Abstract: Due to the imbalance between supply and demand caused by the impractical arrangement of battery swapping cabinets for take-out delivery electric bicycles, some cabinets have low utilization rates, failing to meet timely swapping demands. This study determines the optimal locations and sizes for the battery swapping cabinets. Firstly, the starting and ending points for take- out deliveries are established by clustering the Point of Interest (POI) data. The spatial-temporal distribution of battery swapping demands of electric bicycles is forecasted by simulating the delivery routes taken by delivery riders. A multi-objective model is then devised to minimize the total cost for cabinet operators while ensuring maximum user satisfaction. The NSGA-II algorithm is applied to solve an optimal scheme for the location and size of battery swapping cabinets within Xinxiang City's main urban area. The results show the predicted temporal distribution of battery swapping demand obtained by simulation is close to the actual value. The demand increases sharply at around 11:00, 14:00, 17:00 and 20:00, and the demands at 11:00 and 14:00 is significantly higher than that at 17:00 and 20:00. The delivery route simulation method exhibits high accuracy in predicting battery swapping demand. The site selection scheme of battery swapping cabinets cannot satisfy the interests of both operators and users simultaneously, and the improvement of user satisfaction needs to increase the total cost of operators. Striking a balance between these interests, an optimal plan recommends 26 take-out electric bicycle battery swapping cabinets, which comprise 11 cabinets with an 11-unit capacity, 8 with a 22-unit capacity, and 7 with a 33-unit capacity. The number of the battery swapping cabinets is suggested to increase to 30 according to the construction order of sites 15-7-19-17, resulting in highest user satisfaction. However, continuing to increase the number of cabinets increases operator costs without increasing user benefit.

Key words: urban traffic, locating of battery swapping cabinet, point of interest clustering, delivery route simulation, electric bicycle, penalty cost

摘要： 针对外卖配送电动自行车换电柜布局不合理带来的部分换电柜利用率低与部分换电需求得不到及时满足的供需矛盾问题，本文通过聚类POI(Point of Interest)数据确定外卖配送起止点，并通过仿真模拟外卖骑手配送路径预测外卖配送电动自行车换电需求时空分布，构建换电柜运营商总成本最低和用户满意度最高的多目标换电柜选址定容模型，并以新乡市主城区为例，采用NSGA-II(Non-dominated Sorting Genetic Algorithm II)算法得到换电柜选址定容方案。研究结果表明：仿真模拟得出的换电需求时间分布预测值与实际值基本吻合，换电需求在11:00，14:00，17:00和20:00左右急剧增长，且11:00和14:00左右的换电需求量显著高于17:00和20:00左右的换电需求量，外卖骑手配送路径仿真模拟方法在换电需求预测上具有较高的预测精度；换电柜选址方案不能同时满足运营商和用户利益均为最优，用户满意度的提高需以增加运营商总成本为代价；同时，兼顾运营商和用户利益的新乡市主城区外卖配送电动自行车换电柜最佳建设数量为26，其中，容量为11的换电柜11个，容量为22的换电柜8个，容量为33的换电柜7个；新乡市主城区应按照备选点编号15-7-19-17依次新增换电柜至30个，此时，用户满意度最大，若继续增加换电柜建设数量，只会增加运营商总成本。

关键词: 城市交通, 换电柜选址, POI聚类, 配送路径模拟, 电动自行车, 惩罚成本