基于后验分布信息的SSAE暂态稳定评估模型倾向性修正方法

被引:11
作者
林楠
王怀远
陈启凡
机构
[1] 福州大学电气工程与自动化学院智能配电网装备福建省高校工程研究中心
关键词
深度学习; 暂态稳定评估; 代价敏感; 后验分布信息; 堆叠稀疏自动编码器; 不平衡样本;
D O I
暂无
中图分类号
TM712 [电力系统稳定];
学科分类号
080802 [电力系统及其自动化];
摘要
为了解决样本不平衡带来的评估倾向性问题,从深度学习模型的损失函数出发,分析样本不平衡对评估模型的影响,发现训练过程中的损失函数值能够反映样本的不平衡程度,由此提出基于样本后验分布信息的代价敏感修正方法。通过预先训练获得样本的后验分布信息,引入稳定样本与不稳定样本的损失函数均值比得到修正系数;将修正系数通过代价敏感法修正模型的损失函数,重新对模型进行训练,从而修正模型的评估倾向性。相较于传统方法,该方法从模型的训练机理上量化了样本的不平衡程度,修正系数综合考虑了样本数量与空间分布的不平衡对模型参数的影响,实现了更好的修正效果。IEEE 39节点系统和华东某区域系统的仿真结果验证了所提方法的有效性。
引用
收藏
页码:135 / 141
页数:7
相关论文
共 22 条
[1]
Transient stability assessment model with improved cost-sensitive method based on the fault severity [J].
Wang, Huaiyuan ;
Wang, Qingyin ;
Chen, Qifan .
IET GENERATION TRANSMISSION & DISTRIBUTION, 2020, 14 (20) :4605-4611
[2]
Transfer Learning-Based Power System Online Dynamic Security Assessment: Using One Model to Assess Many Unlearned Faults [J].
Ren, Chao ;
Xu, Yan .
IEEE TRANSACTIONS ON POWER SYSTEMS, 2020, 35 (01) :821-824
[3]
Radial-Based oversampling for noisy imbalanced data classification.[J].Michał Koziarski;Bartosz Krawczyk;Michał Woźniak.Neurocomputing.2019,
[4]
A novel data-driven approach for transient stability prediction of power systems considering the operational variability.[J].Yanzhen Zhou;Qinglai Guo;Hongbin Sun;Zhihong Yu;Junyong Wu;Liangliang Hao.International Journal of Electrical Power and Energy Systems.2019,
[5]
Post-fault prediction of transient instabilities using stacked sparse autoencoder.[J].Mohammed Mahdi;V.M. Istemihan Genc.Electric Power Systems Research.2018,
[6]
The Anatomy of the 2016 South Australia Blackout: A Catastrophic Event in a High Renewable Network [J].
Yan, Ruifeng ;
Nahid-Al-Masood ;
Saha, Tapan Kumar ;
Bai, Feifei ;
Gu, Huajie .
IEEE TRANSACTIONS ON POWER SYSTEMS, 2018, 33 (05) :5374-5388
[7]
Hybrid AC-DC Microgrid: Systematic Evaluation of Control Strategies [J].
Gupta, Ajay ;
Doolla, Suryanarayana ;
Chatterjee, Kishore .
IEEE TRANSACTIONS ON SMART GRID, 2018, 9 (04) :3830-3843
[8]
Intelligent Time-Adaptive Transient Stability Assessment System [J].
Yu, James J. Q. ;
Hill, David J. ;
Lam, Albert Y. S. ;
Gu, Jiatao ;
Li, Victor O. K. .
IEEE TRANSACTIONS ON POWER SYSTEMS, 2018, 33 (01) :1049-1058
[9]
Redundancy-driven modified Tomek-link based undersampling: A solution to class imbalance.[J].Debashree Devi;Saroj kr. Biswas;Biswajit Purkayastha.Pattern Recognition Letters.2016,
[10]
A Review on Ensembles for the Class Imbalance Problem: Bagging-, Boosting-, and Hybrid-Based Approaches [J].
Galar, Mikel ;
Fernandez, Alberto ;
Barrenechea, Edurne ;
Bustince, Humberto ;
Herrera, Francisco .
IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART C-APPLICATIONS AND REVIEWS, 2012, 42 (04) :463-484