电力系统暂态稳定评估的特征选取  

Feature selection for transient stability evaluation in power system

作　　者：向丽萍[1] 王晓茹[1] 王晓红[1] 

机构地区：[1]西南交通大学电气工程学院,四川成都610031

出　　处：《继电器》2008年第6期26-31,共6页Relay

基　　金：国家973重点基础研究发展规划资助项目(G1998010301)

摘　　要：在对IEEE16机系统采用基于支持向量机的暂态稳定分类的特征选择的基础上,建立了IEEE50机453节点的暂态稳定分类初始特征样本集;同样采用基于主成分和遗传算法的方法对IEEE测试系统进行特征选择,通过主成分分析得到32个综合特征;运用遗传算法选取类内类间距离最大的一组综合特征进行分析,选出特征子集;用SVM对所选的特征进行测试,达到较高的预测率;同时,分析所选出的特征子集,比较16机系统与50机系统的异同,使模型更具有泛化性。Feature selection based on SVM is used for IEEE16-machine-system in this paper, which is the exponent of transient stability classifying. Furthermore, the primary feature pattern of IEEE50-machine 453-bus system is built. Principal component analysis（PCA） and genetic algorithm （GA） are used to efficiently reduce the dimension of the primary feature. SVM test result shows that the selected features possess a wonderful recognition capability. Then the similarities and differences between IEEE16-machine system and IEEE50-machine system are compared and analyzed, which are chosen from subset of primary feature sample. Thus model is more widely applicable.

关 键 词：特征选择 IEEE测试系统 遗传算法 支持向量机 

分 类 号：TM712[电气工程—电力系统及自动化]