| 引用本文: | 徐伟,严明辉,戴玉臣,周海锋,陈臣鹏,许庆,郭金鹏.基于因果分解和逆向特征增强时间卷积网络的电力系统静态电压稳定在线评估方法[J].电力自动化设备,2026,46(2):185-193. |
| XU Wei,YAN Minghui,DAI Yuchen,ZHOU Haifeng,CHEN Chenpeng,XU Qing,GUO Jinpeng.Online assessment method of static voltage stability for power system based on causal decomposition and reverse feature enhanced temporal convolutional network[J].Electric Power Automation Equipment,2026,46(2):185-193. |
|
| 本文已被：浏览 247次 下载 51次 |
码上扫一扫！ |
|
|
| 基于因果分解和逆向特征增强时间卷积网络的电力系统静态电压稳定在线评估方法 |
|
徐伟1,2, 严明辉1,2, 戴玉臣1,2, 周海锋1,2, 陈臣鹏3, 许庆3, 郭金鹏3
|
|
1.南瑞集团有限公司(国网电力科学研究院有限公司)，江苏 南京 211106;2.电网运行风险防御技术与装备全国重点实验室，江苏 南京 211106;3.河海大学 电气与动力工程学院，江苏 南京 211100
|
|
| 摘要: |
| 针对现有数据驱动方法因关键变量选择不合理及数据特征挖掘不充分导致的评估性能受限问题，提出基于因果分解和逆向特征增强时间卷积网络的静态电压稳定评估方法。直接将底层量测数据作为输入特征，利用因果分解法辨识出对电压稳定裕度影响较大的变量，根据因果强度缩减原始特征集，形成关键特征子集，降低模型训练难度，提高计算效率；构建逆向特征增强的时间卷积模型，通过构造逆向特征提取支路以及增加多头注意力机制充分挖掘电力系统的数据特征，实现静态电压稳定裕度的评估。在改进的4机2区域系统和改进的IEEE 39节点系统上进行验证，结果表明，该方法能够在保证估计精度的同时，大幅降低输入变量维度，有效提升算法性能。与其他方法相比，所提方法具有更高的评估精度，可有效提升电力系统的风险防控水平。 |
| 关键词: 静态电压稳定 影响因素分析 因果分解 时间卷积网络 数据驱动 |
| DOI：10.16081/j.epae.202507021 |
| 分类号: |
| 基金项目:南瑞集团有限公司科技项目(SGNRGFNKSJS2235580)；国家重点研发计划项目(2022YFB2402705) |
|
| Online assessment method of static voltage stability for power system based on causal decomposition and reverse feature enhanced temporal convolutional network |
|
XU Wei1,2, YAN Minghui1,2, DAI Yuchen1,2, ZHOU Haifeng1,2, CHEN Chenpeng3, XU Qing3, GUO Jinpeng3
|
|
1.NARI Group Corporation(State Grid Electric Power Research Institute),Nanjing 211106, China;2.National Key Laboratory of Grid Operation Risk Defense Technology and Equipment?,Nanjing 211106, China;3.School of Electrical and Power Engineering, Hohai University, Nanjing 211100, China
|
| Abstract: |
| Aiming at the limit assessment performance problem of current data-driven methods caused by unreasonable key variable selection and insufficient data feature mining, an assessment method of static voltage stability based on causal decomposition and reverse feature enhanced temporal convolutional network is proposed. The original measured data are directly used as the input features, and the causal decomposition method is used to identify the variables that have great impact on the voltage stability margin. The original dataset is reduced according to the causal strength, and a subset of key features is formed, which reduces the model training difficulty and improves the calculation efficiency. An inverse feature enhanced temporal convolutional model is built, the data features of power system are fully excavated by the construction of reversed feature extraction branch and the adding of multi-head attention mechanism, and the assessment of static voltage stability margin is realized. The verification is carried out on the modified 4-machine 2-zone system and the modified IEEE 39-bus system, and the results show that the proposed method can significantly reduce the dimensionality of input variables and effectively improve the algorithm performance while guaranteeing the assessment accuracy. Compare with the other methods, the proposed method has higher assessment accuracy, and can effectively enhance the risk prevention and control level of power system. |
| Key words: static voltage stability influencing factor analysis causal decomposition temporal convolutional network data-driven |
