融合多尺度周期性特征的光伏功率短期预测模型

陈万志; 刘昊洋; 王天元

doi:10.19912/j.0254-0096.tynxb.2025-0766

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 >

融合多尺度周期性特征的光伏功率短期预测模型

更新时间：2026-10-15

- 融合多尺度周期性特征的光伏功率短期预测模型
- 融合多尺度周期性特征的光伏功率短期预测模型
- 太阳能学报 2026年47卷第9期页码：327-340
- 作者机构：
- 作者简介：
- 基金信息：
- DOI：10.19912/j.0254-0096.tynxb.2025-0766
  中图分类号： TM615
- 网络首发：2026-10-10，
  
  纸质出版：2026
- 稿件说明：
移动端阅览
陈万志, 刘昊洋, 王天元. 融合多尺度周期性特征的光伏功率短期预测模型[J]. 太阳能学报, 2026,47(9):327-340.

陈万志, 刘昊洋, 王天元. 融合多尺度周期性特征的光伏功率短期预测模型[J]. 2026, 47(9): 327-340.
陈万志, 刘昊洋, 王天元. 融合多尺度周期性特征的光伏功率短期预测模型[J]. 太阳能学报, 2026,47(9):327-340. DOI： 10.19912/j.0254-0096.tynxb.2025-0766.

陈万志, 刘昊洋, 王天元. 融合多尺度周期性特征的光伏功率短期预测模型[J]. 2026, 47(9): 327-340. DOI： 10.19912/j.0254-0096.tynxb.2025-0766.

摘要

针对光伏功率预测中存在的非线性、波动性及多尺度周期性特征建模不足的问题

提出一种融合多尺度周期性特征的光伏功率短期预测模型。首先

对原始数据进行预处理和皮尔逊相关系数特征重构

并利用门控聚合重分配机制优化相关因素交互;其次

通过构建周期-残差双阶段优化机制

借助残差循环预测技术获取残差分量和周期分量

并将残差分量送入Transformer编码器。此外

采用多尺度周期特征融合模块处理周期分量

结合时序适应性的多尺度卷积核以及基于时间维度均值化融合策略提升对周期模式理解能力;最后

对残差分量和周期分量分别建模后的输出结果进行整合得出预测结果。在3个真实数据集上的实验结果表明

所提模型在不同步长任务下均优于对比模型。在光伏数据集1中

相较于次优模型

其MAE和RMSE分别相对平均降低4.10%和3.98%;在光伏数据集2中

两项指标较次优分别相对平均降低3.31%和1.51%;在电力负荷数据集ETTh1中的短步长预测任务中

相较于次优模型

其MAE降低1.66%。此外

模型在多项任务中的决定系数均达到最优水平

证明其在短期光伏预测任务中的优越性能与泛化能力。

Abstract

In response to the problem of limitations in modeling nonlinearity

volatility

and multi-scale periodicity characteristics in photovoltaic power prediction

a short-term photovoltaic power prediction model using multi-scale periodic features was proposed. Firstly

the raw data are preprocessed and the Pearson correlation coefficient features are reconstructed

then cross-feature interactions are optimized through a gated aggregation redistribution mechanism. Secondly

a dual-phase periodicity-residual optimization mechanism is established

where residual components obtained through residual cycle forecasting are processed by Transformer encoders

while periodic components undergo multi-scale periodic feature fusion. The periodic processing module employs temporally adaptive multi-scale convolutional kernels and a time-dimension averaging fusion strategy to enhance periodic pattern recognition. Finally

predictions are generated through integrated outputs from both residual and periodic modeling paths. Experimental evaluations across three real datasets show that the proposed model outperforms comparison models across various horizons. Specifically

in PV Dataset 1

the average MAE and RMSE are reduced by 4.10% and 3.98% respectively compared to the second-best model; in PV Dataset 2

these indicators are reduced by 3.31% and 1.51%; in the short-term tasks of the ETTh1 dataset

the MAE is reduced by 1.66%. Additionally

the model achieves the highest in multiple tasks

verifying its superior predictive performance and generalization ability in short-term photovoltaic power prediction.

关键词

Keywords

references

邹才能, 陈艳鹏, 熊波, 等. 碳中和目标下中国新能源使命[J]. 中国科学院院刊, 2023, 38(1): 48-58.

Feng Y, Gong D Z, Zhang Q W, et al.Evaluation of temperature-based machine learning and empirical models for predicting daily global solar radiation[J]. Energy Conversion and Management, 2019, 198: 111780.

