电力建设2025，Vol.46Issue(10)：88-98,11.DOI:10.12204/j.issn.1000-7229.2025.10.008
基于Transformer-GAT的新型电力系统宽频振荡源定位
Wide-Frequency Oscillatory Source Localization of a New Power System Based on Transformer-GAT
摘要
Abstract
[Objective]To solve the problem of new power system security caused by large-scale new-energy-grid connections from traditional industrial frequency bands to medium-and high-frequency bands,and to promote systematic research on a broadband oscillation source location method based on artificial intelligence,a wide-frequency oscillation source location method based on a transformer and graph attention neural network(GAT)is proposed.[Methods]First,on the substation side,the powerful signal processing and feature extraction capabilities of the transformer network were used to efficiently encode and compress the measurement signals of the power system to ensure the effective transmission of key broadband oscillation information under limited bandwidth conditions and reduce data redundancy.Subsequently,on the main-station side,combined with the compressed signal characteristics and system network topology,the GAT was used to locate the oscillation source.Finally,a four-machine,two-area system with a wind farm was used for verification.[Results]Simulation and experimental results show that the proposed transformer encoder can extract effective features from wide-frequency oscillation signals to realize the compression of substation signals.The proposed GAT model can achieve the accurate positioning of a wide-frequency oscillation source.In the comparison experiments with other algorithms,the GAT model had a lower false alarm cost when locating the oscillation source and maintained a balance between sensitivity and specificity.[Conclusions]Through the collaborative analysis of signal characteristics and network topology,the Transformer-GAT method effectively improves the positioning accuracy and robustness of wide-frequency oscillation sources.It provides technical support for the stable operation of new power systems.关键词
新型电力系统/宽频振荡/振荡源定位/Transformer/时空特性/图注意力神经网络Key words
new power system/wide-frequency oscillation/oscillation source localization/Transformer/spatiotemporal characteristics/graph attention neural network分类
信息技术与安全科学引用本文复制引用
张清源,周波,池建飞,赵妍,陶亮,胡枭..基于Transformer-GAT的新型电力系统宽频振荡源定位[J].电力建设,2025,46(10):88-98,11.基金项目
国家自然科学基金项目(52477178) （52477178）
国网浙江省电力有限公司科技项目(5211HZ240001) （5211HZ240001）
吉林省教育厅科学研究项目(JJKH20240144KJ) This work is supported by the National Natural Science Foundation of China(No.52477178),the Science and Technology Project of State Grid Zhejiang Electric Power Co.,Ltd.(No.5211HZ240001),and the Science Research Project of Jilin Provincial Department of Education(No.JJKH20240144KJ). （JJKH20240144KJ）
