This paper investigates the cost control problem of congestion management model in the real-time power systems. An improved optimal congestion cost model is built by introducing the congestion factor in dealing with the cases of opening the generator side and load side simultaneously. The problem of real-time congestion management is transformed to a nonlinear programming problem. While the transmission congestion is maximum and the adjustment cost is minimum based on the minimax idea and simulated annealing algorithm, and the global optimal solution is obtained. Simulation results show that the improved optimal model can obviously reduce the adjustment cost, and the designed algorithm is safe and easy to implement.