Abstract: Firstly variation rules of intersection capacity and vehicle delay in intersections after left-turning traffic control are analyzed.Then with the decision-making variables of whether each entrance lane of each intersection has left-turning traffic forbiddance and with the goal of minimizing the total travel time of network flows,the bi-level programming model for optimizing left-turning traffic forbiddance system of intersections within a certain areal network is constructed to make the saturation of roads and intersections under the congestion level.The method of using genetic algorithm for the model is discussed.A numerical example demonstrates that the left-turning traffic forbiddance form of each intersection can be determined by the optimization of this model,which can provide scientific foundations for urban left-turning traffic control and management.