摘要
传统的电力能源大数据异常修正方法存在搜索次数过多问题.会造成异常数据辨识结果异常、修正结果不准确。为此,引入低秩模型.改善以上问题。采用低秩模型处理电力能源数据样本.去除样本数据噪声;在离线模式下,通过训练支持向量机对数据样本进行聚类;在联机模式下,利用滑动窗口辨识异常数据;针对单个或多个不相关的异常数据.依据基尔霍夫电流定律完成修正。试验结果表明,与以往的大数据修正方法相比,设计的基于低秩模型的电力能源大数据异常修正方法残差值更低,并且电力负荷修正后.与实际负荷相符。
The traditional correction method for anomaly big data brings out the problem of too many search times,which makes the identification result of abnormal data being iregular and the correction result being inaccurate.Therefore,the low rank model is introduced to solve above problems.Low-rank model is used to process power energy data samples for removing sample data noise.Cluster data samples by training support vector machines in offline mode are conducted.In online mode,the sliding window is used to identify abnormal data for single or multiple unrelated abnormal data.The correction is completed according to Kirchhoff's current law.The experimental resuls show that,compared with the previous big data correction methods,the residual value of the power energy big data anomaly correction method based on the low rank model is lower,and the power load correction is consistent with the actual load.
作者
马草原
MA Caoyuan(State Grid Tianjin Electric Power Company,Tianjin 300000,China)
出处
《自动化仪表》
CAS
2021年第3期90-93,97,共5页
Process Automation Instrumentation
关键词
低秩模型
电力能源
大数据异常
修正方法
支持向量机
电力负荷
基尔霍夫电流定律
聚类
Low rank model
Power supply
Big data exception
Correction method
Support vector machine
Electric load
Kirhhoff's current law
Clusterin
