贝叶斯网络在高炉铁水硅含量预测中的应用

Application of Bayesian Network to Predicting Silicon Content in Hot Metal

应用贝叶斯网络对高炉铁水硅含量进行预测。首先阐述了贝叶斯网络的数学描述,在此基础上给出贝叶斯网络预测公式的一种简化形式。然后建立高炉铁水硅含量的贝叶斯网络预测模型,对山东莱钢1 号高炉在线采集的2 000炉数据进行网络学习,离线预测取得了较好的效果。与神经网络等其他方法相比,它更适合解析高炉过程,而且透明的推理过程对高炉工长判断炉温变化趋势具有指导意义。

A new approach to predict the silicon content in hot metal is based on Bayesian network. Firstly, a kind of abbreviated forecasting formula was proposed after the mathematical basis of Bayesian network had been depicted. Secondly, a Bayesian network model to predict the silicon content in molten iron was created, and the parameters of the model were estimated by processing real time data of No.1 BF in Laiwu Iron and Steel Group Co.. The Bayesian network prediction model has good results. Compared with BP network, Bayesian network is more fitting for BF ironmaking, and most importantly, the inference in Bayesian network is visible, which is of great value to judge how the hot metal temperature changes.