Accurate photovoltaic (PV) power prediction is the key to successful grid integration of PV power plants. However, factors such as solar radiation, climate and geographical conditions can cause frequent fluctuations in PV power generation, posing significant challenges to power prediction. In response to the current demand for large-scale grid integration of PV renewable energy, the significance and classification of PV power prediction are discussed from multiple perspectives. The latest applications of artificial intelligence (AI) technology in the field of PV power prediction are reviewed, including traditional machine learning, deep learning and hybrid methods, and are compared and summarized. Currently, the main types of researches are ultra-short-term and short-term PV power prediction for single PV power stations, and deep learning and hybrid methods are the mainstream prediction methods. Data pre-processing, feature extraction and error compensation are the key factors to improve prediction accuracy. Finally, future trends and research innovations in AI technology for PV power prediction are discussed.