摘要
针对大规模输电网络最优潮流（Optimal Power Flow, OPF）计算中存在的计算复杂、约束冗余等问题，提出一种基于图神经网络（Graph Neural Network, GNN）的高效最优潮流模型简化方法。首先，基于支路潮流计算结果和载流率阈值（如90%）定义关键支路，并构建包含节点特征、边特征与邻接矩阵的电网图结构数据；其次，训练GNN模型实现关键支路状态预测，识别电网中可能发生越限的支路集合；最后，以GNN预测得到的关键支路集合为基础，构建约束显著减少的简化最优潮流（Reduced Optimal Power Flow, ROPF）模型。以自建73节点输电网络为算例进行仿真分析，结果表明：GNN模型在90%载流率阈值下的关键支路预测误差仅为OPF的2.47%，ROPF求解时间较OPF缩短13.2%。
关键词： 最优潮流；图神经网络；关键支路；模型简化；载流率阈值
Abstract
Addressing the challenges of computational complexity and redundant constraints in Optimal Power Flow (OPF) calculations for large-scale transmission networks, this paper proposes an efficient OPF model simplification method based on Graph Neural Networks (GNNs). First, critical branches are defined based on branch power flow calculation results and current carrying capacity thresholds (e.g., 90%), and power grid graph-structured data including node features, edge features, and adjacency matrices are constructed. Second, a GNN model is trained to predict the states of critical branches, identifying the set of branches in the power grid that are likely to violate limits. Finally, based on the critical branch set predicted by the GNN, a Reduced Optimal Power Flow (ROPF) model with significantly fewer constraints is formulated. Simulation analysis is conducted using a self-built 73-node transmission network. Results show that under the 90% loading threshold, the prediction error of key branches using the GNN model is only 2.47% compared to the original OPF, and the solution time of ROPF is reduced by 13.2%.
Key words： Optimal power flow; Graph neural network; Critical branches; Model reduction; Loading threshold
参考文献 References
[1] 鲁宗相, 李昊, 乔颖. 从灵活性平衡视角的高比例可再生能源电力系统形态演化分析[J]. 全球能源互联网, 2021, 4(01): 12-18.
[2] 徐潇源, 王晗, 严正, 等. 能源转型背景下电力系统不确定性及应对方法综述[J]. 电力系统自动化, 2021, 45(16): 2-13.
[3] 王晗, 徐潇源, 严正, 等. 随机源-荷影响下配电网运行不确定性量化理论方法与应用展望[J]. 全球能源互联网, 2022, 5(03): 230-241.
[4] 方晓涛, 严正, 王晗, 等. 考虑源–荷随机–模糊特征的配电网潮流不确定性量化方法[J]. 中国电机工程学报, 2022, 42(20): 7509-7524.
[5] 郝广涛, 韩学山, 梁军, 等. 传统电力系统和高比例可再生能源电力系统调度与控制方法综述[J]. 电力系统及其自动化学报, 2020, 32(09): 10-19.
[6] 陈思宇, 柴庆宣, 李延松, 等. 综合能源系统潮流及最优潮流计算模型与方法综述[J]. 热力发电, 2020, 49(07): 1-12+20.
[7] 卫志农, 裴蕾, 陈胜, 等. 高比例新能源交直流混合配电网优化运行与安全分析研究综述[J]. 电力自动化设备, 2021, 41(09): 85-94.
[8] Pan X, Zhao T, Chen M, et al. Deepopf: A deep neural network approach for security-constrained dc optimal power flow[J]. IEEE Transactions on Power Systems, 2020, 36(3): 1725-1735.
[9] Yang K, Gao W, Fan R. Optimal power flow estimation using one-dimensional convolutional neural network[C]// 2021 North American Power Symposium (NAPS). IEEE, 2021: 1-6.
[10] Nellikkath R, Chatzivasileiadis S. Physics-informed neural networks for ac optimal power flow[J]. Electric Power Systems Research, 2022, 212: 108412.
[11] López-Cardona Á, Bernárdez G, Barlet-Ros P, et al. Proximal policy optimization with graph neural networks for optimal power flow[J]. arXiv preprint arXiv: 2212.12470, 2022.
[12] Wu H, Wang M, Xu Z, et al. Graph attention enabled convolutional network for distribution system probabilistic power flow[J]. IEEE Transactions on Industry Applications, 2022, 58(6): 7068-7078.
[13] Liao W, Bak-Jensen B, Pillai J R, et al. A review of graph neural networks and their applications in power systems[J]. Journal of Modern Power Systems and Clean Energy, 2021, 10(2): 345-360.
[14] Donon B, Donnot B, Guyon I, et al. Graph neural solver for power systems[C]//2019 international joint conference on neural networks (ijcnn). IEEE, 2019: 1-8.
[15] Liu S, Wu C, Zhu H. Topology-aware graph neural networks for learning feasible and adaptive AC-OPF solutions[J]. IEEE Transactions on Power Systems, 2022, 38(6): 5660-5670.
[16] Zimmerman R D, Murillo-Sánchez C E, Thomas R J. MATPOWER: Steady-state operations, planning, and analysis tools for power systems research and education[J]. IEEE Transactions on power systems, 2010, 26(1): 12-19.
[17] 侯磊, 刘金环, 于旭, 等. 图神经网络研究综述[J]. 计算机科学, 2024,51(06): 282-298.
[18] Grattarola D, Alippi C. Graph neural networks in tensorflow and keras with spektral [application notes][J]. IEEE Computational Intelligence Magazine, 2021, 16(1): 99-106.
[19] Yaoshiang H ,Samuel W .The Real-World-Weight Cross-Entropy Loss Function: Modeling the Costs of Mislabeling[J]. IEEE Access, 2020, 84806-4813.
[20] Grigg C, Wong P, Albrecht P, et al. The IEEE reliability test system-1996. A report prepared by the reliability test system task force of the application of probability methods subcommittee [J]. IEEE Transactions on power systems, 1999, 14(3): 1010-1020.