Scheduling optimization, especially public bicycle scheduling analysis, has become a hot topic due to its benefits in social, economic and environmental aspects. The traditional methods mainly focus on the distance among bicycle stations to realize a self-balancing bicycle supply by minimizing the scheduling workload for each region but ignore the origin/destination flow-data and topological features of bicycle station networks. This paper proposes a novel partition model for public bicycle scheduling region. Besides the distance, the proposed model combines dynamic origin/destination flows of bicycle stations with the topological features of them in bicycle networks to construct a similarity matrix. Then scheduling region partition is obtained by using the affinity propagation clustering algorithm. To evaluate the performance of the proposed model, we compare it with the other baseline methods by calculating the correlation coefficient between the regions while the coupling coefficient functions on the real data from Hangzhou public bicycle system. The experimental result shows that the proposed model can achieve the smallest scheduling workload among the scheduling regions.