Generative Adversarial Networks Based Power Quality Data Generation Method and Its Application

The construction of new power systems with more new energy resources connected to the power grid may lead to more power quality problems. In this paper, generative adversarial networks (GANs) method was used to process power quality data restoration. Firstly, the power quality data missing problem was elaborated and the feasibility of GAN in the application of power quality data generation was analyzed. Secondly, the GAN model for power quality data was constructed and the loss function was detail structured to improve the data generation accuracy and enhance the performance of missing data generation, which the processing procedure includes two-dimensional mapping conversion, greyscaling processing, data feature extraction, generative adversarial data training and calibration. Finally, F-test was conducted to verify the effectiveness of the method through experiments on repairing real power quality data.