Production forecasting remains a vital model in Enterprise Resource Planning system. The accuracy of this model directly affects whole production chain process, such as safety inventory quantity, out of stock losses, and timely performance. Nevertheless, the complexity of Enterprise Resource Planning system offering thousands of data across functions creates difficulty for the enterpriser to optimize resources. Thus, a more accurate production forecasting model is needed considering rapid economic growth. In this study, the single forecasting models which are Grey System, Neural Network and Bass Diffusion were discussed. The Grey System is used to forecast the production by few data and utilize accumulated data to reduce noise influence in a certain extent. The Neural Network whose abstraction and simulation of many basic characteristics of human brain can solve the complex nonlinear problem. And the back propagation algorithm of Neural Network does not require any priority formula which can induce rules automatically from the existing data and getting the output data through the internal laws. The Bass Diffusion consists of a simple differential equation that describes the process of how new products get adopted in a population. Therefore, the three kind of combination forecasting models based on Grey System and Neural Network were proposed to improve the accuracy of prediction, which are Parallel Grey Neural Network, Serial Grey Neural Network, and Inlaid Grey Neural Network. Furthermore, the combination of Bass Diffusion and Neural Network was proposed to compare with Serial Grey Neural Network. As the experimental result, it shows that the accuracy of combination forecasting methods is better than the individual one.