Accurate wind power and photovoltaic power outputs forecasting is important to improve the scheduling quality of wind and solar hybrid generation system. The hybrid wind-PV forecasting model based on BP neural network is established to achieve the forecasting of the entire wind power and photovoltaic power station, using the same forecasting model instead of twice forecasting, it can simplify the prediction and improve the prediction precision effectively. The method is validated by wind farm and photovoltaic system data in a coastal islands and the forecast error is calculated and analyzed. The results show the method has high accuracy, which has good academic value and practical value to forecast power for hybrid wind-PV generation system.