电力用户用电数据的异常数据审查和分类

Anomaly Detection and Category of Electrical Utilization Data

在电网的运行过程中,电力自动抄表系统收集的用户用电数据因为受到天气原因,线路故障和系统故障等影响出现偏差和错误。这些偏差和错误是用户用电数据中的异常数据,它们的存在严重影响了电网运行时信息的准确采集和用户用电信息分析。这就要求对用户用电数据进行预处理,在大量的用户用电数据中发现识别出异常数据,进而采用一定的方法对异常数据进行处理和补偿。着眼于自动抄表系统中用户用电数据的数据清理方法研究,对用户用电数据预处理的主要问题进行比较详尽的讨论,提出了用户用电数据预处理的模型和流程方法,采用k近邻法对异常数据进行分类,并利用实际用户用电数据进行实验,得到了初步的成果和经验,对未来智能电网系统中用户用电数据的预处理具有借鉴的作用。

During the grid operation, users＇ consumption data collected by the automatic metric gathering sys- tem may have deviations and errors due to the weather, line failures and system failures. These deviations and errors are anomalies in the data and their presence has seriously affected the accuracy of the information collec- tion and analysis of user consumption. It is urgent to preprocess user consumption data, including identifying anomaly data in the large-scale user consumption data and using certain methods to handle and compensate ab- normal data. This paper focuses on the data cleansing method for user consumption data in an automated met- ric gathering system. It presents a detailed discussion of major problems in user consumption data, builds a model for data preprocessing using k-nearest neighbor method to classify corrupted data and carries experi- ments based on the above methods and model. The preliminary results are presented and concluded, which provides reference value for the future work of user consumption data preprocessing.