力矩(物理)
断层(地质)
计算机科学
故障指示器
控制理论(社会学)
故障检测与隔离
可靠性(半导体)
工程类
陷入故障
算法
信号处理
实时计算
电子工程
观测误差
故障覆盖率
错误检测和纠正
断层模型
距离测量
作者
Mou‐Fa Guo,Siqi Chen,Jian‐Hong Gao,Shilin Huang,Shuyue Lin
标识
DOI:10.1109/tim.2026.3736195
摘要
When locating the single-phase ground (SPG) fault section in the distribution network, an error in fault moment (FM) detection is inevitable, which may lead to the lack of key transient fault information in the input, thereby degrading the accuracy of fault location. To address this issue, an SPG fault section location method that is not affected by the FM error is proposed in this article. In this method, the input covers a relatively long time window and can still fully encompass the key transient fault information even within the range of FM errors. To accommodate this input format, an enhanced training set simulating various FM deviations was constructed to enhance the model’s robustness to FM errors. A physics-guided labeling strategy was proposed, which effectively highlights the key features in the long-time input. The 1D-U-Net model used can achieve rapid inference even when dealing with long-time inputs. To effectively process the long-time inputs and exploit the physics-guided labels, the structure of the model is modified to improve its ability to capture both local transient features and global temporal dependencies. The fault section is determined based on the number of different category labels in the model’s output, and the reliability of the output is quantitatively evaluated through a constructed confidence level. This method has been fully verified through 10-kV full-scale test systems and field data, and an industrial prototype based on Raspberry Pi integrating this algorithm has been developed. The test results demonstrate that the proposed method achieves high accuracy and fast diagnostic speed. Even with an inaccurate captured FM, it can still accurately locate the fault section.
科研通智能强力驱动
Strongly Powered by AbleSci AI