Abstract: The water resources is very short in north of China, the water utilization ratio and crops yield are improved effectively by optimal irrigation schedule of forage crop in deficit irrigation. The RAGA-DP model was established by combining the Real coding based Accelatating Genetic Algorithm(RAGA) with Multi-Dimensional Dynamic Planning(DP), and with which irrigation schedule of silage maize, elymus dahuricus and alfalfa were optimized under deficit irrigation in typical grassland in Xilinguole of Inner Mongolia. The results were preferably by test validating. The problem of water resources distribution at different growth stages were availably solved under the limited water condition and the optimum water supply quantity of three crops were obtained through the relationships between relative yield and supply water amount, between water production function and marginal yield. The model solves the prematurity phenomena and the problem that get into a partial optimization easily but it is difficult to obtain the global optimum value when DP are applyed for optimizing irrigation schedule under the condition of insufficient irrigation.