引用本文:杨楠,候少波,邢超,等.基于动态门控数据融合的 GCN-Transformer 配电网故障区段定位方法[J].电力系统保护与控制,2026,54(10):127-138.
YANG Nan,HOU Shaobo,XING Chao,et al.Fault section location method for distribution networks based on dynamic gated data fusion and GCN-Transformer[J].Power System Protection and Control,2026,54(10):127-138
【打印本页】   【下载PDF全文】   【查看/发表评论】  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被：浏览 112次   下载 11次 本文二维码信息
码上扫一扫！
分享到： 微信 更多
字体:加大+|默认|缩小-
基于动态门控数据融合的 GCN-Transformer 配电网故障区段定位方法
杨楠1，候少波1，邢超2，王灿1，关钦月3，叶学程3，李斯吾3，黄悦华1
1. 梯级水电站运行与控制湖北省重点实验室 (三峡大学)，湖北 宜昌 443002；2. 云南电网有限责任公司电力科学研究院，云南 昆明 650217；3. 国网湖北省电力有限公司经济技术研究院，湖北 武汉 430000
摘要:
在智能电网快速发展背景下，如何有效利用不同测量设备获取的多源数据，以满足不同故障场景下的配电网故障定位需求，对提升含分布式电源配电网的供电可靠性和运行安全性具有重要意义。鉴于此，提出一种基于动态门控数据融合的图卷积神经网络 (graph convolution network, GCN) 与 Transformer 相结合的配电网故障区段定位方法。首先，通过基于掩码感知的动态门控数据融合方法实现同步相量数据与同步波形数据的融合。然后，构建 GCN-Transformer 模型完成故障特征提取与融合，并引入焦点监督对比混合损失函数优化模型。最后，通过全连接分类层实现故障区段定位。仿真结果表明，所提方法在不同故障场景及样本不平衡条件下均表现出良好的故障定位性能。
关键词:  同步相量数据  同步波形数据  数据融合  GCN-Transformer  故障区段定位
DOI：10.19783/j.cnki.pspc.251301
分类号:
基金项目:国家自然科学基金项目资助 (62233006)
Fault section location method for distribution networks based on dynamic gated data fusion and GCN-Transformer
YANG Nan1, HOU Shaobo1, XING Chao2, WANG Can1, GUAN Qinyue3, YE Xuecheng3, LI Siwu3, HUANG Yuehua1
1. Hubei Provincial Key Laboratory for Operation and Control of Cascaded Hydropower Station (China Three Gorges University), Yichang 443002, China; 2. Electric Power Research Institute of Yunnan Power Grid Co., Ltd., Kunming 650217, China; 3. Economic and Technical Research Institute, State Grid Hubei Electric Power Co., Ltd., Wuhan 430000, China
Abstract:
With the rapid development of smart grids, effectively utilizing multi-source data obtained from different measurement devices to meet the fault location requirements under different fault scenarios is of great significance for improving the power supply reliability and operational safety of distribution networks with distributed generation. To this end, a fault section location method for distribution networks based on dynamic gated data fusion and a combination of graph convolution network (GCN) and Transformer is proposed. First, the synchrophasor data and synchronized waveform data are fused through a dynamic gated data fusion method based on mask perception. Then, a GCN-Transformer model is constructed to extract and fuse fault features, and a focal supervised contrastive hybrid loss function is introduced to optimize the model. Finally, the fault section location is achieved through a fully connected classification layer. Simulation results show that the proposed method exhibits strong fault location performance under different fault scenarios and sample imbalance conditions.
Key words:  synchrophasor data  synchronized waveform data  data fusion  GCN-Transformer  fault section location
  • 1
X关闭
  • 1
X关闭