For smart networks to integrate renewable energy sources as efficiently as possible, solar energy (SE) forecasting is essential. Solar power systems can be made more efficient, dependable, and sustainable by accurately predicting solar radiation (Rs). We examine various approaches, including statistical, physical, and hybrid methods, emphasizing the critical function of feature selection (FS) in enhancing forecasting precision. The analysis highlights that hybrid models combining ML techniques with physical models improve forecast accuracy, demonstrating significant reductions in forecasting errors compared to traditional methods. Additionally, deep learning models such as convolutional neural networks (CNNs) and long short-term memory (LSTM) networks are identified as highly effective in capturing complex temporal patterns in Rs data. The review describes current advances in ML, particularly in deep learning, and recommends further exploration of sophisticated hybrid models and novel FS techniques to advance this rapidly developing field.