基于两阶段特征选择的电力系统暂态功角与电压一体化稳定性评估方法  

Two-stage Feature Selection-Based Integrated Transient Angle and Voltage Stability Assessment Method for Power Systems

作　　者：徐艳春 张婧宇[1,2] 张涛 席磊 MI Lu[3] XU Yanchun;ZHANG Jingyu;ZHANG Tao;XI Lei;MI Lu(Hubei Provincial Engineering Technology Research Center of Transmission Line,Yichang 443002,China;College of Electrical Engineering&New Energy,China Three Gorges University,Yichang 443002,China;Department of Electrical and Computer Engineering,Texas A&M University,College Station 77843,USA)

机构地区：[1]湖北省输电线路工程技术研究中心,湖北宜昌443002 [2]三峡大学电气与新能源学院,湖北宜昌443002 [3]得克萨斯农工大学电气与计算机工程学院,美国得克萨斯州卡城77843

出　　处：《智慧电力》2025年第4期11-19,共9页Smart Power

基　　金：国家自然科学基金资助项目(52277108)。

摘　　要：随着电力系统暂态功角与电压稳定性问题耦合加剧,针对高维冗余特征的一体化评估需求凸显。提出了两阶段特征选择方法。首先通过类可分离性评分和皮尔逊相关系数筛选关键特征,消除分类低效与冗余;进而采用二进制竞争搜索算法优化特征子集,结合多任务卷积门控循环网络构建评估模型。算例验证表明,该方法在降低特征维度的同时显著提升了暂态稳定评估效果。With the increasing coupling between transient angle and voltage stability issues in power systems,the demand for the integrated assessment of high-dimensional redundant features has become prominent.This study proposes a two-stage feature selection method:first,key features were screened using class separability scores and Pearson correlation coefficients to eliminate classification inefficiency and redundancy.Subsequently,a binary competitive search algorithm was employed to optimize the feature subset,and the assessment model was constructed combined with a multi-task convolutional gated recurrent network.Case studies demonstrate that this method significantly improves transient stability assessment effectiveness while reducing feature dimensionality.

关 键 词：特征选择 暂态功角稳定 暂态电压稳定 类可分离性 竞争搜索算法 卷积门控循环单元 

分 类 号：TM712[电气工程—电力系统及自动化]