The wind power has the characteristics of intermittent, randomness and uncertainty. In order to make better use of wind power, many scholars use historical data to study the power fluctuation characteristics of wind power generation system, but as for the preparation or built-in wind farms, there are abundant wind speed data but without power data. So how to infer the electric power fluctuations from large numbers of wind speed data that is meaningful to the selection of wind farm related facilities. This paper decomposes the electric and wind power fluctuations in different frequency bands by using wavelet packet method with large numbers of wind speed and electric power data as samples. The rules of the electric and wind power changing with frequency are analyzed in detail and the transfer characteristics from wind power to electric power were drawn finally, which transform the electric power estimation into wind power estimation. This conclusion can provide some references for power fluctuations in preparation or built-in wind farms evaluation and the energy storage device selection to smooth the power fluctuations.