A hybrid artificial intelligence and deep learning architecture for accurate renewable energy forecasting: comprehensive case studies on wind and PV power

• Hybrid wind-PV forecasting via optimized BiGRU for nonlinear uncertainty modeling. • The proposal of SLFHBA with self-learning mechanism and mutation. • A high-precision forecasting model based on SLFHBA-BiGRU is constructed. • The robustness validation of proposed model across multi-scenario tests ensures forecasting stability. • A systematic validation and performance evaluation paradigm is established. This study proposes a novel hybrid forecasting architecture that synergistically integrates artificial intelligence and deep learning techniques. This approach effectively overcomes the limitations of conventional single models in characterizing complex weather-power coupling relationships. First, we propose an enhanced honey badger algorithm based on self-learning factors, which incorporates three key innovations: (1) the good point set principle for population initialization, (2) an adaptive dynamic learning mechanism, and (3) a Gaussian mutation strategy. Comprehensive case studies demonstrate that the proposed algorithm achieves superior performance in iterative optimization compared to existing methods. Second, we develop a short-term forecasting model for wind and photovoltaic power generation, explicitly designed to balance model effectiveness and generalizability. The self-learning honey badger algorithm is employed to optimize model parameters, enabling robust feature extraction from renewable energy output sequences. Finally, extensive validation and analysis are conducted using real-world hybrid power generation datasets combining wind and solar energy. The results show that the proposed model achieves the lowest root mean square error values and consistently exceeds 99% in R2 across diverse weather conditions. These findings provide a scientific foundation for enhancing the economic operation and dispatch of renewable energy generation systems.