FVS-KELM方法在化工过程软测量建模中的应用

Application of FVS-KELM method in soft measuringmodeling of chemical process

针对强非线性、复杂的化工过程软测量建模问题,提出一种基于特征向量选取(FVS)与核极限学习机(KELM)结合的方法。通过核方法将输入数据映射在高维空间,选出一组最大无关的数据向量作为特征空间,并将原始数据投影在该特征空间上,形成新的输入数据,基于KELM建立软测量模型,KELM无须设定网络隐含层节点的数目,以核函数表示未知的隐含层非线性特征映射,具有良好的泛化能力。为验证所提出方法的有效性,将该方法应用于脱丁烷塔的软测量实例建模中。在同等条件下与KELM,KPCA-KELM,FVS-SVM和FVS-LR等方法进行比较,结果表明,FVS-KELM方法具有最高的相关系数和建模精度,在化工生产中有优势。

Aiming at problem of soft measuring modeling of complex chemical process with strong nonlinear feature,a method based on kernel extreme learning machine(KELM)and feature vector selection(FVS)is proposed.The input data is mapped to the high-dimensional space by kernel method,and a group of the largest unrelated data vector is selected as the feature space,and the original data is projected on the feature space to form new input data,and the soft measuring model is established based on KELM.KELM does not need to set the number of network hidden layer nodes,and the kernel function represents the unknown hidden layer nonlinear feature mapping,which has good generalization ability.In order to verify the effectiveness of the proposed method,it is applied to the soft measuring modeling of debutane tower.Under the same conditions,it is compared with the methods of KELM,KPCA-KELM,FVS-SVM,FVS-LR and other methods,the experimental results show that the FVS-KELM method has the highest correlation coefficient and modeling precision,and has advantages in chemical production.