Northwest China has formed into a benign industrial area named "big appellation and big brand". The unique black soil can also be a benefit for wine grape planting. However, the regional flavors of dry red wine vary greatly in the different regions in Northwest China. Potential risk can remain for the sustainable development of the wine industry. In this study, a wine discrimination system was developed to accurately identify the origin and age of dry red wine using machine learning. 23 physical and chemical parameters were determined about the wine color, taste, and aroma. Firstly, 600 samples of dry red wine were collected from the different wineries in Ningxia, Gansu, and Xinjiang areas. The total phenols, total anthocyanin, and titrated acids were then measured during the experiment. Secondly, the Pearson correlation coefficient formula was selected to evaluate the consistency of 23 dry red wine parameters. Subsequently, random forest (RF) was used to calculate the percentage contributions of each parameter for the identification of the origin and age of dry red wines. Finally, an accurate discrimination system was developed to identify the origin and vintage of dry red wines using an artificial neural network (ANN) classifier. The results showed that the eight color-related parameters provided with 31% reference to distinguish the origin and year of dry red wine. Specifically, the wine samples from Ningxia's wineries showed higher yellowness and brick red. The wine samples from Xinjiang's wineries showed a high degree of redness, like the peach red. The wine samples from Gansu's wineries showed higher blueness and purple. Five anthocyanin and seven phenolic-related parameters were provided with 26% and 21% reference, respectively. In terms of taste, Ningxia's wine samples had less astringency, while Gansu's wine samples showed a stronger astringency, and Xinjiang's wine samples tasted the most astringent. The loading analysis demonstrated that the Ningxia wine shared a certain aging potential because the old wine from the Ningxia showed more flavor characteristics. By contrast, the new wine from Xinjiang presented strong flavor characteristics, but the old wine showed weak flavor characteristics, indicating the quick dissipation of flavor as time increased. Similar flavor characteristics were achieved in the old and new Gansu wine, indicating the slow quality loss of Gansu wine during aging. The sensitivity (SEN) and accuracy (CCR) of the wine's origin discriminant model were 98.72% and 98.72% for the wine samples from Ningxia's wineries, respectively, 95.45% and 100% for the wine samples from Xinjiang's wineries, respectively, while 100% and 95.45% for the wine samples from Gansu's wineries, respectively. The correct rate of the wine's age discriminant model for each region was 98.7% for the Ningxia wine, and 100% for the Xinjiang wine and Gansu wine. The discrimination model can be expected to make accurate discrimination of wine origin and age from Northwest China. The finding can provide scientific support for the production of premium wine in the market supervision in these regions.