摘要
精确地计算台区线损、提取用户用电特征实现窃电预测,是精准营销策略制定的关键。然而,台区可再生能源发电的随机性、电动汽车充放电无序性、环境变化等因素使用户用电行为极易突变,导致台区用户用电数据呈现混沌随机特性,无法有效检测用户窃电行为。对此,建立了台区用户用电的时序相关混沌模型,并提取窃电负荷样本与正常样本的特征,使用回归卷积神经网络对窃电样本和正常样本训练学习,获得增强特征分类学习器,以此实现对窃电用户用电预测。通过对某实际电力公司用电数据的测试分析表明,所提方法的计算结果具有较高的精确度。
Accurate calculation of line loss and extraction of users’electricity consumption characteristics are the key to accurate mar-keting strategy.However,the randomness of renewable energy power generation,disorderly charging and discharging of electric vehicles,environmental changes and other factors make the user’s electricity consumption behavior easily mutate,which leads to the chaotic and random characteristics of user’s electricity consumption data in the station area,and can not effectively detect user’s electricity stealing behavior.In order to solve this problem,a time-series correlation chaotic model of power consumption for station area users is estab-lished,and the characteristics of power stealing load samples and normal samples are extracted.The regression convolutional neural net-work is used to train and learn the power stealing samples and normal samples,and the enhanced feature classification learner is ob-tained,so as to realize the power consumption prediction for power stealing users.Through the test and analysis of the power consump-tion data of an actual power company,it shows that the calculation results of the proposed method have high accuracy.
作者
靳海岗
谢振刚
任峰
JIN Haigang;XIE Zhengang;REN Feng(State Grid Shanxi Electric Power Company,Taiyuan Shanxi 030002,China)
出处
《电子器件》
CAS
北大核心
2023年第1期232-237,共6页
Chinese Journal of Electron Devices
基金
国网山西省电力公司科技项目(520531200004)。
关键词
回归卷积神经网络
负荷
混沌
窃电
预测
regression convolution neural network
load
chaos
stealing electricity
forecast
