Photovoltaic Power Forecasting Based on Deep Perception-Interaction Transformer

ZHANG Luxi; SHI Yan; CAO Zhiyuan; ZHANG Yi; YANG Zhenzhen

doi:10.20204/j.sp.2025.08009

English Version Reader LoginJournal Login

您当前的位置：

首页 >

文章列表页 >

Photovoltaic Power Forecasting Based on Deep Perception-Interaction Transformer

更新时间：2025-10-24

- Photovoltaic Power Forecasting Based on Deep Perception-Interaction Transformer
- Vol. 53, Issue 8, Pages: 70-78(2025)
- 作者机构：
  
  国网内蒙古东部电力有限公司供电服务监管与支持中心，内蒙古通辽 0280002.南京邮电大学理学院,江苏,南京,210003
- 作者简介：
- 基金信息：
- DOI：10.20204/j.sp.2025.08009
  CLC：
- Published：2025
- 稿件说明：
移动端阅览
张禄晞,石研,曹志远,张一,杨真真.基于深层感知交互Transformer的光伏功率预测[J].智慧电力,2025,53(8):70-78.

ZHANG Luxi, SHI Yan, CAO Zhiyuan, et al. Photovoltaic Power Forecasting Based on Deep Perception-Interaction Transformer[J]. 2025, 53(8): 70-78.
张禄晞,石研,曹志远,张一,杨真真.基于深层感知交互Transformer的光伏功率预测[J].智慧电力,2025,53(8):70-78. DOI： 10.20204/j.sp.2025.08009.

ZHANG Luxi, SHI Yan, CAO Zhiyuan, et al. Photovoltaic Power Forecasting Based on Deep Perception-Interaction Transformer[J]. 2025, 53(8): 70-78. DOI： 10.20204/j.sp.2025.08009.

摘要

为充分挖掘数据关联性和时序性，提出基于深层感知交互Transformer （DPIT）的光伏功率预测方法。首先，构建改进时间卷积网络以高效提取跨时间、跨变量依赖关系并扩展感受野；其次，构建包含改进倒置Transformer（IIT）、深度感知自注意力（DPSA）及残差自注意力（RSA）的双分支交互学习网络。其中，IIT在倒置维度上集成卷积加性自注意力、前馈网络及层归一化，以降低计算复杂度，同步提升全局学习能力与关键气象特征和多元相关性捕获性能；DPSA和RSA融合深度感知机制与自注意力机制以增强特征学习能力；最后，通过大量仿真实验验证了DPIT模型兼具低计算复杂度与预测高精度优势。

Abstract

To fully exploit data correlations and temporal dependencies

this study proposes a deep perception-interaction transformer (DPIT) based photovoltaic power forecasting method. First

an enhanced temporal convolutional network (TCN) is constructed to efficiently capture cross-temporal and cross-variable dependencies while expanding the receptive field. Second

a dual-branch interactive learning network is established

comprising an improved inverted transformer (IIT)

deep perception self-attention (DPSA)

and residual self-attention (RSA). The IIT integrates convolutional additive self-attention

feedforward networks

and layer normalization in inverted dimensions

significantly reducing computational complexity while enhancing global learning capabilities and improving the capture of critical meteorological features/multivariate correlations. DPSA and RSA synergize deep perception mechanisms with self-attention to strengthen feature learning. Finally，extensive simulations demonstrate that the DPIT model achieves dual advantages of low computational complexity and high forecasting accuracy.

关键词

Keywords

references

马成廉，韩瑞，宋萌清，等.基于多源气象数据的光伏功率预测深度学习架构[J].电网与清洁能源，2025，41（4）：104-112.

黄莉，甘恒玉，刘兴举，等.基于Transformer编码器的超短期光伏发电功率预测[J].智慧电力，2024，52（5）：16-22，59.

陈建军，李良杰，卢泽汉.基于改进LSTM-TCN模型的分布式光伏功率预测研究[J].电子设计工程，2024，32（16）：187-190，195.

IHEANETU K J.Solar photovoltaic power forecasting：A review[J].Sustainability，2022，14（24）：1-31.

谈玲，康瑞星，夏景明，等.融合多源异构气象数据的光伏功率预测模型[J].电子与信息学报，2024，46（2）：503-517.

YANG D，KLEISSL J，GUEYMARD C A，et al.History and trends in solar irradiance and PV power forecasting：A preliminary assessment and review using text mining[J].Solar Energy，2018，168（1）：60-101.

张海涛，李文娟，李雪峰，等.基于变分模态分解和时间注意力机制TCN网络的光伏发电功率预测[J].电测与仪表，2024，61（12）：156-163.

史加荣，殷诏.基于GRU-BLS的超短期光伏发电功率预测[J].智慧电力，2023，51（9）：38-45.

LUO D，WANG X.ModernTCN：a modern pure convolution structure for general time series analysis[C]// The Twelfth International Conference on Learning Representations，Vienna，Austria，2024：1-13.

