The phenomenon of “tidal flow” during the morning peak and evening peak affects the operational efficiency of the urban traffic network. In order to relieve the phenomenon of unbalanced traffic flow and the waste of road resources, this paper proposes a bi-level programming model to optimize the deployment of reversible lanes from the point of traffic network. The upper-level model aims to find the optimal deployment of reversible lanes to minimize the total travel time of the traffic network, while the lower-level model is a network equilibrium model to assign the traffic flow. To address the proposed model, a genetic algorithm which incorporates Frank-Wolfe algorithm is designed. Numerical examples were conducted based on the Sioux Falls network. The results show that the total travel time will decrease greatly when deploying reversible lanes in the traffic network. Moreover, deploying more reversible lanes will play a great part in improving the operational performance of traffic network under an allowable investment. It is concluded that the proposed model can provide theoretical insights for traffic managers to improve the operational efficiency of traffic network.