一个全国规模的电动汽车充电调度框架：基础设施容量约束下的最优迂回路由
A National-Scale EV Charging Scheduling Framework: Optimal Detour Routing Under Infrastructure Capacity Constraints
浏览论文内容
中文总结 AI 辅助
研究在现有充电基础设施下电动汽车长途旅行调度负担与成本，提出基于优化的可扩展框架，分三阶段计算最优充电调度，应用于大量车辆轨迹，能快速生成可行调度并评估电动汽车成本竞争力。
中文摘要 AI 辅助
随着电动汽车使用量的增加，在现有充电基础设施下量化长途旅行的调度负担和经济成本对基础设施规划和政策愈发重要。本文提出了一个基于优化的可扩展框架，使用POLARIS在美国实际充电站和模拟长途个人车辆轨迹上调度电动汽车充电站点。该框架以固定的现有充电网络为输入，在考虑每个站点插头容量约束的同时，最小化每辆车的总迂回和排队成本。方法分三个阶段：通过前向可达性启发式修剪不可行性；在有向无环图上通过动态规划计算每辆车的最优充电调度；通过基于惩罚的启发式方法进行容量感知迭代拥塞解决，在拥塞站点增加迂回成本，并采用先进先出队列回退。应用于POLARIS基于代理的交通模拟框架中约270万个起点 - 终点车辆轨迹，涵盖14260个直流快速充电站和68641个插头，该框架在128核高性能计算集群上不到1.3小时就能生成容量可行的调度，且无需任何商业优化求解器。还进行了三层经济分析来评估电动汽车相对于内燃机车辆在不同场景下的成本竞争力。
英文摘要
As electric vehicle (EV) adoption grows, quantifying the scheduling burden and economic cost of long-distance travel under the existing charging infrastructure becomes increasingly important for infrastructure planning and policy. This paper presents a scalable, optimization-based framework for scheduling EV charging stops along real-world charging stations and simulated long-distance personal vehicle trajectories across the United States using POLARIS. Taking the existing charging network as fixed input, the framework minimizes total detour and queuing costs for each vehicle while respecting plug capacity constraints at each station. The methodology proceeds in three phases: (i) infeasibility pruning via a forward-pass reachability heuristic, (ii) per-vehicle optimal charging schedule computation via dynamic programming on a directed acyclic graph, and (iii) capacity-aware iterative congestion resolution through a penalty-based heuristic that augments detour costs at congested stations, with a first-in, first-out queue fallback. Applied to approximately 2.7M origin--destination vehicle trajectories derived from a 1\% sample of national personal travel demand within the POLARIS agent-based transportation simulation framework and covering 14,260 DC fast charging stations with 68,641 plugs from the Alternative Fuels Station Locator, the framework produces capacity-feasible schedules in under 1.3 hours on a 128-core high-performance computing cluster without requiring any commercial optimization solver. A three-tier economic analysis spanning operational costs, total cost of ownership, and amortized infrastructure investment is conducted to evaluate EV cost competitiveness relative to internal combustion engine vehicles across scenarios.
