[1]郝武帮,段文强,李　 琰,等.基于GNN-FOCOPS的配电网动态集群划分方法[J].郑州大学学报(工学版),2027,48(XX):1-9.[doi:10. 13705 / j. issn. 1671-6833. 2026. 06. 003]
　HAO Wubang,DUAN Wenqiang,LI Yan,et al.A Dynamic Clustering Method for Distribution Networks Based on GNN-FOCOPS[J].Journal of Zhengzhou University (Engineering Science),2027,48(XX):1-9.[doi:10. 13705 / j. issn. 1671-6833. 2026. 06. 003]
点击复制

基于GNN-FOCOPS的配电网动态集群划分方法()
分享到：

《郑州大学学报(工学版)》[ISSN:1671-6833/CN:41-1339/T]

卷:
48
期数:
2027年XX
页码:
1-9
栏目:
出版日期:
2027-12-10

文章信息/Info

Title:
A Dynamic Clustering Method for Distribution Networks Based on GNN-FOCOPS
作者:
郝武帮1,2, 段文强1,2, 李　 琰1,2, 张　 春1,2, 沈东奇1,2
1.云南民族大学 电气信息 工程学院,云南 昆明650504;2云南省无人自主系统重点实验室,云南 昆明650504
Author(s):
HAO Wubang 1,2,DUAN Wenqiang1,2, LI Yan1,2, ZANG Chun1,2, SHEN Dongqi1,2
1. School of Electrical and Information Engineering, Yunnan Minzu University., Kunming Yunnan 650504, China; 2. Yunnan Provincial Key Laboratory of Unmanned Autonomous Systems, Kunming Yunnan 650504, China
关键词:
配电网动态集群划分; 图神经网络; 约束强化学习; FOCOPS; 电压调节能力; 资源多样性熵
Keywords:
distribution network cluster division;  graph neural networks;  constrained reinforcement learning;  FOCOPS;  voltage regulation;  resource diversity entropy
分类号:
TM734 文献标 志码：A doi:
DOI:
10. 13705 / j. issn. 1671-6833. 2026. 06. 003
文献标志码:
TM374
摘要:
高渗透率分布式电源接入使配电网净负荷与潮流方向剧变，静态集群划分难以适应光伏波动与负荷峰谷切换等动态工况。为此，提出一种基于图神经网络与约束强化学习的动态集群划分方法：构建由变权模块度、能量可持续电压调节能力、有功–无功协同调节能力和资源多样性熵组成的综合评价指标体系，并形成GNN–Tree-DP–FOCOPS决策框架，其中GNN提取拓扑与运行特征，Tree-DP对划分结果施加连通性硬约束投影，FOCOPS在运行约束下完成策略优化。IEEE 33节点与123节点算例表明，相比逐时重启粒子群算法与K-Means聚类，所提方法在光伏大发与负荷高峰时段综合指标显著提升，训练后单时刻推理仅需8 ms；在训练拓扑内未见工况下仍保持良好泛化性，光伏骤降场景约束违约率为0%。但跨拓扑迁移时受GNN策略与网络结构耦合影响，综合指标降至0.6561且违约率升至70%，需要重新训练。结果表明，该方法可在训练拓扑内实现面向动态运行的快速、安全集群划分，为高比例新能源接入下的集群自治运行提供技术支撑。
Abstract:
High penetration of distributed energy resources caused drastic variations in net load and power-flow directions in distribution networks, so static clustering did not adapt well to photovoltaic (PV) fluctuations and peak–valley load transitions. To address this issue, a dynamic clustering method based on a graph neural network (GNN) and constrained reinforcement learning was proposed. A comprehensive evaluation index system was first constructed by integrating variable-weight modularity, energy-sustainable voltage regulation capability, active–reactive coordination capability, and resource diversity entropy. Then, a GNN–Tree-DP–FOCOPS decision-making framework was developed, in which the GNN extracted topology- and operation-aware features, Tree-based dynamic programming (Tree-DP) imposed a hard connectivity projection on the clustering result, and first-order constrained optimization in policy space (FOCOPS) optimized the policy under operational constraints. Case studies on the IEEE 33-bus and 123-bus systems showed that, compared with the hourly-restarted particle swarm optimization method and K-means clustering, the proposed method significantly improved the comprehensive index F during high-PV-generation and peak-load periods; the trained policy required only 8 ms for single-step inference. The method also maintained good generalization to unseen operating conditions within the trained topology, achieving a 0% constraint-violation rate under a PV ramp-down scenario. However, because the learned GNN policy was tightly coupled with the training network structure, the performance degraded markedly in cross-topology transfer, where the comprehensive index dropped to 0.6561 and the violation rate increased to 70%, indicating that retraining was required. Overall, the proposed method enabled fast and safe dynamic clustering within the trained topology and provided technical support for cluster-based autonomous operation in distribution networks with high renewable penetration

