Research on the origin of rice wine has significant commercial value for agricultural markets. Inspired by the advantages of fast, non-destructive and high sensitivity of electronic nose (e-nose) in rice wine analysis, this paper applies an e-nose system based on headspace sampling and combines with a convolutional neural network (CNN) to achieve effective identification for rice wine origins. The results show that, first, the odor information of rice wine from 10 origins is obtained by the e-nose system. Second, odor features of the samples from different origins are obtained by the convolutional and pooling layers. Finally, the combination of the e-nose and CNN achieved the best performance, including the accuracy of 98.00%, the kappa coefficient of 97.51%, and the F1-score of 98.00%, in compared with multiple classification models. Effective identification of rice wine origin results is acquired by the e-nose and CNN, providing a tool for quality assessment of food products.