基于决策树和数据驱动的零电量用户筛选方法

None-consumption users filtering algorithm based on decision tree and data-driven methods

随着电力系统中用电客户的增多及客户用电形式的多样化,零电量用户(none-consumption user,NCU)逐渐增多,对零电量用户进行排查耗费了电网公司大量的人力物力。在此背景下,基于电力用户用电信息采集系统(electricity information acquisition system,EIAS)的零电量用户信息,提出了一种零电量用户筛选的数据驱动算法,判断正常零电量用户和异常零电量用户。采用决策树对电力用户用电信息采集系统数据进行分析,确定零电量用户异常类型;对决策树无法辨别的用户类型,通过分析零电量用户计量采集数据和营销数据,提取适用于零电量用户筛选的关键因子,进而构建零电量用户筛选评价体系;在此基础上,采用(criteria importance though intercrieria correlation,CRITIC)法确定关键因子的权重,并采用雷达图法对零电量用户进行筛选分类。以浙江省某供电所管辖下的零电量用户为例对所提出的方法进行说明,并通过现场排查进行校验,结果表明所提出的零电量用户筛选方法具有一定的有效性。

With the increasing of power consumers and diversification of power consumption in power systems,the number of noneconsumption users(NCUs)was also increasing rapidly.Thus,lots of manpower and material resources of power supply companies were arranged to perform troubleshooting on NCUs.Given this background,a data-driven method based on the electricity information of NCUs collected by electricity information acquisition system(EIAS)was proposed to determine the filtering results of normal NCUs and abnormal NCUs.The decision tree was utilized to analyze the electricity data of NCUs,and determine the types of NCUs.The key factors suitable for NCUs filtering were determined based on the original data to filter the NCUs that could not be screened by the decision tree,and the evaluation system for NCUs filtering was constructed.On this basis,CRITIC and radar chart methods were adopted to determine the weights of the key factors and to determine the filtering results of NCUs,respectively.The NCUs power-supplied by an actual power supply station in Zhejiang Province were served for demonstrating the proposed algorithm of NCUs filtering,and the simulation and on-site inspection results showed that the proposed data-driven method was effective for screening out the abnormal NCUs.