TR2021-130

Distribution system fault location analysis using graph neural network with node and link attributes


    •  Sun, H., Kawano, S., Nikovski, D.N., Takano, T., Mori, K., "Distribution system fault location analysis using graph neural network with node and link attributes", IEEE PES Innovative Smart Grid Technologies Conference - Europe (ISGT Europe), October 2021.
      BibTeX TR2021-130 PDF
      • @inproceedings{Sun2021oct,
      • author = {Sun, Hongbo and Kawano, Shunsuke and Nikovski, Daniel N. and Takano, Tomihiro and Mori, Kazuyuki},
      • title = {Distribution system fault location analysis using graph neural network with node and link attributes},
      • booktitle = {IEEE PES Innovative Smart Grid Technologies Conference - Europe (ISGT Europe)},
      • year = 2021,
      • month = oct,
      • url = {https://www.merl.com/publications/TR2021-130}
      • }
  • MERL Contacts:
  • Research Area:

    Electric Systems

Abstract:

This paper presents a graph neural network based fault location method for distribution systems, in which both link attributes and node attributes are considered. The proposed method integrates multiple measurements at different buses with branch parameters at different branches as inputs of the GNN, and transforms fault locations on branches into output features of corresponding connected nodes for the faulted branch. Besides the system topology that can be naturally considered by the GNN, the branch parameters and related regulation and energization statuses are explicitly taken into account as link attributes. Numerical examples are provided to demonstrate the usage of the proposed method.