In recent years, the sales volume of refrigerated food grew rapidly around the world. An efficient refrigerated distribution center is the key for success in the cold chain, and storage assignment is one of the most complicated and time-consuming activities among all processes in a refrigerated distribution center. In reality, locations near the market site are usually chosen due to perishable properties of refrigerated food, but the land values of these locations are much more expensive than those of other locations. Therefore, this study aims at formulating a model on dynamic assignment of storage space for efficiently increasing the utilization of storage spaces in a refrigerated distribution center with automated storage/retrieval system (AS/RS). The dynamic storage assignment is divided into two stages, and the model is constructed using the linear programming (LP) method. The first stage assignment assigns stored food to retailers and takes the overdue food out by analyzing the perishable properties of food according to its expiration time. The distribution of occupied and vacant spaces is identified in this stage. The second stage assignment combines the demand forecast results for various refrigerated items using dynamic grey predictions and the results of the first stage assignment. It continuously assigns different food items to empty spaces according to their forecasted demands. Finally, a simulated case study for an AS/RS practical system is provided to demonstrate the feasibility and the results of applying the proposed model. The results show that the model developed in this study can result in a higher space utilization as compared with the conventional models.