Paper presents a new power quality disturbance classification method based on atomic energy. Design four coherent dictionaries by analyzing characteristic of power quality disturbance, which decrease the calculation work of the Matching Pursuits (MP) algorithm. Use the MP algorithm to extract the fundamental component and disturbance component. By analyzing the relationship of energy which between disturbance component and residual signal, we can identify the disturbance types of power quality. Simulation and real data results confirm the effectiveness of the proposed method, and obtaining accuracy with high signal-to-noise ratios.