Based on the angle and Euclidean distance similarity of the loading matrix, this paper reports on an improved principal component analysis (PCA) modeling method that is successfully applied to identify three different working conditions in the secondary loop of a nuclear power plant (NPP). First, a simulation platform of the secondary loop of the NPP is built in which three kinds of working conditions are set. Second, the multiple-PCA modeling method is used to construct the offline models. Finally, the effectiveness and superiority of the improved method is verified in the simulation platform.