基于向量投影的数据检验PCA方法

DATA VALIDATION METHOD BASED ON PCA WITH VECTOR PROJECTION

提出了一种基于向量投影的数据检验PCA方法。该方法在对测量参数相关性分析的基础上,采用主成分分析进行建模,提取了其中的主要特征信息,实现了高维复杂数据的降维,通过数据重构可对各测量数据进行正确的估计。采用向量投影的方法对重构残差向量进行变换,通过得出的两类结构化残差可对故障仪表进行正确的检测与定位。以国产200MW机组为具体对象,给出了该方法的算例和检验结果。

ABSTRATCT: The using of principal component analysis (PCA) for data validation is expatiated and improved in this paper. Based on the correlation analysis, the PCA model of the sampling data is established to extract the main characteristic.By means of dimension reduction, the sampling data can be reconstructed from the principal component model of them. The reconstruction residual is further transformed by the method of vector projection. Two kind of structured reconstructionresiduals can be used to accurately detect and isolate the fault instrument. The examples of the domestic 200MW unit are given in this paper.