基于改进聚类方式的牵引负荷分类方法

为得到更为准确的牵引负荷分类结果，基于大量的牵引负荷实测数据，提出了一种改进后的自适应模糊C均值聚类方法. 该方法能够自动获取最佳聚类数，以馈线电流带电有效系数、最大值、平均值、95%值以及1～5阶样本矩作为聚类指标对实测牵引负荷进行聚类；然后采用非参数核密度估计方法对牵引负荷概率密度函数进行拟合，得到了每一类馈线电流概率分布模型. 结果表明:聚为一类的牵引负荷特征参数相近、概率分布相似. 

Traction Load Classification Method Based on Improved Clustering Method

In order to obtain more accurate traction load classification, based on a large amount of measured traction load data, an improved fuzzy C-means clustering method is proposed, which can automatically obtain the best classification number. A charged effective coefficient, the maximum value, the average value, the value of 95% and one to five order moments were chosen as clustering indicators to classify feeder current. Then the probability density function of traction loads was fitted using non-parametric kernel density estimation, and the probability distribution model of each feeder current type was obtained. The results show that the characteristic parameters and probability distributions of the traction loads that were clustered together are.