Deep Learning Based Prediction of Renewable Energy Using Hybrid CNN & LSTM Model

Abstract

Renewable energy sources such as solar photovoltaic (PV) and wind power are vital for sustainable development and reducing dependence on fossil fuels, but their power generation is highly intermittent due to changing meteorological conditions like solar irradiance, temperature, wind speed, and cloud cover, which creates challenges in maintaining grid stability and efficient energy management. Accurate forecasting is therefore essential for reliable power system operation, enabling optimal energy dispatch, reserve planning, and cost reduction. Conventional forecasting methods based on physical and statistical models rely on historical and weather data but often fail to capture complex nonlinear relationships and dynamic variations. To overcome these limitations, a hybrid deep learning approach combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks is proposed, where CNN extracts important features from input data and LSTM captures temporal dependencies and long-term patterns in time-series data, improving prediction accuracy. The model is implemented in MATLAB Simulink and evaluated using performance metrics such as Mean Absolute Error (MAE) and Root Mean Square Error (RMSE), with results demonstrating improved accuracy and robustness compared to traditional methods, thereby supporting effective integration of renewable energy into modern power systems.

Introduction

The text explains the importance of renewable energy forecasting and proposes a hybrid deep learning model for improved prediction.

Energy demand is increasing, but traditional fossil fuels cause pollution and are non-renewable. Renewable sources like solar and wind are cleaner but highly variable due to changing weather conditions, making accurate power prediction essential for grid stability and efficient energy management.

To address this, the study proposes a hybrid CNN–LSTM model. The CNN extracts spatial features from meteorological and historical data, while the LSTM captures temporal patterns and long-term dependencies in time-series data. The system follows steps such as data collection, preprocessing, feature extraction, model training, and evaluation.

The model uses inputs like solar irradiance, temperature, wind speed, and past power output. Data is cleaned, normalized, and split into training, validation, and testing sets. The hybrid architecture combines the strengths of CNN (feature extraction) and LSTM (sequence learning) to better handle complex, nonlinear relationships.

Results show that the proposed CNN–LSTM model outperforms traditional models like ANN, CNN, and LSTM alone, achieving lower error rates (MAE and RMSE) and higher prediction accuracy. It accurately tracks real power output under different conditions, demonstrating strong reliability and efficiency.

Conclusion

The increasing integration of renewable energy sources into modern power systems has introduced significant operational challenges due to the intermittent and stochastic nature of solar photovoltaic (PV) and wind energy. Variations in meteorological parameters such as solar irradiance, temperature, and wind speed result in fluctuating power generation, which affects grid stability, reserve management, and economic dispatch. Accurate forecasting of renewable energy output is therefore essential for ensuring reliable system operation and efficient utilization of sustainable resources. This paper presented a hybrid deep learning framework based on Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks for renewable energy prediction. Conventional statistical and physical forecasting methods often struggle to model the nonlinear and time-varying relationships between environmental conditions and power output. Although standalone machine learning models offer improved performance, they typically require manual feature extraction and exhibit limitations in capturing long-term temporal dependencies. The proposed CNN–LSTM architecture integrates spatial feature extraction and temporal modeling within a unified framework. The CNN component automatically learns meaningful correlations among meteorological variables and historical power data, eliminating the need for handcrafted features. The LSTM component effectively captures sequential dependencies and long-term trends in time-series data through its gated memory mechanism. By combining these capabilities, the hybrid model enhances prediction accuracy and robustness compared to standalone CNN or LSTM approaches. Performance evaluation using statistical metrics such as Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) demonstrates the effectiveness of the proposed method in forecasting wind and PV power generation. Improved prediction accuracy contributes to optimized energy scheduling, enhanced grid reliability, and better renewable energy integration within smart grid and microgrid systems. Future work may focus on incorporating attention mechanisms, probabilistic forecasting techniques, and real-time data acquisition to further improve forecasting performance and adaptability.

