摘要
在详细分析窃电用户用电特性的基础上,搭建了用于处理海量用电数据的分布式存储Hadoop平台,分析和改进了适用于并行处理的BP神经网络算法,进而提出了基于人工神经网络的窃电嫌疑分析模型.根据窃电行为将会导致的用电异常数据,以及参考供电行业同业经验,选取了影响窃电嫌疑系数的十二个指标.结合窃电嫌疑分析模型,分析得出无窃电嫌疑、一般窃电嫌疑和重大窃电嫌疑三种情况.最后结合实例对该模型的精确度进行验证.
Based on a detailed analysis of characteristics of users who steal electricity, this paper built a distributed storage Hadoop platform to handle vast amounts of electricity data. Furthermore, BP neural network algorithm which is suitable for parallel processing, was analyzed and advanced. On this basis, the author put forward the analysis model about the suspicion of stealing electricity based on artificial neural network. Twelve indicators of suspect coefficient for stealing electricity have been selected, according to the electricity＇ s abnormal data which will be led by the behavior of power stealing and the experience of businesses of the same trade. Besides, combining with stealing suspects analysis model, the author found out that there were three cases for stealing electricity： no suspicion, general suspicion and great suspicion. Finally, the accuracy of the model was verified with examples.
出处
《河南科学》
2015年第10期1767-1772,共6页
Henan Science
关键词
窃电分析
HADOOP
数据挖掘
BP神经网络
the analysisofstealingelectricity
Hadoop
data mining
BP （Back Propagation） neural network
