应用DBN深度学习算法的电能计量反窃电技术研究  

Research on Anti-stealing Technology of Electric Energy Metering Based on DBN Deep Learning Algorithm

作　　者：刘岩 袁瑞铭 郑思达 杨晓坤 王玉君 LIU Yan;YUAN Rui-ming;ZHENG Si-da;YANG Xiao-kun;WANG Yu-jun(State Grid Hebei North Marketing Service Center(Metric Center),Beijing 102208.China)

机构地区：[1]国网冀北营销服务中心(计量中心),北京102208

出　　处：《计算技术与自动化》2021年第4期151-155,共5页Computing Technology and Automation

基　　金：河北省国网电力公司科技项目(61437062)。

摘　　要：针对窃电问题严重阻碍建立公平、合理的用户秩序的问题,基于云计算的智能电网大数据处理平台SP-PPP(smart power system big data processing platform in cloud environment,SP-DPP),提出了融合自适应加权融合算法和深度置信网络DBN(Deep Belief Networks,DBN)学习算法的反窃电系统,采用DBN逐层贪婪训练算法对大数据进行处理,并利用双层RBM结构,构建出DBN深度学习算法,对获取的电能计量窃电信息进行归一化处理,将获取的宏观高纬度数据信息转换为容易识别和计算的低纬度数据。实验表明,本研究的算法识别率高,稳定性能好。Aiming at the problem of electricity theft seriously hindering the establishment of a fair and reasonable user order,An anti-theft system that combines adaptive weighted fusion algorithm and deep belief network(Deep Belief Networks,DBN)learning algorithm is proposed based on Smart power system big data processing platform in cloud environment(SP-DPP),big data is processed by using DBN layer-by-layer greedy training algorithm,and a DBN deep learning algorithm is constructed by using the double-layer RBM structure,which can normalize the acquired electricity metering information and convert the acquired macro high-latitude data information into low-latitude data that is easy to identify and calculate.Tests show that the algorithm of this study has high recognition rate and good stability.

关 键 词：窃电 SP-DPP 自适应加权融合算法 深度置信网络 逐层贪婪训练算法 

分 类 号：TM561[电气工程—电器]