基于期望最大化算法的电能表误差估计技术  

A novel error estimation methodology for power meters based on the expectation-maximization algorithm

作　　者：陈昊 杜新纲 葛得辉 于海波 CHEN Hao;DU Xingang;GE Dehui;YU Haibo(China Electric Power Research Institute Co.,Ltd.,Beijing 100192,China;State Grid Corporation of China Co.,Ltd.,Beijing 100031,China)

机构地区：[1]中国电力科学研究院有限公司,北京100192 [2]国家电网有限公司,北京100031

出　　处：《中国测试》2023年第3期47-52,共6页China Measurement & Test

基　　金：国家电网公司科技项目(5700-202227226A-1-1-ZN)。

摘　　要：根据能量守恒原理,电能表的计量误差可以通过一种线性模型来估计。该文提出一种基于期望最大化(expectation-maximization,EM)算法的新型误差估计方法,用于评估在运行电能表的误差程度。该方法将电能表误差模型中存在的大量无法观测到的变量作为隐变量对待,并根据最大似然估计(maximum likelihood estimation,MLE)法则,以迭代计算的方式对误差模型中的各种变量进行估计,能有效地对真实环境中存在的各种噪声进行模拟。在实验室仿真数据与工程中的真实数据上的实验结果都表明,该方法对小超差电能表计量误差的估计能达到较高精度,具有工程上的实用性和经济价值。According to the law of conservation of energy,the measurement errors of power meters satisfy a set of equations and can be solved using a linear model.This article proposes a new approach to estimate errors of running power meters based on the expectation-maximization(EM)algorithm.This approach regards many unobservable variables and parameters in the line loss model of power meters as latent variables,then use an iterative procedure to approximately calculate the measurement errors using the maximum likelihood estimation(MLE)methodology.Various noise components in real environment can be modeled efficiently in this approach.Experimental results on both synthetic data and real-world data show that this method can achieve relatively high accuracy levels in error estimation,especially for power meters with small over-tolerance errors which are difficult by other traditional algorithms.Engineering practicability and potentially significant economic value are demonstrated.

关 键 词：电能表 误差估计 期望最大化算法 失准模型 

分 类 号：TB9[一般工业技术—计量学]