A Review of Generative Adversarial Networks and Their Application to Short-Term Load Prediction
Power systems are facing the challenge of multi-source and multi-dimensional data.Deep learning is more suit-able for processing electric power data compared with conventional methods,with powerful capability of dimension reduc-tion,nonlinear fitting and feature extraction.Generative adversarial network (GAN)improves the performance of genera-tor and discriminator through adversarial training,and can effectively predict short-term daily load,distribution network load and electric vehicle load.This paper first outlines the basic concepts of GAN and analyzes its advantages and disadvan-tages.Then it describes four types of GAN-derived models that are widely used in short-term load prediction,and gives a detailed overview of the current status of modern applications of GAN in short-term load prediction.Finally it discusses the future application prospects.
deep learninggenerative adversarial networksshort-term load
万方数据
维普
