Spatiotemporal attention network for ultra-short-term photovoltaic power forecasting considering spatiotemporal correlations and multiple environmental factors

Ming Yang; Zehao Wang; Haipeng Chen; Ming Yang; Zehao Wang; Haipeng Chen

doi:10.3934/energy.2025041

AIMS Energy

2025, Volume 13, Issue 5: 1104-1132. doi: 10.3934/energy.2025041

Previous Article Next Article

Research article

Spatiotemporal attention network for ultra-short-term photovoltaic power forecasting considering spatiotemporal correlations and multiple environmental factors

1.
Shandong University School of Electrical Engineering, Jinan 250100, China
2.
Key Laboratory of Modern Power System Simulation and Control & Renewable Energy Technology, Ministry of Education, Northeast Electric Power University, Jilin 132012, China

Received: 18 June 2025 Revised: 24 August 2025 Accepted: 05 September 2025 Published: 12 September 2025

The accuracy of photovoltaic (PV) power forecasting greatly influences power system operation and control. However, models often fail to simultaneously capture spatial correlations among distributed PV stations and temporal patterns in power time series. To overcome this challenge, we proposed a Spatiotemporal Attention Network (STAN). First, autocorrelation, cross-correlation, and smoothing effects in PV systems were examined, forming a theoretical foundation for prediction. Then, multi-head self-attention was applied to extract spatial features across stations, while a sequence-to-sequence model with global attention captured temporal dependencies. Case studies demonstrated that compared with conventional Convolutional Neural Network-Long Short-Term Memory, STAN reduced MAE by 45.6% and RMSE by 32.8%, effectively enhancing forecasting accuracy and minimizing prediction errors.
- spatiotemporal attention network,
- spatiotemporal correlation,
- PV power prediction,
- attention mechanism
Citation: Ming Yang, Zehao Wang, Haipeng Chen. Spatiotemporal attention network for ultra-short-term photovoltaic power forecasting considering spatiotemporal correlations and multiple environmental factors[J]. AIMS Energy, 2025, 13(5): 1104-1132. doi: 10.3934/energy.2025041

Related Papers:

Abstract

The accuracy of photovoltaic (PV) power forecasting greatly influences power system operation and control. However, models often fail to simultaneously capture spatial correlations among distributed PV stations and temporal patterns in power time series. To overcome this challenge, we proposed a Spatiotemporal Attention Network (STAN). First, autocorrelation, cross-correlation, and smoothing effects in PV systems were examined, forming a theoretical foundation for prediction. Then, multi-head self-attention was applied to extract spatial features across stations, while a sequence-to-sequence model with global attention captured temporal dependencies. Case studies demonstrated that compared with conventional Convolutional Neural Network-Long Short-Term Memory, STAN reduced MAE by 45.6% and RMSE by 32.8%, effectively enhancing forecasting accuracy and minimizing prediction errors.

