基于数据挖掘的电能计量装置异常自动检测方法

Method of Anomaly Automatic Detection for Power Metering Device Based on Data Mining

为维护电能计量装置能够正常运行,提出了基于数据挖掘的电能计量装置运行异常自动检测方法。收集电能计量装置的日用电、历史变量等运行数据,经层次k均值聚类方法对数据进行预处理,并提取数据中的电流、电压、电功率等特征,然后将特征作为卷积神经网络输入,电能装置运行类型作为输出,运用线性归一化对电能计量装置的运行检测类型进行编码,不同的编码对应不同的异常检测结果。实验结果表明,选取15个3×1的过滤器,可确保该方法电能计量装置运行异常自动检测准确性,该方法能够及时准确检测出电能计量装置运行异常情况,实用性较好。

In order to maintain the normal operation of electric energy metering devices,an automatic detection method for abnormal operation of electric energy metering devices based on data mining is proposed.It collects daily power consumption,historical variables and other operating data of electric energy metering devices,preprocesses the data by hierarchical k-means clustering method,extracts the current,voltage,electric power and other characteristics in the data,and then inputs the characteristics as convolutional neural network,takes the operation type of electric energy devices as output,and uses linear normalization to code the operation detection type of electric energy metering devices,different codes correspond to different anomaly detection results.The experimental results show that 153×1 filter can ensure the accuracy of automatic detection of abnormal operation of the electric energy metering device in this method.This method can detect abnormal operation of the electric energy metering device in a timely and accurate manner,with good practicability.