摘要
电力负荷的变化不仅具有潜在周期性特征,外部气象等因素对负荷变化同样具有较大的影响,为提高负荷预测精度,提出一种基于Transformer-CNN融合内部周期性特征和外部气象特征的短期负荷预测方法。首先,采用奇异谱分析(Singular Spectrum Analysis,SSA)对历史负荷序列进行周期性重构,采用快速傅里叶变换(Fast Fourier Transform,FFT)提取典型序列周期,同时采用改进的灰色关联度法计算外部多种气象因素与历史负荷关联度,提取负荷内部7种周期性特征和外部4种气象特征,建立特征集;其次,设计一种多特征融合神经网络,基于CNN的多特征融合网络捕获特征集中潜藏特征与隐藏信息,基于Transformer的时序网络捕获历史负荷数据的时变特性,最终经隐式特征融合网络实现网络融合和短期负荷预测。实验结果表明,提取负荷内外双重特征能够有效提高模型预测精度,对节假日等特殊日期负荷预测的精度提高尤为明显。
The variation in electric load is influenced not only by underlying periodic characteristics but also by external meteorological and other factors.To enhance the accuracy of load forecasting,we propose a short-term load forecasting method that integrates internal periodic features and external meteorological features using the Transformer-CNN model.Firstly,we reconstruct the historical load sequence using Singular Spectrum Analysis(SSA)and extract the typical sequence period via Fast Fourier Transform(FFT).In addition,we use an improved gray correlation analysis to calculate the correlation between external meteorological factors and historical load.Based on this,we extract seven periodic features and four external meteorological features to construct the feature set.Secondly,we design a multi-feature fusion neural network.The CNN-based multi-feature fusion network captures latent features and hidden information within the feature set,while the Transformer-based time-series network captures the time-varying characteristics of historical load data.Finally,we utilize the implicit feature fusion network to achieve network fusion and load prediction.Experimental results indicate that extracting both internal and external load features can effectively improve model prediction accuracy,particularly for special dates such as holidays.
作者
张帅
刘文霞
唐浩洋
马英杰
万海洋
鲁宇
ZHANG Shuai;LIU Wenxia;TANG Haoyang;MA Yingjie;WAN Haiyang;LU Yu(State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources(North China Electric Power University),Beijing 102206,China;Economic and Technological Research Institute of State Grid Jilin Electric Power Co.,Ltd.,Changchun 130022,China)
出处
《华北电力大学学报(自然科学版)》
北大核心
2025年第3期68-75,83,共9页
Journal of North China Electric Power University:Natural Science Edition
基金
国网吉林省电力有限公司科技项目(2021JBGS-03).