References

[1]	Lin H, Gao L, Cui M, et al. (2025) Short-term distributed photovoltaic power prediction based on temporal self-attention mechanism and advanced signal decomposition techniques with feature fusion. Energy 315: 134395. https://doi.org/10.1016/j.energy.2025.134395 doi: 10.1016/j.energy.2025.134395
[2]	Chen Y, Xiao JW, Wang YW, et al. (2023) Regional wind-photovoltaic combined power generation forecasting based on a novel multi-task learning framework and TPA-LSTM. Energy Convers Manage 297: 117715. https://doi.org/10.1016/j.enconman.2023.117715 doi: 10.1016/j.enconman.2023.117715
[3]	Xiang Y, Tang Q, Xu W, et al. (2024) A multi-factor spatio-temporal correlation analysis method for PV development potential estimation. Renewable Energy 223: 119962. https://doi.org/10.1016/j.renene.2024.119962 doi: 10.1016/j.renene.2024.119962
[4]	Shi J, Wang Y, Zhou Y, et al. (2024) Bayesian optimization—LSTM modeling and time frequency correlation mapping based probabilistic forecasting of ultra-short-term photovoltaic power outputs. IEEE Trans Ind Appl 60: 2422–2430. https://doi.org/10.1109/TIA.2023.3334700 doi: 10.1109/TIA.2023.3334700
[5]	Dai X, Liu GP, Hu W (2023) An online-learning-enabled self-attention-based model for ultra-short-term wind power forecasting. Energy 272: 127173. https://doi.org/10.1016/j.energy.2023.127173 doi: 10.1016/j.energy.2023.127173
[6]	Wang L, Liu Y, Li T, et al. (2020) Short-term PV power prediction based on optimized VMD and LSTM. IEEE Access 8: 165849–165862. https://doi.org/10.1109/ACCESS.2020.3022246 doi: 10.1109/ACCESS.2020.3022246
[7]	Wang Y, Yang Q, Xue H, et al. (2022) Ultra-short-term PV power prediction model based on HP-OVMD and enhanced emotional neural network. IET Renew Power Gener 16: 2233–2247. https://doi.org/10.1049/rpg2.12514 doi: 10.1049/rpg2.12514
[8]	Houran MA, Bukhari SMS, Zafar MH, et al. (2023) COA-CNN-LSTM: Coati optimization algorithm-based hybrid deep learning model for PV/wind power forecasting in smart grid applications. Appl Energy 349: 121638. https://doi.org/10.1016/j.apenergy.2023.121638 doi: 10.1016/j.apenergy.2023.121638
[9]	Yang C, Li S, Gou Z (2025) Spatiotemporal prediction of urban building rooftop photovoltaic potential based on GCN-LSTM. Energy Build 334: 115522. https://doi.org/10.1016/j.enbuild.2025.115522 doi: 10.1016/j.enbuild.2025.115522
[10]	Dimitriadis CN, Passalis N, Georgiadis MC (2025) A deep learning framework for photovoltaic power forecasting in multiple interconnected countries. Sustainable Energy Technol Assess 77: 104330. https://doi.org/10.1016/j.seta.2025.104330 doi: 10.1016/j.seta.2025.104330
[11]	Passalis N, Dimitriadis CN, Georgiadis MC (2025) Residual adaptive input normalization for forecasting renewable energy generation in multiple countries. Pattern Recognit Lett 196: 52–58. https://doi.org/10.1016/j.patrec.2025.05.008 doi: 10.1016/j.patrec.2025.05.008
[12]	Qu J, Qian Z, Pei Y (2021) Day-ahead hourly photovoltaic power forecasting using attention-based CNN-LSTM neural network embedded with multiple relevant and target variables prediction pattern. Energy 232: 120996. https://doi.org/10.1016/j.energy.2021.120996 doi: 10.1016/j.energy.2021.120996
[13]	Mirza AF, Shu Z, Usman M, et al. (2024) Quantile-transformed multi-attention residual framework (QT-MARF) for medium-term PV and wind power prediction. Renewable Energy 220: 119604. https://doi.org/10.1016/j.renene.2023.119604 doi: 10.1016/j.renene.2023.119604
[14]	Yang J, He H, Zhao X, et al. (2024) Day-ahead PV power forecasting model based on fine-grained temporal attention and cloud-coverage spatial attention. IEEE Trans Sustainable Energy 15: 1062–1073. https://doi.org/10.1109/TSTE.2023.3326887 doi: 10.1109/TSTE.2023.3326887
[15]	Pan C, Liu Y, Oh Y, et al. (2024) Short-term photovoltaic power forecasting using PV data and sky images in an auto cross modal correlation attention multimodal framework. Energies 17: 6378. https://doi.org/10.3390/en17246378 doi: 10.3390/en17246378
[16]	Zhou Z, Dai Y, Leng M (2025) A photovoltaic power forecasting framework based on Attention mechanism and parallel prediction architecture. Appl Energy 391: 125869. https://doi.org/10.1016/j.apenergy.2025.125869 doi: 10.1016/j.apenergy.2025.125869
