Dynamic degradation identification of solid oxide fuel cell systems based on automatic spectral clustering and neighborhood rough sets

To address the problem that some of the degradation phenomenon states of solid oxide fuel cell systems under dynamics are difficult to be observed directly from the data, this study proposes a dynamic degradation identification method for SOFC systems based on automatic spectral clustering and neighborhood rough sets. First, this method identifies all the health states exhibited by the SOFC system from the dynamic data by using an automatic spectral clustering-based algorithm to determine whether degradation phenomena have occurred. Then, this study extracts the feature variables that lead to changes in the dynamic performance of the SOFC system using a feature extraction method based on neighborhood rough sets, which helps analyze where the anomalies that lead to degradation of the performance may occur and provides a theory to trace the causes of the performance degradation.