Signal-Guided Dual-Predictor Learning for Ultra-Short-Term PV Power Forecasting

Authors

  • Shumei Zhang School of Automation and Electrical Engineering, Lanzhou University of Technology, Lanzhou, China
  • Zhiwen Wang School of Automation and Electrical Engineering, Lanzhou University of Technology, Lanzhou, China; Key Laboratory of Gansu Advanced Control for Industrial Processes, Lanzhou University of Technology, Lanzhou, China
  • Haoxu Wang School of Automation and Electrical Engineering, Lanzhou University of Technology, Lanzhou, China
  • Wannian Wei School of Automation and Electrical Engineering, Lanzhou University of Technology, Lanzhou, China
  • Wei Qi School of Automation and Electrical Engineering, Lanzhou University of Technology, Lanzhou, China
  • Hao Liu School of Automation and Electrical Engineering, Lanzhou University of Technology, Lanzhou, China

DOI:

https://doi.org/10.54097/kkp2fq98

Keywords:

Renewable Energy, Deep Learning, Dual-predictor Learning, Multi-scale TCN, GRU

Abstract

To address the insufficient adaptability of a unified predictor caused by the non-stationarity of photovoltaic (PV) power series, pronounced multi-scale fluctuations, and differences in the predictable structures of decomposed subcomponents, this paper proposes an ultra-short-term PV power forecasting method based on CEEMDAN-VMD and signal-property-guided dual predictors. First, Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) is employed to decompose the original power series, and the resulting Intrinsic Mode Functions (IMFs) are reconstructed into high-, medium-, and low-frequency sequences using K-means clustering. Subsequently, Variational Mode Decomposition (VMD) is applied only to the high-frequency component for secondary decomposition. Furthermore, Sample Entropy (SE), Central Frequency (CF), Energy Proportion (EP), First-order Autocorrelation Coefficient (Lag1ACF), and Fluctuation Index (FI) are used to characterize the properties of the final components, based on which differentiated modeling is performed. Specifically, H1, H2, and Low are modeled using the Regularity Predictor (RP), whereas H3–H6 and Mid are modeled using the Dynamic Predictor (DP), which integrates a multi-scale dilated Temporal Convolutional Network (TCN) with a Gated Recurrent Unit (GRU). The forecasting results of all components are directly summed to obtain the final PV power forecast. Experiments conducted on seasonal datasets from a PV power station in Northwest China showed that, compared with several representative deep learning models, the proposed method reduced the Mean Absolute Error (MAE) by 72.49%, 61.54%, 61.79%, and 44.10% in spring, summer, autumn, and winter, respectively, and reduced the Root Mean Square Error (RMSE) by 79.91%, 68.61%, 77.33%, and 63.16%, respectively, with the coefficient of determination (R²) ranging from 0.9948 to 0.9991. The Wilcoxon test indicated that the performance improvements were statistically significant. On the independent Alice Springs dataset in Australia, the MAE, RMSE, and coefficient of determination (R²) were 0.0793, 0.1308, and 0.9988, respectively, further validating the generalization capability of the model. The results demonstrate that component-level differentiated forecasting based on signal properties can effectively improve the accuracy and model adaptability of ultra-short-term PV power forecasting.

Downloads

Download data is not yet available.

References

[1] Sabadus, A., Blaga, R., Hategan, S. M., & Pop, N. (2024). A cross-sectional survey of deterministic PV power forecasting: Progress and limitations in current approaches. Renewable Energy, 226, Article 120385. https://doi.org/10.1016/j.renene.2024.120385

[2] Limouni, T., Yaagoubi, R., Bouziane, K., Guissi, K., & Baali, E. H. (2023). Accurate one step and multistep forecasting of very short-term PV power using LSTM-TCN model. Renewable Energy, 205, 1010-1024. https://doi.org/10.1016/j.renene.2023.01.079

[3] Wang, M., Rao, C. J., Xiao, X. P., Zhang, Y., & Li, H. (2024). Efficient shrinkage temporal convolutional network model for photovoltaic power prediction. Energy, 297, Article 131295. https://doi.org/10.1016/j.energy.2024.131295

[4] Dai, Y. M., Yu, W. J., & Leng, M. M. (2024). A hybrid ensemble optimized BiGRU method for short-term photovoltaic generation forecasting. Energy, 299, Article 131458. https://doi.org/10.1016/j.energy.2024.131458

