An LSTM-Based Approach for Short-Term Solar Power Forecasting with Diurnal and Intra-Day Variability
DOI:
https://doi.org/10.65780/bima.v1i2.7Keywords:
Solar power forecasting, Long short-term memory, Time-series prediction, Photovoltaic systems, Deep learningAbstract
The increasing penetration of solar photovoltaic (PV) systems into modern power grids demands accurate, reliable short-term power forecasting to ensure operational stability and efficient energy management. However, solar power generation exhibits strong nonlinearity, non-stationarity, and pronounced temporal dependencies, driven by diurnal cycles and rapid environmental variations, which pose significant challenges for conventional forecasting approaches. This study aims to develop an efficient Long Short-Term Memory (LSTM)-based framework for short-term DC power prediction that effectively captures the temporal dynamics of solar power generation while maintaining low computational complexity. The proposed approach utilizes historical power and operational data collected from two utility-scale solar PV plants in India. A comprehensive time-series preprocessing pipeline is applied, including temporal feature extraction, categorical transformation, and Min–Max normalization. Multiple LSTM architectures with varying numbers of hidden units are systematically evaluated to identify an optimal balance between model complexity and predictive performance. Model training is conducted using the Adam optimizer with exponential learning rate decay and early stopping to prevent overfitting. Experimental results demonstrate that the proposed LSTM model with a 25–50 unit configuration achieves the best performance, yielding a test Mean Squared Error of 51.92 and a prediction error of only 0.36%. Visual and quantitative analyses confirm that the model accurately reconstructs diurnal patterns and intra-day fluctuations, with strong generalization capability on unseen data. The findings indicate that a carefully configured LSTM can deliver high forecasting accuracy without relying on complex hybrid architectures or additional weather data, making it suitable for practical solar energy management applications.
Downloads
References
[1] K. Sudharshan, C. Naveen, P. Vishnuram, D. V. S. Krishna Rao Kasagani, and B. Nastasi, ‘Systematic Review on Impact of Different Irradiance Forecasting Techniques for Solar Energy Prediction’, Energies, vol. 15, no. 17, p. 6267, Jan. 2022, doi: 10.3390/en15176267.
[2] G. Moreno, C. Santos, P. Martín, F. J. Rodríguez, R. Peña, and B. Vuksanovic, ‘Intra-Day Solar Power Forecasting Strategy for Managing Virtual Power Plants’, Sensors, vol. 21, no. 16, p. 5648, Jan. 2021, doi: 10.3390/s21165648.
[3] J. Cao, X. Li, H. Zuo, J. Wang, and L. Wang, ‘Complementary Characteristics Between Hydro-Solar-Wind Power Factors in the Upper Yellow River Region During 1979~2018’, Energies, vol. 18, no. 7, p. 1648, Jan. 2025, doi: 10.3390/en18071648.
[4] M. Muzammal Islam et al., ‘Improving Reliability and Stability of the Power Systems: A Comprehensive Review on the Role of Energy Storage Systems to Enhance Flexibility’, IEEE Access, vol. 12, pp. 152738–152765, 2024, doi: 10.1109/ACCESS.2024.3476959.
[5] X. Fu, X. Wu, C. Zhang, S. Fan, and N. Liu, ‘Planning of Distributed Renewable Energy Systems Under Uncertainty Based on Statistical Machine Learning’, Protection and Control of Modern Power Systems, vol. 7, no. 4, pp. 1–27, Oct. 2022, doi: 10.1186/s41601-022-00262-x.
[6] N. Jannah, T. S. Gunawan, S. H. Yusoff, M. S. A. Hanifah, and S. N. M. Sapihie, ‘Recent Advances and Future Challenges of Solar Power Generation Forecasting’, IEEE Access, vol. 12, pp. 168904–168924, 2024, doi: 10.1109/ACCESS.2024.3496120.
[7] E. R. M. Shouman, ‘Solar Power Prediction with Artificial Intelligence’, in Advances in Solar Photovoltaic Energy Systems, IntechOpen, 2024. doi: 10.5772/intechopen.1002726.
[8] A. Bosak, D. Matushkin, L. Davydenko, L. Kulakovskyi, and V. Bronytskyi, ‘Short-Term Forecasting of Photovoltaic Solar Power Generation Based on Time Series: Application for Ensure the Efficient Operation of the Integrated Energy System of Ukraine’, in Power Systems Research and Operation: Selected Problems II, O. Kyrylenko, S. Denysiuk, D. Derevianko, I. Blinov, I. Zaitsev, and A. Zaporozhets, Eds, Cham: Springer International Publishing, 2023, pp. 159–179. doi: 10.1007/978-3-031-17554-1_8.
