Abstract: Most of photovoltaic（PV） generation forecasting models need to take solar irradiance as their input parameters.However,they were difficult to implement in China due to insufficient solar radiation stations available and poor performance of forecasting.After investigating the correlation among PV generation and several meteorological elements through distance analysis,a back propagation（BP） neural network forecasting model was proposed whose input parameters were ambient temperature and humility.Furthermore,in order to adapt sudden weather changes,the future weather type was recognized from forecasted cloud cover by using self-organizing feature map（SOM）.Then,PV power generation in each weather type could be forecasted from its corresponding forecast network.Therefore,the over fitting issue of single network model could be addressed.Comparison experiments were made as opposed to the forecasting model with radiation observation and the one without weather type classification.The experimental results indicate that weather type clustering can significantly improve the precision of power prediction,and that the short-term forecasting model without irradiance with high precision can withstand 50% disturbance with humidity.