基于改进聚类算法的电能计量表故障检测方法

Fault detection method of electric energy meter based on improved clustering algorithm

传统的电能计量装置故障监测方式为周期性现场校验,该方式存在着运行管理粗放不规范、工作效率低、故障发现及排查难度大、监测时效性差等问题。为了解决电能计量故障检测准确性低的问题,提出基于改进聚类算法的电能计量表故障检测方法。该方法采用量子机制改进后的聚类算法,结合自回归积分滑动平均算法清洗数据中异常值并分类数据,将分类数据输入贝叶斯B样条故障检测算法,通过计算故障率完成电能计量表故障检测。实验结果显示,该方法的AUC面积最大,表明检测精度高,并且检出故障点数量与实际一致,检测时间在5 s以内,说明该方法提高了检测准确性的同时,提高了检测效率。

The traditional fault monitoring method for electric energy metering devices is periodic on-site verification,which has problems such as extensive and non-standard operation management,low work efficiency,difficulty in fault detection and troubleshooting,and poor monitoring timeliness.In order to solve the problem of low accuracy in fault detection of electric energy metering,an improved clustering algorithm based fault detection method for electric energy metering meters is proposed.This method adopts a clustering algorithm improved by quantum mechanics,combined with the autoregressive integral moving average algorithm to clean the outliers in the data and classify the data.The classified data is input into the Bayesian B-spline fault detection algorithm,and the fault detection of the electric energy meter is completed by calculating the fault rate.The experimental results show that the AUC area of this method is the largest,indicating high detection accuracy,and the number of detected fault points is consistent with the actual situation.The detection time is within 5 seconds,indicating that this method improves both detection accuracy and detection efficiency.