Dynamic PLS and Dynamic PCR Modelling for Industry Processes

摘要：

The large amount of input and output data obtained from the industry process is not independent to each other in most cases,and even in time serials.In this paper,we present the dynamic Partial Least Squares (PLS) and dynamic Principle Component regression (PCR) modelling for this situation.PLS and PCA are dimensionality reduction modelling techniques.They capture most of the data variation of the industry process,and project the system into a reduced space.More accurate models can be obtained because the surveying noise is ignored in dynamic PLS and PCR modelling.Considering MIMO systems,the direct dynamic PLS and PCR equations are proposed,which can be used in control systems straightforward.The predictive accurate of the two models is compared with ordinary MLR model through an industry modelling example.The validity of dynamic PLS and dynamic PCR modelling is proved.Simulation experiments indicate that,in general,dynamic PLS model performs better than dynamic PCR model.

关键词： partial least square (PLS) MIMO systems score vectors principal component regression (PCR) multiple linear regression (MLR).

作者: Xue-Lian Zhang Guang-Yi Cao

作者单位: Department of Automation,Shanghai Jiao Tong University,Shanghai,China

会议类型: 国际会议

会议名称: International Conference on Modelling,Identification and Control(模拟、鉴定、控制国际会议)

会议地点: 上海

会议语种:英文

在线出版日期: 2008-06-29（万方平台首次上网日期，不代表论文的发表时间）