Aggregated traffic flow models based on the Macroscopic Fundamental Diagram (MFD), also known as multi-reservoir or multi-region MFD models, have been developed for more than a decade for various applications. While being very appealing for simulating traffic dynamics at a city level, the outputs of these models were rarely confronted with real data measurements. Thus, this paper focuses on calibration and validation of an MFD simulation for a city partitioned into multiple reservoirs. The traffic predictions from the MFD simulation (total accumulation and mean speed) are compared with real data from loop and probe sensors. The questions addressed in this study include the influence of the city partitioning, the MFD and average trip length estimation, and the path flow distribution among reservoirs. This study is carried on the network of Lyon, France, composed by around 27,000 links that extend over an urban area of 80 km². Two different partitioning cases are defined and compared, with respectively 5 and 10 reservoirs. Our results notably show that the proper estimation of three elements is critical for accurate traffic state prediction: (i) the total “active” network length of each reservoir, (ii) the regional trip lengths in the reservoirs, and (iii) the path flow distribution at the regional network level. While the network equilibrium found in the 5-reservoir partitioning can be roughly approximated with Wardrop’s principle, the 10-reservoir case is more complex and requires to design ad-hoc optimization process to derive regional path flow distributions that fit the data. The global equilibrium found in this latter case turns out to be hardly predictable with any traffic equilibrium principle.

基于宏观基本原理图（MFD）的聚合交通流模型（也称为多水库或多区域MFD模型）已经开发了十多年，可用于各种应用。尽管在模拟城市级别的交通动态方面非常吸引人，但这些模型的输出很少会遇到实际数据测量的情况。因此，本文着重于对划分为多个水库的城市进行MFD模拟的校准和验证。将来自MFD仿真的流量预测（总累积量和平均速度）与回路和探头传感器的真实数据进行比较。本研究中解决的问题包括城市划分的影响，MFD和平均行程长度估计以及水库之间的径流分布。这项研究是在里昂的网络上进行的，²。定义并比较了两种不同的划分情况，分别有5个和10个储层。我们的结果特别表明，对三个要素的正确估算对于准确的交通状态预测至关重要：（i）每个水库的总“活动”网络长度，（ii）水库中的区域行程长度，以及（iii）路径区域网络级别的流量分配。虽然用Wardrop的原理可以粗略地估计在5个储层分区中发现的网络平衡，但10个储层的情况更为复杂，需要设计临时优化过程以得出适合数据的区域路径流量分布。在后一种情况下发现的全球均衡，用任何交通均衡原理都很难预测。