摘要
针对高炉煤气发生量波动对煤气调度和优化影响的问题,提出一种基于改进BP网络的高炉煤气发生量预测模型.由于影响高炉煤气发生量的因素很多,采用机理和现场实际数据相结合方法分析影响高炉煤气发生量的因素,确定模型输入,采用炼铁高炉现场数据对模型进行训练,利用贝叶斯正则化算法来提高神经网络泛化能力.仿真结果表明,该模型能够准确预测高炉煤气发生量的变化趋势,为制定煤气管网平衡调度策略提供科学的依据和决策支持,有利于减少煤气排放,提高煤气利用率和企业的信息化水平.该预测模型已经成功应用到杭州钢铁集团煤气调度系统中,运行结果验证模型的有效性.
To the problem of fluctuations in the amount of blast furnace gas which affects the gas scheduling and gas balance,the improved back propagation(BP) network was used to build the forecasting model of blast furnace gas output.Since there were a lot of factors affecting the output of blast furnace gas,the input of the model was determined through understanding the operation mechanism of blast furnace gas and the correlation analysis of practical production data.Practical production data of blast furnace was also used to train the model.The Bayesian regularization method was used for improving the generalization ability of artificial neural network.Simulation shows that this model has high accuracy in forecasting the trends of the output of blast furnace gas,to provide the scientific basis and decision support for the gas resources balance and scheduling.As a result,it is useful to reduce the leaking gas and improve the utilization ratio of gas and the information-based degree of the whole enterprise as well.This forecasting model has been successfully applied in the gas scheduling system of Hangzhou Iron Steel Group Company(HISGC) of China and the running results demonstrate the validity of the proposed method.
出处
《浙江大学学报(工学版)》
EI
CAS
CSCD
北大核心
2012年第11期2103-2108,共6页
Journal of Zhejiang University:Engineering Science
基金
国家"863"高技术研究发展计划资助项目(2006AA04Z184)
关键词
高炉煤气发生量
BP网络
预测模型
贝叶斯正则化
泛化能力
output of blast furnace gas
BP network
forecasting model
Bayesian regularization
generalization ability