Abstract: In order to optimize agricultural machinery system, it is necessary to predict operation costs of agricultural machinery. Based on real situation in Chinese state farm, the dynamic prediction model for operation costs of agricultural machinery was proposed. The operation costs of agricultural machinery were composed by depreciation costs, maintenance costs, fuel costs, labor costs, and management costs. The results show that the models have reached high precision, of which the adjusted sum squares of remaining value index and cumulated reapir and maintenance index are reached 0.8367 and 0.8840, respectively. The comparative analyses of the models show that total prediction operation costs are 28585.79 yuan for plough working unit and 23868.42 yuan for rotary tiller working unit, respectively, when these two working units powered by tractor JDT654 No.6 are supposed to finish operation area of 180.28 hm2 and 165.46 hm2 respectively in 2007. The prediction errors are -2.11% and -5.92% respectively, and the prediction error of the total operation costs of the two working units is -3.88%.