[1]周 蕾,秦子健,元晓燕,等.基于改进Wasserstein生成对抗网络和迁移学习的电网运行方式生成方法[J].机械与电子,2026,44(09):103-111.
　ZHOU Lei,QIN Zijian,YUAN Xiaoyan,et al.A Power Operation Mode Generation Method Based on Improved Wasserstein Generative Adversarial Network and Transfer Learning[J].Machinery & Electronics,2026,44(09):103-111.
点击复制

基于改进Wasserstein生成对抗网络和迁移学习的电网运行方式生成方法()
分享到：

《机械与电子》[ISSN:1001-2257/CN:52-1052/TH]

卷:
44
期数:
2026年09期
页码:
103-111
栏目:
电力控制
出版日期:
2026-09-25

文章信息/Info

Title:
A Power Operation Mode Generation Method Based on Improved Wasserstein Generative Adversarial Network and Transfer Learning
文章编号:
1001-2257(2026)09-0103-09
作者:
周 蕾; 秦子健; 元晓燕; 纪祥贞; 孟凡敏; 许晓敏
国网山东省电力公司莱芜供电公司,山东 济南 271100
Author(s):
ZHOU Lei; QIN Zijian; YUAN Xiaoyan; JI Xiangzhen; MENG Fanmin; XU Xiaomin
(Laiwu Power Supply Company,State Grid Shandong Electric Power Company,Jinan 271100,China)
关键词:
梯度惩罚; 生成对抗网络; 迁移学习; 电网运行方式
Keywords:
gradient penalty; generative adversarial network; transfer learning; power system operation mode
分类号:
TM732
文献标志码:
A
摘要:
针对传统典型运行方式生成方法难以满足新型电力系统在线稳定分析与调控运行需求的问题,提出一种基于带梯度惩罚的改进Wasserstein生成对抗网络(WGAN)和迁移学习的电网运行方式智能生成方法。首先,引入梯度惩罚自适应调节机制改进WGAN,并基于潮流方程构建了面向系统运行方式的 WGAN 模型。其次,采用迁移学习策略将源电网模型迁移至目标电网并结合小样本数据进行微调,提高电网运行方式的生成能力与泛化性能。最后,在39节点系统上进行验证。结果表明,所提方法能够有效生成满足电网运行约束、覆盖安全裕度的多样化运行方式样本,在样本质量、分布拟合精度及训练稳定性方面均表现优异。
Abstract:
Conventional methods for generating typical operating modes are inadequate for the online stability analysis and dispatch control requirements of modern power systems.To address this limitation,an intelligent generation method for operating conditions based on an improved Wasserstein Generative Adversarial Network (WGAN) with gradient penalty and transfer learning is proposed.First,an adaptive gradient penalty adjustment mechanism is introduced to enhance the WGAN,and a WGAN model oriented to power system operating modes is constructed from the power flow equations.Subsequently,a transfer learning strategy is adopted to migrate the source grid model to the target grid,where fine tuning with a small-sample data is performed to improve the generation capability and generalization performance.Finally,validation on a 39-bus system demonstrates that the proposed method can effectively generate diverse operating mode samples that satisfy grid operational constraints and cover safety margins,exhibiting excellent performance in sample quality,distribution fitting accuracy,and training stability.

参考文献/References:

[1] 王国春,许洪强,冯长有,等.新一代在线安全分析技术架构及未来展望[J].电力系统自动化,2023,47(24):110-120.

