基于大数据分析的变频器异常检测方法

变频器设备机理复杂、设备监测数据异常样本少,导致在实际应用中基于经验和基于模型的异常检测方法操作性难、常规的数据驱动方法可行性差。针对上述问题,结合密度峰值聚类与时序运行数据的转移特点,提出一种基于大数据分析的变频器异常检测方法。该方法首先通过密度峰值聚类将运行数据聚类,然后将时序过程数据在不同类簇间的动态变化规律用概率表示,并基于聚类结果和转移概率提出异常检测体系框架,最后通过异常案例验证该方法的有效性,表明该方法能够正确识别变频器异常行为。

An Abnormal Detection Method of Frequency Converter Based on Big Data Analysis

In practical,experience-based and model-based abnormal detection methods are difficult to operate,and conventional data-driven methods are poorly feasible,because inverter equipment mechanism is complicated and data abnormal sample is small.Aiming at the above problems,combined with the density peak clustering and the characteristics of time series operation data transfer,a method of inverter abnormal detection based on big data analysis is proposed.Firstly,the running data is clustered by density peak clustering,then the dynamics of the time series process data between different clusters is expressed by probability,and an anomaly detection system framework is proposed based on clustering results and transition probability.Finally,the effectiveness of the method is verified by abnormal cases,which indicates that the method can correctly identify the abnormal behavior of the inverter.