The integrity of the data gathered by the power grid platform is very important to promote the construction of the intelligent management and control platform of the power grid. However, in reality, it is almost impossible to obtain complete data because of the damage of equipment, the failure of data acquisition process, the loss of records and other reasons. In this paper, the structure of the traditional GAN is improved, and a lost data repair model of the power grid platform based on the generative adversarial network is proposed, which provides additional information for the discriminator (D) in the form of "hints" to ensure that the generator (G) generates samples according to the true underlying data distribution. Experiments show that the model can run in an unsupervised environment even the data is incomplete, and can better repair the data of the power grid platform.