applying deep generative methods to generate synthetic data in power systems

The lack of access to reliable databases, as well as the small number and imbalance of databases, is one of the main limitations of using machine learning methods in power systems, which can reduce efficiency and cause distrust in the results obtained from these methods. one of the solutions used to solve this problem is the use of synthetic data generation. two deep generative architectures, generative adversarial network (gan) and variational auto encoder (vae), are currently used to generate synthetic data. due to the novelty and importance of the subject, until now, a comparative study has not been done on the research conducted in this field, in terms of subject classification, with an emphasis on validation methods of synthetic production databases. the purpose of this research is to review the studies done in this field up to now and examine the research trends for the future. in this regard, after introducing the principles of gan and vae deep architectures, the subject of synthetic data generation using the mentioned methods in power systems has been studied comparatively.