摘要
在智能电网技术背景下,智能电表(SM)数据的处理面临诸多挑战,包括大数据量、快速处理要求和数据多样性等问题。为解决这些问题,提出了一种基于分层随机网络模型的智能电表计量数据异常分析框架。首先,引入了一个用于训练的实时电力计量异常检测模型,并采用了一种有效的参数估计方法,以充分利用从SM处收集的大规模数据集。其次,通过设计一个综合处理缺失点和正则化的方法,以解决SM数据处理过程中的两个主要难题。通过融合多个子系统的观测数据,不仅能够更好地进行多变量计数,还能够识别系统的异常状态。实验中,将2500个样本中的80%(2000个)用于训练模型,其余20%(500个)用于测试。实验结果表明,所提出的方法相较于其他方法在使用多个测量值进行异常诊断时,模型表现得更为出色。因此,本研究不仅提供了一种处理SM数据的创新方法,还为实时电力计量异常检测领域提供了有力的支持。
In the context of smart grid technology,the processing of data from smart meters(SM)faces numerous challenges,including large data volumes,fast processing requirements,and data heterogeneity.To address these issues,a framework for anomaly analysis of smart metering data is proposed based on a hierarchical random network model.Firstly,a real-time electric power metering anomaly detection model is introduced for training,utilizing an effective parameter estimation method to fully leverage the extensive dataset collected from SM.Secondly,a comprehensive approach dealing with missing data points and regularization is designed to tackle the two main challenges in SM data processing.By integrating observational data from multiple subsystems,not only can multivariate counting be conducted more effectively,but also the identification of the system's abnormal states can be achieved.In the experiments,80%(2000 samples)of the 2500 samples are used for training the model,and the remaining 20%(500 samples)are used for testing.The experimental results indicate that the proposed method outperforms other approaches,particularly when utilizing multiple measurement values for anomaly diagnosis.Therefore,this study not only offers an innovative method for handling SM data but also provides robust support for the field of real-time electric power metering anomaly detection.
作者
牛任恺
王利赛
郑思达
郭伟
张鑫磊
NIU Renkai;WANG Lisai;ZHENG Sida;GUO Wei;ZHANG Xinlei(Metering Center,Jibei Power Grid Co.,Ltd.,Beijing 100045,China)
出处
《电子器件》
CAS
2024年第5期1337-1343,共7页
Chinese Journal of Electron Devices
基金
国网冀北营销服务中心2022年揭榜制项目研究(B30185220001)。
关键词
分层随机网络
智能电表
计量异常
多变量计数
hierarchical stochastic network
smart meters
measurement anomaly
multivariable counting
