Comparative Analysis of Deep Learning Architectures for Short-Term Solar Power Forecasting

Abstract

The increasing use of renewable energy sources, where solar power, has introduced a challenge in maintaining grid stability due to their intermittent and dependent on nature weather. Accurate forecasting of solar power generation is essential for efficient energy management and reliable power system operation. This paper presents a comparative analysis of traditional statistical and deep learning models for short-term solar power forecasting using a dataset. The ARIMA model is used as a baseline and compared with advanced deep learning models like, Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and a hybrid CNN-LSTM architecture. The models are evaluated using performance metrics such as MAE, RMSE, MAPE, and R², along with computational efficiency. The results show that deep learning models perform better than the traditional method, where the ARIMA model in capturing nonlinear and temporal patterns. Among them, the CNN-LSTM model achieves the highest prediction accuracy, while the GRU model offers a better trade-off between accuracy and computational cost. The findings highlight the importance of advanced data-driven techniques in improving renewable energy forecasting and supporting sustainable energy systems.

Introduction

The text discusses the global transition toward renewable energy sources such as solar and wind power, driven by environmental concerns, government regulations, and technological advancements. Unlike conventional fossil-fuel-based power plants, renewable energy generation depends heavily on weather conditions, making power output variable and difficult to control. Accurate forecasting is therefore essential to maintain grid stability, optimize energy storage and reserve management, support market operations, reduce curtailment, and improve maintenance planning.

Traditional forecasting methods such as ARIMA and SARIMA are simple and mathematically strong but struggle with the nonlinear and complex nature of renewable energy data. Recent research has shifted toward machine learning and deep learning techniques, including RNNs, LSTMs, GRUs, CNN-LSTM hybrids, and transformer-based models, which better capture temporal dependencies and improve forecasting accuracy. However, these advanced models face challenges such as high computational costs, limited reproducibility, and poor generalization across regions.

The literature review highlights the growing importance of renewable energy forecasting due to grid stability challenges caused by the intermittent nature of solar and wind energy. It compares traditional statistical, machine learning, and deep learning approaches, emphasizing the advantages and limitations of each. The study identifies key research gaps, including insufficient benchmarking of models, lack of focus on computational efficiency, and unclear benefits of hybrid CNN-LSTM models.

The proposed methodology adopts a quantitative experimental approach to compare traditional statistical models with deep learning techniques for short-term solar power forecasting. Specifically, it evaluates LSTM, GRU, and hybrid CNN-LSTM models using a real-time solar power generation dataset from India collected at 15-minute intervals. The study aims to determine the best balance between forecasting accuracy and computational efficiency while improving renewable energy integration into modern smart grid systems.

Conclusion

This paper shows the comprehensive comparison of models for solar power forecasting using a real-world dataset. The results, indicate that traditional models such as ARIMA are limited in their ability to capture the nonlinear and dynamic nature of renewable energy data. Where deep learning models shows the effectively learning temporal dependencies and complex relationships among variables. Among the evaluated models, the hybrid CNN-LSTM achieved the highest accuracy due to its ability to combine feature extraction with temporal learning. The GRU model emerged as a practical alternative, offering a good balance between prediction accuracy and computational efficiency, making it suitable for real-time and resource-constrained applications. The LSTM model also showed strong performance, particularly in capturing abrupt variations in solar power generation. Overall, the findings emphasize that deep learning approaches are more suitable for modern renewable energy forecasting tasks. Future work can focus on incorporating larger datasets, real-time deployment, and advanced architectures such as transformer-based models to further enhance forecasting accuracy and scalability.

