首页|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

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

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
  • 维普
系统的量测数据可能受到噪声以及样本类分布不平衡问题的影响,导致基于数据驱动的暂态稳定评估模型性能下降.提出一种针对样本类不平衡的的深度残差网络电力系统暂态稳定评估方法.首先,利用改进过采样技术为滤除噪声的少数类样本构造所需的新样本,改善样本类不平衡问题,并减少噪声的影响;然后,基于深度残差网络构建电力系统暂态稳定评估模型,解决梯度消失导致的模型性能退化问题,提高模型的鲁棒性和准确性;最后,在新英格兰10机39节点和47机140节点系统上的仿真结果表明,所提方法能减小噪声干扰、降低不平衡数据集所带来的影响和减少计算复杂度.
Transient Stability Evaluation Method for Power Systems with Deep Residual Network Considering Class Imbalance of Samples
System measurement data may be affected by noise problems and the class imbalance distribution of samples,resulting in the performance degradation of a data-driven transient stability evaluation model.Therefore,the paper proposes a transient stability evaluation method for the power system with deep residual network focusing on the class imbalance of the samples.Firstly,the improved oversampling technique is used to construct new samples for the minority class samples filtering out noise,improving the class imbalance problem and reducing the influence of the noise.Then the transient stability evaluation model for the power system is built based on the deep residual network to solve the performance degradation caused by the disappearance of gradient,and improve the robustness and accuracy of the model.Finally,the simulation results on the New England 10-machine 39-bus system and 47-machine 140-bus system show that the proposed method can reduce the noise interference,mitigate the impact of unbalanced data sets,and simplify computation complexity.

transient stability evaluationnoise problemclass imbalance distribution of sampleimproved synthetic minority oversampling techniquedeep residual network

刘颂凯、党喜、崔梓琪、杨超、阮肇华、袁铭洋

展开 >

三峡大学电气与新能源学院,湖北宜昌 443002

新能源微电网湖北省协同创新中心,湖北宜昌 443002

国网福建宁德供电公司,福建宁德 352100

暂态稳定评估 噪声问题 样本类分布不平衡 改进合成少数过采样技术 深度残差网络

国家自然科学基金湖北省自然科学基金

6223300662233006

2024

智慧电力
陕西省电力公司

智慧电力

CSTPCD北大核心
影响因子：0.831
ISSN：1673-7598
年,卷(期)：2024.52(1)
  • 18