摘要
为了解决经济新常态下电力需求预测难度较大的问题,本文提出了一种基于大数据条件下的分组负荷预测方法,该方法打破了以往根据行政区或固定行业分类开展预测的思路,通过充分挖掘大数据环境下的用电信息进行分析预测。首先,对基础数据进行辨识不良数据、修补坏数据等预处理,再根据不同行业的用电负荷特性,提取不同行业典型负荷曲线;利用K-means聚类算法将行业内用户级的负荷根据其曲线特性的不同进行聚类,进一步形成分组负荷;然后,选择时间序列法、月间相关算法、影响因素预测法等合适的预测方法对不同组负荷分别进行预测;最后,根据同时率将不同组负荷预测结果叠加得到目标区域的负荷预测值。本文通过对江苏省内南部某区域内实际负荷进行算例验证,结果表明该方法具有可行性及实际意义。
In order to solve the difficult problem of power demand forecasting under the new economic normal,this paper proposes a grouped load forecasting method based on big data.This method analyzes and predicts by fully mining electricity consumption information in a big data environment,breaking the previous thinking of predicting based on administrative regions or fixed industries.First,this paper preprocesses the basic data to identify bad data and repair bad data.Then,this paper extracts the typical load curves of different industries according to the electricity load characteristics of different industries.The paper uses the K-means clustering algorithm to cluster the user-level loads in the industry according to the different curve characteristics to further form the grouped load.After forming the grouped load,the paper selects appropriate forecasting methods such as time series method,monthly correlation algorithm,influencing factor forecasting method,etc.to predict different groups of loads respectively,and finally superimpose the load forecast results of different groups according to the simultaneous rate to obtain the load forecast value of the target area This paper verifies the actual load in a certain area in the southern part of Jiangsu province.The results show that the method is feasible and practical.
作者
史静
南开辉
周琪
谈健
李琥
SHI Jing;NAN Kaihui;ZHOU Qi;TAN Jian;LI Hu(Economic Research Institute of State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 210008 Jiangsu,China;State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 210024 Jiangsu,China)
出处
《电力大数据》
2020年第6期9-16,共8页
Power Systems and Big Data
关键词
负荷预测
大数据
聚类
分组负荷
相关系数
load forecasting
big data
clustering
grouped load
correlation coefficient
