The problem of delivering and fetching wagons at a railway station with mixed-shaped goods operation sites is considered, with a view to minimizing the running time between goods operation sites and waiting time of the locomotive in the planning period as the optimization objective. A general mathematical model for delivering and fetching wagons at a railway station with mixed-shaped goods operation sites has been formulated. The methods of judging and processing reverse of delivering and fetching wagons, dividing batches, and judging number of wagons for batch operation are provided to determine the feasibility of the solution, and an improved simulated annealing algorithm is introduced as our algorithm to the model. Finally, an experimental station is taken as an example to verify the model and algorithm. The results show that simulated annealing algorithm is relatively superior in computational efficiency and result compared with genetic algorithm and tabu search algorithm, the computing time of the algorithm provided can meet the requirements of planning shunting operations in railway station, and the model proposed is universal for other layout forms of GOSs and different operation forms.