Characteristic attribute selection based on feature score criterion and application in voltage stability analysis

After the preliminary dimension reduction of attributes with the FSC(Feature Score Criterion), the improved principal component analysis is applied to classify the dimension-reduction attributes and pick up the principal components, which are then merged as the input of SVM(Support Vector Machine). The node voltage and branch loss are taken as the attributes to obtain the classifier of static voltage stability. Simulative results of IEEE 14-bus and IEEE 300-bus systems show that, each of three FSC kinds can effectively winkle out the attributes with less affection on classification. Although the principal components of classification property are more than those of comprehensive attribute, the massive attributes are significantly reduced. The proposed method improves accuracy and saves memory.