The support cost is the main part of LCC(life cycle cost) of weapon equipment, For the purpose of reasonably forecasting the support cost of weapon equipment, the paper analyses the constitution of equipment support cost and its affecting factors. Given the limitation, complexity and nonlinearity of date quantity of equipment support cost, the forecasting precision of single model is low. So the paper puts forward a combined forecasting model based on grey system theory and BP neural network model to forecast the support cost which are based on nonlinear time series, which put the advantage that grey system theory can easily do the small sample data together with the advantage that BP neural network model can solve the complex and nonlinear problem. A simulation example proves that the total error of the combined model's forecasting result is smaller than that of single model's forecasting result. This can largely improve the forecasting precision of equipment support cost.