The purpose of this paper is to present a bi-level-based optimization model and develop a genetic algorithms (GA)-based method to solve the optimal toll design with elastic demand problem for congestion management and to determine the second-best linked-based optimal toll locations and toll levels simultaneously. The upper-level subprogram is to maximize the total social welfare given certain toll level constraints. The lower level subprogram is a traditional user equilibrium problem with elastic demand. The proposed GA model is applied to the Sioux Falls network, which has 76 links and 24 OD-pairs, assuming homogeneous users. Comprehensive numerical results including solutions achieved under continuous tolling and discrete tolling schemes, tolling on optimized links and tolling on heuristically selected most congested links are carefully presented and compared. The impact of value of time and the elastic demand sensitivity are also comprehensively investigated.