多变量动态模糊偏最小二乘建模方法及其应用

Multivariable Modeling Based on Dynamic Fuzzy PLS and Its Application

针对复杂工业过程存在的多变量、非线性和时变不确定性问题,将动态PLS与模糊建模方法相结合,提出一种基于DFPLS(动态模糊偏最小二乘)的多变量非线性动态建模方法。该方法外部采用动态PLS方法解决多变量高维共线性问题,并描述系统的动态特性;内部采用FCM(模糊c均值聚类)与TSK模糊模型相结合,建立多个子模型的方法来拟合系统的非线性。将本方法应用于氧化铝生产过程中铝酸钠溶液组分浓度的软测量,仿真实验表明该方法预测精度高,泛化能力强,用于铝酸钠溶液组分浓度的在线检测是可行有效的。

A new modeling method of multivariable nonlinear dynamic system based on DFPLS was proposed according to the complex industry process with multivariable,nonlinear and time-varying uncertainties.Outer dynamic PLS algorithm was used to reduce the dimensionality of data and to remove collinearity,and it also described the dynamic character.The inner fuzzy model which was combined with FCM and several TSK fuzzy models was used to capture the nonlinearity.This method was implemented in the process of alumina production to measure the component concentrations of sodium aluminate solution and the simulation results show the modeling method has the merits such as accurate prediction and good generalization ability.