Comparison and Review of the Application of Deep Learning Models in Forecasting Electrical Energy Generation of Photovoltaic Systems with a Focus on LSTM and Hybrid Models

Saifoddin, Amirali; Kalantari, Mobina; Allahrabbi Shirazi, Mohammadali

doi:10.22034/stme.2026.562431.1188

Comparison and Review of the Application of Deep Learning Models in Forecasting Electrical Energy Generation of Photovoltaic Systems with a Focus on LSTM and Hybrid Models

Articles in Press

Document Type : Original Article

Authors

Amirali Saifoddin ¹

Mobina Kalantari ²

Mohammadali Allahrabbi Shirazi ²

¹ School of Energy Engineering and Sustainable Resources, Head of Soft Technologies Institute, College of Interdisciplinary Science and Technology, University of Tehran, Tehran, Iran

² School of Energy Engineering and Sustainable Resources, College of Interdisciplinary Science and Technology, University of Tehran, Tehran, Iran

10.22034/stme.2026.562431.1188

Abstract

Accurate solar energy generation forecasting is one of the main challenges in managing renewable energy systems due to the variable nature of solar radiation, dynamic weather conditions, and climate uncertainties. In this study, a comprehensive systematic review of 33 selected studies published between 2019 and 2025 was conducted to investigate the effectiveness of deep learning models in predicting solar energy generation. The main focus was on long-short-term memory (LSTM) networks and hybrid models, which have been widely used in recent years. The models were compared based on error evaluation indices including root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). The results showed that hybrid models performed better in short-term forecasts, reducing the error by about 2.9%, while models equipped with attention mechanisms provided greater accuracy in medium- and long-term horizons. In addition, the use of multi-source data including local information, satellite imagery, and meteorological data improved the forecast accuracy by about 30%. Finally, the findings indicate a growing trend of using multi-level approaches, integrating diverse data, and making models smarter to improve reliability and sustainability in solar energy resource management.

Keywords

Renewable Energy

Solar Energy Prediction

Long Short-Term Memory Networks

Deep Learning Models

Hybrid Models

Subjects