基于分布式神经网络模型的高炉炉温预测建模  

Predictive Modeling of Blast Furnace Temperature by Using Distributed Neural Network Model

作　　者：崔桂梅[1] 程史 

机构地区：[1]内蒙古科技大学信息工程学院,内蒙古包头014010

出　　处：《钢铁研究学报》2014年第6期27-30,共4页Journal of Iron and Steel Research

基　　金：国家自然科学基金资助项目(61164018);内蒙古自然科学基金资助项目(2012MS0911)

摘　　要：高炉炼铁通常采用铁水Si含量间接反映炉温的变化,模型预测精度低。以影响炉温的6个变量为输入变量,采用基于自组织的分布式RBF神经网络模型分别对铁水温度和铁水Si含量建立预测模型,先用自组织神经网络划分输入输出样本空间,然后对每个子空间建立RBF神经网络子网模型,再使用子网模型对测试样本集的同一个样本点进行预测,并以测试样本点对每一子空间的隶属度为权值,对子网预测值进行加权求和,得到最终预测值。对比使用同一输入变量数据的铁水温度和铁水Si含量的预测模型命中率,研究表明,高炉铁水温度的命中率更高,具有更好的炉温预测效果。Blast furnace iron making usually takes silicon content of molten iron to predict variety of furnace temperature.The predicted accuracy of this kind model is relatively lower.6variables infecting furnace temperature were taken as input variables.Distributed RBF neural network based on self-organizing was respectively used to establish predicted models of silicon content of molten iron and molten iron temperature.At first,RBF self-organizing neural network divided sample space of input and output variables,every sample space was used to establish RBF self-organizing subnet neural network,subnet neural network model was used to predict the same sample points of testing data set,test sample membership for each subspace as the weight,and then weighted sum of subnet predicted value was counted and get the final predicted value.Hit rate of the prediction model of outputs-silicon content of molten iron and molten iron temperature were compared.The results show that hit rate of the molten iron temperature model is higher,so the predicted effect of furnace temperature is better.

关 键 词：铁水温度预测 铁水Si含量预测 分布式建模 自组织神经网络 RBF神经网络 

分 类 号：TF53[冶金工程—钢铁冶金] TP183[自动化与计算机技术—控制理论与控制工程]