In the production process of a flow production line with limited buffer, the work-in-process in the buffer will be backlogged due to the mismatch of production speed between the previous and next processes. To prevent production congestion, some intermediate warehouses for temporary storage of work-in-process are often set up in the production line. Under certain conditions, the work-in-process stranded in the buffer will be transferred to intermediate warehouse. It can be seen that the scheduling problem of limited buffer with temporary intermediate warehouse is more complicated, which increases the difficulty of scheduling. In the food processing industry, the scheduling problem of finished product area of ham sausage production line has the typical characteristics of this type of scheduling problem. To solve this problem, this paper proposes a scheduling optimization method based on reinforcement learning. The state space is established based on the production state of ham sausage lot, and the action space is established based on the movement of the RGV trolley in the production line as well as the start and stop of sterilization equipment. To coordinate the continuous production between processes, a variety of scheduling rules are formulated to achieve the goal of reducing waiting time and shortening total working hours in the process of multiple iterative learning. The simulation experimental results prove the effectiveness of this method in solving the scheduling problem of limited buffer with intermediate warehouse.