A two-stage optimal model is proposed to obtain the peak load shaving and the discharging energy in the vehicle-to-grid scenario. The Latin hypercube sampling method is adopted to sample the starting charging time and arrival state of charge to reduce the computation burden. The charging load of plug-in electric vehicles in 2020 and 2030 in China is simulated and the load profiles are compared with and without optimal control. Cost-benefit analysis in different scenarios are performed. Sensitivity analysis is performed to investigate the influence of several key parameters uncertainties on the cost and benefit of plug-in electric vehicle charging and discharging coordination. Simulation results show that the method proposed can effectively reduce the peak load and smooth the load curve. The installed cost and the battery life loss cost will significantly affect the capacity benefit of the charging and discharging control while the network power price will not.