According to the demands of using power quality disturbance for localization, recognition and classification, this paper proposes a new method of using power quality disturbance for localization and classification based on HHT. Firstly, power quality disturbance signals are transformed by using the HHT algorithm to get instantaneous amplitudes, Hilbert spectra and marginal spectra. Secondly, disturbance signals can be located by using instantaneous frequency to record the beginning time and the ending time of disturbance. Thirdly, characteristic parameters can be extracted from instantaneous amplitude, Hilbert spectrum and marginal spectrum, serving as evidence for using the decision-making tree for recognition and classification. The simulation results show that the margin of error can be reduced by using Hilbert spectrum for localization. By combining the HHT algorithm and the decision-making tree, no extra training is needed in using power quality disturbance for recognition and classification. Other advantages include extracting effective but fewer characteristic parameters, more desirable effects of recognition and classification, as well as its favorable noise-proof capability.