基于特征椭球和支持向量机复合映射的暂态稳定预测模型  

A Transient Stability Prediction Model Based on CELL-SVM Composite Projection

作　　者：杨跃[1] 刘俊勇[1] 黄媛[1] 刘友波[1] 刘挺坚 魏巍[2] 

机构地区：[1]四川大学电气信息学院 [2]四川电力科学研究院

出　　处：《电网技术》2017年第8期2514-2520,共7页Power System Technology

基　　金：国家自然科学基金资助项目(51437003;51207098)~~

摘　　要：针对电力系统暂态稳定预测问题,提出基于特征椭球理论(characteristic ellipsoid,CELL)和支持向量机(support vector machine,SVM)复合映射的方法。通过采集系统受扰后的同步相量测量单元(phasor measurement unit,PMU)量测数据,将其映射至多维样本空间中生成特征椭球。椭球形态的变化能反映出受扰系统在稳定与失稳场景下不同的动态行为,从而识别出系统的暂态稳定性。以包含了系统暂态过程信息的椭球几何属性,如体积、偏心率、中心点、体积变化率等,构造支持向量机评估模型的输入特征,训练所得支持向量机模型能够实现电力系统暂态稳定在线预测。该方法以椭球的动态变化作为暂态稳定判据,利用较少的量测数据就能为暂态稳定识别提供足够的信息量,同时保证较高的准确率。通过IEEE 39节点系统和我国西南某省级区域输电系统的算例测试,验证了方法的可行性和有效性。A combined Characteristic Ellipsoid （CELL） and Support Vector Machine （SVM） method based on composite projection is proposed to build transient stability prediction model. Post-fault measurement data from phasor measurement unit （PMU） are projected on multi-dimensional space to form the CELL. Dynamic behaviors after a disturbance for both stable and unstable situations are identified from CELL shape variation. CELL geometrical properties such as volume, eccentricity, center and volume change rate, containing transient process information, are selected to build SVM input feature set to eventually accomplish transient stability prediction with high accuracy. The proposed method takes CELL dynamic variation as transient stability criterion with less measurement data sufficient for information supply of transient stability identification. Case study on IEEE 39-bus system and regional transmission system demonstrate feasibility of the proposed method.

关 键 词：暂态稳定识别 特征椭球 支持向量机 特征选择 

分 类 号：TM71[电气工程—电力系统及自动化]