基于回归分析和SVM的电压稳定在线监视

Online Monitoring of Voltage Stability Based on Regression Analysis and Support Vecto Machine

提出了一种基于支持向量机(SVM)的电压稳定在线监视模型,模型中采用多元回归分析来解决输入变量的选择问题。该模型首先将输出变量进行对数变换,利用逐步回归法建立最优回归方程以实现特征选择,然后将被选择的输入变量作为支持向量机的输入进行训练预测。被选择的输入变量较少,但包含了原来全部输入变量的大部分信息,可保证模型具有较高的预测精度,又避免了过多的输入引起的问题。实例研究结果验证了该方法的有效性。

An online voltage stability monitoring model based on SVM （Support Vector Machine） is presented by using multivariate regression analysis to select input variables. Logarithmic transformation of the output variable is conducted, and then a optimal regres- sion equation is established by means of stepwise regression method to select features. The selected input variables with low dimension and most information included in the original input variables are used as inputs of SVM model for training and prediction. Hence the higher forecast precision is guaranteed, and the problems induced by over-input are avoided. The results of case study show the validity of the proposed approach.