基于Fisher Score特征选择的电力系统暂态稳定评估方法

针对不同电气输入特征与电力系统暂态稳定关联程度不同以及当输入特征受到干扰时评估准确率明显下降的问题，提出一种基于Fisher Score特征选择的电力系统暂态稳定评估方法。设计一种面向电力系统暂态稳定评估二分类问题的样本特征Fisher Score值计算方案；通过Fisher Score值排序有效区分重要特征与冗余特征、噪声特征与非噪声特征；将选择的电气特征输入不同机器学习模型中进行训练和评估。新英格兰39节点系统和IEEE 145节点系统的仿真结果表明，所提特征选择方案能有效筛选电力系统暂态稳定评估中重要度高的特征，提升了评估模型的预测性能。

Transient stability assessment method for power system based on Fisher Score feature selection

In order to solve the problems of different correlation degrees between different electrical input features and transient stability of power system and the obvious decrease of assessment accuracy rate when the input features are interfered, a transient stability assessment method for power system based on Fisher Score feature selection is proposed. A Fisher Score value calculation scheme of sample features for transient stability assessment binary classification problem of power system is designed. Fisher Score value ranking is used to effectively distinguish important features from redundant features, noise features from non-noise features. The selected electrical features are input into different machine learning models for training and assessment. The simulative results of New England 39-bus system and IEEE 145-bus system show that the proposed feature selection scheme can effectively screen out the features of high importance in the transient stability assessment of power system, and improve the prediction performance of the assessment model.