基于改进的模糊C均值聚类的负荷预处理  

Data processing based on improved fuzzy C-means clustering

作　　者：常鲜戎[1] 孙景文[1] 

机构地区：[1]华北电力大学电气与电子工程学院,河北保定071003

出　　处：《华北电力大学学报（自然科学版）》2014年第1期27-32,共6页Journal of North China Electric Power University：Natural Science Edition

摘　　要：SCADA系统采集的电力负荷数据由于各种原因,会存在一些坏数据。在负荷预测中,必须仔细而合理地对历史负荷数据进行处理。电力负荷预处理应该充分考虑负荷曲线本身的特征,即平滑性和相似性。模糊C均值算法可以较好地进行聚类,但是其由于存在聚类数和初始聚类中心未知的问题,因此提出改进的模糊C均值算法——IFCM,即使用改进K均值算法确定初始聚类中心,引入粒度原理确定最佳聚类数。首先采用IFCM对日负荷曲线进行聚类,产生各类特征曲线;然后计算每个时刻点的方差,根据3σ法则进行坏数据的辨识;最后利用特征曲线对坏数据进行修正。针对四川某电网的实际电力负荷进行分析,表明了模型的实用性。There is a number of bad data in the load database from SCADA, thus the data must be cleaned before it is used to forecast electric load or perform power system analysis. A new method is presented to identify outliers in load data by fully utilizing the features of electrical load curves. Fuzzy C-means clustering algorithm can cluster effectively. But because of the existence of the clustering number and the initial clustering center is unknown, so the fuzzy C- means algorithm （IFCM） is improved, which is used to determine the initial cluster center by improved K-means algo- rithm, and determine the optimal cluster number by the principle of granularity. In this paper, IFCM is first used for clustering on the daily load curve, to produce all kinds of characteristic curve; and then calculates the variance for each time point according to the 3~ rule to identify bad data. Finally thr paper adjusts the bad data by using a charac- teristic curve. By analyzing a certain power grid in Sichuan province the practicability of the model is tested.

关 键 词：负荷预处理 模糊聚类 粒度原理 K均值算法 

分 类 号：TM715[电气工程—电力系统及自动化]