Bitte wählen Sie ihr Lieferland und ihre Kundengruppe
Research of wind power prediction has great significance for power balance and economic operation. This paper combined the ARMA model and the neural network model to establish a wavelet-ARMA model revised by neural network. After decomposing the historical wind power data by wavelet analysis, the high frequency components and low frequency components were separately predicted by ARMA. The reconstructed predictive outcomes were revised by BP neural network. To improve the accuracy, the predictive values of predictive moments were added to the BP neural network. This paper gave a 96 points rolling prediction for a single wind turbine. The accuracy of the results is 84.17%, and the pass rate is 87.50%. Compared to the single predictive methods, they are increased by 2%-4% and 3%-9% respectively in combination predictive model.