HE Y，GAO Q，LIU F，et al.Short-term photovoltaic power forecasting method based on convolutional neural network[J].Energy Reports，2022，8（1）：54-62.

周浩杰，杨建卫，王尊，等.基于LSTM光伏发电功率超短期预测模型研究[J].电源技术，2023，47（6）：785-789.

MA H，ZHANG C，PENG T，et al.An integrated framework of gated recurrent unit based on improved sine cosine algorithm for photovoltaic power forecasting[J].Energy，2022，256（1）：124650.

VASWANI A.Attention is all you need[C]//International Conference on Neural Information Processing Systems，California，USA，2017：1-11.

WU H，XU J，WANG J，et al.Autoformer：decomposition transformers with auto-correlation for long-term series forecasting[J].Neural Information Processing Systems，2021，34（1）：22419-22430.

XU H，PENG Q，WANG Y，et al. Power-load forecasting model based on informer and its application[J]. Energies，2023，16（7）:1-14.

ZHOU T，MA Z，WEN Q，et al.FEDformer：frequency enhanced decomposed transformer for long-term series forecasting[C]//International Conference on Machine Learning，Maryland，USA，2022：1-19.

KITAEV N，KAISER L，LEVSKAYA A.Reformer：The efficient transformer[C]//International Conference on Learning Representa-tions，Ababa，Ethiopia，2020：1-12.

HUANG Z，SHI X，ZHANG C，et al.Flowformer：a transformer architecture for optical flow[C]//European Conference on Computer Vision （ECCV），Tel Aviv，Israel，2022：668-685.

DAO T，FU D，ERMON S，et al.Flashattention：fast and memory-efficient exact attention with io-awareness[C]//The 36th Conference on Neural Information Processing Systems，Louisiana，USA，2022：1-16.

LIU Y，HU T，ZHANG H，et al.iTransformer：inverted transformers are effective for time series forecasting[C]//International Conference on Learning Representations，Vienna，Austria，2024：1-25.

TAO K，ZHAO J，TAO Y，et al.Operational day-ahead photovoltaic power forecasting based on transformer variant[J].Applied Energy，2024，373（1）：1-16.

王瑞，靳鑫鑫，逯静.SVMD-PE-BP-Transformer短期光伏功率预测[J].电网与清洁能源，2024，40（8）：141-150.

KIM J，OBREGON J，PARK H，et al. Multi-step photovoltaic power forecasting using transformer and recurrent neural networks[J].Renewable and Sustainable Energy Reviews，2024，200（1）：1-18.

周伊，肖先勇，赵清华，等.基于组合数据清洗与改进注意力机制的光伏发电功率预测[J].供用电，2024，41（10）：31-37，49.

杨鹏伟，赵丽萍，陈军法，等.基于相似时段匹配与Transformer网络建模的分布式光伏超短期功率预测方法[J].中国电力，2024，57（12）：60-70.

SHAKER A，MAAZ M，RASHEED H，et al.Swiftformer：Efficient additive attention for transformer-based real-time mobile vision applications[C]// IEEE/CVF International Conference on Computer Vision，Paris，France，2023：17425-17436.

ZENG A，CHEN M，ZHANG L，et al.Are transformers effective for time series forecasting?[C]//AAAI conference on Artificial Intelligence，Washington，USA，2023，37（9）：11121-11128.

王育飞，汪弦哲，薛花，等.基于BO-SVM和ISO改进的ENN光伏功率超短期预测模型[J].太阳能学报，2025，46（6）:280-288.

卢晓凯，封军，韩永强，等.GraphMLP-Mixer：基于图-多层感知机架构的高效多行为序列推荐方法[J].计算机研究与发展， 2024， 61（8）: 1917-1929.

ZHANG F，WU P，HU C，et al.CAS-FGvit：an efficient convolutional additive mixer for fine-grained classification[J].The Visual Computer，2025，41（1）：1-11.

YAO T，WANG J，WU H，et al.A photovoltaic power output dataset：multi- source photovoltaic power output dataset with Python toolkit[J].Solar Energy，2021，230：122-130.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Short-term Power Load Forecasting Based on VMD-TCN-BiLSTM-Attention

PvIR-DCANet Based Infrared Image Defect Detection for Photovoltaic Modules

Energy Disaggregation of Commercial Buildings Based on Attention-LSTM

Non-invasive Household Load Decomposition Based on GRNN and Attention Mechanism Model

Image Segmentation Method of Transmission Line Insulator Based on Improved U-Net

Related Author

DAI Jie

LI Guoliang

LIU Yiyan

TIAN Yizhi

CHAI Shaobo

WANG Fan

PAN Wei

LYU Xiaojie

Related Institution

College of Energy and Electrical Engineering，Chang’an University

College of Electric Engineering，Xinjiang University

Tianjin Tianda Qiushi Electric Power High Technology Co.，Ltd.

Guangdong Power Grid Co.， Ltd. Guangzhou Power Supply Bureau

State Grid Shanghai Electrical Power Research Institute

AI问答

⁰