A Method for Generating Subsynchronous Oscillation Data of Power System Based on Wasserstein Generative Adversarial Network

With a large number of power electronic devices connected to the power grid, the problem of subsynchronous oscillation caused by the interaction between the converter and the power grid seriously threatens the safe and stable operation of the power system. The existing research methods of subsynchronous oscillation, especially the artificial intelligence methods, mostly rely on a large amount of oscillation data, but in practice, such data is very scarce. Therefore, this paper proposes a method to generate subsynchronous oscillation data in power systems based on Wasserstein generative adversarial network (WGAN). In order to improve the quality of the generated samples and solve the problem of training instability, the JS distance is replaced by Wasserstein distance. Taking the oscillation data in the direct-driven permanent-magnet wind farm system as a practical example, the generated data are analyzed by dynamic time warping (DTW) and frequency to damping ratio methods. The results show that the data generated by the proposed method is consistent with the characteristics of oscillation data and has certain advantages in data quality.