This paper presents a neural network ensemble framework to improve solar photovoltaic (PV) output forecasting, utilizing meteorological data and Bayesian model averaging to enhance prediction accuracy. It evaluates three types of neural networks (FNN, Elman, and newCF) and demonstrates significant improvements in forecast accuracy compared to traditional methods, particularly for 24-hour forecasts across different seasons. The findings emphasize the importance of effective input selection and ensemble techniques in managing the variability of solar energy generation.