[9] Y. Ma, Y. Huang, and Y. Yuan, ‘The short-term forecasting of distributed photovoltaic power considering the sensitivity of meteorological data’, Journal of Cleaner Production, vol. 486, p. 144599, Jan. 2025, doi: 10.1016/j.jclepro.2024.144599.
[10] M. Zhang, Y. Han, C. Wang, P. Yang, C. Wang, and A. S. Zalhaf, ‘Ultra-short-term photovoltaic power prediction based on similar day clustering and temporal convolutional network with bidirectional long short-term memory model: A case study using DKASC data’, Applied Energy, vol. 375, p. 124085, Dec. 2024, doi: 10.1016/j.apenergy.2024.124085.
[11] K. Küllahcı and A. Altunkaynak, ‘Eigen time series modeling: a breakthrough approach to spatio-temporal rainfall forecasting in basins’, Neural Comput & Applic, vol. 37, no. 6, pp. 4471–4492, Feb. 2025, doi: 10.1007/s00521-024-10864-1.
[12] U. A. Usmani, I. Abdul Aziz, J. Jaafar, and J. Watada, ‘Deep Learning for Anomaly Detection in Time-Series Data: An Analysis of Techniques, Review of Applications, and Guidelines for Future Research’, IEEE Access, vol. 12, pp. 174564–174590, 2024, doi: 10.1109/ACCESS.2024.3495819.
[13] B. Budiman, N. Alamsyah, and R. Y. R. Alamsyah, ‘ACTIVATION FUNCTION IN LSTM FOR IMPROVED FORECASTING OF CLOSING NATURAL GAS STOCK PRICES’, jitk, vol. 10, no. 1, pp. 100–107, Aug. 2024, doi: 10.33480/jitk.v10i1.5258.
[14] S. Malakar et al., ‘Designing a long short-term network for short-term forecasting of global horizontal irradiance’, SN Appl. Sci., vol. 3, no. 4, p. 477, Apr. 2021, doi: 10.1007/s42452-021-04421-x.
[15] Z. Garip, E. Ekinci, and A. Alan, ‘Day-ahead solar photovoltaic energy forecasting based on weather data using LSTM networks: a comparative study for photovoltaic (PV) panels in Turkey’, Electr Eng, vol. 105, no. 5, pp. 3329–3345, 2023, doi: 10.1007/s00202-023-01883-7.
[16] N. Q. Nguyen, L. D. Bui, B. V. Doan, E. R. Sanseverino, D. D. Cara, and Q. D. Nguyen, ‘A new method for forecasting energy output of a large-scale solar power plant based on long short-term memory networks a case study in Vietnam’, Electr Power Syst Res, vol. 199, 2021, doi: 10.1016/j.epsr.2021.107427.
[17] Y. Gaur, V. Patel, N. D. Karelia, V. Shukla, and H. Khatri, ‘Performance of Stacked LSTM for Solar Energy Revenue Prediction’, in Int. Conf. Sustain. Energy Technol. Comput. Intell.: Towards Sustain. Energy Transit., SETCOM, Institute of Electrical and Electronics Engineers Inc., 2025. doi: 10.1109/SETCOM64758.2025.10932443.
[18] D. Sadeghi, A. Golshanfard, S. Eslami, K. Rahbar, and R. Kari, ‘Improving PV power plant forecast accuracy: A hybrid deep learning approach compared across short, medium, and long-term horizons’, Renew. Energy Focus, vol. 45, pp. 242–258, 2023, doi: 10.1016/j.ref.2023.04.010.
[19] J. Yang, S. Zhang, J. Liu, J. Liu, Y. Xiang, and X. Han, ‘Short-term Photovoltaic Power Prediction Based on Variational Mode Decomposition and Long Short-term Memory with Dual-stage Attention Mechanism’, Dianli Xitong Zidonghue, vol. 45, no. 3, pp. 174–182, 2021, doi: 10.7500/AEPS20200226011.
[20] N. Zhou, B. Shang, M. Xu, L. Peng, and G. Feng, ‘Enhancing photovoltaic power prediction using a CNN-LSTM-attention hybrid model with Bayesian hyperparameter optimization’, Glob. Energy Interconnect., vol. 7, no. 5, pp. 667–681, 2024, doi: 10.1016/j.gloei.2024.10.005.
