Abstract: From the perspective of overview,this paper examines the development status, existing difficulties, and main influencing factors of power prediction technology in distributed photovoltaic systems,and outlines the technical route for applying data-driven methods to achieve accurate power prediction. Starting from factors such as spatial correlation,historical output power,and meteorological factors,this paper reviews the current research status of data-driven power prediction in photovoltaic systems, analyzes the corresponding data enhancement, spatio-temporal map information,and feature fusion methods,and discusses the advantages and disadvantages of the technology. Finally,research directions and development suggestions for data-driven methods for power prediction are given.