基于双向长短期记忆生成对抗网络的电力系统次同步振荡数据生成方法  

Subsynchronous Oscillation Data Generation Method of Power System Based on Bidirectional Long Shortterm Memory Generative Adversarial Network

作　　者：薛展豪 陈力 林志颖 张敏 许祖峰 郑宇航 冯双 XUE Zhanhao;CHEN Li;LIN Zhiying;ZHANG Min;XU Zufeng;ZHENG Yuhang;FENG Shuang(State Key Laboratory of Smart Grid Protection and Control,NARI Group Co.,Ltd.,Nanjing,211106,China;School of Electrical Engineering,Southeast University,Nanjing,210096,China)

机构地区：[1]南瑞集团有限公司智能电网保护和运行控制国家重点实验室,江苏南京211106 [2]东南大学电气工程学院,江苏南京210096

出　　处：《智慧电力》2024年第5期60-66,共7页Smart Power

基　　金：国家自然科学基金资助项目(52377084);智能电网保护和控制国家重点实验室资助项目(SGNR0000KJJS2302237)。

摘　　要：针对实际工程中次同步振荡数据缺乏的问题,提出1种基于双向长短期记忆生成对抗网络的电力系统次同步振荡数据生成方法。首先,将双向长短期记忆(BiLSTM)网络引入生成模型和判别模型,充分挖掘振荡数据正向和反向的时间序列信息;然后,将Wasserstein距离引入生成式对抗网络(GAN)模型,解决训练不稳定的问题;最后,提出基于动态时间规整(DTW)的相似性指标及基于频域分析的振荡模态真实性指标,以衡量生成样本质量。算例分析表明,所提方法生成的数据符合振荡数据的特性,且在数据真实性方面具有一定优势。The subsynchronous oscillation data generation method based on bidirectional long short term memory generative adversarial networks(BiLSTM-GAN)is proposed to address the lack of subsynchronous oscillation data in practical engineering.Firstly,the bidirectional long short-term memory(BiLSTM)network is introduced into the generative and discriminative models to fully exploit the forward and backward temporal sequence information of oscillation data.Then,Wasserstein distance is incorporated into the generative adversarial network(GAN)model to solve the problem of training instability.Finally,dynamic time warping(DTW)-based similarity metrics and frequency domain analysis-based authenticity metrics are proposed to assess the quality of generated samples.Case studies demonstrate that the data generated by the proposed method aligns with the characteristics of oscillation data,exhibits certain advantages in data authenticity.

关 键 词：长短期记忆网络 次同步振荡 数据生成 Wasserstein距离 

分 类 号：TM732[电气工程—电力系统及自动化]