Abstract:Solar irradiance is the most direct and significant factor affecting photovoltaic output power, and its prediction accuracy is related to the scheduling and planning of power system departments. Affected by geographic latitude, altitude, and climate complexity, the accuracy of traditional surface irradiance prediction models is low. In this paper, based on the irradiance astronomy model combined with the weather type to predict irradiance, By establishing a geometric model of the sun"s illumination and the Earth"s atmosphere, the calculation formula for extraterrestrial irradiance is solved and the transparency coefficient of the sunny day is corrected, Then according to the weather type and historical irradiance, the cluster analysis is performed, the corresponding weight coefficient is calculated, and the solar irradiance prediction model based on weather clustering is established. Through the verification of the measured data, compared with the traditional sunny model, RLS model, and WT model, the weather clustering model has stronger adaptability, higher accuracy, and smaller error.