【Objective】Rice blast was one of the major diseases of rice,and seriously restricted stable high yield of rice. With the influence of varieties layout,cropping system and climate change in recent years,the prevalence of inter-annual fluctuation was very large. Currently, the rice blast showed a trend towards the middle to high of the rice-producing areas in Yunnan Province. Prediction and forecast played an important role as a pioneer of the guide of prevention and control.【Method】 In order to timely and effective do a good job of rice blast prevention work,we adopted its related meteorological factors and field rice panicle blast disease index,used BP neural network technology,selected Dehong prefecture Mang city as a test point to carry out the prediction and forecast research.【Result】 From the relevance of meteorological factor and forecasting object,there was a strong correlation between various meteorological factors by screening and the disease index, the ideal and actual output values were closer, error curve was consistent, prediction accuracy could satisfy the actual demand.【Conclusion】 The medium-term prediction model of the rice blast disease established by BP neural network was more advantageous,that was no need for a mathematical formula to express in advance,had a higher prediction accuracy. The prediction model was established by 5-9 meteorological data of experimental sites and disease index of rice blast in the field were more objective and reliable, and could do a good job of disease prevention and control.