In a microgrid containing wind power,photovoltaic power,residential load,battery,supercapacitor and electric vehicles,a power smoothing strategy considering the energy storage loss and power prediction error is proposed to smoothen the grid-connected power fluctuation.Firstly,the day-ahead prediction data of grid-connected power were decomposed by wavelet packet,and the power fluctuation was smoothed by battery and ultracapacitor.The optimal regulation model was established to optimize the output of battery and ultracapacitor.Then,a rolling optimization scheduling model based on the model predictive control was established for the error of power day-ahead prediction data,and the error was eliminated by adjusting the orderly charging and discharging of electric vehicles.Finally,a real microgrid simulation model was constructed.The operation results meet the expectations,which shows that the proposed method can effectively reduce the over-charge and over-discharge of energy storage,and improve the ability of the grid-connected power tracking scheduling.