基于深度神经网络的热风炉烟温预测模型  

PREDICTION MODEL OF FLUE GAS TEMPERATURE OF HOT BLAST STOVE BASED ON DEEP NEURAL NETWORK

作　　者：王一男 刘晓志[1] 孙超 王伟兵 李仁华 杨英华[1] Wang Yinan;Liu Xiaozhi;Sun Chao;Wang Weibing;Li Renhua;Yang Yinghua(College of Information Science and Engineering, Northeastern University, Shenyang, Liaoning, 110819;HBIS Group Research Institute, Shijiazhuang, Hebei, 050023;Automation Department, HBIS Group Hansteel Company, Handan, Hebei, 056015)

机构地区：[1]东北大学信息科学与工程学院,辽宁沈阳110819 [2]河钢集团钢研总院,河北石家庄050023 [3]河钢集团邯钢公司信息自动化部,河北邯郸056015

出　　处：《河北冶金》2021年第1期34-36,50,共4页Hebei Metallurgy

摘　　要：数字化工厂是智能制造技术中的重要环节,对被控对象精准建模是对工艺设计、自动化及智能化系统实现精准控制的重要支撑。针对热风炉燃烧状况复杂,很难建立合适的机理模型的实际情况,提出了基于长短期记忆深度网络(LSTM)对热风炉废气温度进行建模,并用L2正则化的方式对网络进行了优化。经河钢邯钢8#高炉实际数据仿真验证,预测模型的动态性能良好。Digital factory is an important part of intelligent manufacturing technology.Accurate modeling of controlled object is an important support for accurate control of process design,automation and intelligent system.In view of the complicated combustion conditions of hot blast stoves,it is difficult to establish a suitable mechanism model.Then a long-term short-term memory depth network(LSTM)was used to model the hot blast stove exhaust gas temperature,and the network was optimized by L2 regularization.The simulation results of 8#BF show that the dynamic performance of the prediction model is good.

关 键 词：热风炉 长短期记忆网络 废气温度 建模 

分 类 号：TF578[冶金工程—钢铁冶金]