摘要
设计同步相量测量条件下,面向发电机节点电压同步相量复平面轨迹规则的暂稳实时评估算法。利用K均值和模糊C均值组合算法在线识别扰动后若干周波内轨迹簇同调与聚类偏移特征,采用分类与回归树抽取所定义特征量规则,以实时轨迹簇的相对位置组合特征值规则而非轨迹状态(如幅值大小)本身来评估未来暂稳状态。实践表明,算法能在8～20周波内不依赖模型地高精度预估未来4 s系统是否稳定及临界机、受扰程度等重要信息,且不依赖于既定扰动模式,侧重于系统对大扰动的泛化响应模式挖掘,回避了主导失稳模式判别,且避免了针对不同故障的多套规则存储。新英格兰10机39节点系统算例验证了方法的有效性。
With synchronized phasor measurement, a real-time assessment algorithm for transient stability based on phasor trajectory clusters rules is presented in this paper. A novel combined algorithm, which was composed of K-mean and fuzzy C-mean methods, was applied on line io recognize trajectories＇ coherency state and deviation features of clusters＇ appearance within several cycles after disturbance. A standard classification and regression tree （CART） method was proposed to abstract knowledge rules from the evolution of pro-classified feature index. Such a group of index of real-time trajectory clusters＇ position, rather than trajectories themselves （amplitude for instance）, was utilized to assess transient stability and disturbance severity level. Detailed analysis shows that the proposed algorithm can predict transient stability accurately with regard to the future 4 s by using data within 8-20 cycles without dependence on power system model or specific nature of the disturbance. Significant information such as the stability state, critical machine, severity level are detected at the same time. In addition the whole process mainly focuses on generalized response pattern recognition of power system, and therefore avoids having to directly determine the controlling unstable mode and different rule sets. New England benchmark case verifies the validity of the method.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2011年第16期32-39,共8页
Proceedings of the CSEE
基金
国家自然科学基金项目(50977059)
教育部国家建设高水平大学公派研究生项目([2009]3012)~~
