A Deep Learning-Based Method for Fault Location in HVDC Lines
Yiheng Tang, Kunlun Han
Guangxi University
阅读操作
确认中在文库中上传 PDF 后可生成中文音频讲解。
摘要与影响
High-voltage direct current (HVDC) systems have a relatively high failure rate, making rapid and accurate fault location critical to their stable operation. Traditional methods based on traveling wave front identification suffer from limitations such as mode aliasing, difficulty in precisely locating wavefront singularities, and the need for manual threshold tuning. To address these issues, this paper proposes an end-toend intelligent fault location method based on a 1D-CNN-GNN-Transformer architecture. This method first applies the Karenbauer transform to decouple transient waveforms and eliminate electromagnetic coupling. Subsequently, a One-Dimensional Convolutional Neural Network(1D-CNN) is used to adaptively extract high-frequency features from single-channel time-domain data, while a Graph Neural Networks (GNN) aggregates spatial-topological interaction information from multi-channel electrical quantities. Finally, a Transformer is employed to capture global long-range dependencies, and distance regression is optimized using the Huber loss. PSCAD/EMTDC simulation results demonstrate that this endto-end model requires no manual threshold tuning, exhibits an extremely low overall localization error rate, and demonstrates excellent noise immunity against high transition resistances.
逐年被引趋势
暂无年度引用数据
关键指标
同类平均 = 1
同领域 · 同年份 · 同类型
Google Scholar 与 OpenAlex 的被引统计范围不同,数值存在差异属正常。
AI 辅助阅读
依据:摘要
可就本文提问;依据不足时会说明。
学术脉络
学科主题
工程Power Systems Fault Detection
HVDC Systems and Fault Protection · Electrical Fault Detection and Protection
参考文献 9
此处列出前 3 条