Wind resources have the characteristics of strong fluctuation

randomness and discontinuity

which lead to the low accuracy of wind power forecasting. In order to reduce the impact of wind power fluctuations on the power grid and improve the ability of power systems to accept and absorb wind power

an improved wind power short-term prediction method and fluctuation based error correction method are proposed. Firstly

the wind power is divided into clusters according to different fluctuation processes. The characteristic curves of different fluctuations are extracted to correct the power values. Secondly

the back propagation neural network optimized by gravitational search algorithm (GSA-BP) is used as the basic prediction method to predict. Then the performance of forecasting errors under different fluctuations is analyzed

and the mapping relationship between forecasting errors and comprehensive meteorological indicators is established. A corresponding wind power error correction model is established for different fluctuation processes. A combination of linear model and GSA-BP nonlinear model is proposed to modify the prediction error. Finally

the power prediction value is added to the prediction error correction value as the final forecast result. The wind power prediction error correction method not only involves conventional factors such as wind speed and direction

but also takes into account the fluctuation of wind power.