Aiming at the infiltration of Connected Autonomous Vehicle (CAV) into the road network, the order evaluation was carried out on the different disturbance of traffic flow caused by different section types of machine-non-interaction situations. Taking Taian West Road in Nanxun as an example, the CAV penetration ratio was introduced, and the traffic flow data at peak times were statistically analyzed in combination with the different interaction situations caused by the section characteristics. A genetic algorithm was proposed to correct the simulation parameters of CAV following model, and a VISSIM simulation platform was built to evaluate the order from two perspectives of efficiency and safety. The results show that the increase of road width and permeability is beneficial to the operation of urban road network, and the non-physical isolation can effectively ensure the safety of road vehicles.