Abstract: Aiming at the nonlinear characteristics of wind power, a combined prediction method for wind power is proposed, which combines gray relational analysis (GRA), improves sparrow algorithm (ISSA), support vector regression (SVR) and error correction model (EC). GRA is used to select the influential factors with a large degree of correlation with wind power as the input of the model. Adaptive weight factor and Levy flight strategy are introduced to improve the performance of traditional SSA, and ISSA-SVR model is established to obtain the initial predicted value. The error correction model is established to get the predicted error value. Finally, the initial predicted error value and the predicted error value are combined with the adder to get the final result. The simulation results show that the average determination coefficient of the model is 0.999 6 and 0.998 5, the average absolute error is 0.226 6 kW and 0.014 6 MW, and the root mean square error is 0.277 7 kW and 0.021 3 MW, respectively, when predicting the wind power of two wind farms. Compared with other traditional models, the prediction accuracy is higher.

Key words: wind power prediction, grey relation analysis, improved sparrow search algorithm, support vector regression, error-corrected model