Abstract: Phasor Measurement Units (PMUs) make it possible to monitor the dynamic behavior of power system in real time because of its synchronization and rapidity. However, due to the complex factors on site, the PMU data can be easily compromised by interferences or hardware malfunctions, resulting in different levels of PMU data loss, which directly affects its application in the power system, and even threatens the safe operation of the system. In order to improve the data quality, a PMU missing data recovery method based on PMU clustering is proposed in this paper. Firstly, a PMU clustering method based on the hierarchical clustering is proposed by analyzing the correlation of different PMUs. Then, an enhanced generative adversarial network data recovery method is constructed by using the long/short-term memory. The proposed method is able to recover the lost data under different disturbances, even under transient conditions, by using the high correlated data as the input of the neural network. The effectiveness of the method is verified by simulation and field data. The results show that the method enables to effectively recover the lost data, improving the PMU data quality to guarantee its applications in the power systems.