EN基于GAN关键样本增强和迁移学习的暂态稳定自适应评估方法
基于GAN关键样本增强和迁移学习的暂态稳定自适应评估方法
- 华北电力大学学报（自然科学版） 2026年53卷第1期 页码：57-72
- 作者机构：
1. 华北电力大学电气与电子工程学院
2. 中国电力科学研究院
- 作者简介：
- 基金信息：国家电网有限公司总部管理科技项目(5100-202355409A-3-2-ZN)
DOI：
中图分类号： TM712;TP18-
纸质出版：2026
- 稿件说明：
移动端阅览
[1]胡杰祥,刘君,李岩松,等.基于GAN关键样本增强和迁移学习的暂态稳定自适应评估方法[J].华北电力大学学报(自然科学版),2026,53(01):57-72.
胡杰祥, 刘君, 李岩松, et al. 基于GAN关键样本增强和迁移学习的暂态稳定自适应评估方法[J]. 2026, 53(1): 57-72.
摘要
目前基于深度学习的暂态稳定评估需要离线生成大量数据对预测模型进行训练，但当训练数据不足，或系统运行场景发生变化时，暂稳评估的准确率将显著下降。为此，本文提出了一种基于生成对抗网络(generative adversarial networks
GAN)关键样本增强和迁移学习的暂态稳定自适应评估方法，选择并增强关键样本，适应电网运行工况变化。首先，针对样本质量问题，量化分析样本与模型训练和暂稳边界的相关性，定义样本重要性指标，据此筛选出关键样本，应用GAN对其进行增强。然后，针对电力系统运行工况的变化问题，设计了暂态稳定自适应评估方案，应用高质量样本集实现自适应评估。当系统运行工况不变时，基于适应电网数据特点的CNN-LSTM并联特征提取器构建了暂稳评估模型。分别利用卷积神经网络(CNN)和长短期记忆网络(LSTM)提取一级局部特征与时序特征，在此基础上融合为二级系统特征，传入后级全连接网络实现暂稳预测。当系统运行工况变化时，设计了自适应最优迁移学习策略，根据电力系统运行场景变化自适应选择最优迁移策略，解决了随运行场景变化程度增大，特征的可迁移性降低的问题。将本文所提出的方法应用于新英格兰10机39节点系统以及我国某省级电网。测试结果表明，关键样本比例增强至10%
预测模型的判错率降低了86.54%
能显著提高少样本条件下模型预测能力。本文所提出的自适应暂态稳定评估方案在电网运行工况不改变时的预测准确率提升至99.26%
且具有一定的泛化能力。当电网运行工况改变时，在目标电力系统的预测准确率达到98%以上。
Abstract
Current deep learning-based transient stability assessment requires the offline generation of large datasets to train the prediction model. However
when the training data is insufficient or the system's operating condition changes
the accuracy of the transient stability assessment markedly declines. This paper introduces an adaptive transient stability assessment method based on generative adversarial networks(GAN) key sample enhancement and transfer learning. This method selectively enhances key samples to accommodate changes in power system operating conditions. To address the issue of sample quality
the correlation between samples and both model training and transient stability boundaries is quantitatively analyzed. A sample importance index is defined
based on which key samples are screened and augmented using GAN. A transient stability adaptive assessment scheme is designed to handle variations in power system operating conditions
using a high-quality sample set for adaptive assessment. When the operating conditions of the system remain unchanged
the transient stability assessment model is constructed based on the parallel CNN-LSTM feature extractor adapted to the characteristics of the grid data. The convolutional neural network(CNN) and long-short-term memory network(LSTM) are employed to extract first-level local features and time-series features
which are fused into second-level system features and passed into the subsequent fully connected network to achieve transient stability prediction. For changing operating conditions
an adaptive optimal transfer learning strategy is designed to select the optimal transfer strategy based on changes in the operating conditions of the power system. This addresses the issue of reduced feature transferability as the degree of changes in operating conditions increases. The proposed method is applied to the New England 10-machine 39-bus system and a Chinese provincial grid. Results indicate that the proportion of key samples to 10% reduces the prediction error rate by 86.54%
significantly improving the model's prediction ability with fewer samples. The proposed adaptive transient stability assessment scheme improves prediction accuracy to 99.26% when the operating conditions of the grid remain unchanged
demonstrating a certain generalization ability. For changed conditions
the prediction accuracy in the target power system exceeds 98%.
