曹敏, 邹京希, 魏龄, 赵旭, 张林山, 李鹏. 基于RBF神经网络的配电网窃电行为检测*[J]. 云南大学学报(自然科学版), 2018, 40(5): 872-878. doi: 10.7540/j.ynu.20170426
引用本文: 曹敏, 邹京希, 魏龄, 赵旭, 张林山, 李鹏. 基于RBF神经网络的配电网窃电行为检测*[J]. 云南大学学报(自然科学版), 2018, 40(5): 872-878. doi: 10.7540/j.ynu.20170426
CAO Min, ZOU Jing-xi, WEI Ling, ZHAO Xu, ZHANG Lin-shan, LI Peng. Detection of power theft behavior of distribution network based on RBF neural network[J]. Journal of Yunnan University: Natural Sciences Edition, 2018, 40(5): 872-878. DOI: 10.7540/j.ynu.20170426
Citation: CAO Min, ZOU Jing-xi, WEI Ling, ZHAO Xu, ZHANG Lin-shan, LI Peng. Detection of power theft behavior of distribution network based on RBF neural network[J]. Journal of Yunnan University: Natural Sciences Edition, 2018, 40(5): 872-878. DOI: 10.7540/j.ynu.20170426

基于RBF神经网络的配电网窃电行为检测*

Detection of power theft behavior of distribution network based on RBF neural network

  • 摘要: 目前的窃电手段隐蔽性强、种类多样,传统的窃电检测方法已经越来越力不从心,不仅准确率低且时效性不高.因此,结合机器学习的方法,提出了一种基于RBF神经网络的窃电行为检测方法.通过对目前常见的窃电方式进行分析,挑选出三相电压中各相电压之间的差值、三相电流中各相电流的差值以及功率因数等参考量作为窃电检测的特征指标,并采用包含特征指标的历史数据来构建基于RBF神经网络的窃电行为检测模型.试验结果表明该方法针对目前常见的窃电方式进行识别的准确率达到94.1%,可以有效地筛选出存在窃电嫌疑的用户.该方法不仅达到了实际应用的精度要求,而且使反窃电技术更加智能化,变被动防窃电为主动防窃电.

     

    Abstract: At present,it has become more and more difficult for the traditional methods to detect various covert power theft behaviors,not only due to low accuracy rate,but also due to high time cost.Therefore,as an application of machine learning theory,this paper presents a detection method of power theft behavior on distribution network based on RBF neural network.Through the analysis on the current common power theft behavior,this paper picks up divergence among three-phase voltage,divergence among three-phase current and divergence among power factor as three important features and build a power theft detection model based on RBF neural network with the historical data containing three important features mentioned above.The experimental result shows that the accuracy rate of this method on distinguishing the current common power theft behavior is up to 94.1%,which means this method could be used on suspecting power consumers who are probably stealing power effectively.This method not only meets the precision requirements of practical application,but also makes the power anti-theft technology more intelligent and thus the power supply enterprises can take the initiative.

     

/

返回文章
返回