动态随机环境下电动汽车共享系统配置和运营联合决策优化

The electric carsharing system is an innovative transportation mode to cope with traffic congestion, environmental pollution and energy crisis. It has been widely used in the real world. However, it is always difficult for users to find cars, parking spaces and charging facilities. It is also hard for operators to manage the system because of high operating costs and low profits. This project considers the impact of dynamic and random factors such as user demand, lease time and charging time. Meanwhile, this project takes into account the strong dependence between configuration and operation in addition to the interests of operators, users and government departments. These considerations are the keys to solve the aforementioned problems. There is no effective method in current studies to comprehensively consider the above factors. Therefore, the following work is focused on: 1) First, we build a time-varying and phase-type fluid queue network model for the electric carsharing system to reveal the mechanism of interaction among users, dispatchers and traffic facilities in a dynamic random environment; 2) Then, we develop a nested Logit model to reveal the user choice mechanism of travel modes and travel routes, which can effectively capture the elastic demand and road congestion; and 3) Finally, we develop a joint decision optimization model of configuration and operation to ensure the effective operation and sustainable development of the electric carsharing system under the dynamic random environment. This model takes into account the user choice behaviors and road congestion constraints to guarantee the interests of the aforementioned three parties. A cloud heuristic algorithm is developed to solve it. The expected achievement can provide a theoretical support for the decision making in electric carsharing systems.

电动汽车共享系统是应对交通拥堵、环境污染和能源危机的创新交通方式，已得到广泛应用，但也面临找车难、停车难、充电难和运营成本高、盈利难等现实问题。本研究掌握动态随机的用户需求、租赁时间和充电时间等环境对系统的冲击，考虑配置和运营两个决策层面的强依赖性，兼顾运营商、用户和政府部门三方利益，是解决上述问题的关键。已有研究缺乏有效方法充分考虑上述因素。为此，项目研究：1.建立随时间变化的电动汽车共享系统位相型(PH)流体排队网络模型，揭示动态随机环境下用户、调度员和交通设施之间的作用机理；2.建立两层嵌套多项Logit模型，揭示用户在不同交通方式和出行路径之间的联合选择机理，捕捉弹性需求和道路拥堵；3.建立考虑用户选择行为和道路拥堵约束的系统配置和运营联合决策优化模型及云启发式混合算法，从三方利益出发确保系统在动态随机环境冲击下的有效运行和可持续发展。预期成果可为电动汽车共享系统决策提供理论支撑。

电动汽车共享系统是应对交通拥堵、环境污染和能源危机的创新交通方式，已得到广泛应用，但也面临找车难、停车难、充电难和运营成本高、盈利难等现实问题。本项目掌握动态随机的用户需求、租赁时间和充电时间等环境对系统的冲击，考虑拥挤道路网络与电动汽车共享系统决策之间的相互作用，考虑配置和运营两个决策层面的强依赖性，兼顾运营商、用户和政府部门三方利益，是解决上述问题的关键，能够促进电动汽车共享系统健康可持续发展。首先，本项目针对电动汽车共享系统提出了一种多类型顾客的闭合排队网络模型。在用户出行行为选择建模上，提出考虑道路拥堵影响的弹性需求logit函数，该弹性函数主要与单位价格、背景交通量、用户选择行为导致增加的车辆数有关。我们证明了排队网络模型解的存在性和唯一性，继而发展了一种复杂度与站点容量、车队规模无关的递推算法对其求解。此外，我们推导了时不变系统平衡态的解析性质，结果表明提高瓶颈队列的吞吐量是提高整个系统吞吐量的关键；在由瓶颈路线连接的起始站停止或放慢电动汽车服务，可以实现用户、运营商和公众的三赢局面。该结论强调了在维护电动汽车共享系统健康可持续发展时，考虑共享电动汽车与交通拥堵之间的相互作用的重要性。然后，我们提出了一个基于排队网络的优化模型，以联合确定综合利益最优的配置和运营决策。为了解决大规模问题，提出了APBC-SPSA算法，该算法综合了交替优化框架、高斯赛德尔法、雅可比法、随机逼近的优点。最后在三个不同规模的网络下进行了数值实验验证了算法的有效性。结果表明：（1）我们提出的排队网络模型模拟电动汽车共享系统交通流是准确和可靠的，平均相对误差在0.55%和5.3%之间。（2）公交用户转移到共享车的比率增加，会对非拥挤城市的电动汽车共享平台带来更多的积极影响，但对拥挤城市的汽车共享平台带来更多的消极影响。项目研究成果可为电动汽车共享系统决策提供理论支撑，实现汽车共享行业的可持续发展。

内容获取失败，请点击重试