基于多源信息和机器学习算法的电能表故障诊断研究  

Research on Electricity Meter Fault Diagnosis Based on Multi-source Information and Machine Learning Algorithm

作　　者：马斌 韩洁琼 龙存玉 王逸晖 杨靖 胡少年 MA Bin;HAN Jie-qiong;LONG Cun-yu;WANG Yi-hui;YANG Jing;HU Shao-nian(State Grid Qinghai Marketing Service Center,Xining 810000 China;Wasion Group Limited,Changsha 410205 China)

机构地区：[1]国网青海省电力公司营销服务中心,青海西宁810000 [2]威胜集团有限公司,湖南长沙410205

出　　处：《自动化技术与应用》2024年第10期60-64,共5页Techniques of Automation and Applications

基　　金：国网青海省电力公司营销服务中心营销项目(63283422000A)。

摘　　要：为提升电能表的自动化管理水平,设计了基于多源信息和机器学习算法的电能表故障诊断方法。首先采集电能表的检定数据、用电数据、运行电压等多源信息,利用一致性判断方法融合所采集的电能表多源信息,然后选取独立成分分析(Independent Component Correlation Algorithm,ICA)方法利用ICA变换处理电能表多源信息融合结果,提取多源信息中包含的电能表故障特征,最后提取的电能表故障特征作为支持向量机的输入,利用人工蜂群算法优化支持向量机,输出电能表故障诊断结果。实验结果表明,该方法可以有效诊断电能表的电池欠压、继电器损坏等不同类型故障,为电能表故障分析提供依据。In order to improve the automation management of energy meters,a fault diagnosis method of energy meters based on multi-source information and machine learning algorithm is designed.Firstly,it collects multi-source information such as calibra-tion data,power consumption data and operating voltage of energy meters,fuses the collected multi-source information of energy meters by using consistency judgment method,then selects ICA method to process the fusion result of multi-source information of energy meters by using ICA transformation,extracts the fault characteristics of energy meters contained in multi-source infor-mation,finally the extracted fault characteristics of energy meters are used as the input of support vector machine,and it uses arti-ficial bee colony algorithm to optimize the Support Vector Machine and output the electric energy meter fault diagnosis results.The experimental results show that the method can effectively diagnose different types of faults of energy meters,such as battery undervoltage and relay damage,and provide a basis for energy meter fault analysis.

关 键 词：多源信息 机器学习 电能表 故障诊断 独立成分分析 支持向量机 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TM933[自动化与计算机技术—控制科学与工程]