针对现有充电设施布局方法难以匹配城市配送车辆运输作业特点的问题，本文提出一种停驻特征驱动的设施布局与功率配置方法。基于车辆停驻信息，识别运营基地并构建包含不同功率等级的网络充电站集合。在此基础上，利用同一车队逐秒能耗数据，构建行程未完成率、平均绕行距离和平均延误时间3项指标，对不同布局方案及站点规模情景下的运行效果进行综合评估。结果表明，本文方法各项指标均优于截流选址（Flow-Capturing Location Model, FCLM）模型和最大覆盖模型；以网络充电站数量n=5为例，相较于FCLM模型，本文方法行程未完成率降低约40%，平均绕行距离与延误时间缩短20%~30%。进一步对比本文方法的两种功率配置方案，60 kW与240 kW功率混合配置方案的运行指标虽略高于全240 kW方案，但差异随n增加而迅速收敛。此外，小时负荷分析显示，功率混合配置方案可显著缓解单站负荷压力，使基地与网络侧峰值负荷相较于FCLM模型分别降低约30%和43%，在提升电网运行友好性的同时，有助于降低设施建设成本。

A dwell-characteristic-driven method for charging-facility siting and power rating was proposed to better align charging-facility planning decisions with the operating patterns of urban delivery fleets. Based on vehicle dwell information, depot locations were identified, and a set of network charging stations with different power ratings was then established. On this basis, second-by-second energy-consumption data from the same fleet was used for evaluation. Three metrics were defined: trip incompletion rate, average detour distance, and average delay time. Operational performance was comprehensively evaluated across alternative layouts and station-number scenarios using these metrics. The results showed that the proposed method outperformed the FCLM (Flow-Capturing Location Model) and maximum coverage models across all metrics. When the number of network charging stations was n=5, compared with the FCLM model, the proposed method reduced the trip incompletion rate by approximately 40%, and shortened the average detour distance and delay time by 20%~30%. Further comparison of the two power configuration schemes of the proposed method revealed that, although the operational indicators of the 60 kW and 240 kW mixed power configuration scheme were slightly higher than those of the all-240 kW scheme, the differences narrowed rapidly as n increased. In addition, hourly load analysis showed that the mixed power configuration scheme could significantly alleviate the load pressure on individual stations, reducing the peak loads at the depot and network sides by approximately 30% and 43%, respectively, compared with the FCLM model. It improved grid-friendliness and helped reduce infrastructure costs.