A Novel Hybrid Deep Learning Framework for Enhanced Solar Power Prediction

Solar energy has emerged as a clean and sustainable alternative to conventional energy sources, making it a preferred choice for widespread integration into power grids. Despite its merits, the inherent unpredictability of solar energy presents a significant challenge, particularly in sectors with critical power demands. This study introduces an innovative hybrid deep learning approach that integrates a recurrent neural network (RNN), transformer (Tx), and long short-term memory (LSTM) network to achieve prediction accuracy. A multi-head selfattention mechanism is employed to capture complex dependencies across different temporal scales, yielding superior results, especially under unpredictable weather conditions. The proposed hybrid model demonstrates exceptional accuracy, outperforming recent traditional methods. The finding of this study emphasizes the reliability of the hybrid approach, facilitating the smooth integration of solar electricity into the power demand.