基于时空注意机制与LSTM的暂态电压稳定评估  

Transient Voltage Stability Assessment Based on Spatial-temporal Attention Mechanism and LSTM

作　　者：刘颂凯 崔梓琪 杨超[1,2] 阮肇华 张磊 袁铭洋[1,2] LIU Songkai;CUI Ziqi;YANG Chao;RUAN Zhaohua;ZHANG Lei;YUAN Mingyang(College of Electrical Engineering and New Energy,China Three Gorges University,Yichang 443002,China;Hubei Provincial Collaborative Innovation Center for New Energy Microgrid,Yichang 443002,China;Ningde Power Supply Company,State Grid Fujian Electric Power Co.,Ltd,Ningde 352100,China)

机构地区：[1]三峡大学电气与新能源学院,宜昌443002 [2]新能源微电网湖北省协同创新中心,宜昌443002 [3]国网福建省电力有限公司宁德供电公司,宁德352100

出　　处：《电力系统及其自动化学报》2024年第6期82-91,共10页Proceedings of the CSU-EPSA

基　　金：国家自然科学基金资助项目(52007103,62233006);湖北省自然科学基金资助项目(2022CFB825);电力系统智能运行与安全防御宜昌市重点实验室(三峡大学)开放基金资助项目(2020DLXY06);梯级水电站运行与控制湖北省重点实验室(三峡大学)开放基金资助项目(2019KJX11)。

摘　　要：针对新型电力系统发展背景下现阶段机器学习算法难以准确判断风电并网系统暂态电压稳定性的问题,提出一种基于时空注意机制与长短期记忆网络的暂态电压稳定评估方法。首先,基于支持向量机对初始单特征评价进行特征粗筛,并采用皮尔逊相关系数法判断剩余特征的相似度;其次,通过主成分分析-加权负荷评价获取与暂态电压稳定情况相关性较高的特征集;然后,通过时空注意机制,量化系统负载节点间的空间耦合关系和风电接入点间的空间相关性对整个系统暂态进程的影响,构建基于时空注意机制与长短期记忆网络的评估模型;最后,在算例上进行仿真分析,结果表明该模型有利于风电并网系统暂态电压稳定性判别准确率的提升,以及减少误判和漏判。In the context of the development of novel power systems,it is difficult for the existing machine learning algo-rithms to accurately judge the transient voltage stability of grid-connected wind power systems.Aimed at this problem,a transient voltage stability assessment method based on spatial-temporal attention mechanism(STAM)and long short-term memory(LSTM)network is proposed.First,the initial single-feature assessment is coarsely screened based on support vector machine,and the Pearson correlation coefficient method is used to judge the similarity of the remaining features.Second,the feature set with a high correlation with the transient voltage stability is obtained through principal component analysis-weighted load evaluation.Third,STAM is used to quantify the effects of the spatial coupling be-tween load buses and the spatial correlation between wind power access points on the transient process of the whole sys-tem,and an assessment model based on STAM and LSTM network is constructed.Finally,the simulation and analysis of an example show that this model is beneficial for improving the judgment accuracy of transient voltage stability and re-ducing the cases of misjudgment and missing judgment.

关 键 词：风电并网系统 特征集 时空注意机制 长短期记忆网络 暂态电压稳定 

分 类 号：TM712[电气工程—电力系统及自动化]