References

[1] Abdelsattar, M., Azim, M.A., AbdelMoety, A. and Emad-Eldeen, A., (2025). Comparative analysis of deep learning architectures in solar power prediction. Scientific Reports, 15(1), p.31729. [2] Abubakar, M., Che, Y., Ivascu, L., Almasoudi, F.M. and Jamil, I., (2022). Performance analysis of energy production of large-scale solar plants based on artificial intelligence (machine learning) technique. Processes, 10(9), p.1843. [3] Abumohsen, M. and Owda, A.Y. (2024) ‘Hybrid machine learning model combining CNN-LSTM-RF for time-series forecasting of solar power generation’, e-Prime – Advances in Electrical Engineering, Electronics and Energy, 9, p. 100636. [4] Al-Ali, E.M., Hajji, Y., Said, Y., Hleili, M., Alanzi, A.M., Laatar, A.H. and Atri, M. (2023). Solar Energy Production Forecasting Based on a Hybrid CNN-LSTM-Transformer Model. Mathematics, 11(3), p.676. [5] Alizadegan, H., Rashidi Malki, B., Radmehr, A., Karimi, H. and Ilani, M.A., (2025). Comparative study of long short-term memory (LSTM), bidirectional LSTM, and traditional machine learning approaches for energy consumption prediction. Energy Exploration & Exploitation, 43(1), pp.281-301. [6] AlKandari, M. and Ahmad, I., (2024). Solar power generation forecasting using ensemble approach based on deep learning and statistical methods. Applied Computing and Informatics, 20(3/4), pp.231-250. [7] Almaghrabi, S., Rana, M., Hamilton, M. and Rahaman, M.S. (2024) ‘Multivariate solar power time-series forecasting using multilevel data fusion and deep neural networks’, Information Fusion, 104, pp. 102180. [8] Alshareef, S.M. and Talari, S. (2023) ‘Investigating the power of LSTM-based models in solar energy forecasting’, Processes, 11(5), p. 1382. [9] Benti, N.E., Chaka, M.D. and Semie, A.G. (2023) ‘Forecasting renewable energy generation with machine learning and deep learning’, Sustainability, 15(9), p. 7087. [10] Benti, N.E., Chaka, M.D. and Semie, A.G. (2023). Forecasting Renewable Energy Generation with Machine Learning and Deep Learning: Current Advances and Future Prospects. Sustainability, [online] 15(9), p.7087. [11] Bessa, R.J., Trindade, A. and Miranda, V. (2021) ‘Probabilistic wind power forecasting: An uncertainty modelling approach’, Electric Power Systems Research, 190, pp. 106–117. [12] Boucetta, L.N., Amrane, Y. and Arezki, S. (2023) ‘Comparative analysis of LSTM, GRU and MLP neural networks for short-term solar power forecasting’, in International Conference on Electrical Engineering and Advanced Technology. IEEE, pp. 1–6. [13] Chaaban, A.K. and Alfadl, N., (2024). A comparative study of machine learning approaches for an accurate predictive modeling of solar energy generation. Energy Reports, 12, pp.1293-1302. [14] Chen, J., Lu, C., Chen, C., Cheng, H. and Xuan, D. (2022). An Improved Gated Recurrent Unit Neural Network for State-of-Charge Estimation of Lithium-Ion Battery. Applied Sciences, 12(5), p.2305. [15] Chodakowska, E. et al. (2024) ‘Analysis of the effectiveness of ARIMA, SARIMA and SVR models in time-series forecasting’, Energies, 17(19), p. 4803. [16] Elsaraiti, M. and Merabet, A. (2022) ‘Solar power forecasting using deep learning techniques’, IEEE Access, 10, pp. 31692–31701. [17] Elsaraiti, M. and Merabet, A., (2022). Solar power forecasting using deep learning techniques. IEEE access, 10, pp.31692-31698. [18] Etxegarai, G., Zapirain, I., Camblong, H., Ugartemendia, J., Hernandez, J. and Curea, O., (2022). Photovoltaic energy production forecasting in a short-term horizon: Comparison between analytical and machine learning models. Applied Sciences, 12(23), p.12171. [19] Gaboitaolelwe, J., Zungeru, A.M., Yahya, A., Lebekwe, C.K., Vinod, D.N. and Salau, A.O., (2023). Machine learning based solar photovoltaic power forecasting: A review and comparison. IEEe Access, 11, pp.40820-40845. [20] Vyas, Mehali, Awadh Kishor Singh, and Nidhi Parmar. \"ANALYZING LANGUAGE IN MULTILINGUAL SPEECH USING DEEP NEURAL NETWORK.\" [21] Gopi, A., Sharma, P., Sudhakar, K., Ngui, W.K., Kirpichnikova, I. and Cuce, E., (2022). Weather impact on solar farm performance: a comparative analysis of machine learning techniques. Sustainability, 15(1), p.439. [22] Guermoui, M., Gairaa, K. and Benkaciali, S. (2025) ‘Hybrid deep learning CNN-LSTM model for forecasting direct normal irradiance’, Scientific Reports, 15, p. 6584. [23] Howlader, A.S., (2025). MACHINE LEARNING APPLICATIONS IN RENEWABLE ENERGY: PREDICTIVE ANALYTICS FOR SOLAR CELL PERFORMANCE OPTIMIZATION AND ENERGY YIELD FORECASTING. American Journal of Scholarly Research and Innovation, 4(01), pp.392-427. [24] IEA (2023) Renewables 2023. Available at: https://www.iea.org/reports/renewables-2023 [25] Khan, S., Mazhar, T., Khan, M.A., Shahzad, T., Ahmad, W., Bibi, A., Saeed, M.M. and Hamam, H., (2024). Comparative analysis of deep neural network architectures for renewable energy forecasting: enhancing accuracy with meteorological and time-based features. Discover Sustainability, 