As the energy sector grapples with increasing demand and the growing integration of power grids with renewable resources, the inherently stochastic nature of renewable energy sources poses a significant challenge. Predicting power generation from renewables offers a potent solution. This research paper delves into the use of machine learning models to forecast power generation from solar panels, a key element in optimizing power system stability and management. We examine the performance of various models, including LASSO, linear regression, SVM, gradient boosting, random forest, XGBoost, and XGBoost with hyperparameter tuning techniques like grid search, random search, and Bayesian optimization, to determine which offers the most accurate and reliable predictions for solar power generation. In the data analytical phase, we employed Isolation Forest for outlier prediction, aiding in data cleaning and providing much improved results. The XGBoost model, when coupled with Bayesian optimization, exhibits superior accuracy and reliability in predicting solar power generation. Our results suggest that the proposed model, XGBoost with Bayesian optimization and outlier prediction, provides the appropriate solution for the accurate forecasting of solar irradiance.