高炉铁水硅含量变动量调控决策的智能推荐模型及应用  

Intelligent recommendation model for reducing silicon deviation fluctuation of hot metal in BF and application

作　　者：韩阳 胡支滨 杨爱民 李杰 张玉柱 HAN Yang;HU Zhi-bin;YANG Ai-min;LI Jie;ZHANG Yu-zhu(College of Metallurgy and Energy,North China University of Science and Technology,Tangshan 063210,Hebei,China;College of Science,North China University of Science and Technology,Tangshan 063210,Hebei,China)

机构地区：[1]华北理工大学冶金与能源学院,河北唐山063210 [2]华北理工大学理学院,河北唐山063210

出　　处：《钢铁》2023年第4期30-38,共9页Iron and Steel

基　　金：国家自然科学基金资助项目(52074126,51974131);唐山市市级科技计划资助项目(22130201G)。

摘　　要：为了保证高炉炉况顺行,提高高炉热制度调控质量,深刻理解铁水硅含量变动量的动力学,以冀南地区某钢铁联合公司3号高炉为研究对象,研究铁水硅含量变动量的波动机理。量化表征降低铁水硅含量变动量的6大决策空间,通过单变量估计方法和复相关系数联合估计方法的对照分析,优选可以兼顾决策变量间耦合协同性的联合估计方法,确定铁水硅含量变动量调控决策变量的响应时间,消除变量间的时滞性;通过Hopfield神经网络、Boltzmann神经网络与Elman神经网络算法的对照分析,优选具有记忆功能的Elman神经网络算法,兼顾变量时序特征,确定变量间的耦合非线性关系,构建高炉铁水硅含量变动量智能预报模型。基于铁水硅含量变动量最小为目标,构建调控决策智能推荐模型;通过遗传算法(GA)和进化策略算法(ES)的对照分析,优选变异程度自适应变化的ES算法,智能求解推荐模型,快速获取最优的维持铁水硅含量变动量最小的调控决策。研究结果表明,响应时间估计算法对照验证了多决策变量联合估计响应时间,产生的样本集兼顾了决策变量间的耦合协同性;预报算法对照验证了Elman神经网络的优越性,目标算法构建的预报模型命中率达到94.10%;智能寻优算法对照验证了ES算法对推荐模型的求解速度和结果显著优于对照算法,并且模型在工业应用的实践中保持了离线测试的优良特性。In order to ensure the smooth operation of the blast furnace,improve the quality of blast furnace thermal regulation,and deeply understand the dynamics of molten iron silicon content variation,the fluctuation mechanism of molten iron silicon content variation was studied with the No.3 blast furnace of an iron and steel complex in southern Hebei as the research object.Quantify the six decision spaces for decreasing the variation of silicon content in molten iron.Through the comparative analysis of single variable estimation method and complex correlation coefficient joint estimation method,the joint estimation method that can give consideration to the coupling and synergism between decision variables is selected to determine the response time of the variable momentum control decision variables of molten iron silicon content and eliminate the time lag between variables.Through the comparative analysis of Hopfield neural network,Boltzmann neural network and Elman neural network algorithm,the Elman neural network algorithm with memory function is selected,and the coupling nonlinear relationship between variables is determined by taking into account the time series characteristics of variables,so as to build an intelligent prediction model for the variation of molten iron silicon content in blast furnace.Based on the goal of minimizing the variation of silicon content in molten iron,an intelligent recommendation model for regulation and decision making is constructed;Through the comparative analysis of genetic algorithm(GA)and evolutionary strategy algorithm(ES),the ES algorithm with adaptive variation degree is selected,the recommendation model is solved intelligently,and the optimal control decision to maintain the minimum variation of molten iron silicon content is quickly obtained.The research results show that the response time estimation algorithm verifies that the sample set generated by joint estimation of response time of multiple decision variables takes into account the coupling and cooperation between deci

关 键 词：铁水硅含量变动量 调控决策 响应时间 反馈网络 进化策略 

分 类 号：TF53[冶金工程—钢铁冶金]