Neural Network Training Method for Power System Stability Assessment Considering Distribution Shifts

LU Liangyuchen; ZHOU Yanzhen; ZENG Hongtai; DU Mingqiu; GUO Qinglai

doi:10.20204/j.sp.2025.10001

English Version Reader LoginJournal Login

您当前的位置：

首页 >

文章列表页 >

Neural Network Training Method for Power System Stability Assessment Considering Distribution Shifts

更新时间：2025-12-19

- Neural Network Training Method for Power System Stability Assessment Considering Distribution Shifts
- Vol. 53, Issue 10, Pages: 1-7(2025)
- 作者机构：
  
  清华大学电机工程与应用电子技术系,北京,100084
- 作者简介：
- 基金信息：
- DOI：10.20204/j.sp.2025.10001
  CLC：
- Published：2025
- 稿件说明：
移动端阅览
卢梁宇宸,周艳真,曾泓泰,杜明秋,郭庆来.考虑分布偏移的电力系统稳定评估神经网络训练方法[J].智慧电力,2025,53(10):1-7.

LU Liangyuchen, ZHOU Yanzhen, ZENG Hongtai, et al. Neural Network Training Method for Power System Stability Assessment Considering Distribution Shifts[J]. 2025, 53(10): 1-7.
卢梁宇宸,周艳真,曾泓泰,杜明秋,郭庆来.考虑分布偏移的电力系统稳定评估神经网络训练方法[J].智慧电力,2025,53(10):1-7. DOI： 10.20204/j.sp.2025.10001.

LU Liangyuchen, ZHOU Yanzhen, ZENG Hongtai, et al. Neural Network Training Method for Power System Stability Assessment Considering Distribution Shifts[J]. 2025, 53(10): 1-7. DOI： 10.20204/j.sp.2025.10001.

摘要

神经网络虽在电力系统稳定评估任务中展现出高效建模与推理能力，但其对输入数据的分布偏移较为敏感，在扰动场景下易出现性能下降。针对现有基于神经网络的稳定评估方法在分布偏移下鲁棒性不足的问题，提出一种基于分布鲁棒优化（DRO）的神经网络训练方法。该方法利用输运成本不确定集，将原问题转化为带Lipschitz正则项的经验风险最小化问题，并提出一种输入梯度正则化的训练方法，能够在不生成对抗样本的情况下通过标准反向传播流程高效实现。算例分析表明，所提方法能够显著降低稳定评估神经网络的Lipschitz常数上界，有效提升了不同场景下的评估准确性。

Abstract

Although neural networks demonstrate efficient modeling and reasoning capabilities in power system stability assessment tasks，they are highly sensitive to distribution shifts in input data and prone to performance degradation under disturbance scenarios. To address the insufficient robustness of existing neural network-based stability assessment methods under distribution shifts，this paper proposes a training approach based on distributionally robust optimization （DRO）. By employing a transport cost uncertainty set，the original problem is transformed into an empirical risk minimization problem with a Lipschitz regularization term. Furthermore，an input gradient regularization training method is introduced，which can be efficiently implemented through standard backpropagation without generating adversarial samples. Case study results show that the proposed method significantly reduces the upper bound of the Lipschitz constant of the stability assessment neural network and effectively improves assessment accuracy across various scenarios.

关键词

Keywords

references

SUN H，GUO Q，SHEN X，et al. Energy internet： redefinition and categories [J ] .Energy Internet，2024，1（1）： 3-8.

PANCIATICI P，BAREUX G，WEHENKEL L. Operating in the fog： security management under uncertainty [J ] .IEEE Power and Energy Magazine，2012，10（5）： 40-49.

杨博，陈义军，姚伟，等. 基于新一代人工智能技术的电力系统稳定评估与决策综述 [J ] .电力系统自动化，2022，46（22）：200-223.

王铮澄，周艳真，郭庆来，等. 考虑电力系统拓扑变化的消息传递图神经网络暂态稳定评估 [J ] .中国电机工程学报，2021，41（7）：2341-2350.

郭梦轩，管霖，苏寅生，等. 基于改进边图卷积网络的电力系统小干扰稳定评估模型 [J ] .电网技术，2022，46（6）：2095-2103.

刘颂凯，党喜，崔梓琪，等. 针对样本类不平衡的深度残差网络电力系统暂态稳定评估方法 [J ] .智慧电力，2024，52（1）：116-123.

刘建锋，姚晨曦，陈乐乐. 基于门控时空图神经网络的电力系统暂态稳定评估 [J ] .电力科学与技术学报，2023，38（2）：214-223.

