主成分分析结合极限学习机的高炉炉温预测模型
|
摘要
炉温控制是高炉过程控制的基础与核心技术,炉温的走势最直接地反应了高炉的运行状况.因此,建立合理的炉温控制模型至关重要.铁水中硅的含量与炉温成正比例关系,而冶炼过程中状态变量、控制变量及入炉基本条件等都会影响炉温,如果考虑全部的相关因素,势必会因信息冗余降低模型的性能.为此,首先采用主成分分析(PCA)方法对多维输入变量进行降维,同时回避了变量间的多重共线性问题.其次,将PCA处理得到的相互独立的主成分用于网络训练,建立了基于极限学习机(ELM)的炉温预测模型,该模型克服了前馈神经网络训练速度慢、容易陷入局部极小的缺点.最后比较了传统的BP学习算法、ELM算法和PCA结合ELM算法的预测效率,试验证明本文算法具有较高的命中率,可以用来指导高炉实际生产.
Furnace temperature control is the basis and core technology of blast furnace process control,and the trend of the furnace temperature is the most direct response to the operation of the high furnace. Therefore,it is very important to establish a reasonable furnace temperature control model. The content of silicon in molten iron is proportional to the furnace temperature,while the state variables,control variables and the basic conditions in the process of smelting will affect the furnace temperature. If all the relevant factors are considered,it is bound to reduce the performance of the model due to information redundancy. For this reason,firstly,principal component analysis( PCA) method was applied to reduce the dimension of the multidimensional input variables to avoid the multiple collinearity between variables simultaneously. Secondly,the independent principal components of PCA processing were used for network training,and a furnace temperature prediction model based on extreme learning machine( ELM) was established. The model overcomes the shortcomings of slow training speed and easy falling into the local minimum of the feedforward neural network. Finally,the prediction efficiency of the traditional BP algorithm,ELM algorithm and PCA combined with the ELM algorithm were compared. Experiments show that this algorithm has a high hit rate,which can be used to guide the actual production of blast furnace.
Furnace temperature control is the basis and core technology of blast furnace process control,and the trend of the furnace temperature is the most direct response to the operation of the high furnace. Therefore,it is very important to establish a reasonable furnace temperature control model. The content of silicon in molten iron is proportional to the furnace temperature,while the state variables,control variables and the basic conditions in the process of smelting will affect the furnace temperature. If all the relevant factors are considered,it is bound to reduce the performance of the model due to information redundancy. For this reason,firstly,principal component analysis( PCA) method was applied to reduce the dimension of the multidimensional input variables to avoid the multiple collinearity between variables simultaneously. Secondly,the independent principal components of PCA processing were used for network training,and a furnace temperature prediction model based on extreme learning machine( ELM) was established. The model overcomes the shortcomings of slow training speed and easy falling into the local minimum of the feedforward neural network. Finally,the prediction efficiency of the traditional BP algorithm,ELM algorithm and PCA combined with the ELM algorithm were compared. Experiments show that this algorithm has a high hit rate,which can be used to guide the actual production of blast furnace.
引文
[1]Gao H,Chen H,Zeng J S,Liu Y.A chaos-based iterated multi step predictor for blast furnace iron making process,AICHE[J],2009,55(4):947-962.
[2]王馨.基于炉热指数和改进支持向量机的高炉炉温预测研究[D],衡阳:华南大学,2011.
[3]石琳,汤佳佳,于涛,等.基于样条变换的非线性PLS的反应高炉炉温的参数预测[J].钢铁研究学报,2013,25(2):21-25.
[4]Shi L,Li Z L,Yu T,et al.Model of Hot Metal Silicon Content in Blast Furnace Based on Principal Component Analysis Application and Partial Least Square[J].Journal of Iron and Steel research International,2011,18(10):13-16.
[5]袁冬芳,赵丽,石琳,等.高炉铁水硅含量的支持向量机预测模型[J].太原理工大学学报,2014,45(5):684-688.
[6]周驰化.模糊专家系统在高炉炉温预测中的研究与应用[D].西南石油大学,2007.
[7]曹长修,张国英.基于炉热指数和BP网络的高炉铁水珪含量预报系统[D],重庆:重庆大学,2008.
[8]Chen J.A predictive system for blast furnaces by in-tegrating a neural network with qualitative analysis[J].Engineering Applications of Artificial Intelligence,2001,14(1):77-85.
[9]邱东,仝彩霞,祁晓钰,等.基于神经网络的高炉铁水硅含量预报模型的研究[J].冶金分析.2009,29(2):49-52.
[10]崔翔鹏,黄洪琼.基于GA优化ELM的船舶交通流预测模型[J].微型机与应用,2017,36(9):15-17,21.
[11]徐琴.混沌理论和极限学习机的物流需求预测模型[J].现代电子技术,2017,40(7):151-154.
[12]樊超,郭亚菲,曹培格,等.基于主成分分析的粮食产量极限学习机预测模型研究[J].2017,42(2):1-5.
[13]Deng C W,Huang G B,Xu J,et al.Extreme learning machines:new trends and applications[J].Science China Information Science,2015,58(2):1-16.
[2]王馨.基于炉热指数和改进支持向量机的高炉炉温预测研究[D],衡阳:华南大学,2011.
[3]石琳,汤佳佳,于涛,等.基于样条变换的非线性PLS的反应高炉炉温的参数预测[J].钢铁研究学报,2013,25(2):21-25.
[4]Shi L,Li Z L,Yu T,et al.Model of Hot Metal Silicon Content in Blast Furnace Based on Principal Component Analysis Application and Partial Least Square[J].Journal of Iron and Steel research International,2011,18(10):13-16.
[5]袁冬芳,赵丽,石琳,等.高炉铁水硅含量的支持向量机预测模型[J].太原理工大学学报,2014,45(5):684-688.
[6]周驰化.模糊专家系统在高炉炉温预测中的研究与应用[D].西南石油大学,2007.
[7]曹长修,张国英.基于炉热指数和BP网络的高炉铁水珪含量预报系统[D],重庆:重庆大学,2008.
[8]Chen J.A predictive system for blast furnaces by in-tegrating a neural network with qualitative analysis[J].Engineering Applications of Artificial Intelligence,2001,14(1):77-85.
[9]邱东,仝彩霞,祁晓钰,等.基于神经网络的高炉铁水硅含量预报模型的研究[J].冶金分析.2009,29(2):49-52.
[10]崔翔鹏,黄洪琼.基于GA优化ELM的船舶交通流预测模型[J].微型机与应用,2017,36(9):15-17,21.
[11]徐琴.混沌理论和极限学习机的物流需求预测模型[J].现代电子技术,2017,40(7):151-154.
[12]樊超,郭亚菲,曹培格,等.基于主成分分析的粮食产量极限学习机预测模型研究[J].2017,42(2):1-5.
[13]Deng C W,Huang G B,Xu J,et al.Extreme learning machines:new trends and applications[J].Science China Information Science,2015,58(2):1-16.