基于记分准则的特征属性选择及其在静态电压稳定分析中的应用  

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

作　　者：丁涛[1] 王雨[1] 顾伟[1] 万秋兰[1] 

机构地区：[1]东南大学电气工程学院,江苏南京210096

出　　处：《电力自动化设备》2012年第10期132-137,共6页Electric Power Automation Equipment

基　　金：国家自然科学基金资助项目(50907008;60974036)~~

摘　　要：在使用记分准则对特征属性进行初步降维后,采用改进的主成分分析法对降维后的属性进行分类,分别提取主成分,把各类主成分合并起来作为支持向量机(SVM)的训练输入。以节点电压和支路损耗为属性,得到静态电压稳定的分类器。对IEEE 14节点和IEEE 300节点系统进行仿真分析,结果表明3种记分准则均能有效剔除对分类影响较小的属性,虽然分类属性比综合属性得到的主成分多,但相对海量属性已大幅降低,所提方法能提高准确度,节约内存。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.

关 键 词：数据挖掘 电压稳定 主成分分析 记分准则 支持向量机 稳定性 

分 类 号：TM712[电气工程—电力系统及自动化]