[2] 舒印彪,张智刚,郭剑波,等.新能源消纳关键因素分析及解决措施研究[J].中国电机工程学报,2017,37(1):1-9.
[3] 庄颖睿,程林,齐宁,等.基于深度时间聚类的微电网典型场景生成方法[J].电力系统自动化,2023,47(20):95-103.
[4] 兰健,郭庆来,周艳真,等.基于生成对抗网络和模型迁移的电力系统典型运行方式样本生成[J].中国电机工程学报,2022,42(8):2889-2900.
[5] 孙宏斌,黄天恩,郭庆来,等.面向调度决策的智能机器调度员研制与应用[J].电网技术,2020,44(1):1-8.
[6] 徐潇源,王晗,严正,等.能源转型背景下电力系统不确定性及应对方法综述[J].电力系统自动化,2021,45(16):2-13.
[7] 赵书强,索璕,许朝阳,等.考虑断面约束的多能源电力系统时序性生产模拟[J].电力自动化设备,2021,41(7):1-6.
[8] Zeng Hongtai,Zhou Yanzhen,Guo Qinglai,et al.A distributed deep reinforcement learning-based approachf or fast preventive control considering transient stability constraints[J].CSEE Journal of Power and Energy Systems,2023,9(1):197-208.
[9] 马光,张伊宁,陈哲,等.含大规模风电的交直流混联系统风险评估方法[J].电网技术,2019,43(9):3241-3251.
[10] 彭显刚,林利祥,刘艺,等.基于纵横交叉拉丁超立方采样蒙特卡洛模拟法的分布式电源优化配置[J].中国电机工程学报,2015,35(16):4077-4085.
[11] 盛四清,关皓闻,雷业涛,等.基于混沌海鸥优化算法的含光伏发电系统负荷模型参数辨识[J].太阳能学报,2022,43(7):64-72.
[12] 陈宇辛,江岳文.考虑变量相关性的改进场景优化法区间潮流计算[J].电力自动化设备,2022,42(3):161-167.
[13] 兰天楷,孙华东,王琦,等.考虑分布式新能源的有源综合负荷模型[J].电工技术学报,2024,39(23):7365-7378.
[14] 李聪,彭小圣,王皓怀,等.基于SDAE深度学习与多重集成的风电集群短期功率预测[J].高电压技术,2022,48(2):504-512.
[15] 陈宁,李法社,王霜,等.基于深度强化学习算法的分布式光伏EV 互补系统智能调度[J].高电压技术,2025,51(3):1454-1463.
[16] 赵鹏杰,吴俊勇,王燚,等.基于深度强化学习的微电网优化运行策略[J].电力自动化设备,2022,42(11):9-16.
[17] 曹磊,孙日高,常正胜,等.电网故障诊断方法综述及发展趋势[J].化工自动化及仪表,2020,47(6):465-470.
[18] 杨凌霄,朱亚丽.基于概率神经网络的高压断路器故障诊断[J].电力系统保护与控制,2015,43(10):62-67.
[19] 李保罗,孙华东,张恒旭,等.基于两阶段迁移学习的电力系统暂态稳定评估框架[J].电力系统自动化,2022,46(17):176-185.
[20] 杜东来,韩松,荣娜.基于时空图卷积网络和自注意机制的频率稳定性预测[J].电工技术学报,2024,39(16):4985-4995.
[21] 邵振国,张承圣,陈飞雄,等.生成对抗网络及其在电力系统中的应用综述[J].中国电机工程学报,2023,43(3):987-1004.
[22] 谭本东,杨军,赖秋频,等.基于改进CGAN 的电力系统暂态稳定评估样本增强方法[J].电力系统自动化,2019,43(1):149-157.
[23] Lan Jian,Guo Qinglai,Zhou Yanzhen,et al.Generation of large-scale convergent power flow samples through a data-driven approach[C]∥2020 IEEE 4th Conference on Energy Internet and Energy System Integration.New York:IEEE,2020:722-726.
[24] Guan Jinyu,Tang Hao,Wang Jiye,et al.AGAN-based fully model-free learning method for short term scheduling of large power system[J].IEEE Transactions on Power Systems,2022,37(4):2655-2665.
[25] 胡家祥,曹迪,胡维昊,等.融入拓扑知识的图神经网络配电网抗差状态估计方法[J].电力系统自动化,2023,47(10):84-97.
[26] 赵冬梅,王闯,谢家康,等.考虑惯量中心频率偏移的自编码器暂态稳定评估[J].电网技术,2022,46(2):662-670.
[27] 周任军,孙嘉赣,张浩,等.考虑双重市场的含风电电力系统双层随机优化调度.电力系统及其自动化学报,2018,30(1):15-21.
[28] 王守相,陈海文,潘志新,等.采用改进生成式对抗网络的电力系统量测缺失数据重建方法[J].中国电机工程学报,2019,39(1):56-64.
[29] 刘云鹏,许自强,和家慧,等.基于条件式Wasserstein生成对抗网络的电力变压器故障样本增强技术[J].电网技术,2020,44(4):1505-1513.
[30] Ren Chao,Xu Yan.Transfer learning-based power system online dynamic security assessment:using one model to assess many unlearned faults[J].IEEE Transactions on Power Systems,2020,35(1):821-824.
[31] 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):2196-2203.
[32] 赵峰,孙宏斌,张伯明.基于电气分区的输电断面及其自动发现[J].电力系统自动化,2011,35(5):42-46.

相似文献/References:

[1]李语心１,赵艳娜２,谢荣理３,等.面向甲状腺结节良恶分类的cRes-GAN算法[J].机械与电子,2020,(04):6.
　,,et al.A cRes-GAN Algorithm for Classification of Benign or Malignant Thyroid Nodules[J].Machinery & Electronics,2020,(09):6.
[2]潘继财.大数据样本与半监督环境下基于生成对抗网络的故障诊断[J].机械与电子,2021,(05):20.
　PAN Jicai.Large Data Samples and Fault Diagnosis Based on Generation Countermeasure Network in Semi-supervised Environment[J].Machinery & Electronics,2021,(09):20.
[3]杨吴奔,苏立鹏,郑建豹,等.基于 AAC-GAN 的起重机缺陷图像生成方法及其在缺陷分类中的应用[J].机械与电子,2025,(09):67.
　YANG Wuben,SU Lipeng,ZHENG Jianbao,et al.Crane Defect Image Generation Method Based on AAC-GAN and Its Application in Defect Classification[J].Machinery & Electronics,2025,(09):67.

备注/Memo

备注/Memo:
收稿日期:2026-04-14
基金项目:国网山东省电力公司科技项目资助(520612250001)
作者简介:周 蕾 (1992-),女,山东菏泽人,硕士,高级工程师,研究方向为负荷预测、电力市场等。
更新日期/Last Update: 2026-10-13