A high-impedance fault location method for resonant grounded distribution networks based on transformer and few-shot learning

High-impedance faults (HIFs) in resonant grounded distribution networks have illegible fault features, which are difficult to accurately be located. Existing HIF location methods rely heavily on the HIF nonlinearity features rather than the deep laws, the performance and adaptability are invalid. This paper proposes a novel HIF location method based on Transformer and Few-Shot Learning (FSL) for the resonant grounded distribution network. A HIF location method is designed to identify the fault feeder accurately and adaptively, in which the improved Transformer-based HIF location model is constructed to explore deep properties of HIFs with high anti-interference ability. Specifically, the FSL is introduced to construct a HIF location migration structure, which can realize the new topology migration with high accuracy and fast training speed only needing a few fault data. This approach ensures strong accuracy and adaptability in various operational conditions and topologies, making it highly effective for practical applications. Finally, numerical simulations based on PSCAD/EMTDC were carried out, which reveals the accuracy and adaptability of the proposed HIF location method.