基于细化用户画像的居民冷暖用电数据预测方法  

Prediction method of electricity consumption changing with temperature based on refined user portrait

作　　者：吴林峰 陈源萍 谭元刚 WU Linfeng;CHEN Yuanping;TAN Yuangang(Marketing Service Center,State Grid Chongqing Electric Power Company,Chongqing 401100,China;Customer Service Branch of Chongqing Guanghui Power Supply Service Co.,Ltd.,Chongqing 401100,China)

机构地区：[1]国网重庆市电力公司营销服务中心,重庆401100 [2]重庆广汇供电服务有限责任公司客户服务分公司,重庆401100

出　　处：《智能计算机与应用》2024年第5期265-269,共5页Intelligent Computer and Applications

摘　　要：为提升电网运营数据应用价值与客户服务精益化管理水平,针对温度变化影响居民用电负荷问题,提出了一种基于细化用户画像+BP神经网络算法的居民冷暖用电数据预测方法。该方法首先利用时序分解对居民随温度变化的日用电量指标进行提取,并基于此细化居民用电行为画像类,随后基于细化类采用BP神经网络对居民冷暖电量进行预测。实验表明,该算法可提升居民冷暖用电量预测准确度,实现了对居民用电隐性动态特征的深度挖掘。In order to improve the application value of power grid operation data and the lean management level of customer service,and aiming at the problem that temperature changes affect residents'electricity load,a prediction method for residents'heating and cooling electricity data based on refined user portrait+BP neural network algorithm is proposed.This method firstly uses time series decomposition to extract the daily electricity consumption index of residents that changes with temperature,and based on this,refines the portrait class of residents'electricity consumption behavior,and then uses BP neural network to predict the heating and cooling power of residents based on the refined class.Experiments show that the algorithm can improve the prediction accuracy of residential heating and cooling electricity consumption,and realize the deep mining of the hidden dynamic characteristics of residential electricity consumption.

关 键 词：时序分解 模糊C均值聚类 BP神经网络 冷暖用电预测 细化用户画像 

分 类 号：TM933[电气工程—电力电子与电力传动]