基于自适应权重特征聚类的用户用电行为分析

ELECTRICITY CONSUMPTION BEHAVIOR ANALYSIS BASED ON ADAPTIVE WEIGHT FEATURE CLUSTERING

智能电表的推广使得监测数据呈指数级增长,但部分大电力公司在需求侧分析和大数据分析方面通常面临极大的挑战,因此,提出一种自适应权重特征K-均值有限传播(adaptive weighted feature K-means finite propagation,AP)聚类算法来分析用户的用电行为。构造一个综合特征集,即负荷曲线的时域、频域和波动特征。将特征集应用到分布式框架中分析客户行为,在局部建模中,考虑负荷曲线的时域和波动特性,采用自适应K-means算法对负荷曲线进行聚类,并采用客观熵权;在全局建模中,引入AP聚类算法,结合加权时域和频域特征得到聚类结果。通过仿真数据的性能测试和大数据集的用电行为分析,验证了所提模型和方法的有效性。

The popularization of smart meters makes the monitoring data increase exponentially.Therefore,utility companies face challenges in demand side analysis and interpretation of big data.This paper proposes an adaptive weighted feature K-means finite propagation(AP)clustering algorithm to analyze the user's electricity consumption behavior.A comprehensive feature set was constructed,which was the time domain,frequency domain and fluctuation characteristics of load curve.The feature set was applied to divide and conquer framework to analyze customer behavior.In the local modeling,considering the time domain and fluctuation characteristics of load curve,the adaptive k-means algorithm was used to cluster the load curve,and the entropy weight was used to objectively weight the load curve.In the global modeling,AP clustering algorithm was introduced,and the weighted time-domain and frequency-domain features were combined to obtain the clustering results.The effectiveness of the proposed model and method was verified by performance test of simulation data and analysis of power consumption behavior of large data sets.