Accurate solar power prediction is crucial for ensuring grid stability and optimizing energy management especially as renewable enegy sources continue to grow in importance. A novel Bayesian-optimized ensemble model that integrates the strengths of two distinct forecasting approaches: SARIMAX (Seasonal AutoRegressive Integrated Moving Average with eXogenous regressors) and LSTM (Long Short-Term Memory) is developed. The SARIMAX model is well-suited for capturing linear temporal trends and seasonality, while the LSTM model is effective in modeling complex nonlinear patterns and dynamics inherent in solar power data. To ensure optimal performance, Bayesian optimization is employed for hyperparameter tuning enhancing both models' forecasting capabilities. The ensemble approach combines the predictions of SARIMAX and LSTM using a weighted average leveraging their complementary strengths. The proposed model is validated using real-world solar power data showing superior forecasting accuracy and lower predcition losses measured in terms of MAE, RMSE compared to standalone SARIMAX and LSTM models. The proposed work highlights the potential of combining traditional statistical models with modern deep learning techniques to achieve more robust and reliable solar power predictions, a critical component for future energy systems.