参考文献/References:

[1].Calero I, Cañizares C, Bhattacharya K, et al. Duck-Curve Mitigation in Power Grids With High Penetration of PV Generation[J]. IEEE Transactions on Smart Grid, 2022, 13(1): 314-329.
[2].MICALLEF A, Spiteri-Staines C, Licari J. Voltage regulation in low voltage distribution networks with unbalanced penetrations of photovoltaics and battery storage systems[J]. IET Smart Grid, 2024, 7(3): 264-276.
[3].Mao M, Wu Z, Xu D, et al. Community-detection-based Approaches for Distribution Network Partition[J]. CSEE Journal of Power and Energy Systems, 2021, 10(5): 1965-1976.
[4].胡雪凯,尹瑞,时珉,等.基于改进粒子群算法的分布式光伏集群划分与无功优化策略[J].电力电容器与无功补偿,2021,42(04):14-21.
[5].HU X K, YIN R, SHI M, et al. Distributed Photovoltaic Cluster Partition and Reactive Power Optimization Strategy Based on Improved Particle Swarm Optimization Algorithm[J]. Power capacitor & reactive power compensation, 2021, 42(4): 14-21.
[6].郑吉祥, 钟俊. 基于节点类型和分区耦合性的复杂网络无功电压快速分区方法[J].电网技术, 2020, 44(1): 223–230.
[7].ZHENG J X, ZHONG J. Rapid reactive voltage partitioning method for complex networks based on node type and partition coupling[J]. Power System Technology, 2020, 44(1): 223-230.
[8].吕文琪,刘子文,董晓霄,等.考虑灵活性资源可调容量和响应速度的集群划分方法[J].电力工程技术,2024,43(05):37-47.
[9].LU W Q, LIU Z W, DONG Y X, et al. A cluster partitioning method considering adjustable capacity and response speed of flexible resources [J]. Electric power engineering technology, 2024, 43(05): 37-47.
[10].冯侃,魏立保,吴兆彬,等.基于集群划分的含高比例户用光伏配电网电压控制研究[J].综合智慧能源,2025,47(02):60-70.
[11].FENG K, WEI L, WU Z, et al. Research on voltage control of distribution networks with high-proportion household photovoltaics based on cluster division [J]. Integrated Intelligent Energy, 2025, 47(2): 60-70.
[12].彭琦,欧阳森,张晋铭,等.考虑多断面源荷特征匹配的高比例分布式光伏接入配电网的集群划分方法[J].南方电网技术, 2025, 19(4): 161-172, 184.
[13].PENG Q, OUYANG S, ZHANG J M, et al. Cluster Partitioning Method for High Proportion Distributed Photovoltaic Access Distribution Network Considering Multi-Section Source-Load Features Matching [J]. Southern Power System Technology, 2025, 19(4): 161-172,184.
[14].陈婧华,张琳娟,卢丹,等.基于改进粒子群优化算法的分布式电源集群划分方法[J].郑州大学学报(工学版),2023,44(05):77-85.
[15].CHEN J H, ZHANG L J, LU D, et al. Cluster Partition Method of Distributed Power Supply Based on Improved Particle Swarm Optimization Algorithm [J].Journal of Zhengzhou University (Engineering Science), 2023, 44(05): 77-85.
[16].王晶,袁懿,邵尹池,等.考虑资源储备度的主动配电网多目标集群划分与电压控制方法[J].中国电力,2023,56(12):69-79.
[17].WANG J, YUAN Y, SHAO Y C, et al. Multi-Objective Cluster Classification and Voltage Control Approach for Active Distribution Network Considering Resource Reserve Degree [J]. Electric Power, 2023, 56(12): 69-79.
[18].易姝娴,王晶,梁伟宸,等.面向农村配电网电压优化控制的自适应动态分区方法[J].电力系统及其自动化学报,2025,37(03):110-119.
[19].YI S X,WANG J,LIANG W C et al. Adaptive Dynamic Partitioning Method for Voltage Optimization Control in Rural Distribution Networks [J]. Proceedings of the CSU-EPSA, 2025, 37(03): 110-119.
[20].段雪松,张晋嘉,李宏博,等.基于改进Louvain和BAAM-ADMM的配电网多目标集群划分和电压协调控制[J].智慧电力,2024,52(10):32-39102.
[21].DUAN X S, ZHANG J J, LI H B, et al. Multi Objective Cluster Partitioning and Voltage Coordination Control in Distribution Networks Based on Improved Louvain Algorithm and BAAM-ADMM [J]. Smart Power, 2024, 52(10): 32-39102.
[22].曹 敬,金玉龙,郑 涛,等.计及分布式电源集群不确定性的配电网分散鲁棒电压控制[J].电力系统保护与控制,2023,51(22):155-166.
[23].CAO J, JIN Y L, ZHENG T, et al. A decentralized robust voltage control method for distribution networks considering the uncertainty of distributed generation clusters [J]. Power System Protection and Control, 2023, 51(22): 155-166.
[24].LI H, SONG K, MENG F, et al. Voltage control strategy of a high-permeability photovoltaic distribution network based on cluster division, Front. Energy Res. 12:1377841. arch, 2024, 12: 1377841.
[25].DING J, ZHANG Q, HU S, et al. Clusters partition and zonal voltage regulation for distribution networks with high penetration of PVs. IET Gener. Transm. Distrib., 12: 6041-6051.
[26].吴 桐,刘丽军,林钰芳,等.基于动态分区的配电网日前优化调度研究[J].电力系统保护与控制,2022,50(15):21-32.
[27].WU T, LIU L J, LIN Y F, et al. Day-ahead optimal dispatch for a distribution network based on dynamic partitioning [J]. Power System Protection and Control, 2022, 50(15): 21-32.
[28].王艺涵,吴俊勇,原梦遥等.面向光伏消纳的配电网集群动态划分方法与优化运行策略[J].电力自动化设备,2025,45(08):122-129.
[29].WU Y H, WU J Y, YUAN M Y, et al. Dynamic partitioning method and optimized operation strategy for distribution network clusters considering photovoltaic integration [J]. Electric Power Automation Equipment, 2025, 45(08): 122-129.
[30].李宗强,赵耀,王云,等.面向有功潮流最优问题的配电网集群划分策略[J].电力工程技术,2024,43(03):151-160.
[31].LI Z Q, ZHAO Y, WANG Y, et al. Distribution network cluster division strategy for active power flow optimization problem [J]. Electric power engineering technology, 2024, 43(03): 151-160.
[32].XU Y, HU P, ZHANG F, et al. Distributed Generation Cluster Division Method Considering Frequency Regulation Response Speed. Applied Sciences. 2024; 14(6): 2432.
[33].ZHANG Y, VUONG Q, ROSS K. First order constrained optimization in policy space [J]. Advances in Neural Information Processing Systems, 2020, 33: 15338-15349.

