基于特征指标降维及熵权法的日负荷曲线聚类方法

Daily Load Curve Clustering Method Based on Feature Index Dimension Reduction and Entropy Weight Method

日负荷曲线聚类是负荷建模背景下分析负荷特性的基础。针对现有聚类方法在聚类质量、聚类效率等方面的不足,综合运用模糊C均值及熵权法原理提出一种基于特征指标降维及熵权法的日负荷曲线聚类方法。首先提取日负荷率、日峰谷差率、日最大利用时间等7类降维特征指标替代各采样点负荷数据作为聚类输入;其次,引入熵权法自适应配置各特征指标的权重系数;最后,采用特征加权的模糊C均值聚类算法对用电日负荷曲线进行聚类。采用所提方法对某地区日负荷曲线进行聚类分析,算例结果表明该方法在运行效率、鲁棒性、聚类质量等方面具有一定的优越性,聚类结果能真实有效地反映负荷的实际用电特性。

Daily load curve clustering is the basis of analyzing load characteristics under the background of load modeling. In view of the shortcomings of existing clustering methods in terms of clustering quality and clustering efficiency and based on the principle of fuzzy C-means and entropy weight method, a method of daily load curve clustering based on dimensionality reduction of feature index and entropy weight method is proposed. Firstly, seven kinds of feature indices of dimensionality reduction such as daily load rate, daily peak-to-valley difference rate and daily maximum utilization time are extracted and taken as the clustering input to replace the load data of each sampling point. Secondly, the entropy weight method is introduced to configure the weight coefficient of each feature index adaptively. Finally, the power load curves are clustered with the feature-weighted fuzzy C-means(FW-FCM) clustering algorithm. The daily load curves of a region are clustered using the proposed method. Results show that the method has certain advantages in operating efficiency, robustness, clustering quality and so on. Moreover, the clustering results can reflect actual power consumption characteristics of the load truly and effectively.