基于特征集构建的工业企业电负荷预测模型研究  

Research on Industrial Enterprise Electricity Load Forecasting Model Based on Feature Set Construction

作　　者：张然 黄宸 ZHANG Ran;HUANG Chen(CATARC Automotive Industry Engineering,Tianjin 300300,China)

机构地区：[1]中汽研汽车工业工程(天津)有限公司,天津300300

出　　处：《自动化应用》2024年第24期116-118,121,共4页Automation Application

摘　　要：提前获取用户用电量有助于维持电网可靠性与规划制定调度策略,因此建立准确的电负荷预测模型具有重要意义。建立了一种基于数据集构建的工业企业电负荷预测方法。在特征维度上,通过引入聚类算法,对采集数据有效分类,并在各分类中使用相关性分析方法完成特征筛选。在时间维度上,计算基于余弦距离的相似性判断指标,遴选与预测日特征相似的历史数据构成历史相似日用电量,并作为特征输入预测模型。使用重构的特征集,与基础预测模型比较,可减小预测期间1.9%~5.1%的误差。Obtaining user electricity consumption in advance helps maintain the reliability of the power grid and formulate scheduling strategies,therefore establishing an accurate electricity load forecasting model is of great significance.This paper establishes a method for industrial enterprise electricity load forecasting based on dataset construction.In the feature dimension,clustering algorithms are introduced to effectively classify the collected data,and correlation analysis methods are used to complete feature screening in each classification.In temporal dimension,calculate the similarity judgment index based on cosine distance,select historical data with similar features to the predicted day to form historical similar daily electricity consumption,and use it as the feature input for the prediction model.Compared with the basic prediction model,using the reconstructed feature set can reduce the error during the prediction period by 1.9%to 5.1%.

关 键 词：电负荷 预测模型 K-MEANS聚类 数据集 

分 类 号：TM715[电气工程—电力系统及自动化]