基于DBSCAN密度聚类和长短期记忆网络的同调机群辨识方法  

Identification method of coherent generator groups based on DBSCAN density clustering and long short-term memory network

作　　者：闫旭 薛易 相东昊 YAN Xu;XUE Yi;XIANG Donghao(State Grid Zhejiang Electric Power Co.,Ltd.,Chunan Power Supply Company,Hangzhou 310000,China;School of Electrical Engineering,Heilongjiang University of Science and Technology,Harbin 150000,China;State Grid Shandong Electric Power Co.,Ltd.,Dongying Power Supply Company,Dongying 257000,China)

机构地区：[1]国网浙江省电力有限公司淳安县供电公司,杭州310000 [2]黑龙江科技大学电气与控制工程学院,哈尔滨150000 [3]国网山东省电力公司东营供电公司,山东东营257000

出　　处：《黑龙江电力》2021年第5期377-383,390,共8页Heilongjiang Electric Power

基　　金：黑龙江省高等教育教学改革研究项目(项目编号:SJGY20200643)。

摘　　要：为了满足大电网暂态稳定在线分析的时效性与精度要求,提出了基于DBSCAN密度聚类和长短期记忆网络的同调机群快速辨识方法。基于广域量测数据,利用长短期记忆网络对电压相量的实部时序轨迹和虚部时序轨迹并行超实时预测,得到时序节点电压轨迹以提升长短期记忆网络预测精度;构建轨迹偏移特征平面,提取电压相量轨迹的时序变化特征,是同调机群快速辨识的重要前提;基于DBSCAN密度聚类方法对时序变化特征进行聚类分析,得到最终分群结果,并通过扩展等面积准则(EEAC)对同调辨识结果进行验证。通过新英格兰39节点系统验证了所提同调机群辨识方法的准确性。In order to meet the accuracy requirements of on-line transient stability analysis of large power grid,a fast identification method of coherent generator groups based on density-based spatial clustering of applications with noise(DBSCAN)density clustering and long short-term memory network(LSTM)is proposed.Based on the wide area measurement data,the real and imaginary part time series trajectories of voltage phasor are predicted in parallel super real time by using the LSTM,and the time node voltage trajectories are obtained to improve the prediction accuracy of the LSTM.Constructing the trajectory offset feature plane and extracting the temporal variation characteristics of the voltage phasor trajectory is an important premise for the rapid identification of coherent clusters.Based on DBSCAN density clustering method,the clustering analysis of time-series variation characteristics is carried out to obtain the final clustering results,and the coherence identification results are verified by extended equal area criterion(EEAC).The accuracy of the coherent cluster identification method is verified by the IEEE-39 bus system.

关 键 词：同调机群 电压相量轨迹 DBSCAN密度聚类 长短期记忆网络 

分 类 号：TM762.3[电气工程—电力系统及自动化]