High Impedance Fault Detection Method Based on Vmd and Transformer in Distribution Networks

In forest and grassland distribution networks, high impedance faults (HIF) frequently occur. Due to the weak fault characteristics and their similarity to other normal disturbances, these faults are often difficult to detect. If such faults are not detected and cleared in time, they could ignite vegetation, causing wildfires and posing significant safety hazards. Hence, this study introduces a HIF identification approach utilizing Variational Mode Decomposition (VMD) and the Transformer architecture. By performing time-frequency analysis on transient zerosequence voltage during HIF and other typical disturbances through VMD, the outcomes are input into the Transformer model for training and evaluation, enabling precise and reliable fault detection. The experimental findings demonstrate that this approach enhances the precision of HIF identification while successfully differentiating it from comparable scenarios, including load variations, unloaded line operations, and transient currents. Additionally, it maintains high detection accuracy and robustness even under data loss and noisy environments in complex power systems.