摘要
为解决量测数据缺失时电力系统暂态稳定评估模型泛化能力不足的问题,基于多向循环神经网络和XGBoost算法,提出一种针对数据缺失的电力系统暂态稳定评估方法。首先使用多向循环神经网络修复缺失数据;然后采用完整的数据集对XGBoost模型进行训练;最后基于SHAP理论量化不同输入特征对模型输出结果的影响。此外,还提出了一种模型更新机制,在系统工况发生改变时对模型进行持续更新。在新英格兰10机39节点系统上仿真结果表明,所提方法相较于传统方法具有更好的数据修复能力,能显著提高暂态稳定评估性能。
To solve the problem of insufficient generalization capability of the power system transient stability assessment model when measurement data is missing,an assessment method for power system transient stability with missing data is proposed on the basis of multi-directional recurrent neural networks(M-RNN)and the XGBoost algorithm.First,M-RNN is used to repair the missing data.Then,the XGBoost model is trained with a complete data set.Finally,based on the SHAP theory,the influence of different input characteristics on the model output is quantified.In addition,a model update mechanism is proposed to update the model continuously when the system’s operation condition is changed.The simulation results of a New England 10-machine 39-bus system show that compared with the traditional methods,the proposed method has a better data repair capability and can significantly improve the performance of transient stability assessment.
作者
张雅婷
刘颂凯
张磊
刘聪
刘书池
崔梓琪
ZHANG Yating;LIU Songkai;ZHANG Lei;LIU Cong;LIU Shuchi;CUI Ziqi(College of Electrical Engineering and New Energy,China Three Gorges University,Yichang 443002,China)
出处
《电力系统及其自动化学报》
CSCD
北大核心
2023年第3期59-68,共10页
Proceedings of the CSU-EPSA
基金
国家自然科学基金资助项目(52007103)。
关键词
缺失数据
暂态稳定安全
多向循环神经网络
XGBoost算法
估计误差
missing data
transient stability and safety
multi-directional recurrent neural networks(M-RNN)
XG⁃Boost algorithm
estimation error
