【作者】 冯婷；

【导师】 刘祥官；

【作者基本信息】 浙江大学 ， 运筹学与控制论， 2008， 硕士

【摘要】 高炉炼铁是钢铁工业的上游主体工序,作为国民经济支柱产业的重要组成部分,对钢铁工业的发展与节能降耗都有十分重要的作用。高炉冶炼过程是一个高度复杂的过程,其运行机制往往具有非线性、时滞、高维、大噪声、分布参数等特性,导致很难建立起准确有效的高炉炉温预测控制模型。非参数回归是非参数统计理论中的重要组成部分,在计量经济、交通、医学等领域得到了广泛应用。非参数回归中,回归函数形式的任意性和自变量与因变量分布的少限制,很好地解决了经典统计理论中模型及参数的假定与实际背离造成模型设定误差的问题,使得模型能更加准确地反映实际问题的变化情况。本文选取《包钢6#高炉(2500m~3)冶炼专家系统》在线采集的数据,首先对铁水含硅量[Si]的自相关性进行分析,证明了铁水含硅量[Si]序列存在较强的线性自相关。然后通过相关系数和灰关联熵的计算,综合分析了所选取的高炉冶炼过程中的19个参数与高炉铁水含硅量[Si]之间的关联度。本文第4章利用偏最小二乘回归方法,对参数进行综合降维,最大可能地提取参数中与铁水含硅量[Si]变化相关的信息,减少参数中夹杂的冗余信息,从而使综合变量能充分反映铁水含硅量[Si]的变化。在此基础上,对得到的三个综合变量和铁水含硅量[Si]建立广义加性(GAM)模型,通过非参数光滑函数的迭代得到它们的局部近似函数关系。在探求综合变量与铁水含硅量[Si]局部关系的基础上,第5章通过遗传算法的全局搜索和非参数回归中正交序列估计方法,找到了最能表征铁水含硅量[Si]变化的参数组合,用数据事实证明了之前关联度分析结论的正确性,并建立了最优的高炉炉温预测的非参数回归模型。第6章中,将非参数回归与高炉冶炼的混合控制偏微分方程结合,得到了炉温预测控制的变系数回归模型,分析了料速LS、风量FQ、喷煤PM和透气性FF四个参数与铁水含硅量[Si]的局部线性关系,用权重描述了当前炉各个参数对铁水含硅量[Si]影响的大小和方向,为炉温预测之后的控制奠定理论基础。更多还原

【Abstract】 As the main upper procedure of metallurgical industry,Blast Furnace(BF) ironmaking is an important component of steel industry in national economy,which plays a significant role in energy saving and technical development of the whole industry.The ironmaking process is highly complicated,whose operating mechanism is characteristic of nonlinearity,time lag,high dimension,big noise and distribution parameters etc,thus makes it difficult to model the process accurately and effectively.Nonparametric regression is an important part of Nonparametric Statistical Theory.It is widely used in econometrics,traffic system and clinical statistics etc.In nonparametfic regression,the form of regression function is discretional,and there is little restriction on the form of regression function and the distribution of independent and dependent variables,which well accommodates the problem of deviation between model assumptions and real data.The current work uses data collected from BF No.6(2500 m~3) in Baotou Iron & Steel Group Co.to identify the model.The autocorrelation of[Si]series was analyzed and strong correlation was detected.Correlation coefficients and gray relation entropy between monitored process variables and silicon content in hot metal were also discussed.Section 4 deals with the problem of dimension reduction of model parameters based on partial least squares(PLS).By performing PLS redundancy is reduced and the most useful information in input variables is extracted to reflect the fluctuation of silicon content.A generalized additive model(GAM),which gets the local approximated function relation via an iterative process of nonparametric smooth function,was constructed using the three variables selected from PLS to predict the silicon content.On the basis of above analysis,section 5 uses genetic algorithms and orthogonal sequence estimation method to find the best parameters combination to indicate the fluctuation of silicon content.Simulation results prove the correctness of the previous analysis on relations between process variables and silicon content.A optimal nonparametric regression model for prediction of silicon content was constructed and good result was obtained.By Combined the nonparametric regression and the hybrid control partial differential function of BF ironmaking,a varying-coefficient regression model for predictive control of blast furnace hot metal temperature was given in section 6.It analyses the local linear connection between parameters like speed of materials LS, wind blasted FQ,coal injected PM and the permeability index FF and the output variable silicon content.A weight matrix is used to describe the influence that each parameter to silicon content in hot metal.Thus a theoretic basis of predictive control is established.更多还原