应用改进粒子群优化的模糊均值聚类算法的暂态稳定机组分群方法  

A Unit Clustering Approach Used in Transient Stability Analysis Based on Improved PSO-FCM Algorithm

作　　者：张园园[1] 龚庆武[1] 陈道君[1] 刘栋[1] 

机构地区：[1]武汉大学电气工程学院,湖北省武汉市430072

出　　处：《电网技术》2011年第9期92-98,共7页Power System Technology

摘　　要：针对模糊C-均值聚类方法(fuzzy C-means,FCM)应用于机组分群时存在易收敛于局部最优值的问题,提出了改进的粒子群优化(particle swarm optimization,PSO)的模糊C-均值聚类算法(PSO-FCM)用于机组分群的方法,并阐述了分群算法中关键参数的选取方法。为充分利用FCM多特征量分析的优点,同时引入了功角、角速度作为分群特征量,提出了利用同调性指标自适应确定分群数目的方法,分群时采用约1个摇摆周期的数据分析。理论分析和仿真结果表明,所提方法能够取得一致、稳定的分群结果,效果优于传统的模糊聚类方法。To remedy the defect of fuzzy C-means （FCM） that it is sensitive to initial value and when it is used to group the generation units the solution is likely to converge on local optimal value, an improved particle swarm optimization （PSO）-FCM algorithm is proposed, and an approach to select the key parameters in the grouping algorithm is expounded. To take full advantage of multi-character analysis of FCM, the angle and angular velocity are simultaneously led in as clustering characters, so the problem of selecting recombination coefficient in recombination angle method can be avoided, and a method to adaptively determine the clustering number by coherent norm, which utilizes fully fuzzy partition matrix and enhances the flexibility of grouping, is proposed. Trajectories data around a swing cycle are analyzed for grouping, which can track the dynamic changes of generator grouping. Theoretical analysis and simulations results show that the proposed algorithm can achieve consistent and stable grouping result that is better than the traditional fuzzy clustering algorithm.

关 键 词：粒子群优化 模糊C.均值聚类 机组动态分群 同调性指标 

分 类 号：TM712[电气工程—电力系统及自动化]