基于改进的PCA-LS-SVM的软测量技术及应用

针对工业过程中某些重要过程变量难以实现实时在线检测和高维数据处理的问题,提出了将主元分析与改进的最小二乘支持向量机相结合的软测量建模方法,建立了催化裂化主分馏塔柴油凝固点的软测量模型。最小二乘支持向量机与标准支持向量机相比,失去了"稀疏性",最小二乘支持向量机的稀疏化方法解决了这一难题;主元分析方法的引入,有效地提高了最小二乘支持向量机软测量模型的精度和泛化能力。应用结果表明,该改进的PCA- LS-SVM方法具有学习速度快、跟踪性能好以及泛化能力强等优点。

Soft Sensor Modeling and Application Based on Improved PCA-LS-SVM

In order to solve the problem that it is difficult to measure some important process variables on lime at real times,and to handle da- ta with high dimensions,a method of soft sensor is proposed based on the integration of both PCA and improved LS-SVM.A soft sensor of fluid- ized catalytic cracking unit(FCCU)host fractionator's light diesel freezing point is set up.Compared with SVM,sparseness is lost in the LS- SVM.The method which can overcome this drawback is presented.The introduction of the method of PCA contributes to the distinct improve- ment of precision and generalization ability of the soft sensor model based on LS-SVM.The application results show that the proposed method features with high learning speed,good approximation and well generalization ability.