In order to distinguish the soft and hard shadow types of PV modules effectively, a new method of judging the shadow types based on power prediction is proposed. Firstly, the empirical mode decomposition (EMD) is taken to decompose the light intensity and the trend data is excavated to provide the application fundamentals for gray model (GM). Then, according the IAFSA-GM and rolling data update mode to predict the output power of PV module, the shadow fault types can be determined. Finally, On the basis of shadow type identification, a novel method based on the internal equivalent parameters model and measured data is proposed to further distinguish the severity of rigid shadows. The effectiveness and correctness of the proposed method is tested by the simulation and experimental results.