was proposed,which has the big data analysis adopted to perform both data fusion and interpolation on lime kiln’s multi-source historical data,and the multiple linear regression equation adopted to analyze the spatial factors’influence on the combustion temperature,as well as has the features extracted through the time slide window.On this basis,the long short term memory neural network(LSTM)algorithm was employed to construct multi-factor model,including having the adaptive motion estimation algorithm adopted to optimize the model.The experimental results show that,compared with the single factor LSTM model,the multi-factor LSTM model can effectively improve temperature prediction accuracy of the lime kiln,including adjusting process parameters in advance according to the predicted value and realizing local temperature prediction there.

关 键 词：温度预测 长短期记忆神经网络 石灰窑 多元线性回归 多因素 自适应运动估计算法 

分 类 号：TH162[机械工程—机械制造及自动化] TQ115[化学工程—无机化工]