Fault location in high-voltage direct-current grids based on panoramic voltage features and vision transformers
Bo Ren,Yongjie Luo,Niancheng Zhou,Qianggang Wang
DOI: https://doi.org/10.1016/j.ijepes.2024.109958
IF: 5.659
2024-03-28
International Journal of Electrical Power & Energy Systems
Abstract:This paper presents a novel fault location method designed for high-voltage direct-current (HVDC) grids equipped with quick-action protections and circuit breakers. The method relies on a very short data window of single-ended voltage signals, which allows the acquisition of necessary measurements before the ultra-fast fault-isolation stage. Through fault analysis of the simulated four-terminal HVDC grid, it is observed that the fault distance induces a unique transient response in the voltage signal. However, due to the complex mathematical representation of loop components and the multiple fault factors involved, a data-driven strategy is adopted for fault distance estimation. Based on this strategy, four signal processing techniques, including Gramian angular field, recurrence plot, Markov transition field, and continuous wavelet transform, are used to extract panoramic voltage features. Additionally, four advanced lightweight variants of vision transformers (ViT), including MobileViT, EfficientFormer, Efficient-Model, and EdgeViT, are considered for constructing the regression model. The optimal combination of panoramic voltage features and regression models for configuring the locator is determined by statistical evaluation. The results of semi-physical experiments based on real-time digital simulators demonstrate that the proposed method, first, can accurately estimate the fault distance, second, possesses robustness against high transition resistance, low sampling frequency, and strong noise interference, and third, is practically feasible.
engineering, electrical & electronic