关键词
Keywords
references
肖湘宁.新一代电网中多源多变换复杂交直流系统的基础问题[J].电工技术学报，2015,30(15):1-14.XIAO Xiangning.Basic problems of the new complex AC-DC power grid with multiple energy resources and multiple conversions[J].Transactions of China Electrotechnical Society,2015,30(15):1-14 (in Chinese).
周孝信，陈树勇，鲁宗相，等.能源转型中我国新一代电力系统的技术特征[J].中国电机工程学报，2018,38(7):1893-1904.ZHOU Xiaoxin,CHEN Shuyong,LU Zongxiang,et al.Technology features of the new generation power system in China[J].Proceedings of the CSEE,2018,38(7):1893-1904 (in Chinese).
马爱平.新型电力系统助力实现“双碳”目标[N].科技日报，2021-10-21(3).MA Aiping.New power systems help achieve “dual carbon” goals[N].Science and Technology Daily,2021-10-21(3) (in Chinese).
汤涌，卜广全，易俊.印度“7.30”、“7.31”大停电事故分析及启示[J].中国电机工程学报，2012,32(25):167-174.TANG Yong,BU Guangquan,YI Jun.Analysis and lessons of the blackout in indian power grid on July 30 and 31,2012[J].Proceedings of the CSEE,2012,32(25):167-174(in Chinese).
杨悦，李国庆.基于VSC-HVDC的海上风电小干扰稳定控制[J].电工技术学报，2016,31(13):101-110.YANG Yue,LI Guoqing.The small signal stability control of offshore wind farm based on VSC-HVDC [J].Transactions of China Electrotechnical Society,2016,31(13):101-110 (in Chinese).
陈厚合，王长江，姜涛，等.基于端口能量的含VSC-HVDC的交直流混合系统暂态稳定评估[J].电工技术学报，2018,33(3):498-511.CHEN Houhe,WANG Changjiang,JIANG Tao,et al.Transient stability assessment in hybrid AC/DC systems with VSC-HVDC via port energy[J].Transactions of China Electrotechnical Society,2018,33(3):498-511 (in Chinese).
ZHANG Yuchen,XU Yan,DONG Zhaoyang,et al.A hierarchical self-adaptive data-analytics method for power system short-term voltage stability assessment[J].IEEE Transactions on Industrial Informatics,2019,15(1):74-84.
XU Yan,ZHANG Rui,ZHAO Junhua,et al.Assessing short-term voltage stability of electric power systems by a hierarchical intelligent system[J].IEEE Transactions on Neural Networks and Learning Systems,2016,27(8):1686-1696.
叶瑞丽，刘瑞叶，刘建楠，等.直驱风电机组风电场接入后的电力系统暂态稳定计算[J].电工技术学报，2014,29(6):211-218.YE Ruili,LIU Ruiye,LIU Jiannan,et al.Transient stability calculation of power system integrated with direct-drive wind farm with permanent magnet synchronous generators[J].Transactions of China Electrotechnical Society,2014,29(6) :211-218 (in Chinese).
ZHU Lipeng,LU Chao,SUN Yuanzhang.Time series shapelet classification based online short-term voltage stability assessment[J].IEEE Transactions on Power Systems,2016,31(2):1430-1439.
DASGUPTA S,PARAMASIVAM M,VAIDYA U,et al.PMU-based model-free approach for real-time rotor angle monitoring[J].IEEE Transactions on Power Systems,2015,30(5):2818-2819.
汤涌，姚伟，王宏志，等.电网仿真分析与决策的人工智能方法[J].中国电机工程学报，2022,42(15):5384-5406.TANG Yong,YAO Wei,WANG Hongzhi,et al.Artificial intelligence techniques for power grid simulation analysis and decision making[J].Proceedings of the CSEE,2022,42(15):5384-5406 (in Chinese).
汤奕，崔晗，党杰.基于继承思想的时变性电力系统暂态稳定预测[J].中国电机工程学报，2021,41(15):5107-5119.TANG Yi,CUI Han,DANG Jie.Transient stability prediction of time-varying power systems based on inheritance [J].Proceedings of the CSEE,2021,41(15):5107-5119 (in Chinese).