[5] Xiong, B. Y., Chen, Y. T., Chen, D. L., Wang, H., & Li, J. (2025). Deep probabilistic solar power forecasting with Transformer and Gaussian process approximation. Applied Energy, 382, Article 125294. https://doi.org/10.1016/j.apenergy.2025.125294

[6] Liu, Y. F., Liu, Y., Cai, H. H., Zhang, L., & Wang, Q. (2023). An innovative short-term multihorizon photovoltaic power output forecasting method based on variational mode decomposition and a capsule convolutional neural network. Applied Energy, 343, Article 121139. https://doi.org/10.1016/j.apenergy.2023.121139

[7] Wang, L. N., Mao, M. X., Xie, J. L., Lv, Z., & Zhang, H. (2023). Accurate solar PV power prediction interval method based on frequency-domain decomposition and LSTM model. Energy, 262, Article 125592. https://doi.org/10.1016/j.energy.2022.125592

[8] Wu, T., Hu, R. F., Zhu, H. Y., Chen, Y., & Liu, X. (2024). Combined IXGBoost-KELM short-term photovoltaic power prediction model based on multidimensional similar day clustering and dual decomposition. Energy, 288, Article 129770. https://doi.org/10.1016/j.energy.2023.129770

[9] Liu, L., Zhang, J. Q., & Xue, S. B. (2025). Photovoltaic power forecasting: Using wavelet threshold denoising combined with VMD. Renewable Energy, 249, Article 123152. https://doi.org/10.1016/j.renene.2025.123152

[10] Zhou, Z. D., Dai, Y. M., & Leng, M. M. (2025). A photovoltaic power forecasting framework based on attention mechanism and parallel prediction architecture. Applied Energy, 391, Article 125869. https://doi.org/10.1016/j.apenergy.2025.125869

[11] Xue, S. Y., & Li, L. (2026). Photovoltaic power forecasting based on secondary decomposition strategy and hybrid model. Scientific Reports, 16, Article 12915. https://doi.org/10.1038/s41598-026-12915-8

[12] Yang, Z. H., Li, J. D., Wang, H. T., & Zhang, Y. (2025). An Informer model for very short-term power load forecasting. Energies, 18(6), Article 1523. https://doi.org/10.3390/en18061523

[13] Wang, S., Yan, S., Li, H. J., & Zhang, X. (2024). Short-term prediction of photovoltaic power based on quadratic decomposition and residual correction. Electric Power Systems Research, 236, Article 110948. https://doi.org/10.1016/j.epsr.2024.110948

[14] Wang, L. S., Liu, Y. H., Li, T. S., & Xie, X. (2020). Short-term PV power prediction based on optimized VMD and LSTM. IEEE Access, 8, 165849-165862. https://doi.org/10.1109/ACCESS.2020.3022632

[15] Li, Y., Song, L., Zhang, S., Wang, Y., & Chen, H. (2023). A TCN-based hybrid forecasting framework for hours-ahead utility-scale PV forecasting. IEEE Transactions on Smart Grid, 14(5), 4058-4071. https://doi.org/10.1109/TSG.2023.3261984

[16] Pei, J. X., Liu, N., Shi, J. Q., Wang, H., & Li, Y. (2025). Tackling the duck curve in renewable power system: A multi-task learning model with iTransformer for net-load forecasting. Energy Conversion and Management, 326, Article 119458. https://doi.org/10.1016/j.enconman.2025.119458

[17] Li, M. K., Sun, L. T., & Sun, Y. T. (2026). Long-term and short-term photovoltaic power generation forecasting using a multi-scale fusion MHA-BiLSTM model. Energies, 19(2), Article 411. https://doi.org/10.3390/en19020411

[18] Soltani, I., & Shahnia, F. (2026). Hybrid CNN-LSTM modeling with attention mechanism for solar power forecasting. Discover Applied Sciences, 8, Article 145. https://doi.org/10.1007/s42452-026-07341-2

Downloads

Published

29-09-2026

Issue

Vol. 22 No. 3 (2026)

Section

Articles

License

Copyright (c) 2026 Journal of Computing and Electronic Information Management

Creative Commons License

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.