Abstract:In order to solve the problem of difficult positioning after a short-circuit fault in a wind farm collector line, a new fault section positioning method based on Spectral-Grouping-based Mode Decomposition- Wigner-Ville Distribution (SGMD-WVD) time-frequency spectrum combined with knowledge distillation (KD) is proposed. Simulation analysis shows that the SGMD-WVD time-frequency spectrum has a corresponding relationship with the fault section. The fault section of the collector line can be located by mining the relationship between the spectrum line and the fault section with the help of a deep learning algorithm. The KD algorithm can be used to obtain a network with a high recognition rate and is easy to deploy to edge devices, which helps to promote the deployment of deep learning fault diagnosis networks to the field. Simulations show that the method proposed in this paper has good adaptability to multi-branch, mixed short-line networks such as collector lines, and the positioning is less affected by transition resistance, noise and fault phase angle.