CHEN Y，TAN Y，DEKA D. Is machine learning in power systems vulnerable? [C ] //2018 IEEE international conferen ce on communications，control，and computing technologies for Smart Grids （Smart Grid Comm）.Aalborg.Denmark，2018： 1-6.

ZHANG Z Y，LIU M X，SUN M Y，et al. Vulnerability of machine learning approaches applied in IoT-based smart grid：a review [J ] . IEEE Internet of Things Journal，2024，11（11）：18951-18975.

GOODFELLOW I J，SHLENS J，SZEGEGY C. Explaining and harnessing adversarial examples [C ] //International Conference on Learning Representations （ICLR 2015）.San Diego，USA，2015.

MADRY A，MAKELOV A，SCHMIDT L，et al. Towards deep learning models resistant to adversarial attacks [C ] //International Conference on Learning Representations （ICLR 2018）.Vancouver，Canada，2018.

CHEN Y，TAN Y，ZHANG B. Exploiting vulnerabilities of load forecasting through adversarial attacks [C ] //Proceedings of the tenth ACM international conference on future energy systems.Phoenix，United States，2019： 1-11.

ROSS A，DOSHI-VELEZ F. Improving the adversarial robustness and interpretability of deep neural networks by regularizing their input gradients [C ] //Proceedings of the AAAI conference on artificial intelligence，New Orleans，United States，2018，32（1）.

SHALEV-SHWARTZ S，BEN-DAVID S. Understanding machine learning： from theory to algorithms [M ] .Cambridge： Cambridge University Press，2014.

ESFAHANI P M，KUHN D. Data-driven distributionally robust optimization using the Wasserstein metric [J ] .Mathematical Programming，2018，171（1）： 115-166.

CRANKO Z，SHI Z，ZHANG X，et al. Generalised lipschitz regularisation equals distributional robustness [C ] //International Conference on Machine Learning.PMLR，Vienna，Austria.2021： 2178-2188.

SCAMAN K，VIRMAUX A. Lipschitz regularity of deep neural networks： analysis and efficient estimation [C ] //Proceedings of the 32nd International Conference on Neural Information Processing Systems （NIPS'18）.Red Hook，USA，2018： 3839-3848.

VERSHYNIN R. High-dimensional probability： an introduction with applications in data science [M ] .Cambridge： Cambridge University Press，2018.

韩英铎，陆超，宋文超，等. 电力系统在线安全稳定计算分析技术综述与展望 [J ] .中国电机工程学报，2024，44（17）：6733-6761.

陈晓华，王志平，吴杰康，等. 基于IHHO-SVM的电能质量扰动信号识别方法 [J ] .浙江电力，2023，42（8）：115-124.

KUNDUR P，PASERBA J，AJJARAPU V，et al. Definition and classification of power system stability IEEE /CIGRE joint task force on stability terms and definitions[J ] .IEEE Transactions on Power Systems，2004，19（3）： 1387-1401.

SCHÄFER B，GRABOW C，AUER S，et al. Taming instabilities in power grid networks by decentralized control [J ] .The European Physical Journal Special Topics，2016，225（3）： 569-582.

KINGMA D P，BA J. Adam： a method for stochastic optimization [C ] //Proceedings of the 3rd International Conference on Learning Representations （ICLR 2015）.San Diego，USA，2015.

CHIAM D，LIM K H，LAW K H. Multi-level signal decomposition for power quality disturbance classification [C ] //MATEC Web of Conferences，Miri，Sarawak，Malaysia，2023，377： 1021.

WU Z X，ZHOU X X. Power system analysis software package （PSASP）-an integrated power system analysis tool [C ] //POWERCON'98.1998 International Conference on Power System Technology.Beijing，China，1998，1： 7-11.

AMJADY N，MAJEDI S F. Transient stability prediction by a hybrid intelligent system [J ] .IEEE Transactions on Power Systems，2007，22（3）： 1275-1283.

RESHEF D N，RESHEF Y A，FINUCANE H K，et al. Detecting novel associations in large data sets [J ] .Science，2011，334（6062）： 1518-1524.

SZEGEDY C，ZAREMBA W，SUTSKEVER I，et al. Intriguing properties of neural networks [C ] //Proceedings of the 2nd International Conference on Learning Representations （ICLR 2014）.Banff，Canada，2014.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Ultra-short-term Wind Power Multi-step Forecasting Based on DT-KAN-iTransformer

Related Author

YAN Renwu

WU Huimin

LI Peiqiang

Related Institution

School of Electronic，Electrical Engineering and Physics，Fujian University of Technology

College of Electrical and Information Engineering，Hunan University

AI问答

⁰