Yang X D, Xu C B, Zhang Y B, et al.Real-time coordinated scheduling for ADNs with soft open points and charging stations[J]. IEEE Transactions on Power Systems, 2021, 36(6): 5486-5499.

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.

李翰章, 冯江涛, 王鹏程, 等. 基于TDE-SO-AWM-GRU的光伏发电功率预测模型[J]. 中国电力, 2024, 57(12): 41-49.

王玲芝, 李晨阳, 刘婧, 等. 基于GRO-SSA-LSTM的短期光伏发电功率预测[J]. 太阳能学报, 2025, 46(2): 401-409.

陈船宇, 熊国江, 方厚康, 等. 基于MODWT-CEEMDAN-LSTM的短期光伏功率区间预测模型[J]. 太阳能学报, 2025, 46(2): 416-424.

Yu L A, Ma Y M, Ma M Y.An effective rolling decomposition-ensemble model for gasoline consumption forecasting[J]. Energy, 2021, 222: 119869.

王开艳, 杜浩东, 贾嵘, 等. 基于相似日聚类和QR-CNN-BiLSTM模型的光伏功率短期区间概率预测[J]. 高电压技术, 2022, 48(11): 4372-4388.

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

Liu Y, Hu T G, Zhang H R, et al.iTransformer: inverted transformers are effective for time series forecasting[C]//Proceedings of the Twelfth International Conference on Learning Representations. 2024: 1-25.

Nie Y Q, Nguyen N H, Sinthong P, et al.A time series is worth 64 words: long-term forecasting with transformers[C]//The Eleventh International Conference on Learning Representations. 2023.

张研, 景超, 王慧民, 等. 基于周期注意力机制的中长期光伏发电功率预测[J]. 太阳能学报, 2024, 45(10): 298-308.

李逸航, 肖辉, 易纯, 等. 基于STL分解和TPA机制的光伏功率区间预测[J]. 太阳能学报, 2024, 45(12): 22-29.

张辉, 颜星雨, 毛建旭, 等. 面向源网荷的智能化数据协同推断技术研究综述[J]. 自动化学报, 2025, 51(11): 2387-2411.

Chen X Y, Han L, Ye H J, et al.SOFTS: efficient multivariate time series forecasting with series-core fusion[C]//Advances in Neural Information Processing Systems 37. Vancouver, BC, Canada, 2024: 64145-64175.

Hu X Y, Lin S S, Lin W W, et al.CycleNet: enhancing time series forecasting through modeling periodic patterns[C]// Advances in Neural Information Processing Systems 37. Vancouver, BC, Canada, 2024: 106315-106345.

赵如意, 王晓辉, 郑碧煌, 等. 基于特征优化和混合改进灰狼算法优化BiLSTM网络的短期光伏功率预测[J]. 电网技术, 2025, 49(1): 209-222.

Dong J X, Liu Y, Long M S, et al.TimeXer: empowering transformers for time series forecasting with exogenous variables[C]//Advances in Neural Information Processing Systems 37. Vancouver, BC, Canada, 2024: 469-498.

Zeng A L, Chen M X, Zhang L, et al.Are transformers effective for time series forecasting?[C]//Proceedings of the Thirty-Seventh AAAI Conference on Artificial Intelligence and Thirty-Fifth Conference on Innovative Applications of Artificial Intelligence and Thirteenth Symposium on Educational Advances in Artificial Intelligence. Washington, DC, USA, 2023: 11121-11128.

Zhuang W, Li Z H, Wang Y, et al.GCN-informer: a novel framework for mid-term photovoltaic power forecasting[J]. Applied Sciences, 2024, 14(5): 2181.

Wu H X, Hu T G, Liu Y, et al.TimesNet: temporal 2D-variation modeling for general time series analysis[C]//Proceedings of the Eleventh International Conference on Learning Representations. kigali, Rwanda, 2023: 6423-6445.

Chen M X, Lai Q X, Liu M H, et al.SCINet: time series modeling and forecasting with sample convolution and interaction[C]//Advances in Neural Information Processing Systems 35. New Orleans, Louisiana, USA, 2022: 5816-5828.

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

时间嵌入和注意力机制的超短期风速预测混合模型研究

融合经验模态分解与特征增强的Transformer-TCN光伏功率预测模型

基于多时序特征融合的TBiLSTM混合网络光伏发电功率预测

基于时频域分解-增强-融合Transformer的多因素光伏发电功率预测

基于CEEMD-SE-PSR-BP的短期风速预测