Power quality disturbances identification is the important procedure for improving the power quality, and online application has actual value. An efficient method for power quality disturbances identification is presented in this paper. Wavelet decomposition is used for extracting the features of various disturbances, and decision tree in data mining is used for identifying the disturbances. For online application, sliding window model and one-pass scan algorithms for wavelet decompositions are used. This method has low cost in memory and run time, it can identify different disturbances in high accuracy and less time. Simulation experiment using several typical disturbances, swell, sag, interrupt, harmonic, transient impulse, transient oscillation, show the effectiveness of proposed method.