N the context of global carbon neutrality, the efficient utilization and development of clean and renewable energy have become a crucial development strategy for countries around the world today. As the most common clean and renewable energy source, the development prospects and importance of solar energy are self-evident. In the exploration of solar energy development, photovoltaic power generation technology is the most common and mature. The proportion of electricity provided by photovoltaic power generation technology in the world’s total power generation has been increasing year by year. However, photovoltaic power generation is unstable, and large-scale grid connection has a huge impact on the safe and stable operation of regional power grids. Efficient photovoltaic power generation prediction can not only improve the grid connection capability and safety, but also effectively reduce wasted solar energy. This article explores in depth the classification of photovoltaic power generation related predictions based on the current environment, mainly divided into four types: classification according to the prediction process, classification according to the time scale, classification according to the size of the spatial range, and classification according to the prediction modeling method; The focus is on analyzing the photovoltaic power generation prediction models established under two different methods: physical prediction and intelligent algorithms, and studying their respective advantages and disadvantages. This article has certain theoretical value for the large-scale application of the photovoltaic industry.