Hybrid-GNN: Dynamic Regional Reconstruction and Topology Verification of Main Distribution Network by Integrating GraphSAGE and R-GCN

Junjie  Yin; Guoquan  Yuan; Lei  Wei; Zhen  Zhang; Huiguang  Xu

doi:10.6180/jase.202608_31.026

Hybrid-GNN: Dynamic Regional Reconstruction and Topology Verification of Main Distribution Network by Integrating GraphSAGE and R-GCN

Junjie Yin¹, Guoquan Yuan¹, Lei Wei¹, Zhen Zhang¹, and Huiguang Xu²This email address is being protected from spambots. You need JavaScript enabled to view it.

¹State Grid Jiangsu Electric Power Co., Ltd. Information Telecommunication Branch, Nanjing 210024, China

²Nari-Tech Nanjing Control Systems Ltd., Nanjing 211100, China

Received: August 30, 2025
Accepted: September 30, 2025
Publication Date: February 15, 2026

Copyright The Author(s). This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are cited.

Download Citation: ||https://doi.org/10.6180/jase.202608_31.026

Traditional distribution network region reconstruction techniques struggle to cope with frequent changes in network topology and dynamic equipment adjustments, leading to segmentation delays and errors. This paper proposes a novel hybrid graph neural network (Hybrid-GNN) framework that integrates the neighborhood sampling and aggregation mechanism of GraphSAGE with the multi-relation convolution mechanism of R-GCN to simultaneously address the dual tasks of dynamic region reconstruction and topology verification. This model constructs a heterogeneous graph containing real-time attributes and multiple electrical connection types and adapts to dynamic changes using scalable embedding fusion techniques. Through neighborhood sampling and relation-sensitive convolution, node representations are gradually refined, achieving real-time and accurate region segmentation and topology verification. Experiments demonstrate that the Hybrid-GNN maintains high node consistency ( 90.5% ) under 30% topological perturbation, achieves anomaly localization accuracy of 91.2%, and efficiently processes networks with tens of thousands of nodes (within 1200 ms ) in edge environments, significantly outperforming traditional methods and single graph models. Its innovation lies in the simultaneous modeling of structural connectivity and multi-relation semantics, combined with scalable fusion techniques to adapt to dynamic grid conditions, providing important support for smart distribution systems.

Keywords: Power Distribution Network; Dynamic Region Partitioning; Heterogeneous Graph Embedding; GraphSAGE Aggregation; R-GCN Convolution

