Abstract: Based on the photovoltaic power output and radiation data in every 15 minutes from Pu’an Moshe photovoltaic power station in Guizhou province and the meteorological observation data in 2020,the variation of photovoltaic power output and the impact of meteorological factors on photovoltaic power were analyzed.The prediction model of photovoltaic power output was established and carried out intra-month forecast test by using CFSv2 model data.The results showed that the photovoltaic power output reached the lower values in the morning and evening and peaks in the noon.The value of photovoltaic power output was the largest in spring,followed by summer,and the smallest in winter.The key meteorological factors that impact photovoltaic power were total solar radiation and sunshine duration,and their correlation coefficients were both above 0.9.The results of linear regression prediction model test of five combinations showed that the prediction model had a better prediction performance based on the average temperature,maximum temperature and daily range,while the model with single meteorological factor did not perform well.In order to increase the prediction accuracy of photovoltaic power output,the data from forecast model during extended period could be used to make the rolling forecast of photovoltaic power output according to the service demand.