Large-scale electric vehicle (EV) random access to the power grid, the load peak-valley difference will become larger, seriously affect the stable operation of the power grid. In this paper, a V2G based bi-level optimal scheduling model for EV charging and discharging is proposed. The upper model takes the minimum variance of the total load as the objective function, and the lower model takes the increase of user participation and the maximization of user revenue as the objective function. The multi-population genetic algorithm is used to analyze the model, and the results show that the model can not only smooth the load fluctuation, effectively reduce the load peak-valley difference, but also maximize the economic benefits of users participating in V2G service.