A novel KICA–PCA fault detection model for condition process of hydroelectric generating unit

Abstract

Fault detection and diagnosis of hydroelectric generating unit (HGU) have significant importance to the security of hydropower plant and the power system. In recent years, many fault detection methods based on spectral characteristic of vibration signals have been published. However, some faults cannot be effectively recognized just with spectral features for condition process monitoring of HGU. Thus, this study presents a novel fault detection model based on kernel independent component analysis and principal component analysis (KICA–PCA) monitoring model for condition process of HGU. Each of the condition processes is equivalent to a multivariate statistical process monitoring (MSPM). KICA–PCA model of the specific MSPM is trained by normal condition process data at first. Then, confidence limits of two monitoring indices (Hotelling’s T2 statistic and SPE statistic) of the trained KICA–PCA model are used to monitor the same condition process and detect fault online. Moreover, the proposed monitoring model is applied to a real condition process of HGU. Compared to ICA–PCA and PCA monitoring model, the proposed model has superior performance in fault detection.