The increasingly severe energy and environmental issues have made “carbon peaking and carbon neutrality” become one of the important strategic goals for national development. Promoting the widespread adoption of clean, pollution-free, and energy-efficient electric logistics vehicles is a crucial step toward achieving this goal. However, compared to traditional fuel logistics vehicles, electric logistics vehicles have some distinct characteristics, logistics companies face several challenges in their daily delivery operations. On one hand, the relatively low mile range of electric logistics vehicles necessitates rational planning of charging activities within the delivery process. On the other hand, time-of-use electricity pricing and diverse operational strategies employed by charging stations introduce substantial variations in charging costs across different times and stations. This further complicates the charging decision-making process in vehicle route planning. Specifically, in the selection of charging stations, it is necessary to consider various influencial factors, including real-time traffic conditions, geographical location, charging speed, station scale, charging costs, and the operational status (e.g., busy or idle) of charging piles. Subsequently, the charging priorities of stations will be dynamically adjusted, so that charging stations can be rationally employed during the delivery to minimize the total delivery costs. Furthermore, unlike the relatively short refueling process of traditional fuel logistics vehicles, charging electric logistics vehicles during delivery consumes a certain amount of time (including queuing time and charging time). The delivery services have to be paused while charging, and prolonged queuing and charging time result in missing the delivery time window expected by customers. This not only reduces delivery efficiency, but also negatively affects customer satisfaction. These aforementioned challenges significantly hinder the pace of companies’ green logistics transformation. To address the challenges of low delivery efficiency and reduced customer satisfaction associated with electric logistics vehicles, this study comprehensively considers the delivery tasks and real-time information of charging stations, and proposes an Electric Vehicle Routing Problem with Time Window and Partial Recharge (EVRP-TWPR) model to minimize delivery costs while maximizing customer satisfaction. In particular, a wide range of influencing factors (such as vehicle load, queuing time, charging time, expected delivery time, and notably, varying charging prices across different times and stations) are considered in the proposed model. To solve the model, a multi-objective ant colony algorithm is employed, along with an improved method for updating the pheromone matrix based on non-dominated sorting weights. In this algorithm, feasible solutions are sorted during iterations using a fast non-dominated sorting technique, and these rankings are utilized as weights to update the pheromone matrix, which is then aggregated through random selection. This enables the proposed ant colony algorithm to handle multi-objective scenarios effectively. Furthermore, to better align the solving algorithm with the characteristics of electric logistics vehicles and charging stations, this work innovatively designs a charging station insertion mechanism and a charging capacity feedback correction mechanism in response to the temporal and spatial disparities in on-route charging prices. These mechanisms enable electric logistics vehicles to select better charging stations and optimize their charging decisions by jointly considering the remaining battery capacity and the future charging prices across charging stations. Specifically, on-route charging decisions involve determining whether recharging is necessary during delivery or returning to the distribution center for charging. If recharging is required, on-route charging decisions must be made regarding when and where to charge, as well as the optimal charging amount (i.e., the charging duration). Finally, this study conducts an empirical study on delivery tasks of Company A in City C, Fujian Province. More specifically, the site information and real-time electricity prices from public charging stations in City C were collected to calculate important model parameters, such as charging queue waiting time and charging probability.