The drought and flood disaster has features with wide influence,high frequency and hard to forecast. Based on the basic principle of grey model and the complementary advantage of Markov state transition matrix,the grey Markov prediction residuals optimization model was established. We chose the annual precipitation during 60 years in Zhengzhou as historical data and fixed the abnormal sequence,then established the prediction model of drought and flood. Using mean-standard deviation drought and flood classification method,the precipitation thresholds of drought and waterlogging were obtained. Thereby taking fixed upper and lower catastrophe point,the date disaster sequences of drought and waterlogging were built. At last,the sequence was modeled and solved. The result shows that the optimized prediction models of grey Markov are simpler and more accurate than the basic grey model and can be used as a reference for the disaster prevention and mitigation of Zhengzhou City.