5(1), p.533. [26] Uddyalok Chakraborty, D. Thilagavathy, Suresh Kumar Sharma and Awadh Kishore Singh, “Hybrid Deep Learning with Alexnet Feature Extraction and Unet Classification for Early Detection in Leaf Diseases”, ICTACT Journal on Soft Computing Vol. 14, No. 3, pp. 3255-3262, 2024. [27] Khan, S.Z., Muzammil, N., Ghafoor, S., Khan, H., Zaidi, S.M.H., Aljohani, A.J. and Aziz, I., (2024). Quantum long short-term memory (QLSTM) vs. classical LSTM in time series forecasting: a comparative study in solar power forecasting. Frontiers in Physics, 12, p.1439180. [28] Kim, E., Akhtar, M.S. and Yang, O.B. (2023) ‘Designing solar power generation forecasting methods using time-series algorithms’, Electric Power Systems Research, 216, p. 109073. [29] Lara-Benítez, P., Carranza-García, M., Luna-Romera, J.M. and Riquelme, J.C., (2023). Short-term solar irradiance forecasting in streaming with deep learning. Neurocomputing, 546, p.126312. [30] Ledmaoui, Y., El Maghraoui, A., El Aroussi, M., Saadane, R., Chebak, A. and Chehri, A., (2023). Forecasting solar energy production: A comparative study of machine learning algorithms. Energy Reports, 10, pp.1004-1012. [31] Li, H., Wu, Q., Wang, J. and Li, Y. (2022) ‘Graph neural networks for multi-site wind power forecasting’, Renewable Energy, 190, pp. 286–295. [32] Lu, K. et al. (2025) ‘CT-PatchTST: Channel-time patch transformer for long-term renewable energy forecasting’, arXiv preprint, arXiv:2501.08620. [33] Markovics, D. and Mayer, M.J., (2022). Comparison of machine learning methods for photovoltaic power forecasting based on numerical weather prediction. Renewable and Sustainable Energy Reviews, 161, p.112364. [34] A. K. Singh, B. Kadhiwala and R. Patel, \"Hand-written Hindi Character Recognition - A Comprehensive Review,\" 2021 2nd Global Conference for Advancement in Technology (GCAT), Bangalore, India, 2021, pp. 1-5, doi: 10.1109/GCAT52182.2021.9587554. [35] Mellit, A.M.P.V.L.A., Pavan, A.M. and Lughi, V., (2021). Deep learning neural networks for short-term photovoltaic power forecasting. Renewable Energy, 172, pp.276-288. [36] Nam, N.B., Ogliari, E., Leva, S., Pafumi, E., Alberti, D. and Duong, M.Q., (2025). Comparative analysis of conformal prediction techniques and machine learning models for very short-term solar power forecasting. Energy and AI, p.100573. [37] Nguyen, H.N., Tran, Q.T., Ngo, C.T., Nguyen, D.D. and Tran, V.Q., (2025). Solar energy prediction through machine learning models: A comparative analysis of regressor algorithms. PloS one, 20(1), p.e0315955. [38] Olcay, K., Tunca, S.G. and Özgür, M.A., (2024). Forecasting and performance analysis of energy production in solar power plants using long short-term memory (LSTM) and random forest models. IEEE Access. [39] Rajendran, G., Raute, R. and Caruana, C. (2025). A Comprehensive Review of Solar PV Integration with Smart-Grids: Challenges, Standards, and Grid Codes. Energies. [online] [40] Saeed, F. and Aldera, S. (2024) ‘Adaptive renewable energy forecasting using PCA–Transformer architecture’, IEEE Access, 12, pp. 1–15. [41] Salman, D., Direkoglu, C., Kusaf, M. and Fahrioglu, M. (2024) ‘Hybrid deep learning models for solar power forecasting’, Neural Computing and Applications, 36(16), pp. 9095–9112. [42] Sedai, A., Dhakal, R., Gautam, S., Dhamala, A., Bilbao, A., Wang, Q., Wigington, A. and Pol, S., (2023). Performance analysis of statistical, machine learning and deep learning models in long-term forecasting of solar power production. forecasting, 5(1), pp.256-284. [43] Teixeira, R. et al. (2024) ‘Advancing renewable energy forecasting: A comprehensive review’, Energies, 17(14), p. 3480. [44] Theocharides, S., Theristis, M., Makrides, G., Kynigos, M., Spanias, C. and Georghiou, G.E., (2021). Comparative analysis of machine learning models for day-ahead photovoltaic power production forecasting. Energies, 14(4), p.1081. [45] Tsai, W.C., Tu, C.S., Hong, C.M. and Lin, W.M., (2023). A review of state-of-the-art and short-term forecasting models for solar PV power generation. Energies, 16(14), p.5436. [46] Wang, Y., Zhang, N., Chen, Q. and Yang, D. (2024) ‘Weather maps as tokens: Transformers for renewable forecasting’, arXiv preprint, arXiv:2511.13935. [47] Widodo, D.A., Iksan, N. and Udayanti, E.D. (2021) ‘Renewable energy power generation forecasting using deep learning’, IOP Conference Series: Earth and Environmental Science, 700(1), p. 012026. [48] Ying, C., Wang, W. and Yu, J. (2023) ‘Deep learning for renewable energy forecasting: A taxonomy and systematic review’, Journal of Cleaner Production, 384, p. 135414. [49] Zameer, A. et al. (2023) ‘Short-term solar energy forecasting using LSTM and GRU’, PLOS ONE, 18(10), e0285410. [50] Zhu, X., Li, X. and Wu, Y. (2021) ‘Short-term solar irradiance forecasting using satellite images and deep learning’, Solar Energy, 218, pp. 48–58.

Copyright

Copyright © 2026 Pooja Ajaybhai Patel, Ashka Nishant Bhalodia . This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.