The extensive progresses in computer science and communication technology in recent decades facilitate the development of the connected and automated vehicles (CAV). Since the emergence of the concept, the commercialization of CAV has been looked forward to providing an effective tool to the regulation of the freeway re-organizing traffic flow who normally initiate the evolvement of the congestion. To analyse the benefits of the CAV on traffic dispersion, the re-organizing traffic in the work zone and the incident-affected zone (under emergency services) were adopted as two cases of non-recurrent congestion, and the microscopic simulations were conducted on the basis of various car-following models and lane-change models. Furthermore, collaborative instances were added to the traditional traffic dynamic models to emulate the motions of the CAV. Trajectories data extracted from NGSIM open-access database were applied to calibrate the Bayes-classifier-based lane-change prediction model in order to better emulate the human drivers’ lane-change decision and to assist the CAV’s collaborations. With the increasing percentage of the CAV, the traffic congestion on the aforementioned bottlenecks were significantly mitigated. While CAV are proved to be capable of facilitate the cooperative lane-changes, they were also trained to refuse the lane-change request if there would be great impact on the target lanes. Although the lane-changes would inevitably impact the target lanes owing to the increasing densities and the disturbances during the lane-change motions, the simulation results showed that CAV are capable of minimizing the negative effects for the entire traffic system’s perspective.