Key points are not available for this paper at this time.
As a kind of green, clean and renewable energy, wind power generation has been widely utilized in various countries in the world. With the rapid development of wind energy, it is also facing prominent problems. Because wind power generation is intermittent, unstable and stochastic, it has caused serious difficulties for power grid dispatch. At present, the important method to solve this problem is to predict wind speed and wind power. Grey model is suitable for uncertain systems with poor information and needs less operation data, so it can be used for wind speed and wind power prediction. However, the traditional grey system model has the disadvantage of low prediction accuracy. Therefore, firstly the GM (1,1) for wind speed prediction is improved by background value optimization in this paper. In order to comprehensively reveal the inherent uncertainty of wind speed random series, the fractional order grey system models with different orders are constructed. Secondly, in order to overcome the shortcoming of single grey model, each grey model is effectively united, and a combination prediction model based on neural network is presented. The two NWP outputs, i.e. ECMWF and GRAPES-MESO, have been added to the prediction model for reducing the uncertainty. The structure parameters of the neural network are optimized by trial and error. Thirdly, the support vector regression model is established to fit the scatter operation data of wind speed-power, and the parameters of the model are optimized by the particle swarm algorithm. Then the power prediction value is obtained by the fitted wind speed-power relationship and the corresponding result of the grey combination model for wind speed prediction. Finally, wind speed and wind power are predicted based on the actual operation data. In addition, the prediction model based on ARIMA is also constructed as a benchmark model. The results show that the proposed grey combination model improves the prediction accuracy.
Zhang et al. (Tue,) studied this question.