Abstract:

Significant temporal variability exists in the charging demand of ride-hailing vehicles， but the traditional static location methods tend to cause problems such as insufficient coverage during peak peirod and resource waste during off-peak period. To address these issues， a location optimization model for ride-hailing vehicle charging stations that considers temporal variability was proposed. First， an in-depth analysis of the temporal distribution of charging demand was conducted based on Nanjing's ride-hailing vehicle operation data. The whole day was divided into three periods： peak， shoulder and off-peak periods. Differentiated weights and service time thresholds were set according to the proportions of demand and user sensitivity. Second， a dual-objective model was constructed with the goals of "minimizing total deadhead time cost" and “minimizing uncovered penalty in high-difficulty areas”， and the difficulty of ride-hailing was introduced as a priority adjustment factor. Constraints such as station service capacity and regional balance were established. Finally， by integrating the local fine exploitation ability of Artificial Bee Colony （ABC） and the global exploration ability of Grey Wolf Optimizer （GWO）， an Improved artificial Bee colony-grey Wolf hybrid Optimization algorithm （IBWO） was proposed to solve the proposed model， and adaptive parameter adjustment and population restart mechanismes were also added to avoid the local optimum. Using the core area of Nanjing as a case study and optimally selecting 100 stations from 1 465 candidate locations， the proposed model achieved 100% demand coverage across all periods， with uncovered demand in high-difficulty areas at 0. After 100 iterations， compared with the optimal results of Genetic Algorithm （GA）， Particle Swarm Optimization （PSO）， GWO and ABC for solving the proposed model， the optimal fitness value of IBWO is 16 180.1， which is 35.1% lower than that of the best-performing ABC （24 929.8）； the number of stations of IBWO is reduced by 50.7% compared with ABC （203）； the calculation time of IBWO is 334.87 s， which is 14.5% shorter than that of ABC （391.75 s）. The proposed model can accurately adapt to temporal fluctuations in charging demand， and the proposed IBWO offers fast calculation speed and avoids local optimum. This paper establishes a quantitative correlation framework between charging demand temporal dynamics and location decision-making， improves the infrastructure location model under multi-period constraints. They can provide accurate decision-making basis for the planning of urban electric ride-hailing vehicle charging facilities， reduce construction costs by more than 50% while ensuring all-weather service quality， and contribute to the efficient coordination of transportation and energy systems under “carbon peaking and carbon neutrality” goal.

Key words: electric ride-hailing vehicle, charging station location, temporal variability, grey wolf algorithm, fairness constraint

摘要：

针对网约车充电需求存在显著的时间差异性，而传统静态选址方法易导致高峰期覆盖不足以及低峰期资源浪费的问题，本文提出一种考虑时间差异性的网约车充电站选址优化模型。首先，基于南京市网约车运营数据深入分析充电需求的时间分布规律，将全天划分为高峰期、平峰期和低峰期3个时段，并依据需求占比与用户敏感度设定差异化权重及服务时间阈值；其次，构建以“总空驶时间成本最小化”与“高难度区域未覆盖惩罚最小化”为目标的双目标模型，引入打车难易度作为优先级调节因子，并建立站点服务能力和区域均衡等约束；最后，融合人工蜂群算法（ABC）的局部精细搜索与灰狼（GWO）算法的全局探索能力，提出改进的人工蜂群-灰狼混合算法（IBWO）求解上述模型，并加入自适应参数调整与种群重启机制以避免局部最优。实验结果表明，以南京市核心区域为算例，从1 465个候选位置中优化筛选出100个站点，本文模型实现了全时段100%需求覆盖，且高难度区域未覆盖需求为0。在迭代100次后，与遗传算法（GA）、粒子群优化（PSO）算法、GWO、ABC中求解本文模型的最佳结果相比，IBWO的最佳适应度值为16 180.1，较次优的ABC （24 929.8）降低了35.1%；建站数量较ABC （203个）减少了50.7%；计算时间为334.87 s，较次优的ABC（391.75 s）缩短了14.5%。本文模型能精准适配充电需求的时间波动，且提出的IBWO计算速度快，能避免局部最优。本文建立了充电需求时间动态与选址决策的量化关联框架，完善了多时段约束下的基础设施选址模型，可为城市电动网约车充电设施规划提供精准的决策依据，在降低50%以上建设成本的同时，保障全天候服务质量，助力“双碳”目标下交通能源系统的高效协同。

关键词: 电动网约车, 充电站选址, 时间差异性, 灰狼算法, 公平性约束