Abstract: The main problem of the public bicycle rental system is the difficulty of the user access to the bicycle and the rental station needs staff on duty at the peak hours. In order to avoid scheduling process determined by dispatcher empirically blindly, improve the scheduling scientifically, shorten the operation time and reduce the cost, so as to better meet the rental demand of the users, a method is proposed to predict the demand of single public bicycle rental station based on improved Elman Neural Network. The effectiveness of the proposed method is proved by comparing the predicted results with the actual demand.

Key words: urban transport, public bicycle system, demand predict of single station, Elman neural network

摘要： 公共自行车租赁系统目前存在的主要问题是高峰时段用户存取车困难，站点需要工作人员值守。为了提高城市公共自行车调度的科学性、缩短调度时间、降低调度成本，避免调度过程中调度员凭经验确定各个站点需求量的盲目性，从而更好地服务租赁者，满足其出行需求。建立了基于改进的Elman神经网络的公共自行车单站点需求量预测模型。通过仿真实验将改进模型和已有模型的预测结果与实际需求量进行对比，证明了提出的预测方法有效可行。

关键词: 城市交通, 公共自行车租赁系统, 单站点需求量预测, Elman神经网络