基于多分类融合模型的智能电能表故障预测  

Prediction on fault classification of smart meters based on multi-classification integration model

作　　者：陈叶 韩彤 魏龄 于秀丽[2] 李鑫雄 Chen Ye;Han Tong;Wei Ling;Yu Xiuli;Li Xinxiong(Electric Power Research Institute of Yunnan Power Grid Co.,Ltd.,Kunming 650217,China;School of Automation,Beijing University of Posts and Telecommunications,Beijing 100876,China;Key Laboratory of CSG for Electric Power Measurement,Kunming 650217,China)

机构地区：[1]云南电网有限责任公司电力科学研究院,昆明650217 [2]北京邮电大学自动化学院,北京100876 [3]南方电网公司电能计量重点实验室,昆明650217

出　　处：《电测与仪表》2022年第11期162-168,共7页Electrical Measurement & Instrumentation

摘　　要：由于智能电能表功能的丰富多样,随之而来的是设备故障类型及故障率的不断增加,如何准确地判断智能电能表的故障类型,提高故障表的检修效率,对保障智能电能表的安全稳定运行十分重要。文中提出一种基于多分类融合模型的智能电能表故障预测算法。针对智能电能表故障进行多维度分析及故障类型筛选;通过欠采样和过采样相结合的混合采样方式解决数据集中类不平衡问题,构建分类预测模型所需数据;利用基础分类算法的组合获取最优融合算法,在公共数据集上验证了所提算法的有效性,融合后的准确率较基础分类模型有稳定提升,以近年来电网系统中实时采集的智能电能表故障数据为基础,进行了基础模型与融合后算法模型的实验对比,结果表明文中所提的多分类融合算法模型在故障预测的准确率和可靠性上有明显的提升。Due to the rich and diverse functions of smart meters,the fault types and failure rates of equipments are gradually increasing.It is very important to ensure the safe and stable operation of smart meters that how to accurately determine the fault types of smart meters and improve the maintenance efficiency of fault meters.In this paper,a fault prediction algorithm of smart meter is proposed,which is based on multi-classification integration model.Multi-dimensional analysis and fault type selection are carried out for smart meter fault;the problem of class imbalance in data set is solved by the combination of under sampling and over sampling,and the data required for classification and prediction model is constructed.The basic classification algorithms are combined to obtain the best integration algorithm.The effectiveness and accuracy of the proposed algorithm are verified on the public data set.Finally,the comparative experiments based on the real-time fault data of smart meters collected in recent years are made between the basic model and the integration algorithm model.The results show that the multi-classification integration algorithm model proposed in this paper has a significant improvement in the accuracy and reliability of fault prediction.

关 键 词：智能电能表故障 混合采样 多分类算法 模型融合 

分 类 号：TM93[电气工程—电力电子与电力传动]