A new method to identify the transient power quality disturbance based on FCM is proposed. This recognition method is implemented hierarchically. The first layer can judge whether transient oscillation disturbance is included in the signal. The second layer judges whether transient oscillation pulse is contained in the signal. The third layer judges whether the signal contains the magnitude of the disturbance and has a comprehensive judgment on the specific type of complex disturbances. Through combination with the ensemble empirical mode decomposition (EEMD) and singular value decomposition method, effective feature vectors can be extracted hierarchically, which is used as the input of FCM. In this way, the optimized classified matrix and clustering centers are obtained. Calculating the Euclidean distance between the unknown-sample samples and the known-sample ones, the disturbance type is identified. The simulation result indicates that this method is accurate and feasible.