基于相对主元分析的动态数据窗口故障检测方法

Dynamic Data Window Fault Detection Method Based on Relative Principal Component Analysis

传统主元分析方法量纲标准化后常常无法进行有效的主元提取。基于相对主元分析故障检测方法,其控制限只与保留的主元个数和置信度有关,因而不能满足时变、非稳定的过程监控要求。针对这些问题,本文提出了基于相对主元分析的动态数据窗口故障检测方法,该方法通过引入一个权值将传统构建控制限的方法和动态数据窗口相结合。最后应用到风力发电系统故障检测中,有效地检测出故障,并减少了误报情况。

In traditional principal component analysis(PCA),because of the neglect of the influence of dimension standardization,it was difficult to extract principal components(PCs) effectively.The fault detection method based on relative principal component analysis(RPCA),its control limit is related to the number of PCs and confidence.For these problems,a dynamic data window method based on RPCA is proposed in this paper.The proposed method combined the traditional control limit and dynamic data window by introducing a weight.Finally,it is applied to wind power generation system,can detect failures effectively and reduce the rate of false alarm.