基于神经网络预测模型的高炉喷煤操作模式优化

针对高炉冶炼过程中喷煤量通常根据炉长经验总结而出、过程建模困难、生产过程工艺指标波动较大的问题,提出了一种喷煤量操作模式匹配的智能优化方法。该方法以高炉喷煤量为研究对象,以某钢厂高炉大量的实际工业运行数据为基础,首先对不可实时检测变量[Si]含量建立了BP神经网络预测模型,以对当前[Si]含量进行预测;然后基于专家标准挑选出优良模式集,采用模糊C均值聚类的方法对优良模式集进行分类;最后以综合工艺指标评判为前提,对当前输入条件采用基于相似度的方法,从优良操作模式库中寻出与当前输入条件距离最小的操作模式,从而完成模式匹配,进而实现喷煤过程优化控制,对高炉冶炼喷煤过程具有很好的指导作用。

Operational-Pattern Optimization in Blast Furnace PCI Based on Prediction Model of Neural Network

There are many problems in the blast furnace smelting process, such as the difficulty of processing modeling, the large fluctuation of production process indicators, and the coal injection quantity is always according to furnace foreman′s experience. An intelligent optimization method is proposed for the operation pattern matching of the coal injection quantity. This method takes the coal injection quantity of the blast furnace as the object of the research, and is based on a lot of practical industrial operation data of a steel mill. First, BP neural network prediction model is established to forecast the current ［Si］ content, which is not a real-time monitoring variable. After that, the excellent model set is selected based on expert standards, the fuzzy C-means clustering method is used to classify the excellent model set. At last, the mode of operation with the current distance minimum input conditions is searched out from the excellent model library, which is on the premise of comprehensive process indicator to judge and based on similarity method for the current input condition. The selected mode of operation can realize pulverized coal injection process optimization control by completing the pattern matching, and this has a very good guidance on coal injection in the blast furnace smelting process.