The threshold of extreme precipitation is defined by the method of percentile method in this paper. Based on Bayesian theory, probabilistic forecasting of extreme precipitation is discussed and we do the Bayesian extreme precipitation experiment and verification. Experiments which are based on observational and modal extreme precipitation threshold show the following conclusions. False prediction of extreme precipitation is proved to be more when the threshold is based on the observational extreme precipitation. The forecast of main rain band is stronger. False prediction is less while the threshold is based on the modal extreme precipitation and the early warning to the heavy rain is obvious as well. By contrast, the probabilistic forecasting based on the prior probability is smaller. The verification result shows that the accuracy of the extreme precipitation is improved by the Bayesian method, but the false prediction ratio has increased.