A Model for Completing The Transformer-user Linked Data Based on Variational Autoencoder and Generative Adversarial Networks

Complete data is important for the stable operation of the power system, however, in the collection process, due to various reasons, the phenomenon of missing data often occurs.In order to make up for the missing data, this paper proposes a Missing data Imputation by Grouped Variational Autoencoder Generative Adversarial Network(MI-GVAEGAN).The model employs the idea of variational inference on incomplete datasets to generate estimates of missing data, and uses discriminator adversarial training generative network to complete the missing data. Finally, it is compared with commonly used missing complementation algorithms under conditions of different missing rates of data.