基于神经网络的高炉铁水硅含量预报模型的研究

Study on prediction model of silicon content in blast furnace liquid iron based on neural network

根据RBF神经网络具有收敛速度快和全局优化的特点,建立了RBF网络模型,并将其应用对高炉铁水硅含量预报。监于铁水硅含量与炉缸温度之间的密切相关性,通过铁水硅含量来间接地反映炉内温度变化。采用MATLAB中的Newrbe函数进行函数逼近,对高炉一段连续时期内正常生产的数据的归一化处理后进行训练和仿真,提高了铁水硅含量预报的命中率。高炉冶炼运用先进的RBF人工神经网络预报模型,能预报铁水硅含量的高低,判断炉温走势,实现炉温调控,有利于节能降耗,并可监测多个主要控制对象,为高炉操作提供指导。

On the basis of the characteristics of Radial Basis Function（RBF） neural network including rapid convergence rate and global optimization, a RBF neural network was developed and employed in the prediction of silicon content in blast furnace liquid iron. Because the correlation between silicon content in liquid iron and hearth temperature was close, the temperature change in furnace could be reflected indirectly by the content of silicon in liquid iron. Newrbe in MATLAB was used for function approximation, and then the data of normal production during a continuous period in blast furnace were normalized for training and simulation. The results showed that the hit rate of prediction for the content of silicon in liquid iron was improved. The use of advanced RBF artificial neural network pre- diction model in blast furnace could predict the content of silicon in liquid iron to estimate the trend of hearth temperature and realize temperature control, which was advantageous to energy saving and consumption reducing. Moreover, several major control objects could be monitored, providing the guide for blast-furnace operation.