Taking full advantages of the complementary characteristics of the wind power, the solar power and the energy storage devices, the wind, PV and energy storage (wind-PV-ES) co-generation system with reasonable capacity configuration can get stable power output, thus improving the power quality and significantly reducing the impact of power volatility on the grid. In the paper, based on historical data of wind and solar resources, the sum of the peak-vale slopes, power output fluctuation rates and active power deviation rates are chosen as evaluation indexes for the large-scale wind-PV-ES generation system. The probability distribution of the evaluation indexes and the probability change rate of the indexes in the range of fluctuation thresholds during the study period are analyzed by a statistical method. On this basis, an optimal capacity configuration method, which aims to smooth the output by maximizing the probability that the requirements for power change rate of co-generation system are satisfied, is put forward. Finally, the proposed method is verified by an example.