Abstract: The demand for total power of agricultural machinery is a complicated non-linear system, whose developmental changes have dual trends of increase and fluctuation. This paper established metabolic GM(1,1) models based on grey prediction model firstly. The authors also analyzed the merits as well as defects of BP neural network and grey prediction method, then combined the two methods to establish a combined metabolic estimation model named serial grey neural network, and estimated the total power demand of agricultural machinery of China. Case study validated the effectiveness of the proposed model.