摘要
焦炉加热过程具有非线性、强耦合、大滞后等特点。以包钢6号焦炉为背景,火道温度难于在线实时、准确测量,致使控制效果不太理想。为了提高焦炉立火道温度的预测精度,提出了基于深度学习长短时记忆(Long Short-Term Memory,LSTM)网络的焦炉立火道温度预测模型。首先对现场采集的海量工况数据进行预处理和特征提取,然后利用相关分析法归纳出影响炉温的关键变量,最后建立基于时间序列的焦炉立火道温度LSTM预测模型。结果表明:LSTM预测模型与传统BP网络算法相比,预测精度更高、误差更小,可为焦炉的优化、控制奠定良好的基础,以保证焦炭质量、降低能耗并提高产量。
The heating process of coke oven has nonlinear feature, strong coupling feature and large lag. It is difficult to measure the temperature of flue on line and in real time, resulting in poor control effect. In order to improve the prediction accuracy of temperature, a model of predicting the coke oven flue temperature based on depth learning Long Short-Term Memory(LSTM) was proposed. First of all, massive working data collected in the field was preprocessed to extract the feature. And then, the related analysis methods were used to conclude key variables influencing the furnace temperature. Finally, the LSTM prediction model of temperature of coke oven flue based on time series was established. The results prove that the LSTM prediction model has higher prediction accuracy and lower error than that of traditional BP neural network algorithm, so this model can lay a good foundation for optimization and control for the coke oven. Meanwhile, coke quality can be guaranteed and the energy consumption can be reduced and the output can be increased.
作者
李爱莲
张帅
LI Ai-lian;ZHANG Shuai(Information Engineering Institute,Inner Mongolia University of Science and Technology,Baotou Inner Mongolia 014010,China)
出处
《计算机仿真》
北大核心
2020年第6期466-470,共5页
Computer Simulation
基金
内蒙古自治区自然科学基金资助(2016MS0610)
内蒙古科技大学产学研合作培育基金项目(PY-201512)。
关键词
焦炉
火道温度预测
深度学习
长短时记忆
Coke oven
Flue temperature prediction
Deep learning
Long short-term memory
