融合经验模态分解与特征增强的Transformer-TCN光伏功率预测模型

张海清; 刘航; 唐鑫; 杨孟辑; 于曦; 李代伟

doi:10.19912/j.0254-0096.tynxb.2025-0685

欢迎来到能源电力期刊网！

English Version 期刊登录

您当前的位置：

首页 >

文章列表页 >

融合经验模态分解与特征增强的Transformer-TCN光伏功率预测模型

更新时间：2026-10-15

- 融合经验模态分解与特征增强的Transformer-TCN光伏功率预测模型
- 融合经验模态分解与特征增强的Transformer-TCN光伏功率预测模型
- 太阳能学报 2026年47卷第9期页码：203-211
- 作者机构：
- 作者简介：
- 基金信息：
- DOI：10.19912/j.0254-0096.tynxb.2025-0685
  中图分类号： TP18
- 网络首发：2026-10-10，
  
  纸质出版：2026
- 稿件说明：
移动端阅览
张海清, 刘航, 唐鑫, 等. 融合经验模态分解与特征增强的Transformer-TCN光伏功率预测模型[J]. 太阳能学报, 2026,47(9):203-211.

张海清, 刘航, 唐鑫, et al. 融合经验模态分解与特征增强的Transformer-TCN光伏功率预测模型[J]. 2026, 47(9): 203-211.
张海清, 刘航, 唐鑫, 等. 融合经验模态分解与特征增强的Transformer-TCN光伏功率预测模型[J]. 太阳能学报, 2026,47(9):203-211. DOI： 10.19912/j.0254-0096.tynxb.2025-0685.

张海清, 刘航, 唐鑫, et al. 融合经验模态分解与特征增强的Transformer-TCN光伏功率预测模型[J]. 2026, 47(9): 203-211. DOI： 10.19912/j.0254-0096.tynxb.2025-0685.

摘要

针对单一深度学习模型难以兼顾长短期依赖与局部特征波动的问题

提出一种基于经验模态分解（EMD）、Transformer和时间卷积网络（TCN）的双通道耦合预测模型。模型利用EMD将原始光伏序列分解为一组本征模态函数（IMF）分量

针对不同平稳性的IMF分量

分别采用基于残差网络改进的Transformer模型（Enresformer）和基于门控网络改进的TCN模型进行预测

并集成预测结果。引入误差修正机制以进一步提升预测精度。基于多个实际光伏数据集的对比实验表明

所提模型在预测精度方面显著优于当前主流的深度学习模型和耦合模型

为功率预测提供了一种高效且可靠的解决方案。

Abstract

Photovoltaic （PV） power generation

as a crucial source of clean energy

requires accurate power forecasting to facilitate efficient power dispatch. PV power is highly influenced by meteorological factors and exhibits significant intermittency and volatility. Traditional time series forecasting models often fail to capture the dynamic correlations among meteorological factors

leading to limited forecasting accuracy. Meanwhile

single deep learning architectures struggle to effectively capture the intricate interplay between multi-scale temporal dependencies and localized feature variations. To address these challenges

this paper proposes a dual-channel hybrid forecasting model integrating empirical mode decomposition （EMD）

Transformer

and temporal convolutional network （TCN）. EMD is employed to decompose the raw PV power signal into a set of intrinsic mode functions （IMFs）. The stationarity of each IMF component is quantified using sample entropy （SE）

guiding the selection of tailored predictive models for components with distinct stationarity characteristics. An enhanced residual transformer （Enresformer）

which incorporates residual networks

and a gated network-enhanced TCN model are employed to forecast these components

with subsequent ensembling of their outputs. An error correction mechanism is introduced to further refine prediction accuracy. Comparative experiments based on multiple real-world PV datasets show that the proposed model significantly outperforms existing deep learning and hybrid models in terms of forecasting accuracy

providing an efficient and reliable solution for PV power prediction in practical power stations.

关键词

Keywords

references

Siami-Namini S, Tavakoli N, Namin A S.The performance of LSTM and BiLSTM in forecasting time series[C]//2019 IEEE International Conference on Big Data (Big Data). Los Angeles, CA, USA, 2020: 3285-3292.

Shih S Y, Sun F K, Lee H Y.Temporal pattern attention for multivariate time series forecasting[J]. Machine Learning, 2019, 108(8): 1421-1441.

Vaswani A, Shazeer N, Parmar N, et al.Attention is all you need[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach, California, USA, 2017: 6000-6010.

Zhou H Y, Zhang S H, Peng J Q, et al.Informer: beyond efficient transformer for long sequence time-series forecasting[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35(12): 11106-11115.

温廷新, 郭晓赛. 基于特征组合的ECA-TCN光伏发电功率预测模型[J]. 太阳能学报, 2024, 45(12): 94-100.

李青. 基于注意力时间卷积神经网络的光伏功率概率预测[J]. 太阳能学报, 2025, 46(2): 326-332.

Anggraeni W, Yuniarno E M, Rachmadi R F, et al.A hybrid EMD-GRNN-PSO in intermittent time-series data for dengue fever forecasting[J]. Expert Systems with Applications, 2024, 237: 121438.

冯建铭, 希望·阿不都瓦依提, 蔺红. 基于聚类SABO-VMD和组合神经网络的短期光伏发电功率预测[J]. 太阳能学报, 2025, 46(2): 357-366.

徐楚天, 沈良朵, 班文超, 等. 基于CEEMDAN-LSTM风暴潮潮位预测分析研究[J]. 太阳能学报, 2024, 45(11): 578-585.

Huang N E, Shen Z, Long S R, et al.The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis[J]. Proceedings: Mathematical, Physical and Engineering Sciences, 1998, 454(1971): 903-995.

汤健, 柴天佑, 丛秋梅, 等. 基于EMD和选择性集成学习算法的磨机负荷参数软测量[J]. 自动化学报, 2014, 40(9): 1853-1866.

孔祥玉, 李闯, 郑锋, 等. 基于经验模态分解与特征相关分析的短期负荷预测方法[J]. 电力系统自动化, 2019, 43(5): 46-56.

王希, 孙娜, 苏浩, 等. 基于样本熵双分解和SSA-LSTM的超短期风速预测[J]. 太阳能学报, 2025, 46(3): 611-618.

Richman J S, Moorman J R.Physiological time-series analysis using approximate entropy and sample entropy[J]. American Journal of Physiology Heart and Circulatory Physiology, 2000, 278(6): H2039-H2049.

He K M, Zhang X Y, Ren S Q, et al.Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, NV, USA, 2016: 770-778.

崔京港, 王芳, 叶泽甫, 等. 基于QD和因果注意力TCN的光伏功率区间预测[J]. 太阳能学报, 2024, 45(3): 488-495.

Dong J, Zhang Y L, Hu J.Short-term air quality prediction based on EMD-transformer-BiLSTM[J]. Scientific Reports, 2024, 14: 20513.

Laouafi A, Laouafi F, Boukelia T E.An adaptive hybrid ensemble with pattern similarity analysis and error correction for short-term load forecasting[J]. Applied Energy, 2022, 322: 119525.

郭文凯, 王果, 闵永智, 等. 基于FOX-VMD联合小波阈值去噪的短期光伏功率预测研究[J]. 太阳能学报, 2025, 46(6): 260-270.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于注意力时间卷积神经网络的光伏功率概率预测

基于聚合模态分解和TCN-BiGRU的光伏功率预测模型

CloudSwinNetLite：轻量级地基云图细粒度分割网络

基于PM-iTransformer-TCN和贝叶斯优化的超短期风电功率预测方法

融合多尺度周期性特征的光伏功率短期预测模型