高速公路的运营和管理优化方案往往需要依据瓶颈段的通行能力制定。由于高速公路瓶颈段交通流特性的复杂性，目前所采用的瓶颈段通行能力通用计算公式受很多假设条件的限制，其精度不高且误差较大。文中尝试构建一种高速公路瓶颈段通行能力计算方法，并进行实测验证。首先，采用无人机航拍和Tracker软件动态识别，获取高速公路分合流区跟驰和换道数据共680组，得到位置、速度、加速度、车头间距等参数；然后，根据获得的交通流参数建立了考虑驾驶人特性的换道规则模型和概率模型，并对经典的GHR跟驰模型进行参数标定；最后，运用分区思想构建了基于元胞自动机理论的高速公路车流仿真应用模型。依据GHR跟驰模型来刻画车辆的加速度需求，并采用车辆换道次数和小时交通量两个指标验证模型的有效性，其中，平均换道次数误差率为12.06%，平均交通流量误差率为3.19%。结果表明，该模型可以有效地计算高速公路分合流区的通行能力，同时也可应用于变速车道长度等道路参数设计。在本研究案例中，对广州市北环高速公路岑村立交段进行交通流仿真测试，模拟其道路线型特征、车辆到达情况和交通流运作机制，得到该区段分流区与合流区的通行能力依次为5 456和5 253 pcu/h，该分流区与合流区变速车道长度的优化设计值分别是125和200 m。文中所构建的车流元胞模型可为高速公路分合流区的微观车流仿真和道路参数设计提供科学依据，有利于提升高速公路的服务水平和运营质量。

The operation and management optimization schemes of freeway always need to be formulated according to the traffic capacity of bottleneck sections. Due to the complexity of the traffic flow characteristics of the bottleneck sections of freeway, the current general calculation formulas of the traffic capacity are limited to many assumptions, and the accuracy is not high and the error is large. This paper attempted to construct a calculation method of traffic capacity in freeway bottleneck sections and verify it by actual measurement. Firstly, UAV technology and software Tracker were used for aerial video recording and dynamic identification and 680 groups of traffic flow data for car-following and lane-changing on freeway was obtained, including position, speed, acceleration, headway and other parameters. Then, the probability models and rule models of lane-changing considering driver characteristics were established according to the obtained traffic flow parameters, and the classical car-following model GHR was calibrated. Finally, a simulation application model of freeway traffic flow based on cellular automaton theory was constructed by the partitioning method, the acceleration requirements of following vehicles were characterized with GHR model, and the effectiveness of the model was verified by the indicators of vehicle lane-changing times and hourly traffic volume with error rates of 12.06% and 3.19%, respectively. The results show that this model can effectively calculate the capacity of freeway diverging and merging areas and design the road parameters such as the length of speed-change lane. In this case, a traffic flow simulation test was conducted on Cencun interchange sections of Guangzhou Northring Freeway, in which the road geometry characteristics, vehicle arrival conditions and traffic flow operation mechanism were simulated. The capacity of the diverging and merging areas is 5 456 and 5 253 pcu/h, respectively, and the optimization design values for the length of speed-change lanes in the diverging and merging areas are 125 and 200 m, respectively. The cellular automaton model of traffic flow constructed in this paper provides scientific basis for microscopic traffic flow simulation and road parameters design of freeway diverging and merging areas, and helps to improve the level of service and operational quality.