This study demonstrates the effectiveness of Artificial Neural Networks (ANNs) in predicting shortwave downwelling radiation in Muscat, a region with high solar potential. The model successfully forecasts solar radiation under both clear-sky and all-sky conditions, providing valuable insights for optimizing photovoltaic (PV) system performance. Accurate predictions enable energy producers to better manage generation and storage, improving grid reliability and reducing costs associated with renewable energy integration. However, atmospheric variability, including cloud cover and dust storms, presented challenges, as the model’s performance was slightly lower under all-sky conditions. This suggests the need for further refinement in handling meteorological factors like clouds and aerosols. Including additional variables, such as wind speed and humidity, may enhance the model’s predictive accuracy. By leveraging a five-year dataset from NASA MERRA-2, the study offers an insight of solar radiation trends in Muscat, providing key insights for energy planners and policymakers. Comparisons between clear-sky and all-sky conditions illustrate the variability of solar radiation, informing more resilient energy system designs. The study concludes that advanced machine learning models like ANN can significantly improve solar radiation forecasting, which is essential for optimizing PV systems and integrating solar energy into grids in regions like Oman, contributing to global climate change mitigation efforts.