基于核慢特征回归与互信息的常压塔软测量建模
发布时间：2018-09-12 15:07
【摘要】：针对工业过程的非线性及动态特性,提出了一种新的慢特征回归软测量方法。该方法首先通过添加时延数据构造动态数据集,利用互信息最大化准则筛选变量从而减少信息冗余的影响。同时该方法在慢特征分析的基础上引入核函数扩展,加强模型处理非线性数据的能力,并将获得的核慢特征用于回归建模。核慢特征分析通过分析样本的变化,提取具有缓慢变化特征的成分,可以有效地刻画工业过程的变化趋势,提升回归模型精度。最后该方法的有效性在常压塔常顶油干点与常一线初馏点的软测量模型中得到了验证。
[Abstract]:Based on the nonlinear and dynamic characteristics of industrial processes, a new slow feature regression soft sensing method is proposed. In this method, the dynamic data set is constructed by adding delay data, and the influence of information redundancy is reduced by filtering variables by using mutual information maximization criterion. At the same time, the kernel function expansion is introduced on the basis of the slow feature analysis, and the ability of the model to deal with nonlinear data is enhanced, and the kernel slow feature obtained is used in regression modeling. Kernel slow feature analysis can effectively describe the changing trend of industrial process and improve the precision of regression model by analyzing the change of samples and extracting the components with slow changing characteristics. Finally, the effectiveness of the method is verified in the soft sensing model of the dry point and the initial distillation point of the constant top oil in the atmospheric tower.
【作者单位】： 华东理工大学化工过程先进控制与优化技术教育部重点实验室;中国石油天然气股份有限公司独山子石化研究院;
【基金】：国家自然科学基金项目(21676086,21406064)~~
【分类号】：O212.1;TE624.2
，
本文编号：2239412
[Abstract]:Based on the nonlinear and dynamic characteristics of industrial processes, a new slow feature regression soft sensing method is proposed. In this method, the dynamic data set is constructed by adding delay data, and the influence of information redundancy is reduced by filtering variables by using mutual information maximization criterion. At the same time, the kernel function expansion is introduced on the basis of the slow feature analysis, and the ability of the model to deal with nonlinear data is enhanced, and the kernel slow feature obtained is used in regression modeling. Kernel slow feature analysis can effectively describe the changing trend of industrial process and improve the precision of regression model by analyzing the change of samples and extracting the components with slow changing characteristics. Finally, the effectiveness of the method is verified in the soft sensing model of the dry point and the initial distillation point of the constant top oil in the atmospheric tower.
【作者单位】： 华东理工大学化工过程先进控制与优化技术教育部重点实验室;中国石油天然气股份有限公司独山子石化研究院;
【基金】：国家自然科学基金项目(21676086,21406064)~~
【分类号】：O212.1;TE624.2
，
本文编号：2239412
本文链接：https://www.wllwen.com/kejilunwen/yysx/2239412.html
