As one of the renewable energy power generation methods, wind power generation shows a high growth trend. However, while wind power is connected to the grid,the volatility and instability of wind power make the power system produce a lot of uncertain fluctuations, which greatly affects the power quality and jeopardizes the stable operation of the power system. Therefore, high wind speed forecasting accuracy can provide a solid basis for grid management,, improve the power system's ability to consume wind power, and ensure the safety and stabilization of the power system. In order to solve the problem of inaccurate prediction caused by the non-linearity and unsteadiness of wind speed series, this paper proposes a Fractal Ensemble Empirical Mode Decomposition (FEEMD)-Permutation Entropy (PE)-Sparrow Search Algorithm (SSA)-Error Back Propagation (BP) neural network method for short-term wind speed prediction. This method first uses the fractal dimension-based ensemble empirical modal decomposition (FEEMD) to decompose the original wind speed in order from high to low frequency; then calculates the entropy value of each component, and merges the components with similar entropy values to effectively reduce the computation; finally, the new sub-series are predicted by SSA-BP model, and the predicted value of the merged new sub-sequence is accumulated to obtain the final wind speed prediction result. The simulation study shows that the proposed prediction model is not only fast and accurate, but also suitable for short-term wind speed prediction.