快速路入口匝道瓶颈宏观交通流模型

基于入口匝道汇入方式与基本图形态, 提出了一种调整型元胞传输模型; 增加了入口匝道状态变量以追踪入口匝道交通状态, 定义了新的入口匝道汇入规则; 将双通行能力基本图引入到调整型元胞传输模型中, 以适应不同交通状态下通行能力的变化; 将单纯形法与遗传算法相结合, 提出了混合多目标参数优化方法; 建立了3个仿真场景, 评价调整型元胞传输模型与混合多目标参数优化方法的效果。仿真结果表明: 在预测入口匝道上游主线拥堵发生与结束时间方面, 与经典元胞传输模型相比, 调整型元胞传输模型将时间预测准确性分别提升了22.3、10.8 min; 在模拟入口匝道汇入段主线拥堵传播与消散方面, 调整型元胞传输模型模拟结果更加符合实际的传播与消散规律; 在模拟试验路段早发性失效交通特性方面, 调整型元胞传输模型对于拥堵前最大流量与拥堵后消散流量的拟合误差在4%以内, 小于经典元胞传输模型; 在模型仿真精度方面, 调整型元胞传输模型各项评价指标均优于经典元胞传输模型, 前者的仿真速度误差为10.42 km·h-1, 较后者降低了25.4%;与传统的遗传算法相比, 混合多目标参数优化方法的总计算次数更少, 参数标定过程总耗时缩短了29.3%。 

Macroscopic traffic flow model of expressway on-ramp bottlenecks

Based on on-ramp merging mode and fundamental diagram form, an adjusted cell transmission model (CTM) was proposed. On-ramp state variables were introduced to track the traffic state of on-ramps, and new on-ramp merging rules were defined.The dual capacity fundamental diagram was introduced to the adjusted CTM in order to adapt to the varying capacities under different traffic conditions. The nelder-mead method and genetic algorithm were combined, and a hybrid multi-objective parameter optimization method was proposed. Three simulation scenarios were established, the performances of adjusted CTM and the hybrid multi-objective parameter optimization method were evaluated. Simulation result shows that for the prediction of the occurrence time and ending time of congestion on the upstream of on-ramp, compared with the original CTM, the adjusted CTM improves the accuracy by 22.3 and 10.8 min, respectively. For the simulation of the propagation and dissipation of congestion at the on-ramp merging section, the result of adjusted CTM is closer to the actual propagation/dissipation rules.As for the simulation of the early-onset breakdown traffic characteristic on the test segment, the fitting errors of adjusted CTM for the maximum pre-queue flow and queue discharge flow are below 4%, which are less than the values of original CTM. In the term of model simulation accuracy, compared with the original CTM, the various indexes of adjusted CTM are better, the simulated speed error of the former is 10.42 km·h-1, which is 25.4% lower than the value of the latter. Compared with the traditional genetic algorithm, the hybrid multi-objective parameter optimization method can reduce the total calculation times, and the total consumed time of parameter calibration shortens by 29.3%.