Demand-side data augmentation method based on generative adversarial networks

This article proposes a method for data augmentation of demand side data based on generative adversarial networks. Firstly, the model structure and algorithm training iteration process of generating adversarial networks were studied, and appropriate nonlinear activation functions, loss functions, and optimizers were selected to establish a generative adversarial network model. After normalizing and preprocessing the original demand side data, the model was trained and a load curve was generated to fill in the missing daily load data. In multiple real-world scenarios, the effectiveness of the model was evaluated based on the Pearson correlation coefficient, verifying the feasibility and accuracy of using generative adversarial networks to enhance demand measurement data.