Performance Analysis of Different Machine Learning Techniques in Solar Energy Forecasting

Solar power forecasting is crucial for optimizing energy management and integrating renewable energy into power grids. This work compares two DL method, and Transformer—to determine their effectiveness in predicting solar photovoltaic (PV) power generation. Using publicly available meteorological datasets, the models were trained and calculated through Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> score. Findings indicate LSTM models capture sequential patterns effectively, improving accuracy but requiring longer training times. The Transformer model demonstrates superior performance in handling long-term dependencies while maintaining high accuracy, though at a higher computational cost. These models have practical applications in grid optimization, battery storage management, and smart energy systems. The study concludes that deep learning techniques, particularly LSTM and Transformer models, significantly enhance solar power forecasting accuracy. Future research can explore hybrid models combining LSTM and Transformer architectures for improved performance and efficiency. This project contributes to advancing renewable energy forecasting, supporting the transition toward sustainable power systems.