In order to cope with the quicken pace of market, predict the market development in advance and formulate corresponding production scheduling plans, and enable companies to occupy a first-mover advantage in competition. The current forecasting algorithm based on gray neural network usually requires continuous and much sample data to accurately predict product demand. Aiming at the problem of low accuracy and poor reliability of prediction results of small data nonlinear systems, a grey neural network prediction method coupled with genetic algorithm is proposed. Firstly, the gray neural network model for forecasting product order demand is established based on the gray model and neural network. Secondly, the electric locomotive product is taken as an example to demonstrate the prediction performance of the model. Lastly, the genetic algorithm is used to iteratively optimize the network weights and thresholds of the mode to solve the premature convergence and improve the global optimization capability in the prediction process. The results show that the accuracy and robustness of the optimized product prediction model are improved, which verifies the feasibility of the designed method.