相似文献/References:

[1]郑红,罗俞建,凌侃,等.多模态跨尺度特征融合的药物靶标亲和力预测[J].郑州大学学报(工学版),2026,47(XX):1.[doi:10.13705/j.issn.1671-6833.2026.04.001]
　ZHENG Hong,LUO Yujian,LING Kan,et al.Multimodal Cross-scale Feature Fusion for Drug-Target Affinity Prediction[J].Journal of Zhengzhou University (Engineering Science),2026,47(XX):1.[doi:10.13705/j.issn.1671-6833.2026.04.001]
[2]韩继辉,石玉鹏,黄子奇,等.结构-特征协同防御的图神经网络[J].郑州大学学报(工学版),2027,48(XX):1.[doi:10.13705/j.issn.1671-6833.2026.04.010]
　HAN Jihui,SHI Yupeng,HUANG Ziqi,et al.A Graph Neural Network for Structure-Feature Collaborative Defense[J].Journal of Zhengzhou University (Engineering Science),2027,48(XX):1.[doi:10.13705/j.issn.1671-6833.2026.04.010]

备注/Memo

备注/Memo:
收稿日期:2026-02-20;修订日期:2026-04-01基金项目:国家自然科学基金资助项目(62062068)作者简介:郝武帮(1971— ) ,男,云南德宏人,云南民族大学副教授,主要从事电力工程、工业控制研究,E-mail:ayesurmt@139. com。
更新日期/Last Update: 2026-06-12