基于大数据分析的电力运行数据异常检测示警方法

Alarm method of power operation data anomaly detection based on big data analysis

传统的数据检测方法易受电力系统环境变化的影响,难以对加时窗数据进行检测,降低检测准确率。为此,该文利用大数据分析技术估计电力运行数据的最大似然值等信息,设计新的电力运行数据异常检测示警方法,从根本上提高检测准确率。根据电力运行数据异常检测示警原理,对不同时窗中的子序列进行聚类处理,确定每个时窗中的异常数据;通过提取单数据、多数据特征量,用转移概率序列表示电力运行数据动态变化情况,在完成正常数据与异常数据间模糊特征聚类的基础上,采用大数据分析方法计算电力运行均值和方差,完成最大似然值估计,通过似然比建立异常情况检测与示警流程,对数据异常情况进行检测示警。实验结果表明,在油温和环境温度变化的情况下,所提方法的检测准确率较高,且示警过程耗时较少,证明该方法整体有效性较高。

The traditional data detection method is easily affected by the change of power system environment,so it is difficult to detect the data of the overtime window,which reduces the detection accuracy.Therefore,this paper uses big data analysis technology to estimate the maximum likelihood value and other information of power operation data,and designs a new power operation data anomaly detection warning method to fundamentally improve the detection accuracy rate.According to the warning principle of abnormal detection of power operation data,the sub-sequences in different windows are processed by clustering,and the abnormal data in each window is determined.By extracting single data,characteristic data,with the transition probability sequence dynamic change of electric power operation data,fuzzy characteristics between normal data and abnormal data clustering,on the basis of the big data analysis method to calculate power running mean and variance,complete the maximum likelihood estimation,build anomalies by likelihood ratio detection and warning process,warning for detecting data anomalies.The experimental results show that the proposed method has higher detection accuracy and less warning time under the condition of oil temperature change.