| 引用本文: | 徐艳春,左豪杰,张涛,席磊,吕密.基于改进SK-BiLSTM的自适应多尺度暂态电压稳定评估[J].电力自动化设备,2025,45(9):208-215. |
| XU Yanchun,ZUO Haojie,ZHANG Tao,XI Lei,Mi LU.Self-adaptive multi-scale transient voltage stability assessment based on improved SK-BiLSTM[J].Electric Power Automation Equipment,2025,45(9):208-215. |
|
| 本文已被：浏览 1657次 下载 814次 |
码上扫一扫！ |
|
|
| 基于改进SK-BiLSTM的自适应多尺度暂态电压稳定评估 |
|
徐艳春1,2, 左豪杰1,2, 张涛1,2, 席磊1,2, 吕密3
|
|
1.湖北省输电工程技术研究中心(三峡大学)，湖北 宜昌 443002;2.三峡大学 电气与新能源学院，湖北 宜昌 443002;3.德克萨斯农工大学 电气与计算机工程系，美国 卡城 77843
|
|
| 摘要: |
| 为评估电力系统暂态电压稳定性，提高电压失稳状态下失稳节点/区域的划分能力，提出一种基于改进选择性核卷积神经网络与双向长短期记忆网络(SK-BiLSTM)的自适应多尺度暂态电压稳定评估策略。深入分析暂态电压时空特性，揭示电力系统拓扑结构信息与动态时序电气量测数据信息在暂态电压稳定评估中的重要性；利用图注意力网络将拓扑结构信息与动态时序电气量测数据信息有效融合，基于此，构建基于改进SK-BiLSTM的自适应多尺度暂态电压稳定评估模型，通过在选择性核卷积神经网络中引入渐进式分组卷积机制高效提取局部和全局信息，并借助BiLSTM进一步增强时序电气量测数据信息的表征能力，在含有风电机组的IEEE 39、IEEE 118节点系统和河北省某区域电网中进行验证。结果表明，该评估模型具有较高的暂态电压稳定性评估精度，显著提升了电压失稳状态下的失稳节点/区域划分能力，并且有良好的泛化能力。 |
| 关键词: 暂态电压稳定 电压失稳节点/区域 多尺度卷积操作 深度学习 图注意力网络 |
| DOI：10.16081/j.epae.202504016 |
| 分类号: |
| 基金项目:国家自然科学基金资助项目(52277108) |
|
| Self-adaptive multi-scale transient voltage stability assessment based on improved SK-BiLSTM |
|
XU Yanchun1,2, ZUO Haojie1,2, ZHANG Tao1,2, XI Lei1,2, Mi LU3
|
|
1.Hubei Provincial Engineering Technology Research Center of Transmission Line, China Three Gorges University, Yichang 443002, China;2.College of Electrical Engineering and New Energy, China Three Gorges University, Yichang 443002, China;3.Department of Electrical and Computer Engineering, Texas A&M University, College Station 77843, USA
|
| Abstract: |
| To assess transient voltage stability and improve the ability to identify unstable nodes/regions in power system under voltage instabiligy state, a self-adaptive multi-scale transient voltage stability assessment strategy based on the improved selective kernel convolutional neural networks and bidirectional long short-term memory networks(SK-BiLSTM)is proposed. A detailed analysis of the transient voltage’s spatiotemporal characteristics highlights the importance of power system topology structure features and dynamic time-series electrical measurement data in transient voltage stability assessment. A graph attention network is used to effectively integrate the topology structure features and dynamic time-series electrical measurement data. Based on this, a self-adaptive multi-scale transient voltage stability assessment model is constructed based on the improved SK-BiLSTM. By introducing a progressive grouped convolution mechanism in selective kernel convolutional neural networks, both local and global features are efficiently extracted, while BiLSTM further enhances the representation ability of time-series electrical measurement data. The model is validated by the IEEE 39-bus and IEEE 118-bus systems, as well as a regional power grid in Hebei province, with wind power units included. Results show that the assessment model achieves high accuracy in transient voltage stability assessment and significantly improves the ability to identify unstable nodes/regions under voltage instability state, and has good generalization ability. |
| Key words: transient voltage stability voltage instability nodes/regions multi-scale convolution operation deep learning graph attention networks |
