基于数据挖掘的异常用电行为诊断方法研究

Research on diagnosis method of abnormal power consumption behavior based on data mining

针对传统诊断方法过于依赖人工,而导致诊断误差率高的问题,提出了基于数据挖掘的异常用电行为诊断方法。依据数据挖掘过程,分析用电异常行为欠电流、欠电压、扩差、移相特征。利用相似性度量划分数据集,初始化模糊加权指数,初始化隶属度矩阵并更新,计算目标函数,根据计算结果,确定数据对象种类,由此设计异常用电嫌疑用户筛选流程。采用k-means聚类算法,设计诊断分析流程,并研究对窃电和漏电等异常用电行为的闭环诊断机制。由实验结果可知,该方法最高误差率为10%,最低为0,具有良好的诊断效果。

Aiming at the problem that traditional diagnosis methods rely too much on manual labor,leading to high diagnosis error rate,a data mining based diagnosis method for abnormal electricity consumption behavior is proposed.According to the data mining process,the characteristics of abnormal electricity consumption,undercurrent,undervoltage,spreading,and phase shifting are analyzed.The data set is divided by the similarity measure,the fuzzy weighting index is initialized,the membership degree matrix is initialized and updated,the objective function is calculated,and the data object type is determined according to the calculation result,thereby designing a screening process for suspected users of abnormal electricity consumption.Using kmeans clustering algorithm,design the diagnosis analysis process,and study the closedloop diagnosis mechanism for abnormal electricity consumption behaviors such as electricity theft and leakage.It can be seen from the experimental results that the highest error rate of this method is 10%,and the lowest is 0,which has a good diagnostic effect.