摘要
针对现有短期电力负荷预测计算方法中计算量大、训练时间长等问题,本文提出了基于宽度学习系统的短期电力负荷预测方法。采集预测区域影响电力负荷因素的历史数据,首先通过归一法对数据进行预处理,利用领域粗糙集分类算法提取对负荷预测重要度较高的属性,防止非重要影响因素输入模型造成信息冗余,减少负荷预测算法的计算量;然后建立基于宽度学习的短期电力负荷预测模型,将样本输入模型进行训练,对输出结果进行评估,根据评估结果对短期电力负荷预测模型进行增量计算,无须重建整个预测模型,减少负荷预测训练时间;最后将计算结果进行反归一化,获得最终预测结果。本文通过仿真实验对比验证了所提方法可以有效预测短期电力负荷,并且缩短预测时间,提高了预测精度。
This paper proposes a short-term electricity load forecasting method based on a wide learning system,aiming to solve problems such as high computational complexity and long training times in existing methods.Historical data of factors influencing electricity load in the forecasting area are collected.Firstly,data is preprocessed using normalization.The rough set classification algorithm in the domain is then utilized to extract attributes.7035050 with high importance for load forecasting.This helps prevent non-critical factors from causing information redundancy in the model input,reducing the computational load of the load forecasting algorithm.Next,a short-term electricity load forecasting model based on wide learning is established.Samples are input into the model for training,and the output results are evaluated.Based on the evaluation results,incremental calculations are performed on the short-term electricity load forecasting model,avoiding the need to rebuild the entire forecasting model and reducing the training time.Finally,the calculated results are reverse-normalized to obtain the final prediction.Through simulation experiments,this paper demonstrates that the proposed method can effectively forecast shortterm electricity load,shorten the prediction time,and improve prediction accuracy.
作者
刘扬
王佳祯
汪晓东
张威
李鑫
LIU Yang;WANG Jiazhen;WANG Xiaodong;ZHANG Wei;LI Xin(Tongxiang Power Supply Company,State Grid Zhejiang Electric Power Co.,Ltd.,Jiaxing 314599,Zhejiang,China;Haiyan Power Supply Company,State Grid Zhejiang Electric Power Co.,Ltd.,Jiaxing 314399,Zhejiang,China)
出处
《电力大数据》
2023年第7期23-31,共9页
Power Systems and Big Data
关键词
宽度学习
归一化
邻域粗糙集
特征提取
负荷预测
broad learning system
normalization
the neighborhood rough set
feature extraction
load forecasting
