交通运输工程与信息学报

【背景】在地铁站接驳场景下，共享单车在缓解“最后一公里”出行压力方面具有显著优势，但为平衡地铁站周边站点间供需而产生的调度车辆空驶、绕行等行为会带来额外碳排放，从而与城市低碳目标相悖。【目标】针对地铁站周边共享单车供需失衡问题，构建一个考虑碳排放约束的共享单车调度优化模型，在保证运营收益的同时约束调度过程中的碳排放水平。【方法】以带容量约束的车辆路径问题为基础，引入“供需缺口弥补量”概念，用以刻画调度后地铁接驳骑行替代高碳基准出行方式所带来的减排效益，并要求调度带来的碳减排量不低于调度车辆产生的碳排放量，采用强化学习驱动的自适应大邻域搜索算法（RL-ALNS）进行求解并开展关键参数敏感性分析。【数据】基于上海市杨浦区29个地铁站2022年早高峰时段的共享单车开关锁数据，并结合高德API获取的地铁站间的驾车距离与行驶时间数据进行实证分析。【结果】模型在早高峰时段实现综合效益15 682.68元，较调度前提升15.24%。与无约束情形相比，引入碳排放约束使调度排放降低29.57%且综合效益更优，实现了经济和环境的双重优化；敏感性分析证实了模型在不同惩罚因子与碳排放约束强度下的自适应调节能力与鲁棒性。【结论】该模型能有效平衡经济收益与碳排放，为地铁接驳场景下的低碳调度提供决策支持。【应用】可推广至其他城市共享单车系统，帮助运营企业制定低碳调度策略，从而促进城市交通系统的可持续发展。

[Background] Bike-sharing systems play a significant role in alleviating "last-mile" travel pressure in metro feeder scenarios. However， rebalancing operations to address supply–demand mismatches often involve empty trips and detours by dispatch vehicles， generating additional carbon emissions that contradict urban low-carbon goals. [Objective] To address supply–demand imbalances around metro stations， a bike-sharing scheduling optimization model with carbon emission constraints is constructed. This model aims to constrain carbon emissions during scheduling while ensuring operational profitability. [Method] Based on the Capacitated Vehicle Routing Problem (CVRP)， the concept of the "fulfilled supply-demand gap" is introduced to characterize the mitigation benefits achieved by substituting high-carbon baseline travel modes with metro-feeder cycling post-scheduling. The model requires that the carbon reduction realized through scheduling is not lower than the emissions generated by dispatch vehicles， and it is solved using a Reinforcement Learning-driven Adaptive Large Neighborhood Search (RL-ALNS) algorithm accompanied by sensitivity analyses of key parameters. [Data] Empirical analysis utilizes bike-sharing lock/unlock data from 29 metro stations in Yangpu District， Shanghai， during the 2022 morning peak， combined with inter-station driving distance and travel time data from the Gaode API. [Result] The model achieves a total profit of 15682.68 yuan during the morning peak， 15.24% higher than that before scheduling. Compared with the unconstrained scenario， introducing carbon emission constraints significantly reduces dispatch emissions by 29.57% while achieving better comprehensive benefits， realizing dual economic and environmental optimization. Furthermore， sensitivity analyses confirm the adaptive adjustment capability and robustness of the model under various penalty factors and carbon-constraint intensities. [Conclusion] The model effectively balances economic returns with carbon emissions， providing decision support for low-carbon scheduling in metro feeder scenarios. [Application] This approach can be extended to other cities to assist operators in formulating low-carbon scheduling strategies， thereby promoting sustainable urban transportation development.