The accurate prediction of PV power generation can provide reference for grid smooth operation. Due to the influence of complex weather on the photovoltaic field,it is difficult to accurately predict the quantity of photovoltaic power generation. According to the characteristics of climate change and characteristics of GRNN neural network, A prediction method of the photovoltaic power generation by training neural network GRNN is proposed based on historical weather data,simulation results show that,without distinguishing between similar - day circumstances, the prediction method with GRNN neural networks has less relative error rate and lower absolute error rate than the method with BP neural network. The results show that the GRNN neural network can realize the accurate prediction of photovoltaic power generation without need to determine the weather conditions of the same day, which has practical application value.