References

[1] H. Wang, Z. Lei, X. Zhang, B. Zhou, and J. Peng, “A review of deep learning for renewable energy forecasting,” Energy Conversion and Management, vol. 198, p. 111799, 2019. [2] M. Q. Raza, M. Nadarajah, and C. Ekanayake, “On recent advances in PV output power forecasting,” Solar Energy, vol. 136, pp. 125–144, 2016. [3] S. Alessandrini, L. Delle Monache, and J. McCandless, “Short-term wind power forecasting based on numerical weather prediction,” Renewable Energy, vol. 156, pp. 188–200, 2020. [4] A. Mellit and S. A. Kalogirou, “Artificial intelligence techniques for photovoltaic applications: A review,” Prog. Energy Combust. Sci., vol. 34, no. 5, pp. 574–632, 2008. [5] M. Husein and I. Chung, “A survey on short-term wind and solar power forecasting methods,” Energy Reports, vol. 9, pp. 1013–1030, 2023. [6] G. E. Box, G. M. Jenkins, and G. C. Reinsel, Time Series Analysis: Forecasting and Control, 4th ed. Hoboken, NJ, USA: Wiley, 2008. [7] Y. Dou, J. Wang, and X. Zhang, “Limitations of traditional statistical models for renewable energy forecasting,” Frontiers in Energy Research, vol. 11, 2023. [8] A. Kaur and L. Kaur, “Hybrid CNN–LSTM model for wind energy forecasting,” Int. J. Energy Res., vol. 45, no. 10, pp. 14728–14740, 2021. [9] M. Abedin, S. F. Nizami, and M. Alam, “Deep learning-based hybrid model for solar energy prediction using CNN and LSTM,” IEEE Access, vol. 9, pp. 165432–165445, 2021. [10] Riyas, P., Lakshmanan, S.A. Optimal tuning of PI based LF for three-phase SRF PLL synchronization system using pity beetle algorithm under grid abnormalities. Sci Rep 15, 18891 (2025). https://doi.org/10.1038/s41598-025-03530-6 [11] P. Riyas and S. A. Lakshmanan, \"Comparative Analysis of Algorithms for the Optimum Placement of PMUs in Power Systems,\" 2023 IEEE IAS Global Conference on Renewable Energy and Hydrogen Technologies (GlobConHT), Male, Maldives, 2023, pp. 1-7, doi: 10.1109/GlobConHT56829.2023.10087710 [12] Riyas, P., Lakshmanan, S., A Comprehensive Analysis of Three-Phase PLL-Based Grid Synchronization Methods, (2024) International Review of Electrical Engineering (IREE), 19 (6), pp. 432-445.doi:https://doi.org/10.15866/iree.v19i6.25016 [13] J. Li, Y. Zhao, and H. Sun, “Deep learning-based short-term photovoltaic power forecasting: A review,” Applied Energy, vol. 350, p. 121713, 2024. [14] A. Verdone, “A review of solar and wind energy forecasting: From single-site to multi-site predictions,” Renewable and Sustainable Energy Reviews, vol. 195, p. 113040, 2025. [15] F. Aksan et al., “Comparative analysis of machine learning models for PV power prediction,” Energies, vol. 18, no. 22, p. 5980, 2025. [16] A. Hassan et al., “Multi-label deep learning for photovoltaic power forecasting,” Scientific Reports, vol. 15, pp. 1–14, 2025. [17] M. Nandy et al., “CNN–LSTM deep learning framework for renewable energy forecasting,” E3S Web of Conferences, vol. 426, 2025. [18] N. Guo et al., “Advanced hybrid deep learning model for short-term wind power forecasting,” Renewable Energy, vol. 224, pp. 1–15, 2026.

Copyright

Copyright © 2026 Mr. Riyas P P, Musammil Hamza, Anoop P, Muhammed Swalih V, Muhammed Jasil P, Shabeeb Rahman P. 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.