| 引用本文: | 张逸,赵微,李传栋,张加忠.基于图卷积网络的电网电压暂降评估[J].电力自动化设备,2026,46(5):191-198,207 |
| Zhang Yi,Zhao Wei,Li Chuandong,Zhang Jiazhong.Graph convolutional network based voltage sag assessment of power grid[J].Electric Power Automation Equipment,2026,46(5):191-198,207 |
|
| 摘要: |
| 传统的电压暂降评估方法在电网运行方式变化时需重新进行海量故障仿真，无法满足实时性需求。针对上述问题，提出了一种基于图卷积网络的电压暂降评估方法。根据已有运行方式下电压暂降故障仿真数据构建电压暂降特征图并将其作为模型输入；搭建图卷积网络的评估模型，并基于网络拓扑关系挖掘电网运行方式、故障信息与节点电压暂降之间复杂的非线性映射；利用搭建好的模型评估机组启停和负荷水平等运行方式改变后的各节点电压暂降水平。通过对IEEE 39节点系统以及华东某地区的实际网架进行实例分析，证明了所提方法的准确性，且所提方法能有效缩短98.8 %的评估时间，大幅提高了电网电压暂降评估效率。 |
| 关键词: 电压暂降评估 运行方式变化 数据驱动 图卷积网络 贝叶斯优化 |
| DOI：10.16081/j.epae.202512025 |
| 分类号:TM711 |
| 基金项目:福建省科技计划引导性项目(2020H0009) |
|
| Graph convolutional network based voltage sag assessment of power grid |
|
Zhang Yi1, Zhao Wei1, Li Chuandong2, Zhang Jiazhong1
|
|
1.College of Electrical Engineering and Automation, Fuzhou University, Fuzhou 350108, China;2.College of Mechanical and Electrical Engineering, Fujian Agriculture and Forestry University, Fuzhou 350100, China
|
| Abstract: |
| Traditional assessment methods require extensive fault simulations to be redone whenever there are changes in the operating modes of the power grid, which cannot meet the real-time requirements. To address these issues, a graph convolutional network(GCN) based voltage sag assessment method is proposed. A voltage sag feature graph is constructed from the existing simulation data during voltage sag faults under current operating modes, which serves as the input for the model. A GCN based assessment model is built to excavate the complex nonlinear mapping between the power grid operating mode, fault information, and node voltage sag based on the network topology relationship. Then the developed model is employed to assess voltage sag level of each node after the change of operating condition such as generator unit start-stop and load level variations. Through case studies of the IEEE 39-bus system and the actual grid in a region of East China, the accuracy of the proposed method is demonstrated, and the assessment time can be effectively reduced by 98.8 %,which significantly improves the efficiency of power grid voltage sag assessment. |
| Key words: voltage sag assessment operating mode change data-driven graph convolutional network Bayesian optimization |
