基于二维组合属性决策树的暂态稳定评估

Transient Stability Assessment Based on 2D Combined Attribute Decision Tree

提出基于二维组合属性决策树的电力系统暂态稳定评估方法。与单属性决策树不同的是,该文采用的二维组合属性决策树自动对二维属性进行线性组合,并采用基于基尼(Gini)指数最小的二维组合属性的分裂准则,开发分裂直线的反向搜索方法,有效提高了决策树拟合多维属性空间非线性分类边界的能力。单机无穷大系统、IEEE 39节点系统和中国某省级电网的算例研究表明,新方法生成的暂态稳定评估决策树的规模更小,精度更高,泛化能力更强,而且可挖掘出电力系统变量间一些重要的物理关系,具有更强的可解释性和应用潜力。

A method for the power system transient stability assessment based on 2D combined attribute decision tree was proposed.Different from single attribute decision trees, the 2D trees combine two attributes as one automatically to improve its classifying ability in case of multi-dimension and nonlinear border among classes.In the proposed 2D tree,Gini coefficient based 2D combined splitting criterions was adopted and search method for splitting line in an inverse direction was developed.Case study was carried out in one-machine infinite bus system,IEEE 39-bus system and a real provincial system. Results show that,2D combined attribute decision trees generated by the proposed method are smaller in structure, have the higher accuracy and generalization ability than that generated by the single attribute decision trees.Meanwhile, some important relationship between physical variables in power systems can be mined by the proposed method. Advantages mentioned above make that the proposed method is potential for real-life power system applications.