基于改进多分类器的用户电表采集数据修复方法

Restoration Method for User Electricity Meter Collection Data Based on Improved Multiple Classifiers

用户电表位于配电网末端,是开展新型电力系统新兴业务的关键环节。受电表故障、信道噪声等因素影响,用户电表采集数据存在缺失、错误等异常情况,进而影响配电台区“源网荷储”控制的准确性。为解决传统用户电表采集数据修复方法中存在的时序变化规律挖掘不足、异常值修复误差大的问题,提出一种基于改进多分类器的用户电表采集数据修复方法,从而改进多分类器的结构,提取异常数据中的完整区块进行多分类器模型训练,并对用户电表采集数据进行分类。在此基础上,通过变分自编码器学习分类数据的真实变化规律,采用分类集合方式生成修复数据。最后,以某小区用户电表为例进行仿真,得出在异常数据为60%情况下的修复误差为2.8%。该结果表明,所提方法与长短期记忆网络、生成对抗网络相比,具有更好的异常数据修复效果。

The user electricity meters are located at the end of the distribution network,which are the key link to develop the emerging business of Energy Internet.Due to the influence of electricity meter failure,channel noise and other factors,there are some abnormal situations such as missing and error in the data collected by the user electricity meter,which further affects the control accuracy of the“source,grid,load and storage”in the distribution station area.In order to solve the problems of insufficient mining of time sequence change rule and large error of anomaly repair in traditional restoration methods for user electricity meter collection data,a restoration method for user electricity meter collection data based on improved multiple classifiers is proposed to improve the structure of multiple classifiers,extract the intact block of abnormal data for multi-classifier model training,and classify the user electricity meter collection data.On this basis,a variational autoencoder is used to learn the real change rule of classified data,and a set classification method is used to generate the restoration data.Finally,taking the user electricity meters of a community as a simulation case,the restoration error is 2.8%when the abnormal data is 60%.The results show that the proposed method has better abnormal data restoration effect than the long short-term memory network and the generative adversarial network.