高级检索

基于MST-GCN与动态图注意力机制的配网接地故障快速定位研究

Research on Fast Localization of Grounding Faults in Distribution Networks Based on MST-GCN and Dynamic Graph Attention Mechanism

  • 摘要: 配网接地故障测定时,仅以单点故障特征为依据完成定位,缺乏关联性,导致输出的定位效率较低。为此提出对基于MST-GCN与动态图注意力机制的配网接地故障快速定位方法的设计和研究。定义描述配网实时故障状态,通过幅频波动变化提取故障暂态时域特征。以MST-GCN对暂态时域特征聚合处理,根据聚合特征和动态图注意力机制对覆盖区域内的接地故障点关联定位,完成基础性辨识。采用Adam对辨识目标优化,修正定位偏差,输出最终定位结果。实验结果表明:MST-GCN与动态图注意力机制配网接地故障快速定位方法针对多组接地故障定位的时延在0.2~0.3s之间,综合效率显著强化,具有优越的性能与价值。

     

    Abstract: When determining grounding faults in distribution networks, only single point fault characteristics are used as the basis for localization, lacking correlation, resulting in low efficiency in output localization. Therefore, the design and research of a fast fault location method for distribution network grounding based on MST-GCN and dynamic graph attention mechanism are proposed. Define and describe the real-time fault status of the distribution network, and extract the transient time-domain characteristics of the fault through amplitude frequency fluctuations. MST-GCN is used to aggregate transient time-domain features, and based on the aggregated features and dynamic graph attention mechanism, the grounding fault points within the coverage area are associated and located, completing fundamental identification. Using Adam to optimize the identification target, correct the positioning deviation, and output the final positioning result. The experimental results show that the MST-GCN and dynamic graph attention mechanism for fast positioning of grounding faults in distribution networks have a delay of 0.2-0.3s for multiple sets of grounding fault positioning, significantly enhancing the overall efficiency and possessing superior performance and value.

     

/

返回文章
返回