Photovoltaic power station output power has a great influence to power grid scheduling, but photovoltaic power station output power is random and not controllable, and affected by the intensity of solar radiation and meteorological factors. For reasonable use of photovoltaic power generation systems, a BP neural network forecasting mode based on weather forecast information is proposed in the paper to predict PV system output power. The influential factors of PV station output power is determined through correlation analysis. The training and prediction model are processed combined with historical data and meteorological factors. A new forecasting model- double BP neural network model is proposed in this paper. The photovoltaic power station predicted results and the measured values are compared. The results show that the method can effectively improve the predictive accuracy of PV station output power and has good reference value and practical value to power generation planning.