电力多维指标聚合在用户窃电智能分析中的应用  

Application of Power Multidimensional Index Aggregation in Intelligent Analysis of User Stealing Electricity

作　　者：杨婧 宋强 陈庆辉 YANG Jing;SONG Qiang;CHEN Qing-hui(Guizhou Power Grid Co.,Ltd.,Guiyang 550002 China)

机构地区：[1]贵州电网有限责任公司,贵州贵阳550002

出　　处：《自动化技术与应用》2023年第7期170-173,共4页Techniques of Automation and Applications

摘　　要：用户窃电行为分析过程中,由于判断指标的差异,导致智能分析结果的F-Measure值较小,用户窃电智能分析准确率较差,为此提出电力多维指标聚合在用户窃电智能分析中的应用。选取与窃电分析相关的电力特征数据,建立多维聚合判断指标。结合数据修复、数据降维等多种方法对用电数据进行预处理,基于RBF神经网络构建窃电行为检测模型,获取窃电智能分析结果,利用萤火虫寻优算法,获取模型全局寻优结果。仿真实验结果表明提出的用户窃电智能分析方法,将F-Measure值提升了14.56%与22.43%,有效提升了用户窃电智能分析准确率。In the process of user electricity stealing behavior analysis,due to the difference of judgment indicators,the F-Measure value of intelligent analysis results is small,and the accuracy rate of user electricity stealing intelligent analysis is poor.Therefore,it is proposed to aggregate multi-dimensional power indicators in user electricity stealing intelligent analysis.It selects the power characteristic data related to electricity stealing analysis,and establishes multi-dimensional aggregated judgment indicators.Combined with data repair,data dimensionality reduction and other methods to preprocess the electricity consumption data,builds an electricity stealing behavior detection model based on RBF neural network,obtain the intelligent analysis results of electricity stealing,and use the firefly optimization algorithm to obtain the global optimization results of the model.The simulation experiment results show that the proposed method for intelligent analysis of user electricity stealing improves the F-Measure value by 14.56% and 22.43%,and effectively improves the accuracy of user electricity stealing intelligent analysis.

关 键 词：聚合 窃电分析 智能诊断 萤火虫寻优算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]