王怀远，陈启凡.基于代价敏感堆叠变分自动编码器的暂态稳定评估方法[J].中国电机工程学报，2020,40(7):2213-2220.WANG Huaiyuan,CHEN Qifan.A transient stability assessment method based on cost-sensitive stacked variational auto-encoder[J].Proceedings of the CSEE,2020,40(7):2213-2220 (in Chinese).
张若愚，吴俊勇，李宝琴，等.基于迁移学习的电力系统暂态稳定自适应预测[J].电网技术，2020,44(6):2196-2205.ZHANG Ruoyu,WU Junyong,LI Baoqin,et al.Self-adaptive power system transient stability prediction based on transfer learning[J].Power System Technology,2020,44(6):2196-2205 (in Chinese).
BATISTA G E,PRATI R C,MONARD M C.A study of the behavior of several methods for balancing machine learning training data[J].ACM SIGKDD Explorations Newsletter,2004,6(1):20-29.
FERNÁNDEZ A,GARCÍA S,MARÍA J,et al.A study of the behaviour of linguistic fuzzy rule based classification systems in the framework of imbalanced data-sets[J].Fuzzy Sets and Systems,2008,159(18):2378-2398.
李宝琴，吴俊勇，强子玥，等.基于改进DCGAN的电力系统暂态稳定增强型自适应评估[J].电力系统自动化，2022,46(2):73-82.LI Baoqin,WU Junyong,QIANG Ziyue,et al.Enhanced adaptive assessment on transient stability of power system based on improved deep convolutional generative adversarial network[J].Automation of Electric Power Systems,2022,46(2):73-82 (in Chinese).
李宝琴，吴俊勇，李栌苏，等.基于主动迁移学习的电力系统暂态稳定自适应评估[J].电力系统自动化，2023,47(4):121-132.LI Baoqin,WU Junyong,LI Lusu,et al.Adaptive assessment of power system transient stability based on active transfer learning [J].Automation of Electric Power Systems,2023,47(4):121-132 (in Chinese).
TONEVA M,SORDONI A,COMBES R,et al.An empirical study of example forgetting during deep neural network learning[C].The International Conference on Learning Representations,New Orleans,USA,2019.
PAUL M,GANGULI S,DZIUGAITE G K.Deep learning on a data diet:finding important examples early in training[J].Advances in Neural Information Processing Systems,2021,34:20596-20607.
SHI Zhongtuo,YAO Wei,TANG Yong,et al.Bidirectional active transfer learning for adaptive power system stability assessment and dominant instability mode identification[J].IEEE Transactions on Power Systems,2023,38(6):5128-5142.
王铮澄，周艳真，郭庆来，等.面向变化场景的输电断面极限评估主动迁移学习方法[J].中国电机工程学报，2023,43(15):5732-5745.WANG Zhengcheng,ZHOU Yanzhen,GUO Qinglai.Active transfer learning for TTC assessment of transmission interfaces with changing operating scenarios[J].Proceedings of the CSEE,2023,43(15):5732-5745 (in Chinese).
YOSINSKI J,CLUNE J,BENGIO Y,et al.How transferable are features in deep neural networks?[C].Advances in neural information processing systems,Montreal,Canada,2014.
GOODFELLOW I,POUGETABADIE J,MIRZA M,et al.Generative adversarial networks[J].Communications of the ACM,2020,63(11):139-144.
时纯，刘君，梁卓航，等.基于GAN和多通道CNN的电力系统暂态稳定评估[J].电网技术，2022,46(8):3191-3202.SHI Chun,LIU Jun,LIANG Zhuohang,et al.Transient stability assessment of power system based on gan and multi-channel CNN[J].Power System Technology,2022,46(8):3191-3202 (in Chinese).
0
浏览量
0
下载量
0
CSCD
- Meta-XML下载
- BibTeX下载
- Endnote下载
- RefMan 下载
- NoteExpress下载
- NoteFirst下载
关联资源
相关文章
相关作者
相关机构
AI问答- 技术支持由北京北大方正电子有限公司提供 京ICP备09064830号-19
京公网安备11010802024621 - 本系统建议在Chrome、 IE9+ 以上版本浏览器阅读本站内容，360浏览器请切换至极速模式
- Cookies帮助我们提供服务并提供个性化体验。使用本网站，即表示您同意我们使用Cookies