[17]	Zhao P, Hu W, Cao D, et al. (2025) Causal mechanism-enabled zero-label learning for power generation forecasting of newly-built PV sites. IEEE Trans Sustainable Energy 16: 392–406. https://doi.org/10.1109/TSTE.2024.3459415 doi: 10.1109/TSTE.2024.3459415
[18]	Ahmadi M, Samet H, Ghanbari T (2020) Series arc fault detection in photovoltaic systems based on signal-to-noise ratio characteristics using cross-correlation function. IEEE Trans Ind Inf 16: 3198–3209. https://doi.org/10.1109/TⅡ.2019.2909753 doi: 10.1109/TⅡ.2019.2909753
[19]	Meng L, Yang X, Zhu J, et al. (2024) Network partition and distributed voltage coordination control strategy of active distribution network system considering photovoltaic uncertainty. Appl Energy 362: 122846. https://doi.org/10.1016/j.apenergy.2024.122846 doi: 10.1016/j.apenergy.2024.122846
[20]	Benavides D, Arévalo P, Villa-Á vila E, et al. (2024) Predictive power fluctuation mitigation in grid-connected PV systems with rapid response to EV charging stations. J Energy Storage 86: 111230. https://doi.org/10.1016/j.est.2024.111230 doi: 10.1016/j.est.2024.111230
[21]	Jebli I, Belouadha FZ, Kabbaj MI, et al. (2021) Prediction of solar energy guided by pearson correlation using machine learning. Energy 224: 120109. https://doi.org/10.1016/j.energy.2021.120109 doi: 10.1016/j.energy.2021.120109
[22]	Yang M, Jiang Y, Guo Y, et al. (2025) Ultra-short-term prediction of photovoltaic cluster power based on spatiotemporal convergence effect and spatiotemporal dynamic graph attention network. Renewable Energy 255: 123843. https://doi.org/10.1016/j.renene.2025.123843 doi: 10.1016/j.renene.2025.123843
[23]	Yu H, Chen S, Chu Y, et al. (2024) Self-attention mechanism to enhance the generalizability of data-driven time-series prediction: A case study of intra-hour power forecasting of urban distributed photovoltaic systems. Appl Energy 374: 124007. https://doi.org/10.1016/j.apenergy.2024.124007 doi: 10.1016/j.apenergy.2024.124007
[24]	Wang J, Ye L, Ding X, et al. (2024) A novel seasonal grey prediction model with time-lag and interactive effects for forecasting the photovoltaic power generation. Energy 304: 131939. https://doi.org/10.1016/j.energy.2024.131939 doi: 10.1016/j.energy.2024.131939
[25]	Boucetta LN, Amrane Y, Chouder A, et al. (2024) Enhanced forecasting accuracy of a grid-connected photovoltaic power plant: A novel approach using hybrid variational mode decomposition and a CNN-LSTM model. Energies 17: 1781. https://doi.org/10.3390/en17071781 doi: 10.3390/en17071781
[26]	Fu H, Zhang J, Xie S (2024) A novel improved variational mode decomposition-temporal convolutional network-gated recurrent unit with multi-head attention mechanism for enhanced photovoltaic power forecasting. Electronics 13: 1837. https://doi.org/10.3390/electronics13101837 doi: 10.3390/electronics13101837
[27]	Jurado M, Samper M, Rosés R (2023) An improved encoder-decoder-based CNN model for probabilistic short-term load and PV forecasting. Electr Power Syst Res 217: 109153. https://doi.org/10.1016/j.epsr.2023.109153 doi: 10.1016/j.epsr.2023.109153
[28]	Saffari M, Khodayar M, Jalali SMJ, et al. (2021) Deep convolutional graph rough variational auto-encoder for short-term photovoltaic power forecasting. 2021 International Conference on Smart Energy Systems and Technologies (SEST), 1–6. https://doi.org/10.1109/SEST50973.2021.9543326
[29]	Liu W, Mao Z (2024) Short-term photovoltaic power forecasting with feature extraction and attention mechanisms. Renewable Energy 226: 120437. https://doi.org/10.1016/j.renene.2024.120437 doi: 10.1016/j.renene.2024.120437
[30]	Tong J, Xie L, Fang S, et al. (2022) Hourly solar irradiance forecasting based on encoder—decoder model using series decomposition and dynamic error compensation. Energy Convers Manage 270: 116049. https://doi.org/10.1016/j.enconman.2022.116049 doi: 10.1016/j.enconman.2022.116049

Reader Comments

Your name:*

Email:*
© 2025 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)