In this paper, a mesoscopic dynamic model is utilized to depict the acceleration behaviour of congested traffic flow by overcoming the speed averaging drawback. The model is developed by both considering the oversaturation phenomenon and improving the computational efficiency on a previously proposed link model. Link exit function formulation, discretisation on time dimension, definition of capacity constraint rules for over-saturated states and uniformly accelerated speed assumption that allows a realistic representation of flow dynamics is made while setting out the model. Computation of link flows is performed regarding the acceleration of vehicles that validates the consistency of flow propagation with speed. In the presence of step-ups and step-downs on speed, adaptation of flow propagation is simultaneous, relative to the time lag defined to discretise time dimension. The iterative structure of the model enables convergence to any target performance criteria with the coded algorithm.