Abstract: Although the traditional BP neural network has been widely used in power prediction,it has low accuracy for forecasting wind and solar power generation which has strong random fluctuations. this paper proposes an optimized GA-BP network model based on the CEEMD method. First,the CEEMD method is used to decompose the original data into predictable components;second,the forecast result sets of each component are aggregated and averaged to obtain the final result. The effect of this model is verified using historical examples of wind and solar power generation in the energy system of BadenWürttemberg,Germany. The results of comparison with other prediction models show that the model proposed in this paper can improve the forecast accuracy of wind and solar power generation whether it is a day-ahead forecast or an ultra-shortterm forecast.