[1] J. Wang, L. Yao, K. Liu, F. Cheng, J. Xu, and J. Wang, (2024) “Dynamic network partitioning method of distribution networks considering regional autonomy" Power Syst. Technol:
[2] Z. Cheng, M. Min, M. Liwang, L. Huang, and Z. Gao, (2021) “Multiagent DDPG-based joint task partitioning and power control in fog computing networks" IEEE Internet of Things Journal 9(1): 104–116. DOI: https: //doi.org/10.1109/JIOT.2021.3091508.
[3] L. Ma, L. Wang, and Z. Liu, (2021) “Topology identification of distribution networks using a split-EM based data-driven approach" IEEE Transactions on Power Systems 37(3): 2019–2031. DOI: https: //doi.org/10.1109/TPWRS.2021.3119649.
[4] D. Deka, V. Kekatos, and G. Cavraro, (2023) “Learning distribution grid topologies: A tutorial" IEEE Trans actions on Smart Grid 15(1): 999–1013. DOI: https: //doi.org/10.1109/TSG.2023.3271902.
[5] X. Liang, M. A. Saaklayen, M. A. Igder, S. M. R. H. Shawon,S.O.Faried,andM.Janbakhsh,(2022)“Planning and service restoration through microgrid formation and soft open points for distribution network modernization: A review" IEEE Transactions on Industry Applications 58(2): 1843–1857. DOI: https: //doi.org/10.1109/TIA.2022.3146103.
[6] T.M.Aljohani,A.Saad,andO.A.Mohammed,(2021) “Two-stage optimization strategy for solving the VVO problem considering high penetration of plug-in electric vehicles to unbalanced distribution networks" IEEE Trans actions on Industry Applications 57(4): 3425–3440. DOI: https: //doi.org/10.1109/TIA.2021.3077547.
[7] S. Mondal, S. D. Manasi, K. Kunal, Z. Zeng, S. S. Sapatnekar, et al., (2022) “A unified engine for accelerating GNN weighting/aggregation operations, with efficient load balancing and graph-specific caching" IEEE Trans actions on Computer-Aided Design of Integrated Circuits and Systems 42(12): 4844–4857. DOI: https: //doi.org/10.1109/TCAD.2022.3232467.
[8] C. Xue, L. Lin, Y. Huang, and X. Wang, (2024) “Graph neural network-based anomaly detection for human cyber physical systems" Journal of the Chinese Institute of Engineers 47(8): 977–984. DOI: https: //doi.org/10.1080/02533839.2024.2407294.
[9] Y. Pei, J. Yang, J. Wang, P. Xu, T. Zhou, and F. Wu, (2023) “An emergency control strategy for undervoltage load shedding of power system: A graph deep reinforcement learning method" IET Generation, Transmission &Distribution 17(9): 2130–2141. DOI: https: //doi.org/10.1049/gtd2.12795.
[10] X. Gu, T. Liu, S. Li, X. Yang, and X. Cao, (2023) “Identification of vulnerable nodes in power grids based on graph deep learning algorithm" IET generation, trans mission & distribution 17(9): 2015–2027. DOI: https: //doi.org/10.1049/gtd2.12783.
[11] P. Li, Z. Zhong, Y. Zhao, C. Shao, Y. Sui, and R. Sun, (2025) “Power-GNN: a graph over-sampling method to mitigate power-law distribution in graph neural net works" Applied Intelligence 55(7): 1–16. DOI: https: //doi.org/10.1007/s10489-025-06421-5.
[12] L. Waikhom andR. Patgiri, (2023) “A survey of graph neural networks in various learning paradigms: methods, applications, and challenges" Artificial Intelligence Re view56(7): 6295–6364. DOI: https: //doi.org/10.1007/s10462-022-10321-2.
[13] B. Huang andJ. Wang, (2022) “Applications of physics informed neural networks in power systems-a review" IEEE Transactions on Power Systems 38(1): 572–588. DOI: 10.1109/TPWRS.2022.3162473.
[14] X. Zhang, Y. Liu, J. Duan, G. Qiu, T. Liu, and J. Liu, (2021) “DDPG-based multi-agent framework for SVC tuning in urban power grid with renewable energy re sources" IEEE Transactions on Power Systems 36(6): 5465–5475. DOI: https: //doi.org/10.1109/TPWRS.2021.3081159.
[15] Y.Huo,P.Li,H.Ji,J.Yan,G.Song,J.Wu,andC.Wang, (2021) “Data-driven adaptive operation of soft open points in active distribution networks" IEEE Transactions on Industrial Informatics 17(12): 8230–8242. DOI: https: //doi.org/10.1109/TII.2021.3064370.
[16] F. Yan, M. Zhang, and Z. Shi, (2021) “Dynamic partitioning of urban traffic network sub-regions with spatiotemporal evolution of traffic flow" Nonlinear Dynamics 105(1): 911–929. DOI: https: //doi.org/10.1007/s11071-021-06448-6.
[17] Y. Fang, X. Li, R. Ye, X. Tan, P. Zhao, and M. Wang, (2023) “Relation-aware graph convolutional networks for multi-relational network alignment" ACM Transactions on Intelligent Systems and Technology 14(2): 1–23. DOI: https: //doi.org/10.1145/3579827.
[18] W. Li, H. Chen, G. Wang, L. Song, Y. Xia, Z. Wang, and K. Jia. “Many-Objective Optimization Evolutionary Algorithm Based on Dynamic Region Par titioning”. In: 2024 4th International Conference on Consumer Electronics and Computer Engineering (ICCECE). IEEE. 2024, 59–63. DOI: https: //doi.org/10.1109/ICCECE61317.2024.10504213.
[19] X. Wang, D. Bo, C. Shi, S. Fan, Y. Ye, and P. S. Yu, (2022) “A survey on heterogeneous graph embedding: methods, techniques, applications and sources" IEEE transactions on big data 9(2): 415–436. DOI: https: //doi.org/10.1109/TBDATA.2022.3177455.
[20] Y.Chen,F.Chen,Z.Wu,Z.Chen,Z.Cai,Y.Tan,andS. Wang,(2025)“Heterogeneousgraphembeddingwithdual edge differentiation" Neural Networks 183: 106965. DOI: https: //doi.org/10.1016/j.neunet.2024.106965.
[21] M. Anitha and K. Sherly. “Customer Churn Prediction Using GraphSAGE Model with Degree Based Sampling and Max Pooling Aggregation”. In: Inter national Conference on Computing, Communication, Security and Intelligent Systems. Springer. 2024, 103–117.
[22] P.SinghandB.Raman.“GraphNeuralNetworks:Ex tending Deep Learning to Graphs”. In: Deep Learning Through the Prism of Tensors. Springer, 2025, 423–482.
[23] A. Inaolaji, A. Savasci, S. Paudyal, and S. Ka malasadan, (2023) “Distributed optimal power flow in unbalanced distribution grids with non-ideal communication" IEEE Transactions on Industry Applications 59(5): 5385–5397. DOI: https: //doi.org/10.1109/TIA.2023.3283236.
[24] T. ZhaoandJ. Wang, (2021) “Learning sequential distribution system restoration via graph-reinforcement learning" IEEE Transactions on Power Systems 37(2): 1601–1611. DOI: https: //doi.org/10.1109/TPWRS.2021.3102870.
[25] T. N. Nguyen, B.-H. Liu, N. P. Nguyen, B. Dumba, and J.-T. Chou, (2021) “Smart grid vulnerability and defense analysis under cascading failure attacks" IEEE Transactions on Power Delivery 36(4): 2264–2273. DOI: https: //doi.org/10.1109/TPWRD.2021.3061358.
[26] S. Xiao, H. Lin, J. Wang, X. Qin, and S. Wang, (2024) “Multi-Relation Augmentation for Graph Neural Net works" IEEE Transactions on Emerging Topics in Computational Intelligence 8(5): 3614–3627. DOI: https: //doi.org/10.1109/TETCI.2024.3371214.
[27] W. Ai, Y. Liu, C. Wei, T. Meng, H. Shao, Z. He, and K. Li, (2025) “MFLM-GCN: Multi-relation fusion and latent-relation mining graph convolutional network for entity alignment" Knowledge-Based Systems: 113974. DOI: https: //doi.org/10.1016/j.knosys.2025.113974.
[28] T. Tao, Q. Wang, Y. Ruan, X. Li, and X. Wang, (2023) “Graph Embedding with Similarity Metric Learning" Symmetry 15(8): 1618. DOI: https: //doi.org/10.3390/sym15081618.

Latest Articles